Heat transfer fluid reclamation and maintenance prediction model

Abstract

Various embodiments are directed to estimating or predicting one or more reclamation treatment option recommendations for a heat transfer fluid system. Some embodiments use a computer model and user interface functionality for estimating and presenting a specific reclamation process to use for a given heat transfer fluid sample or customer. Additionally or alternatively, various embodiments estimate or predict other associated features, such as a fluid life expectancy (FLE), an amount of waste removed, or cost for using a particular reclamation treatment on a particular heat transfer fluid. Some embodiments additionally or alternatively predict or estimate how much of a heat transfer fluid is needed and the percentage of a heat transfer fluid system that is reclaimed or will be reclaimed given a particular reclamation treatment.

Description

HEAT TRANSFER FLUID RECLAMATION AND MAINTENANCE PREDICTION MODEL BACKGROUND

[0001] Heat transfer fluid (HTF) is a specialized fluid used in thermal systems to transfer heat from one location to another. It is commonly employed in industrial processes such as heating, cooling, and thermal energy storage. For instance, these fluids can act as an intermediary to receive or provide thermal energy to one portion of a process or system and transfer and / or store thermal energy to another portion of the process or system. HTFs are designed to have high thermal conductivity and stability over a wide temperature range.

[0002] When heat transfer fluids are not performing up to expectations, there can be detnmental impacts to the system, such as adverse effects to machine or equipment life, and / or system production rates. When exposed to thermal stress, all HTFs are subject to decomposition processes, which is reflected in a change in physical and chemical properties. While these decomposition processes are mainly triggered by thermal stress, other factors may additionally or alternatively influence the properties of HTF such as oxidation, contamination, radical polymerization, and / or molecular fracturing. The changes of these properties and others may negatively impact the heat transfer system so that actions must be taken to ensure the continuous operation of the heat transfer system.BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

[0003] The present technology7is described in detail below with reference to the attached drawing figures, wherein:

[0004] FIG. 1 is a block diagram of an illustrative system architecture, according to some embodiments;

[0005] FIG. 2 is a screenshot of an example user interface page illustrating different reclamation treatment option recommendations, according to some embodiments;

[0006] FIG. 3 is a screenshot of an example user interface page illustrating current fluid condition and predicted fluid condition given a particular reclamation treatment option, according to some embodiments;

[0007] FIG. 4 is a schematic diagram illustrating how different on-site processes of reclamation can be predicted or estimated, according to some embodiments;

[0008] FIG. 5 is a screenshot of an example user interface page illustrating parameters used to make reclamation predictions, according to some embodiments;

[0009] FIG. 6 depicts a diagram of one or more example neural networks that are trained to generate multiple decision statistics indicative of various predictions, according to some embodiments;

[0010] FIG. 7 is a flow diagram of an example process for generating a model prediction of at least one reclamation treatment option for a given heat transfer fluid sample, according to some embodiments;

[0011] FIG. 8 is a flow diagram of an example process for training a machine learning model to predict a reclamation treatment option, according to some embodiments;

[0012] FIG. 9 is a block diagram of a computing environment in which aspects of the present technology7are implemented within, according to some embodiments; and

[0013] FIG. 10 is a block diagram of a computing device in which aspects of the present disclosure are implemented within, according to some embodiments.DETAILED DESCRIPTION

[0014] The subject matter of the present invention is described with specificity herein to meet statutory7requirements. However, the description itself is not intended to limit the scope of this patent. Rather, the inventors have contemplated that the claimed subject matter might also be embodied in other ways, to include different steps or combinations of steps similar to the ones described in this document, in conjunction with other present or future technologies. Moreover, although the terms “step” and / or “block” may be used herein to connote different components of methods employed, theterms should not be interpreted as implying any particular order among or between various steps herein disclosed unless and except when the order of individual steps is explicitly described.

[0015] As described above, changes of HTF properties and others may negatively impact the HTF system. Accordingly, HTF maintenance is a critical aspect of HTF system operations. Two main service solutions exist for such maintenance. The first solution is a complete or substantial system refill. A system refill refers to the process of draining and disposing of the entire volume (or a substantial portion, typically between 20-100%) of the spent HTF in a system and replacing it with new fluid. This is typically done when the HTF has thermally degraded to the point where its performance is significantly impaired. However, there are various disadvantages to this solution. First, a system refill typically requires shutting down the HTF system to drain the old fluid and refill it with new fluid. This downtime can be very costly for industrial operations that rely on continuous processes, which not only leads to lost productivity or revenue, but fault machinery and equipment. For example, the process of cooling down and then reheating equipment for a complete system refill induces thermal cycling, which stresses materials and can lead to cracks, wear, or failure in components designed for continuous operation. In another example, restarting machinery after a shutdown can cause mechanical and operational issues. Components may seize up or fail to restart smoothly, leading to additional maintenance needs and potential damage. In yet another example, during the draining and refilling process, there is a high risk of introducing contaminants into the system, which affect the quality and performance of the new fluid and potentially damage sensitive components.

[0016] Moreover, any residual contaminants left in the system from the old fluid can mix with the new fluid, potentially leading to quicker degradation and reducing the overall efficacy of the refill. Disposing of large volumes of spent HTF also contributes to environmental waste, which must be managed according to regulator^' standards. This process has environmental implications and increases the carbon footprint of the operation. Producing new HTF also involves energy and raw materials, adding to the environmental impact. Frequent refills mean more frequent consumption of these resources.

[0017] Another service solution for HTF maintenance is reclamation. In the context of heat transfer fluid maintenance, "reclamation" refers to the process of restoring or rejuvenating used HTF to extend its useful life and maintain its performance characteristics. Reclamation involves treating the used fluid to remove contaminants, restore its thermal properties, and improve its overall condition. Reclamation can include methods such as fdtration, chemical treatment, distillation, and other purification techniques. There are various advantages of reclamation over system refills. Not only is reclaiming fluid generally less expensive than purchasing new fluid for refilling, but there are other various technical and operational advantages of reclamation. For example, avoiding the shutdow n and restart cycles associated with a complete system refill reduces thermal and mechanical stress on equipment, prolonging its life. Reclamation processes, especially mobile or on-site reclamation, can often be performed without shutting down the system. This allows continuous operation and avoids the costly downtime associated with a complete system refill. Consequently, there is no or less thermal cycling, which reduces the stress of materials so there are no cracks, wear, or failure in components designed for continuous operation. There is no need to restart machinery with mobile reclamation, which means there is less likely to be mechanical and operational issues. Components may be less likely to seize up or fail to restart so there is no additional maintenance required and less potential damage. In yet another example, there is a lower risk of introducing contaminants into the system, since there is no refill, which keeps the quality and performance of the fluid high if reclaimed appropriately. Reclamation processes can be controlled to ensure that contaminants are effectively removed, whereas complete refills can introduce new contaminants if not carefully managed.

[0018] Moreover, reclamation can typically be performed more quickly than a complete system refill, which involves draining, cleaning, and refilling the system. This results in less disruption to operations. There are also various environmental benefits. Reclaiming and reusing HTF reduces the amount of w aste fluid that must be disposed of, which is beneficial for the environment. By extending the life of the existing fluid, reclamation reduces the need for new fluid production, conserving raw materials andenergy used in manufacturing. Reclamation reduces the risks associated with handling and disposing of large volumes of hazardous waste, improving workplace safety.

[0019] However, not all reclamation treatment options are the same and they do not offer the same technical benefits. For example, a first type of reclamation -stationary reprocessing -requires the system’s fluid to be pumped out of the heat transfer system and transported to the stationary reprocessing system. Hence, the production system cannot continue to be operated for the reprocessing period. Ideally, this method is useful in times the system cannot be operated as a result of maintenance or revision measures, which requires extensive preliminary logistical planning. Stationary' systems have a higher efficiency and usually deliver a higher quality of the reprocessed heat transfer fluid, which can lead to a longer lifetime of the filling. In addition, the time that is needed to reprocess the fluid could be quite short if the right stationary system is used for the reprocessing procedure. These are the benefits compared to mobile reclamation.

[0020] Mobile Reclamation (MR) offers a number of advantages in terms of handling, since the heat transfer medium can be processed on site using the bypass method, which eliminates the need for transporting the used HTF to a separate reclamation site. This method can be used while a facility (e.g., a manufacturing plant) is in operation and does not interrupt the ongoing production process, which can lead to a significant reduction in costs. The disadvantage is that mobile systems do not have the same performance that a stationary system can offer. This can limit the reprocessing quality. In addition, in some instances the reprocessing takes place in batches, and the reprocessed product is fed back into the plant with the product that is still to be processed, so that the reprocessing is relatively slow and intensive of energy consumption. Also, a higher volume of refreshing fluid may be required. Mobile reclamation should be done if economic and technical possibilities are more beneficial than partial or full replacements. Existing technologies fail to accurately estimate which reclamation treatment options are the best to use for a given customer or heat transfer fluid, and they do not estimate post-treatment fluid quality and performance. Depending on the method (e.g., temperature ranges and pressures) used during either stationary ormobile reclamation process, the resulting HTF fluid can vary greatly in lifetime and fluid properties making the resulting lifetime prediction complicated.

[0021] Various embodiments of the present disclosure provide one or more technical solutions that have technical effects in light of the technical problems as described above, as well as other problems, as described herein. Specifically, various embodiments generally leverage the reclamation fluid maintenance option by using a computer model and user interface functionality for predicting or estimating a specific reclamation process to use for a given sample or customer, which is an alternative to complete or partial system refill with anew fluid. Additionally or alternatively, various embodiments estimate post-reclamation treatment fluid quality and performance. Such functionality allows for comparison against various reclamation fluid treatment options in order to choose which is best for the customer system.

[0022] Various embodiments have the technical effect of improved Human-Computer Interaction (HCI). Various embodiments include user-friendly interfaces (as illustrated in FIG. 2 and FIG. 3) that present complex data in a clear and accessible manner. These interfaces allow users to easily compare different reclamation fluid treatment options, understand current and predicted fluid conditions, and make informed decisions efficiently since all of the information is summarized at a single user interface page. This enhances human-computer interaction by providing intuitive visual representations and interactive elements that facilitate user engagement and comprehension. In other words, the interfaces, such as graphical user interfaces (GUI) simplify the complex task of HTF monitoring and maintenance. The interface is designed with user experience (UX) principles in mind, ensuring that users, regardless of their technical expertise, can easily navigate the system. Some features include Visual Dashboards: Real-time data visualization through graphs, charts, and trend lines helps users quickly understand fluid conditions and performance metrics. In another example, these interfaces allow for increased user interaction performance. For example, as illustrated in FIG. 2 and 3, the arrangement of dials (e.g., the dial 204 of FIG. 2) and an entire map of reclamation process prediction (e.g., FIG. 4) result in more efficient user interaction, such as less drilling, paging, clicking, and less data input. This is because the user can, for example, look at a dial to quickly derive a HTF lifeexpectancy, cost, and fluid condition, which would otherwise require additional drilling to different pages of a user interface of existing technologies or manual data entry, such as various equations into a spreadsheet. Moreover, user interfaces, such as those illustrated in FIG. 2, FIG. 3, and FIG. 4 allow clear navigation in that there are well-organized menus and easy-to-find options reduce the cognitive load on users, making it straightforward to access different functionalities.

[0023] Another technical effect is reduced burden of user input and input / output (I / O) to a computing device. Various embodiments reduce the burden of user input by automatically pulling required HTF system data (e.g., historical customer data, including historical reclamation treatment options, and / or pre-computed degradation models) from a database or other data store. Existing technologies require users to input extensive data manually themselves (e.g., to a spreadsheet). This manual input process typically involves users entering various parameters, measurements, and historical data related to the HTF system into the system interface. However, various embodiments do not require users to manually input extensive data. Instead, they rely on the system or model to fetch and utilize relevant historical and / or real-time data from such data store. In other words, one technical solution is the use of a model that uses an existing data store (e.g., training data). The technical effect is that instead of writing or accessing a disk or other storage device several times during I / O (which places unnecessary wear and tear on the storage device) due to users having to repetitively and manually enter in spreadsheet information, particular embodiments reach out to the storage device a single time (or fewer times) during model inference / prediction. This automation streamlines the user experience and minimizes the potential for input errors. The use of pre-built prediction models means users do not need to manually calculate or enter detailed parameters for each maintenance scenario. The system generates reclamation treatment options and predictions based on available data, which reduces the user's input requirements. For example, a user need only input an identifier, such as a HTF sample ID and / or a customer ID. Various embodiments may then map such ID to the customer's sample, which is then analyzed using the model and historical data to predict viable reclamation treatment options for the customer’s HTF system.

[0024] Various embodiments also have the technical effect of reduced processor load and increased processing speed. This is because various embodiments use the model and data source described above (e.g., historical customer data 114, degradation formula(s) 110, laboratory analysis data 112, and / or reclamation service provider data 116 of FIG. 1). This reduces the need for intensive real-time processing because much of the heavy computational lifting has already been done (e.g., during training). When a user queries the system, for example, the response is generated using pre-built models and existing data, leading to a lighter processor load and faster processing times. By leveraging pre-built prediction models, the system quickly generates fluid maintenance reclamation options. This eliminates the need for complex, multi-step calculations each time a prediction is required, thus speeding up the processing. These pre-computed models essentially act as shortcuts or templates that encapsulate the results of intensive computational processes performed in advance. When a user queries the system or otherwise requests a reclamation treatment option recommendation, various embodiments quickly retrieve and apply these pre-computed models, significantly reducing the need for intensive real-time processing. As a result, the processor load is reduced because the system does not have to perform the heavy computational lifting from scratch for each user query. Furthermore, since much of the processing work has already been done beforehand, the response time is faster because the system can generate responses using pre-built models and existing data. For example, in some embodiments, these prediction models are based on historical data, statistical analysis, and machine learning algorithms that have been trained to predict various outcomes related to fluid quality and reclamation treatment options. By using pre-built prediction models, the system eliminates the need for complex, multi-step calculations or analyses each time a prediction is required. Instead, the system can apply these pre-built models directly to the user's query, resulting in faster processing times.

[0025] Another technical effect is energy savings via efficient system operation. Various embodiments recommend or estimate viable or optimal reclamation treatment options, which means that operators will reclaim the HTF. Reclaiming and reusing HTF avoids the energy-intensive processes of producing new fluid and disposing of spent fluid. This reduction in the frequency of fluid replacement cycles translates to energysavings not only in the production of new fluids but also in the logistics and operational disruptions associated with complete system refdls.

[0026] Other technical effects include simplified software development and reduced hardware requirements. The model’s design to integrate seamlessly with the data store (e.g., of historical customer data and / or degradation models) as described above simplifies software development. Developers can build on the existing infrastructure without needing to create new systems from scratch, reducing both the complexity and time required for development. For example, in some embodiments the existing infrastructure includes standardized data formats, established APIs (Application Programming Interfaces), and well-defined workflows, which streamline the development process. In this way, developers can focus their efforts on enhancing or extending the existing infrastructure, rather than dealing with the intricacies of building everything from scratch. For example, in some embodiments, historical customer data and fluid analysis results are stored in a structured format such as JSON (JavaScript Object Notation) or XML (extensible Markup Language). Developers leverage these standardized data formats to ensure consistency and compati bility across different components of the system. They do not need to spend time designing new data schemas or formats from scratch. This standardization facilitates seamless integration between various modules or components of the reclamation prediction system, allowing data to be easily shared and processed.

[0027] In another example, the existing infrastructure in some embodiments provides established APIs for accessing and manipulating data stored in the system. These APIs define clear interfaces and protocols for interacting with different functionalities or sen ices. For instance, the Fluid Genius™ online database offers APIs for querying historical fluid analysis data, retrieving degradation models, or submitting new samples for analysis. Developers can utilize these APIs to retrieve relevant data, perform calculations, and generate predictions without needing to develop the underlying logic or functionality themselves. By leveraging established APIs, developers can focus on implementing higher-level prediction algorithms or refining the user interface, rather than worrying about the intricacies of data retrieval or processing. With respect to workflows, there are predefined workflows for collectingfluid samples, conducting laboratory7analysis, generating degradation models, and recommending maintenance actions based on the results. Developers can build upon these well-defined workflows to streamline the development of the reclamation prediction system. They can map their new features or enhancements onto existing workflows, ensuring consistency and coherence with established practices. This alignment with existing workflows helps maintain usability7and familiarity7for users, as they can follow familiar procedures when interacting with the system.

[0028] Moreover, some embodiments (e.g., the user interfaces in FIG. 2 and FIG. 3) employ unified user interfaces, which consolidates various functionalities — such as fluid condition monitoring, maintenance option comparisons, reclamation option recommendation, and cost analysis — into one platform. This simplifies the overall software architecture and reduces the need for multiple hardware components to support different functions.

[0029] Another technical effect is enhanced reliability7and reduced error rate relative to existing technologies. As described above, existing technologies require arduous manual user input into, for example, a spreadsheet to incorporate historical data and perform basic calculations. However, these technologies do not utilize models for reclamation predictions. Various embodiments use a data store (e.g., a large historical dataset) that contains information about past fluid analyses, reclamation treatments, and / or their outcomes. For example, in a machine learning context, various past samples are labeled with an identifier that indicates the reclamation treatment for that sample and its post-treatment condition, which indicates whether the reclamation treatment option was successful. Robust statistical methods, such as machine learning algorithms or regression models, are applied to this dataset in some embodiments to identify patterns, correlations, and predictive factors related to fluid quality and reclamation effectiveness. By analyzing a diverse range of historical data points, various embodiments train on or leam complex relationships between various parameters and accurately predict the outcome of different reclamation treatments. For example, the system may identify that certain combinations of fluid properties, such as viscosity, acidity, high boilers, and carbon residue, are strongly correlated with the success ofspecific reclamation techniques. Based on this insight, it can recommend the most suitable treatment option for a given set of fluid conditions.

[0030] In some embodiments, the predictive system continuously incorporates the most recent in-service sample data from HTF systems into its predictive models. By integrating real-time data, such as fluid analysis results and operational parameters, the system ensures that predictions and maintenance recommendations are based on the latest information available. This real-time data integration reduces the likelihood of errors caused by outdated or inaccurate data inputs. It allows the system to adapt to changing conditions and provide timely, relevant guidance to users. For instance, if a sudden change in fluid viscosity is detected during routine monitoring, the system can promptly adjust its predictions and recommend appropriate maintenance actions, such as adjusting reclamation parameters or scheduling additional sample analysis.

[0031] Consider a scenario where the predictive system is tasked with recommending the optimal reclamation treatment for an HTF system experiencing a gradual increase in viscosity and acidity over time. Here is how error rate reduction is achieved in this context: the system leverages historical data on similar HTF systems with varying levels of viscosity and acidity. Using advanced statistical methods, such as regression analysis or neural networks, it constructs predictive models that map fluid properties to reclamation outcomes. As new in-service samples are collected and analyzed, the system updates its predictive models with the latest data. If the viscosity and acidity levels of the current sample deviate significantly from historical trends, the system identifies this anomaly and adjusts its predictions accordingly.

[0032] To further reduce the risk of errors, some embodiments incorporate error mitigation strategies, such as outlier detection and data validation checks. If any inconsistencies or anomalies are detected in the input data, the system flags them for review by a human operator or applies corrective measures automatically. By combining accurate prediction modeling with real-time data integration and / or error mitigation strategies, the predictive system minimizes the likelihood of errors in recommending reclamation treatment options. This enhanced reliability instills confidence in customers and encourages adoption of the reclamation process as a cost-effective alternative to complete refills.

[0033] Overall, these technical effects collectively enhance the efficiency, reliability, and user-friendliness of HTF system maintenance, offering significant improvements over traditional fluid replacement technologies. Various embodiments not only optimize operational processes but also provides a robust framework for longterm maintenance planning and decision-making.

[0034] FIG. 1 is a block diagram of an illustrative system architecture 100 in which some embodiments of the present technology are employed. Although the system 100 is illustrated as including specific component types associated with a particular quantity, it is understood that alternatively or additionally other component types may exist at any particular quantity. In some embodiments, one or more components may also be combined. It is also understood that each component or module can be located on the same or different host computing devices. For example, in some embodiments, some or each of the components within the system 100 are distributed across a cloud computing system (e.g., the computer environment 900 of FIG. 9). In other embodiments, the system 100 is located at a single host or computing device (e.g., the computing device 1000 of FIG. 10). In some embodiments, the system 100 illustrates executable program code such that all of the illustrated components and data structures are linked in preparation to be executed at run-time.

[0035] System 100 is not intended to be limiting and represents only one example of a suitable computing system architecture. Other arrangements and elements can be used in addition to or instead of those shown, and some elements may be omitted altogether for the sake of clarity. Further, many of the elements described herein are functional entities that may be implemented as discrete or distributed components or in conjunction with other components, and in any suitable combination and location. For instance, the functionality of system 100 may be provided via a software as a service (SAAS) model, e g., a cloud and / or web-based senice. In other embodiments, the functionalities of system 100 may be implemented via a client / server architecture. In some embodiments, the components of the system 100 are communicatively coupled via one or more networks. The network(s) can be any suitable network, such as a Local Area Network (LAN), a Wide Area Network (WAN), the internet, or a combination of these, and / or include wired, wireless, or fiber optic connections. In general, network(s)can represent any combination of connections (e.g., APIs or linkers) or protocols that will support communications between the components of the system 100.

[0036] The system 100 is generally directed to heat transfer fluid reclamation and maintenance prediction via one or more reclamation models 108. At a first time, heat transfer fluid sample data 112 is received or determined. The heat transfer fluid sample data 102 represents raw data from sensors, lab reports, and / or historical records. In some embodiments, this includes temperature, pressure, flow rates, chemical composition, and / or lab analysis results. The data 102 is received or determined according to one or more suitable methods. There are multiple ways heat transfer fluid (HTF) data and parameters can be received and analyzed. For example, traditional sample analysis may be engaged in. In these aspects, a customer sends a HTF sample to a lab. The customer first collects a sample of the HTF from their system using a clean, sealed container or bag. The sample is then shipped to a laboratory specializing in HTF analysis. At the lab, the sample undergoes a series of tests to measure various parameters such as viscosity, acidity, contaminant levels, boiling points, and carbon residue. The results from the lab are uploaded to a central database or other dataset represented by 102. In an illustrative example, a customer collects an HTF sample from their system and sends it to the lab. Different sensors and equipment lab performs a detailed analysis, measuring viscosity, boiling points, contaminant levels, and / or other parameters. The lab uploads the results to the Fluid Genius™ database, which the reclamation model(s) 108 accesses to provide re recommendations.

[0037] In an illustrative example of what the different sensors and equipment can be used, viscometers and rheometers are used. For instance, a capillary viscometer measures the time it takes for a HTF to flow through a capillary tube under the influence of gravity. The viscosity is calculated based on this flow time. A rotational viscometer measures the torque required to rotate a spindle at a constant speed while immersed in the fluid. The resistance to rotation correlates with the viscosity7. A falling ball viscometer measures the time it takes for a ball to fall through the fluid under gravity. The viscosity is determined based on the ball's descent time. In another example, boiling point measurements can be taken by Distillation Apparatus and Thermogravimetric Analyzer (TGA). The distillation apparatus separates componentsof the fluid based on their boiling points. The fluid is heated, and the temperature at which each fraction evaporates is recorded. TG measures the weight loss of a fluid sample as it is heated. The boiling points of the various components can be identified based on the temperatures at which significant weight losses occur. For instance, an ASTM D86 Distillation Apparatus is used to determine the boiling range characteristics of petroleum products, including HTFs, by distilling the sample and recording the temperatures at which various fractions are collected.

[0038] With respect to low boilers and high boilers measurement. Gas Chromatography (GC) is utilized in some embodiments. Gas Chromatography (GC) separates the components of the HTF based on their volatility. Low boilers (more volatile components) and high boilers (less volatile components) are detected and quantified as they elute from the GC column at different times. Mass Spectrometry (MS) is often coupled with GC (GC-MS) to provide detailed molecular identification of the components based on their mass-to-charge ratio.

[0039] Additionally or alternatively, on-site sampling and analysis may be performed to obtain the heat transfer fluid sample data 102. Technicians take samples during maintenance. For example, field sampling may be performed where during routine maintenance or a scheduled visit, technicians collect samples directly from the HTF sy stem on-site. Technicians may use portable lab equipment to perform initial tests and analyses on-site. These can include viscosity’ measurements, pH testing, and contaminant checks. Results are recorded and can either be uploaded immediately to the central database or data structure represented by 102 or analyzed further in a more comprehensive lab setting.

[0040] Additionally or alternatively, HTF fluid sample / data 102 represents realtime (or near real-time) sensor data. Sensors are installed at key points in the HTF system to continuously monitor parameters such as temperature, pressure, flow- rate, viscosity, and / or chemical composition. For example, Temperature Sensor Resistance Temperature Detectors (RTDs) are used to measures temperature by correlating the resistance of the RTD element with temperature. In another example, Type K Thermocouple are used to measure temperature based on the voltage generated at the junction of two different metals. Widely used for their wide temperature range and fastresponse time. Pressure sensors, such as strain gauge pressure transducers measures pressure by converting the deformation (strain) of a diaphragm into an electrical signal. Commonly used for high-precision pressure measurements. In another example, turbine flow meters measure flow rate by detecting the rotational speed of a turbine placed in the fluid flow. Viscosity sensors are also used in some embodiments to measure viscosity by detecting the damping effect of the fluid on a vibrating element (e.g., a tuning fork or a vibrating rod). In another example, Infrared (IR) Spectroscopy Sensors are used to measure the chemical composition of the fluid by analyzing the absorption of infrared light at different wavelengths, which is used for real-time monitoring of fluid contaminants and chemical properties.

[0041] These sensors transmit data in real-time to a central processing unit or directly to a cloud-based database. The data can be analyzed in real-time by the reclamation model(s) 108, providing immediate feedback and allowing for dynamic adjustments to the system if needed. Additionally or alternatively, hybrid approaches may exist, such as a combination of real-time sensors and periodic sampling. In these embodiments, a detailed baseline analysis is performed using samples sent to a lab, providing comprehensive data on the HTF’s condition. Real-time sensors continuously monitor the HTF and send data to the central system. Regular samples are still sent to the lab for detailed analysis, ensuring the accuracy of sensor data and catching any issues that sensors might miss.

[0042] The preprocessing module 104 takes the heat transfer fluid sample data 102 as input and preprocesses and / or otherwise converts such raw data. For example, in some embodiments, the preprocessing module 104 removes noise, handles missing values, and corrects any inconsistencies in the data. In some embodiments, this involves techniques like interpolation, imputation, or deletion of invalid entries. Some embodiments additionally or alternatively normalize the data by standardizing the data to ensure all features have a similar scale, which helps in the efficiency and performance of downstream algorithms. Normalization adjusts the range of the data to a standard scale, often [0, 1] or [-1, 1], Methods include min-max normalization and Z-score normalization. Standardization (e.g., Z-score Normalization) transforms data to have a mean of 0 and a standard deviation of 1. This is particularly useful for algorithms thatassume normally distributed data. Log transformation applies the logarithm function to reduce the impact of outliers and compress the range of the data. Quantization converts continuous values into discrete bins, reducing the precision of the data.

[0043] The encoder 106 takes the preprocessed data 102 (i.e., the output of the preprocessing module 104) as input and encodes such data into a machine-readable format. For example, in some embodiments the encoder 104 performs numeric encoding, such as binary encoding that represents data using binary numbers (0s and Is), commonly used for digital data. In another example, integer encoding maps categorical data to integer values. For example, mapping "cat" to 1, "dog" to 2, etc. In another example, in some embodiments, the encoder 106 performs one-hot encoding, which represents categorical data as binary vectors where only one element is "hot" (1), and the rest are "cold" (0). One-hot encoding transforms categorical values (e.g., particular words or tokens in 102) into a binary vector representation. Each category is represented by a unique binary7vector with all elements set to 0 except for one element, which is set to 1. This encoding ensures that the categorical data is represented in a way that can be fed into machine learning algorithms without implying any ordinal relationship between categories.

[0044] In some embodiments, the encoder 106 alternatively or additionally represent machine learning model encoders, such as transformer encoders. Transformers are a type of deep learning model used primarily for natural language processing (NLP). They use self-attention mechanisms to weigh the influence of different parts of the input data 102 in some embodiments. The encoder part of a transformer takes the input data 102 and processes it through multiple layers of selfattention and feedforward neural networks. Each layer includes a Self-Attention Mechanism, which computes the importance of each word in the context of all other words in the sequence. For example, specific parameters may (e.g., high boilers and low boilers and their values) may be determined to be more important for reclamation option predictions at 118 and so these words are given higher weights for attention. A feedforward neural network then applies a series of transformations to the data to extract higher-level features. Positional encoding of transformers adds informationabout the position of words in the sequence to the embeddings, as the self-attention mechanism does not inherently capture positional information.

[0045] Other encoding techniques include label encoding (maps categorical labels to integer values. Useful for ordinal data where there is a natural order), hashing (e.g., feature hashing maps features to a fixed-size vector using a hash function. Useful for handling large datasets with high-dimensional feature spaces. Other encoding techniques include the use of embeddings, such as word embeddings. Word embeddings represent words as dense vectors in a continuous vector space, capturing semantic relationships. Examples include Word2Vec, GloVe, and FastText. In deep learning models, embeddings can be learned (learned embeddings) as part of the training process. These embeddings capture relationships and structures in the data.

[0046] In some embodiments, the encoder 106 performs feature engineering, which creates new features or parameters from existing data that can help in analysis. For example, creating a feature that represents the rate of change of viscosity over time. A “parameter” as described herein refers to a particular attribute, feature, element, or other characteristic that makes up. is measured from, or is otherwise associated with a heat transfer fluid. For example, a parameter can be or include a high boiler, low boiler, moisture control, acid number, or other attributes. Specific parameters are described in more detail below. A “parameter value” (also described herein as “values” associated with parameters) represent the particular numbers or specific parameter information. For example, a particular heat transfer fluid can include an “acid number” parameter with a value of 0.50 mg KOH / g. Particular parameter values are associated with a particular reclamation treatment option, post-treatment quality, waste disposal prediction, and / or life expectancy of the reclaimed HTF, as described in more detail below.

[0047] In some embodiments, the after the encoder 1 6 encodes the data, feature extraction techniques are performed. For example, statistical methods, such as calculating statistical properties of the data such as mean, median, variance, and standard deviation can be performed.

[0048] Additionally or alternatively. Principal Component Analysis (PCA) can be performed to reduce the dimensionality of the data while retaining the mostimportant features, helping to eliminate noise and redundant information. In an illustrative example of extracting Specific Low Boilers and High Boilers parameters, Gas Chromatography (GC) Data Processing can be performed. Various embodiments first obtain chromatogram data from GC analysis, which shows peaks corresponding to different boiling components. Various embodiments then identify and quantify peaks corresponding to low boilers (early eluting compounds) and high boilers (late eluting compounds). A feature extractor then Extracts features or parameters such as peak area, peak height, retention time, and the ratio of peak areas of low boilers to high boilers.

[0049] One or more reclamation models 108 then take, as input, the encoded heat transfer fluid sample data 102 (i.e., the output of the encoder 106), and data from the degradation formula(s) 110, laboratory’ analysis data 112, historical customer data 114, and / or reclamation service provider data 116 to generate vanous decision statistics (e.g., classifications, regression scores, etc.), such as the reclamation treatment option(s) 118, the post-treatment fluid qualify 122, the waste disposal 124, and / or the life expectancy 120. As described above and in some embodiments, computing devices that host the degradation formula(s) 110, the laboratory analysis data 112, historical customer data 114, and reclamation service provider data 116 include established APIs so that the reclamation model(s) 108 can manipulate the data stored in the system. These APIs define clear interfaces and protocols for interacting with different functionalities or services. For instance, the Fluid Genius™ online database offers APIs for querying historical fluid analysis data, retrieving degradation model(s) 110, or submitting new samples for analysis. Developers can utilize these APIs to retrieve relevant data, perform calculations, and generate predictions without needing to develop the underlying logic or functionality themselves.

[0050] In some embodiments, the one or more reclamation model(s) 108 represents any suitable model functionality, such as supervised learning (e g., using logistic regression, using back propagation neural networks, using random forests, decision trees, etc.), unsupervised learning (e.g., using an Apriori algorithm, using K-means clustering), semi-supervised learning, reinforcement learning (e.g., using a Q-leaming algorithm, using temporal difference learning), a regression algorithm (e.g., ordinary least squares, logistic regression, stepwise regression, multivariate adaptiveregression splines, locally estimated scatterplot smoothing, etc.), an instance-based method (e.g.. k-nearest neighbor, learning vector quantization, self-organizing map, etc.), a regularization method (e.g., ridge regression, least absolute shrinkage and selection operator, elastic net, etc.), a decision tree learning method (e.g., classification and regression tree, iterative dichotomiser 3, C4.5, chi-squared automatic interaction detection, decision stump, random forest, multivariate adaptive regression splines, gradient boosting machines, etc.), a Bayesian method (e.g., naive Bayes, averaged one-dependence estimators, Bayesian belief network, etc.), a kernel method (e.g., a support vector machine, a radial basis function, a linear discriminate analysis, etc.), a clustering method (e.g., k-means clustering, expectation maximization, etc.), an associated rule learning algorithm (e.g., an Apriori algorithm, an Eclat algorithm, etc.), an artificial neural network model (e.g., a Perceptron method, a back-propagation method, a Hopfield network method, a self-organizing map method, a learning vector quantization method, etc.), a deep learning algorithm (e.g., a restricted Boltzmann machine, a deep belief network method, a convolution network method, a stacked auto-encoder method, etc.), a dimensionality reduction method (e.g., principal component analysis, partial lest squares regression, Sammon mapping, multidimensional scaling, projection pursuit, etc.), an ensemble method (e.g., boosting, bootstrapped aggregation, AdaBoost, stacked generalization, gradient boosting machine method, random forest method, etc.), a Large Language Model (LLM), a transformer, a Small Language Model, and / or any suitable form of machine learning algorithm.

[0051] The reclamation treatment option(s) 118 is a decision statistic representative of one or more reclamation treatment recommendations that a user or operator should take when reclaiming their HTF. For example, one option may be filtration. Filtration is used to remove solid particles and sludge from the HTF. This process typically employs mechanical filters with varying pore sizes to capture contaminants. For instance, coarse filtration removes larger particles and debris using filters with larger pore sizes (e.g., 50-100 microns). Fine filtration utilizes finer filters (e.g., 5-25 microns) to remove smaller particles and sludge. In another example, centrifugation involves using centrifugal force to separate contaminants based on their density. Heavier particles and sludge are forced to the outer edge, while cleaner fluidremains in the center. For instance, batch centrifugation processes a batch of HTF at a time, suitable for smaller systems or periodic maintenance. In another example, vacuum dehydration is another reclamation treatment option that removes water and low-boiling contaminants from the HTF by lowering the pressure to create a vacuum. This reduces the boiling point of water and other volatile compounds, allowing them to be evaporated and removed. Other options additionally or alternatively include distillation (separates components of the HTF based on their boiling points. The fluid is heated, causing lower boiling point components to evaporate and be collected separately from the higher boiling point components), adsorption (involves passing the HTF through a bed of adsorbent material (e.g., activated carbon, silica gel) that captures and removes contaminants through surface adhesion), chemical treatment (uses specific chemicals to neutralize acids, precipitate out contaminants, or stabilize the HTF), electrostatic separation (uses an electric field to attract and remove charged particles and contaminants from the HTF), and / or the like. In some embodiments, a combination of the above methods is used to achieve optimal fluid reclamation. For example, filtration might be combined with vacuum dehydration and adsorption to comprehensively remove various types of contaminants. For example, multi-stage filtration and dehydration uses sequential stages of coarse and fine filtration followed by vacuum dehydration.

[0052] In an illustrative example, an HTF system shows increased viscosity and the presence of sludge, moisture, and acidic degradation products. The reclamation treatment option(s) 118 may thus include the following: coarse filtration (to remove the larger sludge particles), vacuum dehydration (to remove moisture and low-boiling components), chemical treatment (to neutralize acidic components), and activated carbon adsorption (to remove organic contaminants and stabilizes the fluid). The HTF is restored to near-original properties, improving its performance and extending its useful life.

[0053] The laboratory analysis data 112 represents any data store or component (e.g., database server) that contains laboratory analysis data. This data is the result of detailed testing and evaluation of heat transfer fluid (HTF) samples that have been processed through reclamation treatments. These analyses are performed to assess theeffectiveness of the reclamation processes and to determine the current condition of the fluid. The results from these analyses provide useful data that can be used to make informed predictions about future fluid performance and to recommend appropriate reclamation treatment options. For example, samples of HTF are collected from various points in the system after reclamation treatments have been applied. These samples represent the reclaimed fluid's current state. The samples are then securely transported to a laboratory equipped for HTF analysis. Various embodiments use different sensors and other equipment as described with respect to the raw data 112. For example, using distillation apparatus or thermogravimetric analyzers, some embodiments determine the boiling ranges of different fluid components. Some embodiments employ gas chromatography (GC), inductively coupled plasma optical emission spectroscopy (ICP-OES), and Fourier transform infrared spectroscopy (FTIR) to identify and quantify contaminants.

[0054] Historical customer data 114 refers to the accumulated records of one or more customer’s heat transfer fluid (HTF) system over time. This data captures the long-term performance, maintenance history, operational conditions, and previous analytical results of the HTF in use. For instance, dates and details of past maintenance activities, such as fluid top-ups, filtration, and reclamation treatments is included in 114 in some embodiments. In some embodiments, the historical customer data 114 additionally or alternatively includes records of system downtimes, issues encountered, and repairs made, operating parameters such as temperature, pressure, and flow rates over time of a customer’s system, historical logs from real-time sensors (if applicable) capturing daily operational conditions, measurements of HTF performance indicators like heat transfer efficiency, energy consumption, and system output, records of any performance degradation overtime and its correlation with operating conditions, results from past fluid analyses, showing changes in fluid properties such as viscosity, acidity, contamination levels, and chemical composition, trends in these properties over time, helping to predict future fluid behavior.

[0055] Historical customer data 114 is used to identify long-term trends and patterns in fluid performance and degradation, and ultimately the decision statistics at 120, 118, 122 and / or 124. It helps in building predictive models by providing acomprehensive dataset that captures the evolution of the fluid’s properties and the system's performance over time. This data is useful for understanding the context of current fluid conditions and making informed predictions about future maintenance needs. Thus, historical customer data 114 captures trends and patterns over months or years or other time period. It includes various types of information such as maintenance history, operational conditions, and past analytical results.

[0056] The degradation formula(s) 110 represent one or more degradation formulas. A degradation formula is a mathematical model or set of equations that describe how the properties of a heat transfer fluid (HTF) change over time due to various factors such as thermal stress, oxidation, contamination, and operational conditions. These formulas are based on empirical data and scientific understanding of the chemical and physical processes that cause fluid degradation. For example, a High Boiling Compounds (HBC) degradation formula is the formation and accumulation of high boiling compounds, which are heavy fractions that result from the thermal decomposition of the HTF. An example HBC degradation formula is as follows:HBC(t)=HBCo+ki • f(T(i),P(r),O(T))dTwhere HBCo is the initial concentration, ki is a rate constant, and f (T,P,0) is a function of temperature T. pressure P. and oxygen exposure O over time r.

[0057] In some embodiments, the current state of the HTF, including recent lab analysis results and / or real-time sensor data, is fed into the degradation formulas. Historical customer data 114 provides a context for the current state and helps refine the predictions in some embodiments. The degradation formulas predict how the HTF properties will change over time if no reclamation is performed. This involves integrating the formulas over a specified future period. The degradation formula(s) 110 simulate different reclamation scenarios by modifying the inputs (e.g., reduced contaminants, lowered temperature) and recalculating the degradation rates.

[0058] The reclamation service provider data 116 refers to the information or records obtained from companies or sendees that specialize in reclaiming and restoring heat transfer fluids (HTF). This data includes details about the processes used, the condition of the HTF before and after reclamation, the effectiveness of various treatment methods, and operational parameters during the reclamation process.Reclamation process data includes details about the specific reclamation techniques used (e.g., filtration, centrifugation, vacuum dehydration). Parameters such as temperature, pressure, flow rates, and duration of the reclamation process. Pre- and post-reclamation fluid condition includes measurements of HTF properties before and after reclamation, including viscosity, boiling points, contaminant levels, and chemical composition. It also includes comparative data showing the improvement in fluid properties due to reclamation.

[0059] Operational logs are records of the operational conditions during reclamation, such as system uptime, any issues encountered, and corrective actions taken. It also includes logs of equipment settings and adjustments made during the reclamation process. Effectiveness metrics include quantitative measures of how effective the reclamation process was in restoring the HTF’s properties, such as percentage reduction in contaminants, percentage improvement in viscosity, and overall enhancement of fluid performance. In some embodiments, operational logs include pump curves or other equipment operating measurement. Monitoring pump workloads over time could be used for monitoring fluid lifetime and be used in place of some laboratory data.

[0060] In some embodiments, the reclamation service provider data 116 is used to calibrate predictive models (e.g., the reclamation model(s) 108). By incorporating real-world data from reclamation services, the predictive models used by the HTF Reclamation Calculator can be calibrated for accuracy in some instances, though not required. This may help ensure that the predictions and recommendations are based on proven reclamation outcomes. By analyzing operational logs and effectiveness metrics, the system can identify best practices and optimize the reclamation processes for different scenarios. Examples of Reclamation Service Provider Data 116 include-filtration process data where, for example, the reclamation process data includes multistage filtration (coarse and fine). The parameters include coarse filtration at 50 microns, fine filtration at 5 microns. The Pre-reclamation condition includes Viscosity: 15 cP and Contaminants: 200 ppm of solid particles. The Post-Reclamation Condition includes; Viscosity: 10 cP and Contaminants: 20 ppm of solid particles. Theeffectiveness metrics include a 90% reduction in solid particle contaminants and a 33% improvement in viscosity.

[0061] As illustrated in FIG. 1, the reclamation model(s) 108 take, as input, an encoded representation of the heat transfer fluid sample data 102 (e.g., parameters representing a current state of a user’s HTF), the degradation formula(s) 110, the laboratory analysis data 112, the historical customer data 114 and / or the reclamation service provider data 116 to generate one or more decision statistics indicative of model predictions, such as life expectancy, reclamation treatment option(s) 118, posttreatment fluid quality, and / or waste disposal 124.

[0062] In some embodiments, the reclamation model(s) 108 simulate various reclamation scenarios to assess the effectiveness of different treatment options to generate the reclamation treatment option(s) 118. Various embodiments modify input parameters (e.g., temperature, filtration method, chemical additives) to evaluate their impact on HTF properties. Various embodiments perform optimization and recommendation by identifying the optimal reclamation treatment via comparing the predicted outcomes of different scenarios. Various embodiments then generate reclamation treatment option recommendations based on criteria such as minimizing contaminants, improving viscosity, and extending HTF life.

[0063] In an illustrative example, various embodiments receive and process the degradation formulas, such as those for High Boiler Compound (HBC) accumulation, Low Boiler Compound (LBC) depletion, and carbon residue. The Laboratory Analysis Data 112 may include the following: Current viscosity: 12 cP, Low boilers: 4%, High boilers: 14%, Carbon residue: 1.2%, Total acid number (TAN): 0.7 mg KOH / g, and Moisture content: 300 ppm. The Historical Customer Data 114 includes the maintenance history (annual fdtration. biannual chemical treatment), operational data (e.g., average temperature 320°C, pressure 5 bar) and previous analysis (e.g., gradual increase in viscosity, consistent contaminant levels). The reclamation sendee provider data 116 includes previous reclamation (e.g., Fine fdtration, vacuum dehydration, chemical treatment), effectiveness metrics (e.g.. 80% reduction in contaminants. 30% improvement in viscosity), and process parameters (e.g., fdtration at 10 microns, vacuum dehydration at 60°C, pressure 100 mbar, duration 4 hours).

[0064] Some embodiments then combine the most recent lab data 112 with historical data 114 and reclamation service provider records 116 to perform feature extraction. Feature extraction includes extracting features such as current fluid properties, historical trends in viscosity and contaminant levels, and effectiveness of past reclamation treatments. Various embodiments then perform predictive modeling by using degradation formulas 110 to predict future changes in HTF properties. Some embodiments then train a machine learning model (e g., random forest regressor) on the integrated dataset to predict the impact of various reclamation treatments.

[0065] Consider the following simulation scenarios where embodiments simulate different reclamation processes. For example, the different processes may be as follows: Scenario 1: Fine filtration and vacuum dehydration. Scenario 2: Chemical treatment for acid neutralization and moisture removal. Scenario 3: Combined treatment with filtration, dehydration, and chemical stabilization

[0066] Various embodiments then optimize and provide corresponding recommendation. For example, some embodiments evaluate the predicted outcomes for each scenario, e.g., Scenario 1: Predicted viscosity 10 cP. high boilers 10%, carbon residue 0.8%; Scenario 2: Predicted viscosity 11 cP, high boilers 12%, carbon residue 1.0%; Scenario 3: Predicted viscosity 9 cP, high boilers 8%, carbon residue 0.6% The reclamation treatment option(s) 118 thus include a recommendation for Scenario 3 for optimal fluid performance and longevity.

[0067] In some embodiments, the reclamation model(s) 108 additionally or alternatively generates life expectancy 120, which refers to a decision statistic indicative of a prediction of an expected duration that a heat transfer fluid will last or be good in a system for. In some embodiments, such calculation is based on the reclamation treatment option(s) 118 that have been provided or recommended to a user. For example, Post-Reclamation Condition Data may indicate that the viscosity improved from 15 cP to 10 cP, the High Boilers reduced from 18% to 10%, and the carbon residue reduced from 1.5% to 0.8% based on a measured change in a sample post reclamation. The Historical Customer Data 114 may indicate that previous maintenance included annual filtration and biannual chemical treatments and that past degradation was viscosity increased by 1 cP per year on average. The ReclamationSen ice Provider Data 116 indicates that filtration pically extends HTF life by 25%, and dehydration by another 15%. The predictive modeling combines the improved HTF properties with historical degradation rates and the efficiency of the reclamation processes.

[0068] Various embodiments then simulate future conditions by using degradation formula(s) 110 to simulate how the improved HTF will degrade under standard operating conditions (e.g., average temperature of 320°C, pressure of 5 bar). Various embodiments then adjust the degradation rates to reflect the reduced contaminant levels and restored viscosity. Using this information as input, the life expectancy 120 may then be calculated. For example, the historical customer data 114 suggests atypical HTF life of 3 years before reclamation. Various embodiments adjust such life expectancy - given the improvements, various embodiments predict a 40% increase in life expectancy (25% from filtration and 15% from dehydration), resulting in an additional 1.2 years. The model 108 thus predicts that the HTF, post-reclamation, will last approximately 4.2 years before requiring another reclamation or replacement. Therefore, by combining post-reclamation condition data, historical customer data, and reclamation service provider data, and applying degradation formulas, the model 108 can accurately predict the life expectancy of the HTF. This comprehensive approach ensures that maintenance and reclamation schedules are optimized, extending the fluid’s useful life and improving overall system performance.

[0069] In some embodiments, the reclamation model(s) 108 generate a decision statistic indicative of a prediction of a post-treatment fluid quality for each recommended / determined reclamation treatment option at 118. Various embodiments first define reclamation treatments by listing the specific reclamation treatments recommended (e.g., those in the reclamation treatment option(s) 118) including their parameters (e.g., filtration pore size, dehydration temperature and pressure, types of chemical additives). Various embodiments then simulate heat transfer fluid treatment effects by using historical data (e.g., historical customer data 114) and empirical results to model the expected changes in HTF properties for each treatment option. Various embodiments also apply degradation formulas 110 adjusted for the specific treatment parameters to estimate post-treatment properties.

[0070] The reclamation model(s) 108 then responsively predict post-treatment fluid quality 122 by calculating the expected post-treatment values for key HTF properties such as viscosity, boiling points, contaminant levels, and carbon residue. Some embodiments use statistical models and / or machine learning algorithms trained on historical reclamation data to refine these predictions. Various embodiments then compare the predicted post-treatment qualities of HTF for each recommended reclamation treatment option. Some embodiments then optimize the recommendations based on the predicted improvements and the specific goals (e.g., extending fluid life, improving heat transfer efficiency).

[0071] Consider the following example, the current HTF condition may be: Viscosity: 15 cP; High Boilers: 18%; Low Boilers: 3%; Carbon Residue: 1.5%; Total Acid Number (TAN): 0.7 mg KOH / g; Moisture Content: 400 ppm. The following represent different recommended reclamation treatments: Option 1: Fine Filtration and Vacuum Dehydration (e.g., Filtration at 10 microns; Dehydration at 60°C and 100 mbar for 4 hours); Option 2: chemical treatment for acid neutralization and moisture removal (e.g., the addition of neutralizing agents and use of desiccants for moisture removal; Option 3: Combined Treatment (Filtration, Dehydration, and Chemical Stabilization), such as filtration at 5 microns, dehydration at 50°C and 50 mbar for 6 hours, and addition of antioxidants.

[0072] Various embodiments then simulate the treatment effects as follows: Option 1: fine Filtration and Vacuum Dehydration Historical Data and Empirical Results. Similar treatments show 70% reduction in high boilers, 50% reduction in carbon residue, and 80% reduction in moisture. The post-Treatment fluid Quality Prediction 122 is Viscosity: 11 cP; High Boilers: 10%; Low Boilers: 3%; Carbon Residue: 0.8%; TAN: 0.6 mg KOH / g; Moisture Content: 80 ppm.

[0073] Option 2 is chemical treatment for acid neutralization and moisture removal. The historical data and empirical results, for example, indicate that neutralizing agents typically reduce TAN by 50%, and desiccants reduce moisture by 90%. The post-treatment fluid quality prediction 122 may thus be: Viscosity: 14 cP; High Boilers: 16%; Low Boilers: 3%; Carbon Residue: 1.4%; TAN: 0.35 mg KOH / g; Moisture Content: 40 ppm.

[0074] Various embodiments then compare these treatment options, where for example, option 1 is associated with a significant reduction in viscosity, high boilers, carbon residue, and moisture. And option 2 is associated with a moderate improvement in viscosity, high boilers, and carbon residue but excellent reduction in TAN and moisture. The final recommendation is option 2. In this example, this option provides the best overall likely enhancement of fluid properties, ensuring extended fluid life and improved system performance. By simulating the effects of recommended reclamation treatments using historical data, empirical results, and degradation formulas, the model can accurately predict the post-treatment quality of the HTF. This allows for informed decision-making, ensuring the selection of the most effective reclamation treatment to optimize the performance and longevity of the heat transfer fluid.

[0075] The waste disposal 124 refers to a decision statistic indicative of predicting waste parameters based on the reclamation treatment option(s) 118. For example, the reclamation model(s) 108 may predict a quantity of waste that needs to be disposed based at least in part on the model(s) 108 generating a first decision statistic indicative of the reclamation treatment option(s) 118.

[0076] Various embodiments first specify the parameters for each recommended reclamation treatment (i.e., the reclamation treatment option(s) 118), such as distillation. Various embodiments use historical data and empirical results to estimate the efficiency of each treatment in removing contaminants, reducing high boilers, and other properties.

[0077] For each treatment option, the reclamation model(s) 108 calculate the expected amount of contaminants and degraded components removed from the HTF. The reclamation model(s) 108 then use degradation formula(s) 110 to predict the quantity of waste generated based on the initial condition of the HTF and the effectiveness of the treatment. The reclamation model(s) 108 additionally quantify the types and amounts of waste generated, such as solid particles, sludges, moisture, and chemical byproducts. The reclamation model(s) 108 compare the predicted waste quantities for different reclamation options. Various embodiments optimize the recommendations to minimize waste while achieving the desired HTF quality improvement.

[0078] In an illustrative example of the waste disposal 124, the current conditions are as follows: Current HTF Condition: Viscosity: 15 cP; High Boilers (HBC): 18%; Low Boilers (LBC): 3%; Carbon Residue: 1.5%; Total Acid Number (TAN): 0.7 mg KOH / g; Moisture Content: 400 ppm. The recommended reclamation treatments in 118 may be as follows: Option 1: Fine Filtration and Vacuum Dehydration, which includes Filtration at 10 microns, Dehydration at 60°C and 100 mbar for 4 hours. Option 2 includes Chemical Treatment for Acid Neutralization and Moisture Removal, which includes addition of neutralizing agents, and use of desiccants for moisture removal. Option 3 is combined Treatment (Filtration, Dehydration, and Chemical Stabilization), where Filtration is at 5 microns, Dehydration at 50°C and 50 mbar for 6 hours, and addition of antioxidants. Option 4 is topping off. Option 5 is venting. Option 6 is reclamation. Option 6 is flush replacement. Topping off refers to adding fresh HTF to the system to dilute degraded fluid and restore some fluid properties without fully replacing the system's contents. Venting is the releasing of volatile components (e.g., low boilers or moisture) from the HTF system, often to maintain pressure stability and reduce acidity or contamination. Reclamation is treating the degraded HTF (e g., through filtration, distillation, or other methods) to remove contaminants and restore fluid quality, often done on-site or off-site. Flush-replacement is completely draining the system, optionally flushing it to remove residual contaminants, and refilling with fresh HTF to fully restore system performance.

[0079] In some embodiments, the reclamation model(s) 108 then perform simulation or prediction of waste Generation: for Option 1 : Fine Filtration and Vacuum Dehydration, where the estimated efficiency is 70% reduction in high boilers, 50% reduction in carbon residue, and 80% reduction in moisture. The following predicted waste generated for option 1 is as follows: Predicted High Boilers: 0.7xl8%=12.6% where 0.7: represents the fraction of the system being reclaimed (e.g., 70% of the total system volume is processed during reclamation), 18% represents the initial concentration of high boilers (HB) in the fluid, measured in the current HTF sample. This value comes from laboratory analysis or on-site sensors in some embodiments.12.6% is the total high boiler waste generated during reclamation, as a percentage ofthe reclaimed volume. In various embodiments, there are similar calculations for carbon residue and moisture, for example.

[0080] Various embodiments then quantify the waste. For example, for option 1, the total waste is a combination of solid waste (e.g., high boilers, carbon residue, corrosion particulates, and debris) liquid waste (e.g., moisture), and / or any intermediate states, such as anything between a thick liquid / sludge and a solid. Solid Waste: 12.6% high boilers + 0.75% carbon residue = 13.35%. Liquid Waste: 320 ppm moisture. Various embodiments then compare the different options and quantities predicted in such options. For example, option 1 generates 13.35% solid waste and 320 ppm liquid waste. Option 2 generates minimal solid waste but 360 ppm liquid waste and 0.35 mg KOH / g acid waste. And option 3 generates 15.3% solid waste and 360 ppm liquid waste and 0.35 mg KOH / g acid waste. Option 3 is the most comprehensive in improving HTF quality but generates the highest amount of waste. Option 1 provides a good balance between improving HTF qualify and minimizing waste.

[0081] For the final recommendation, if minimizing waste is a priority while still achieving significant improvement in HTF qualify, Option 1 (Fine Filtration and Vacuum Dehydration) is recommended in some embodiments. By integrating current HTF condition data, historical and empirical results, degradation formulas, and reclamation service provider data, the model(s) 108 can predict the quantify of waste generated for each reclamation treatment option. This allows for informed decisionmaking that balances the improvement of HTF qualify with the management of waste disposal.

[0082] The user interface component 126 is generally responsible for causing presentation of one or more user interface elements that indicate the life expectancy 120, the reclamation treatment option(s) 118, the post-treatment fluid qualify 122, and / or the waste disposal 124. The user interface component 126 may comprise one or more applications or services on a user device, across multiple user devices, or in the cloud. For example, in one embodiment, the user interface component 126 manages the presentation of content to a user across multiple user devices associated with that user. Based on content logic, device features, associated logical hubs, inferred logical location of the user, and / or other user data, presentation component may determine onwhich user device(s) content is presented, as well as the context of the presentation, such as how (or in what format and how much content, which can be dependent on the user device or context) it is presented and / or when it is presented.

[0083] In some embodiments, the user interface component 126 generates (or causes generation of) user interface features. Such features can include interface elements (such as graphics buttons, sliders, menus, audio prompts, alerts, alarms, vibrations, pop-up windows, notification-bar or status-bar items, in-app notifications, or other similar features for interfacing with a user), queries, and prompts. In some embodiments, the user interface component 126 generates structured data, tagged data, or otherwise causes presentation of structured or tagged data that was previously unstructured, semi-structured, or untagged. For example, in some embodiments the user interface component 126 causes presentation of tagged data (e.g., fluid condition score being “fair” or parameter value being within a “normal” range), which may have been previously unstructured or otherwise been in a different form (e.g., existed only in natural language form) than the output provided by the user interface component 126. In this way. some embodiments convert input data to an output that is different than the input.

[0084] FIG. 2 is a screenshot of an example user interface page 200 illustrating different reclamation treatment option recommendations, according to some embodiments. In some embodiments, the page 200 represents what is caused to be presented by the user interface component 126 of FIG. 1.

[0085] The page 200 includes multiple reclamation treatment option recommendations 202 (refill), 212 (dilute), 222 (mobile reclaim), and 232 (bleed & feed). In some embodiments, each of these recommendations represent the reclamation treatment option(s) 118 of FIG. 1. Each of these reclamation treatment option recommendations include additional datasets, including dials 204, 214, 224, and 234, which represents heat transfer fluid life expectancy (e.g., the life expectancy 120 of FIG. 1). Other datasets include user interface elements 206, 216, 226, and 236, which represent when heat transfer fluid system maintenance was last performed. Other datasets include user interface elements 208, 218, 228, and 238, which represent the projected or predicted end of life for a particular heat transfer fluid when acorresponding reclamation treatment option is chosen. Other datasets include user interface elements 210, 220, 230, and 240, which represents the estimated cost to incorporate a corresponding reclamation treatment option recommendation. For example, regarding the refill reclamation treatment option 202, it is recommended that it will cost $1,303,000 to incorporate into a heat transfer fluid system as indicated by the user interface element 210.

[0086] Given a user’s heat transfer fluid sample data (e.g., 102 of FIG. 1), various embodiments, such as the reclamation model(s) 108, generate a decision statistic indicative of a prediction of multiple reclamation treatment option recommendations 202, 212, 222, and 232. The "refill" reclamation treatment option recommendation 202 refers to the process of draining and disposing of a portion of the degraded heat transfer fluid (HTF) and replacing it with new HTF product. This option involves either a partial or complete replacement of the HTF, depending on the severity of degradation and the desired restoration of fluid quality7. The “dilute” reclamation treatment recommendation option 212 refers to the process of partially draining a portion of the degraded heat transfer fluid (HTF) from the system and replacing it with new or rejuvenated fluid. This dilution process aims to lower the concentration of degraded components such as high boilers, low boilers, and contaminants, thereby improving the overall quality of the fluid without the need for a complete system refill. The bleed & feed reclamation treatment option recommendation 232 refers to a maintenance process where a small, controlled amount of the degraded heat transfer fluid (HTF) is continuously or periodically removed (bleed) from the system and replaced (feed) with fresh or reclaimed fluid. This method helps maintain the quality7of the HTF by gradually diluting contaminants and degradation products with new fluid over time.

[0087] Various embodiments compute or calculate any of the reclamation treatment options described herein in any suitable manner based on at least one of: a customer’s current heat transfer fluid sample, historical customer data, a degradation formula, existing reclamation service provider data, or a laboratory analysis of a reclaimed fluid. For example, the inputs and feature engineering can include the following: Current HTF Sample (Xcun-ent), which includes parameters like viscosity,TAN (Total Acid Number), moisture content, low boilers, high boilers, insoluble solids, etc. Other inputs include Historical Customer Data (Xhistoricai), which contains data on past fluid conditions and corresponding treatment outcomes. Other input features include Degradation Formula (D), which is a mathematical or empirical models predicting the degradation of HTF over time or under certain conditions. Other input features include Reclamation Service Provider Data (Xservice), which includes details on the effectiveness of various reclamation treatments. Other input features include Laboratory Analysis of Reclaimed Fluid (Xiab), which are results from lab tests on reclaimed fluids, providing empirical effectiveness data for various treatments.

[0088] This raw data is transformed into a feature vector F (e.g., an input tensor) that includes Xcurrent features or current HTF properties, aggregated historical data, which is the average or weighted average of past fluid conditions and outcomes, degradation rates, which are the predicted rates of degradation using D, and Effectiveness metrics derived from Xservice and Xhb. In some embodiments, the model M takes the feature vector F and outputs a recommendation score for each treatment option T={T1,T2,....Tn}, where the model equation is: Y=M(F;0), and where Y is the vector of scores for each treatment option and theta 0 are the model parameters (weights) learned during training. Each element yi in the vector Y represents a score for the corresponding treatment option T; The highest score indicates the most recommended treatment. The equation for the treatment recommendation is as follows: Trecommended = argmax(Y). This function selects the treatment option with the highest score.

[0089] In an illustrative example, a current HTF Sample may include the following: a Xcurrent = {Viscosity=15 cP, TAN-0.7 nig KOH / g. Moisture=400 ppm; Low Boilers=5%; High Boilers=18%; Insoluble Solids = 0.3%. The historical data is aggregated data indicating typical outcomes for similar initial conditions. The Degradation Formula D predicts an annual increase in TAN by 0.2 mg KOH / g and viscosity by 2 cP. The Reclamation Service Data indicates to "Dilute Treatment reduces TAN by 20%, High Boilers by 10%". The Laboratory Analysis confirms that previous "Bleed & Feed" reduced high boilers by 30%.

[0090] The Feature Vector F combines normalized current HTF conditions, degradation predictions, historical success rates, and reclamation service metrics. Suppose the model M predicts the following scores: yi=0.75 for "Dilute Treatment"; y2=0.65 for "Complete Refill"; ys=0.85 for "Bleed & Feed" using the equation:Trecommended = argmax( {0.75,0.65,0.85 } ) = T3The argmax function identifies the index of the highest score in Y. This means the model recommends the "Bleed & Feed" treatment option, as it has the highest score (0.85) among the evaluated options. The machine learning model M thus utilizes the feature vector F derived from a customer's current HTF sample, historical data, degradation formulas, and additional sen ice data to score and recommend reclamation treatment options. The option with the highest score, indicating the most likely successful outcome, is recommended. This predictive process is guided by a supervised learning model trained on labeled input-output pairs, leveraging positive and negative samples to improve accuracy and reliability.

[0091] In some embodiments, however, reclamation treatment options need not use a machine learning model but can be predicted via any suitable statistical model. This approach relies on using historical data and expert knowledge to define thresholds and rules for decision-making. For example, in some embodiments, the model uses conditional logic or rules to recommend the best reclamation treatment based on the input data. For instance, if High Boilers are above a certain level (HB_ > Threshold), consider "Complete Refill" represented as the following equation:If Xcurrent [HB] > 15%, recommend Complete RefillIn another example, if both Low Boilers (LB) and HB are above certain levels, suggest "Bleed & Feed.” For instance, if the percentage of low boilers (LB) in the current heat transfer fluid (HTF) sample is greater than 5%, and the percentage of high boilers (HB) is greater than 10%, then recommend the "Bleed & Feed" treatment option.

[0092] In an illustrative example of the entire process, Given Data, the current HTF sample may be as follows: Xcun-ent = {Viscosity=14 cP,TAN=0.6 mg KOH / g; Moisture=300 ppm; Low Boilers=6%; High Boilers=16%; Insoluble Solids=0.2. Historical Data indicates that "Bleed & Feed" is 85% effective when both LB and HB are above 5% and 10%, respectively. The degradation formula predicts that HB willincrease by 1 % per year and TAN by 0.1 mg KOH / g per year. The Reclamation Service Data shows "Dilute Treatment" reduces TAN by 30% but has limited effect on HB and LB. Laboratory Analysis confirms that previous "Complete Refill" significantly improves fluid condition. In other words, the Historical Data indicates that "Bleed & Feed" is 85% effective when both LB and HB are above 5% and 10%, respectively. The Degradation Formula predicts that HB will increase by 1% per year and TAN by 0.1 mg KOH / g per year. The Reclamation Service Data shows "Dilute Treatment" reduces TAN by 30% but has limited effect on HB and LB. The Laboratory Analysis confirms that previous "Complete Refill" significantly improves fluid condition.

[0093] Various embodiments then apply conditional rules, such as those described above: Check High Boilers: HB=16%>ThresholdHB=15% (which means that the score is incremented to “complete refill”); Check Low Boilers and High Boilers: LB=6%,HB=16%, which indicates that the system should recommend "Bleed & Feed" (if LB>5%, HB>10%). Some embodiments also determine that historical success rate for "Bleed & Feed" is 85% under similar conditions. Accordingly, based on the above rules, the model may recommend "Bleed & Feed" as it meets the condition for both LB and HB levels, and historical data suggests a high success rate. For example, if the percentage of low boilers (LB) in the current heat transfer fluid sample is greater than 5%, and the percentage of high boilers (HB) is greater than 10%, then the recommended treatment option is 'Bleed & Feed'."

[0094] In some embodiments, each of the reclamation treatment options 202, 212, 222, and 232 are ranked based on particular factors, such as fluid life expectancy, estimated cost to incorporate a corresponding reclamation treatment options, or the like, where the user interface page 200 indicates such ranking (e.g., by ordering them in a hierarchical listing view where the higher up the option, the higher the ranking). A scoring system can be implemented to quantify these factors, allowing for a comparative ranking. Key factors for ranking may include the following, fluid life expectancy (FLE), which is the improvement in fluid life expectancy post-treatment, indicating how long the HTF can continue to function effectively before another treatment is needed. Other factors include the estimated cost (EC), which is the total cost associated with each treatment option, including the cost of new fluid, disposal,operational costs, and any downtime. Other factors include effectiveness €, which is the overall effectiveness of the treatment in reducing contaminants, improving viscosity, and restoring fluid quality. Other factors include downtime (D), which is the amount of system downtime required for the treatment, if applicable.

[0095] To rank the options, some embodiments use a weighted scoring method:Score = w i / FLE ■ '2xCC1I ws^ E ■ '4zD1where wi,W2.W3,W4 are the weights assigned to each factor based on their importance, andEC1and D1are the inverse of the cost and downtime to reflect that lower costs and downtime are preferred.

[0096] The "fluid life expectancy dials" 204, 212, 222, and 232 refer to visual indicators or gauges displayed on the user interface page 200. These dials provide a graphical representation of the predicted remaining lifespan of the heat transfer fluid (HTF) after a specific reclamation treatment option is applied. The dials show how long the fluid is expected to remain effective before reaching a critical level of degradation, requiring another reclamation treatment or complete replacement. They help users quickly assess the impact of different reclamation treatments on the fluid's lifespan. By displaying the projected fluid life expectancy for each treatment option, the dials allow for easy comparison, enabling users to select the most effective treatment based on expected longevity. These dials aid in decision-making by providing a clear, visual summary of how each treatment option will extend the HTF's life, helping to balance costs, efficiency, and maintenance schedules. A dial might show the current fluid life expectancy as a baseline, indicating the expected remaining lifespan w ithout additional treatment.

[0097] In some embodiments, the Fluid Life Expectancy (corresponding to the dials 204, 214, 224, and 234) are calculated in any suitable manner. For example, the model 108 uses a combination of current fluid condition data, historical degradation trends, and the expected impact of selected reclamation treatments. The FLE represents the estimated remaining useful life of the heat transfer fluid before it reaches a critical level of degradation that necessitates replacement or further reclamation. The Factors influencing the FLE calculation are as follows in some embodiments: current fluidcondition, such as Key parameters include viscosity, high boilers (HBC), low boilers (LBC), carbon residue, and total acid number (TAN). Additionally or alternatively, factors such as degradation rate. These rates are derived from historical data and degradation formulas, indicating how quickly the fluid properties deteriorate under current operational conditions. Additionally or alternatively, effectiveness of reclamation treatment may be used, which refers to the predicted improvement in fluid properties post-treatment, which can slow down the degradation rates. Additional or alternative factors include operational data, which includes factors such as temperature, pressure, and flow rates that influence the degradation process.

[0098] In some embodiments, the FLE is estimated using the following formula:>where Critical Level is the threshold value at which the HTF is considered no longer viable (e.g., viscosity or TAN reaches a critical limit), Current Level is the current measurement of the key parameter (e g., current viscosity7or TAN), the Degradation Rate is the rate at which the parameter degrades over time, and AFLEireatment is the additional lifespan gained from the selected reclamation treatment.

[0099] In an illustrative example of calculating the FLE, Current Fluid Condition: Viscosity: 15 cP (critical limit: 20 cP); High Boilers (HBC): 18% (critical limit: 25%); Total Acid Number (TAN): 0.7 mg KOH / g (critical limit: 1.0 mg KOH / g); Degradation Rates: Viscosity Increase: 1 cP / year; HBC Increase: 1% / year; TAN Increase: 0.1 mg KOH / g / year; Reclamation Treatment: Bleed & Feed; Improvement in TAN: 20% reduction; New TAN Degradation Rate: 0.1 * (1 - 0.2) = 0.08 mg KOH / g / year.

[0100] Based on this information, the calculation is as follows: The TAN Based FLE Calculation: Critical Level: 1.0 mg KOH / g; Current Level: 0.7 mg KOH / g; Degradation Rate: 0.08 mg KOH / g / year:Without treatment AFLErreatment would be 0. With Bleed & Feed treatment, AFLE reatment extends the FLE by 1 year due to an improved degradation rate:0 3FLE Projected=0.08 + 1 = 3.75 = 1 = 4.75 yJears.

[0101] The fluid life expectancy dials, such as 234, would then be set based on this calculation. If the scale of the dial ranges from 0 to 10 years, for example, with 0 being the immediate need for replacement and 10 being the maximum possible life extension, the dial might be set at 4.75, indicating that the fluid is projected to last another 4.75 years with the Bleed & Feed treatment.

[0102] In some embodiments, the data entry for the maintenance performed user interface elements 206, 216, 226, and 236 is generated based on manual user input. For example, when a user registers their heat transfer fluid and system, various embodiments receive an indication that the last time the user performed maintenance on their system was October 2023, as indicated in all of the user interface elements 206, 216, 226, and 236.

[0103] In some embodiments, projected end of life for a heat transfer fluid corresponding to the user interface elements 208, 218, 228, and 238 represents or is calculated based on the life expectancy 120 of FIG. 1 and / or the FLE corresponding to the dials 204, 214, 224, and / or 234. The projected end of life (EOL) refers to the estimated time at which the heat transfer fluid (HTF) will require replacement or significant reclamation based on the selected treatment option. This estimation helps in planning maintenance and avoiding unexpected system downtimes. Factors for projected EOL calculation includes the following in some embodiments, current fluid condition, such as viscosity, high boilers (HBC), low boilers (LBC), carbon residue, and other key indicators of fluid degradation. Other factors include degradation rates, which are rates at which these parameters degrade over time, determined through empirical data and degradation formulas (e.g., 110), effectiveness of reclamation treatment, which is the anticipated improvement in fluid properties following a specific reclamation treatment, which can extend the fluid's useful life, and operational data, which includes temperature, pressure, and other operational conditions that influence degradation rates.

[0104] The projected EOL can be calculated using a formula that incorporates the current condition of the HTF, degradation rates, and the effectiveness of the selected reclamation treatment:EOLprojected C lll Snl EOL+AEOLlreatmentwhere EOLprojected is the projected EOL. current EOL is the remaining lifespan based on cunent fluid conditions without any further treatment and AEOLireatment is the additional lifespan gained from the chosen reclamation treatment option, which improves the fluid quality and slows down the degradation rate.

[0105] In an illustrative example, Current Fluid Condition: Viscosity: 15 cP (critical limit: 20 cP), High Boilers (HBC): 18% (critical limit: 25%), Low Boilers (LBC): 5% (critical limit: 10%), Carbon Residue: 1.5% (critical limit: 2%), Total Acid Number (TAN): 0.7 mg KOH / g (critical limit: 1.0 mg KOH / g), Degradation Rates: Viscosity Increase: 1 cP / year, HBC Increase: 1.5% / year, LBC Increase: 0.5% / year, Carbon Residue Increase: 0.1% / year, TAN Increase: 0.1 mg KOH / g / year. Treatment Option: Bleed & Feed, Improvement in Viscosity: 20% reduction. Improvement in HBC: 25% reduction, Improvement in LBC: 30% reduction, Improvement in Carbon Residue: 20% reduction, Improvement in TAN: 15% reduction.

[0106] Current EOL basedyears; and AEOLireatment is an improvement due to Bleed & Feed treatment (20% reduction in TAN, slowing degradation). The New TAN Degradation Rate =0.1 x(l-0.2)=0.08 mg KOH / g / year. The New EOL with imp1roved TAN is as follows:100.0 °8'7= 3.75 yJears. The AEOLireatment is 3.75-3=0.75 years. The EOLprojected= 3+0.75=3.75 years. Thus, after applying the Bleed & Feed treatment, the projected end of life of the HTF is extended to 3.75 years from the original 3 years. This calculation helps in selecting the most effective reclamation treatment option and planning future maintenance.

[0107] The estimated option cost corresponding to user interface elements 210, 220, 230, and 240 may be calculated in any suitable manner. The estimated option cost for a reclamation treatment in the context of the HTF Reclamation Calculator model involves calculating the total expenses associated with implementing a particular treatment option. In some embodiments, this includes costs related to materials, labor, equipment, and / or any potential system downtime. For example, there may be Material Costs (MC), which include new HTF fluid cost (e.g., cost of purchasing new fluid to replace degraded fluid), reclamation chemicals (costs for chemicals used in the treatment process (e.g.. neutralizers, antioxidants)). Additionally or alternatively, theremay be Labor Costs (LC), which include technician time (cost associated with technicians performing the reclamation process, including setup, operation, and cleanup), and consultation fees (if applicable, fees for expert consultation or specialized services). Additionally or alternatively there may be Equipment Costs (EC), such as equipment rental / purchase (costs for renting or purchasing equipment required for reclamation (e.g., fdtration units, vacuum dehydrators)), and maintenance and calibration, which are costs for maintaining and calibrating the equipment. Additionally or alternatively, there may be Disposal Costs (DC), such as waste disposal (costs for disposing of removed contaminants, degraded fluid, and other waste products), and environmental fees (fees associated with environmental regulations and waste disposal compliance). Additionally or alternatively, there may be Downtime Costs (DTC). such as loss of production, which are costs associated with system downtime if the process requires shutting down the system. Additionally or alternatively there may be other operational costs, such as any additional costs incurred due to extended operational hours or temporary solutions during downtime.

[0108] The total estimated cost can be calculated using the following equation:Estimated Option Cost (EOC) = MC+LC+EC+DC+DTC

[0109] For example, where the treatment option is a dilute treatment, the MC is New HTF Fluid Cost: $10,000 (for 20% of the system volume) and Reclamation Chemicals: $2,000. The Labor Costs (LC) is Technician Time: $3,000 (3 technicians for 2 days) and Consultation Fees: $1,000. The Equipment Costs (EC) is Equipment Rental / Purchase: $5,000 (filtration and dehydration units) and Maintenance and Calibration: $500. The Disposal Costs (DC) are Waste Disposal: $2,000 (handling and disposal of contaminated fluid) and Environmental Fees: $1,000. The Downtime Costs (DTC) include Loss of Production: $10,000 (1 day of downtime) and Additional Operational Costs: $500. Accordingly, the calculation is as follows:EOC = 12,000+4,000+5,500+3,000+10,500 = 35,000 USD

[0110] The estimated option cost provides a comprehensive financial estimate for implementing a particular reclamation treatment option, helping users make informed decisions based on cost-benefit analyses. In this example, the dilute treatment option has an estimated total cost of $35,000.

[0111] FIG. 3 is a screenshot of an example user interface page 300 illustrating current fluid condition and predicted fluid condition given a particular reclamation treatment option, according to some embodiments. In some embodiments, the user interface page 300 is caused to be displayed in response to clicking on any of the user interface elements of the page 200 of FIG. 2. For example, in response to receiving an indication that a user selects one of the reclamation treatment option recommendations 202, 212. 22. and / or 232, the page 300 is displayed. In some embodiments, the PostReclamation fluid score 304 represents and is calculated as described with respect to the post-treatment fluid quality 122 of FIG. 1.

[0112] The user interface page 300 includes a current fluid condition dial 302, a post-reclamation fluid dial 304, and existing or predicted parameter values 306. The "Current Fluid Condition" refers to the present state of the heat transfer fluid (HTF) (e.g., the heat transfer fluid sample data 102) based on various measurable properties or parameters. These properties indicate the fluid's performance and degradation level, helping determine the need for maintenance or reclamation treatments. In some embodiments, the parameters used to assess the current fluid condition include viscosity, acid number, moisture content, low boilers, high boilers, and insoluble solids.

[0113] Viscosity is a measure of the fluid's resistance to flow. High viscosity can indicate degradation due to the accumulation of high boiling compounds or contaminants, affecting heat transfer efficiency. Acid Number (Total Acid Number -TAN) refers to the quantity of acidic components in the fluid, measured in mg KOH / g. An increase in TAN indicates oxidation or contamination, leading to potential corrosion and fluid degradation. Moisture Content is the amount of water present in the HTF, usually measured in ppm (parts per million). Moisture can cause corrosion and reduce the thermal stability of the fluid. Low Boilers (LBC) refer to volatile compounds in the HTF that have a low boiling point. High levels of low boilers can lower the flash point and increase vapor pressure, posing safety hazards. High Boilers (HBC) refer to heavy, non-volatile compounds that have a high boiling point. High levels of high boilers can increase viscosity and form deposits, reducing heat transfer efficiency. Insoluble Solids are indicative of solid particles suspended in the fluid, often resulting from degradationor contamination. These solids can cause fouling, clogging, and reduced efficiency in the heat transfer system.

[0114] To calculate the current fluid condition, in some embodiments, the parameters are measured and analyzed to determine the overall state of the HTF. For example, various embodiments collect samples of the HTF from the system. Laboratory analyses may then be performed to measure viscosity, TAN, moisture content, low boilers, high boilers, and insoluble solids. Various embodiments then compare the measured values against industry standards, critical limits, or other thresholds established for each parameter. Various embodiments then identify which parameters exceed acceptable levels, indicating degradation or contamination.

[0115] A scoring system can be used to quantify the severity of degradation for each parameter. For example, a scale from 0 to 100 can be used, with 100 indicating severe degradation. Various embodiments then combine the scores from each parameter to form an overall condition index. This index provides a summary view of the HTF's condition. Weighting factors can be applied to give more importance to certain critical parameters, such as viscosity and TAN.

[0116] Consider the following example for calculating the current fluid condition, suppose the following measurements are obtained from the HTF: Viscosity: 14 cP (critical limit: 20 cP); TAN: 0.6 mg KOH / g (critical limit: 1.0 mg KOH / g); Moisture Content: 300 ppm (critical limit: 500 ppm); Low Boilers: 4% (critical limit: 10%); High Boilers: 15% (critical limit: 25%); Insoluble Solids: 0.2% (critical limit: 0.5%).

[0117] To score these: Viscosity7Score: (14 / 20) * 100 = 70; TAN Score: (0.6 / 1.0) * 100 = 60; Moisture Score: (300 / 500) * 100 = 60; Low Boilers Score: (4 / 10) * 100 = 40; High Boilers Score: (15 / 25) * 100 = 60; Insoluble Solids Score: (0.2 / 0.5) * 100 = 40. The overall condition index might be calculated as an average or weighted sum of these scores, depending on the importance of each parameter.

[0118] ‘‘Post-Reclamation Fluid” corresponding to the dial 304 refers to the predicted condition of the heat transfer fluid (HTF) after undergoing a specific reclamation treatment option. This prediction involves estimating the improvement in fluid properties or parameters such as viscosity7, acid number, moisture content, lowboilers, high boilers, and insoluble solids, based on the effectiveness of the chosen treatment. In some embodiments, to predict the condition of the HTF after a particular reclamation treatment, the model 108 uses historical data, empirical results, and degradation formulas, such as 100. The improvement in each parameter is calculated based on the known effectiveness of the treatment in removing or reducing contaminants and degradation products.

[0119] In some embodiments, the predicted value of a specific fluid parameter after reclamation, Ppost , is calculated using the following general equation:Ppost P current (ExP current)

[0120] Where Ppost is the predicted post-reclamation value of the parameter, Pcurrent is the current value of the parameter before treatment, and E is the effectiveness factor of the reclamation treatment, representing the percentage reduction or improvement in the parameter. Consider the following example, Current Fluid Condition: Viscosity: 15 cP; Acid Number (TAN): 0.7 mg KOH / g; Moisture Content: 400 ppm; Low Boilers: 5%; High Boilers: 18%; Insoluble Solids: 0.3%; Reclamation Treatment: Filtration and Dehydration; Effectiveness Factors:; Viscosity Reduction (E_viscosity): 20%; TAN Reduction (E_TAN): 30%; Moisture Reduction (E_moisture): 50%; Low Boilers Reduction (E_LBC): 25%; High Boilers Reduction (E_HBC): 35%; Insoluble Solids Reduction (E_solids): 40%. Accordingly, the viscosity post-reclamation, for example, may be calculated as follows:Ppost. viscosity = 15 - (0.20* 15) = 15-3 = 12 cP

[0121] To estimate the overall Fluid Condition after a specific reclamation treatment (i.e., corresponding to 304), individual fluid properties such as viscosity, acid number (TAN), moisture content, low7boilers, high boilers, and insoluble solids are assessed. These properties are critical indicators of fluid quality and performance. The overall Fluid Condition can be summarized or quantified using a composite score or index that takes into account the improvement in each property.

[0122] For each property (e.g., viscosity, TAN), some embodiments calculate a score representing the quality or acceptability of the fluid post-treatment. This score can be based on how close the property is to an optimal or acceptable range. Some embodiments assign weights to each property based on its importance to the overallfluid performance. For example, viscosity might be weighted more heavily than moisture content if it is considered more critical to system performance. Some embodiments then combine the weighted scores of all properties to form an overall Fluid Condition score reflected in the dial 304. This score provides a single value that summarizes the fluid's quality and suitability' for continued use.

[0123] In some embodiments, the overall Fluid Condition score, Fcondition (as indicated in dial 304) is finally calculated using the following equation:where n is the number of fluid properties considered, w, is the weight assigned to the i-th property, and Pi is the score for the z-th property, reflecting its condition postreclamation.

[0124] FIG. 4 is a schematic diagram illustrating how different on-site processes of reclamation can be predicted or estimated, according to some embodiments. In some embodiments, FIG. 4 represents a screenshot of a user interface page presented to a user in order to visually summarize all reclamation site predictions. In some embodiments, the predictions of FIG. 4 represent or are included in the reclamation treatment option(s) 118 of FIG. 1, and / or the waste disposal predictions of 124. The on-site process of reclamation can be estimated for the customer’s and sendee provider’s benefit, as illustrated in FIG. 4. The treatment option requires some new fluid 404 o be purchased in advance, which the model(s) 108 can estimate for the customer along with the amount of their system being reclaimed. For the service provider, the waste needing disposal can be estimated as well at 418.

[0125] Given more data collected on HTF Reclamation in the future, the potential duration of fluid treatment could also be shown (e.g., 2-4 weeks on-site). To predict the potential duration of fluid treatment, such as the reclamation of heat transfer fluid (HTF), various factors are considered in some embodiments, including the volume of fluid to be treated, the capacity7of the reclamation equipment, the rate of fluid processing, days for onsite safety training (e.g., 1 day), equipment setup time (e g., 1 to 2 days), time to shutdown / disassemble equipment (e.g., 1 to 2 days) and / or any operational constraints. For example, the treatment duration can be estimated using an equation that accounts for these factors, such as:Total Fluid Volume to be Treated (MT) Treatment Duration (days) = Processing Rate (MT / day)where Total Fluid Volume to be Treated (MT) is the volume of HTF that needs to be reclaimed and the Processing Rate (MT / day) is the rate at which the reclamation equipment can process the HTF, expressed in metric tons per day.

[0126] In an illustrative example, given that the Total System Volume is 100 MT (total capacity of the HTF system), System Reclaimed Percent is 69% (portion of the system's HTF undergoing reclamation), and the Processing Rate is 3 MT / day (rate at which the reclamation unit can process the fluid), various embodiments calculate the volume to be treated by first, calculating the total volume of fluid being reclaimed: Total Fluid Volume to be Treated-! 00MTx0.69=69MT

[0127] Then embodiments estimate the Treatment Duration by using the total fluid volume to be treated and the processing rate: Treatment Duration= 69 MT / 3MT per day = 23 days. Thus the estimated duration for the reclamation treatment is approximately 23 days. This is true if the fluid is not recycled back into the system. The HTF system may have to be shut down because eventually it would run out of fluid. This is not necessarily true for Mobile Reclamation.

[0128] The HTF producer 402 represents the source or manufacturer of the heat transfer fluid. The producer supplies HTF, which is needed to replenish the system after the reclamation process. This ensures the availability of new fluid 404 to replace the degraded portion removed during reclamation.

[0129] Element 406 indicates the amount of predicted new HTF needed to replenish the system after removing degraded fluid. In some embodiments, this quantity (31 metric tons) is calculated or predicted based on the percentage of fluid being reclaimed and the volume of the system. It ensures that the system operates efficiently post-reclamation with optimal fluid quality. The "31 MT New Fluid Required" refers to the amount of new heat transfer fluid (HTF) needed to replace the degraded portion removed during the reclamation process. To calculate the required amount of new fluid, the following equation is used in some embodiments:New Fluid Required (MT) = Total System Volume (MT) x Reclamation Percentage - Recovered Fluid Volume (MT)where Total System Volume (MT) is the total volume of HTF in the system before reclamation, Reclamation Percentage is the percentage of the total system volume being reclaimed, and Recovered Fluid Volume (MT) is the volume of fluid that is cleaned and returned to the system during the reclamation process. This can be calculated as the difference between the reclaimed fluid and the waste removed.

[0130] In an illustrative example of predicting 406, Total System Volume: 100 MT (as per the image and system capacity); Reclamation Percentage: 69% (the percentage of the system's HTF undergoing reclamation); Waste Removed: 31 MT (the amount of degraded fluid and contaminants removed). Various embodiments then calculate the reclamation volume:Reclamation Volume = Total System Volume x Reclamation Percentage = 100MT x 0.69 = 69MT

[0131] Various embodiments then estimate recovered fluid volume. Assuming the waste removed is equal to the degraded portion of the fluid (which is not recovered), the volume of reclaimed fluid:Recovered Fluid Volume = Reclamation Volume - Waste Removed = 69MT -31MT = 38MT.Various embodiments then calculate the new fluid required, which is the amount of new fluid needed to replace the waste removed:New Fluid Required = Reclamation Volume - Recovered Fluid Volume New Fluid Required = 69MT - 38MT = 31MT.

[0132] This calculation shows that after reclaiming 69% of the system volume and removing 31 MT of waste, 31 MT of new HTF is needed to replenish the system to maintain optimal performance. The "31 MT New Fluid Required" 406 is estimated based on the total volume of the system, the extent of the reclamation process, and the amount of degraded fluid removed. This equation provides a straightforward method to calculate the volume of new fluid necessary to restore the system's HTF to its desired quality and volume.

[0133] The mobile reclaimer 408 is a mobile unit that comes to the customer's site to perform the reclamation process. The mobile reclaimer 408 processes the used fluid, removing contaminants and restoring the fluid's properties. This unit allows foron-site reclamation without transporting the fluid to an external facility. The prediction 410 indicates the predicted percentage of the total system volume that undergoes reclamation. In this case. 69% of the 100 MT system is predicted as being returned as reclaimed fluid.

[0134] In some embodiments, this percentage in 410 is predicted or calculated based on the system's total capacity (100 MT system 414), the condition of the fluid, and / or the reclamation treatment's objectives. For example, the reclamation percentage as indicated in 410 can be calculated using the following equation:Volume to be reclaimed (MT)Reclamation Percentage (%) = ( Total system volume (MT) ) X 1where volume to be reclaimed (MT) is the volume of HTF identified for reclamation based on its degraded condition, and total system volume (MT) is the total volume of HTF in the system.

[0135] In an illustrative example of such calculation in 410, the total system volume is 100 MT (total capacity of the HTF system), assuming that the criteria for selecting the fluid to be reclaimed include factors such as high boilers, low boilers, viscosity, and / or other degradation indicators. Based on these criteria, 69 MT of fluid is identified as needing treatment. Various embodiments then identify the volume to be reclaimed - based on fluid analysis, it is determined that 69 MT of the HTF has degraded beyond acceptable levels and requires reclamation. Some embodiments then calculate the reclamation percentage using the identified volume for reclamation and the total system volume:Reclamation Percentage (%) = (69 ('MT')) X 100 = 69%

[0136] This calculation shows that 69% of 100MT system can be returned as reclaimed fluid. The service pulls out X amount of the degraded HTF system, reclaims it, then returns it back into the HTF system to blend. Then another X amount is removed, reclaimed, and returned. This is repeated until a preset target of degraded byproducts is achieved. The 69% in this instance is indicative that the end of life customer system can be reclaimed. 69% of the fluid can be blended with virgin HTF fluid. If reclamation is tried a higher percentage, then the resulting system will have more degradation and shorter lifetime than the target. In other words, 100% of thesystem is processed during the reclamation process - fluid is removed, reclaimed, and returned to the system in cycles. The 69% refers to the final volume of fluid that can be reclaimed and reintegrated into the system, meeting degradation, and lifetime targets. The remaining volume (31% in this case) represents waste that cannot be reclaimed to the target quality and needs to be replaced with virgin HTF fluid. The reclamation process involves iterative removal, reclamation, and reintroduction of the fluid. The goal is to reclaim as much fluid as possible while achieving acceptable quality levels (e.g., targets for low boilers, high boilers, viscosity, etc.). The "69%" here represents the reclaimed fraction of the total system volume that meets quality and lifetime standards, not the fraction of the system that is physically processed.

[0137] The customer plant in 414 represents the industrial facility or plant where the HTF is used in the heat transfer system. This is the location where the reclamation process is carried out. The system's total capacity' is 100 metric tons (indicated in 414), indicating the scale of operations. The used fluid 414 Refers to the HTF currently in the system, which has degraded over time due to thermal and chemical stresses. This fluid contains contaminants such as high and low boilers, moisture, and insoluble solids, which need to be removed to restore fluid quality. The reclaimed fluid 416 represents the fluid that has been treated and cleaned through the reclamation process. The reclaimed fluid 416 is returned to the system, now free of contaminants and with improved properties, enhancing the efficiency and lifespan of the HTF.

[0138] The predicted waste removal 418 refers to the predicted total amount of waste material removed from the system during reclamation for a given predicted reclamation treatment option. Predicted waste removal 418 includes predicted degraded fluid components, contaminants like low and high boilers, moisture, and insoluble solids. Proper disposal of this waste is crucial for environmental compliance and operational safety. In some embodiments, waste removal 418 represents the predicted waste disposal 124 as illustrated in FIG. 1.

[0139] In some embodiments, the total waste removed as indicated in 418 is calculated using the following equation:Waste Removed (MT) = Total System Volume (MT) xReclamation Percentage x Degraded Fractionwhere Total System Volume (MT) is the total volume of HTF in the system, Reclamation Percentage is the percentage or amount of usable material returned to the customer or recovered through reclamation and Degraded Fraction is the fraction of the reclaimed fluid that is considered waste, based on the concentration of contaminants and degraded components.

[0140] In an illustrative example of calculating 418, the given data may be as follows: The 100MT system has 50% good HTF chemistry in it and the rest is degraded byproducts. If 31% of the bad material / byproducts are removed, then the remaining blend is able to reach the spec required reclaimed spec (<0.5% CCR and / or 2% HBs). Thus 69% is reclaimed for the purposes of the customer and 31% virgin HTF is needing to be purchased.

[0141] As described above, the EOL 420 refers to the estimated time at which the heat transfer fluid (HTF) (e.g., the reclaimed fluid 416) will require replacement or significant reclamation a second based on the selected treatment option. In some embodiments, the EOL 420 represents the user interface elements 208, 218, 228, or 238 of FIG. 2.

[0142] FIG. 5 is a screenshot of an example user interface page 500 illustrating parameters used to make reclamation predictions, according to some embodiments. In some embodiments, the page 500 represents a complex spreadsheet with several embedded functions. The page 500 shows a detailed dataset including calculations from the "Heat Transfer Fluids Reclamation Calculator v 1.5" of the left panel 502 and "HTF Reclamation Prediction Model vl.5” of the right panel 508. Each section represents specific aspects of the heat transfer fluid (HTF) system and its reclamation process. Below is an analysis of each dataset and an explanation of how the predictions can be made.

[0143] In some embodiments, the information within the left panel 502 represents or is included within existing system parameters of one or more databases (e.g., the historical customer data 114, the reclamation service provider data 116, or laboratory analysis data 112) or manually edited by a technical service representative or other personnel. For example, some of the data under system information 504 and / orthe mobile reclamation information 506 are measurements pulled from the most recent in-service sample analysis in the laboratory analysis data 112 and some additional testing prior performing reclamation fluid treatment.

[0144] The “System Information” 504 refers to parameter values associated with a customer’s heat transfer fluid system. This includes the following: a “Fill Volume” value, which is the total volume of HTF in the system, given (e.g., 100 Metric Tons). The “density” of the HTF (e.g.. 24°C) helps in calculating the mass of the fluid in different conditions. The “On-stream Time” refers to the operating hours of the system per year, which is useful for estimating wear and degradation rates. “Severity Factor MR” is a factor indicating the severity of the operating conditions on the fluid. This affects the degradation rate. The “Operating Temperature” is the temperature at which the system operates, useful for determining degradation rates. The “Leakage Rate” refers to the rate at which fluid is lost from the system per year, expressed as a percentage.

[0145] The “Max. Low Boiler / High Boiler” refers to the maximum allowable percentages for low boilers and high boilers in the system before they are considered problematic. The “Max. CR% Limit” refers to the maximum allowable carbon residue percentage. The “Time Interval” refers to the time interval over which measurements or predictions are made.

[0146] The “Mobile Reclamation Information” 506 refers cunent parameters associated with a particular heat transfer fluid. For example, “Current Low Boilers (SIMDIS)” refers to the current percentage of low boilers in the fluid. “Current High Boilers (SIMDIS)” refers to the current percentage of high boilers. “Total PHT & TP (GC)” refers to Partially Hydrogenated Terphenyls (PHT) and Terphenyls (TP). These are HTF chemistries to be kept in a customer system. Everything else is considered a high boiler or low boiler. “Current Carbon Residue” is the current percentage of carbon residue. “Target Carbon Residue” is the desired level of carbon residue postreclamation. “Fluid lifetime rated on CR or HB” is the estimated fluid lifetime based on carbon residue or high boilers.

[0147] The right panel in 508 is the “HTF Reclamation Prediction Model” output values. The data shown in the right panel 508 are all generated values predictingthe Reclamation process and quality of the HTF fluid after treatment is completed. In some embodiments, such output values in the right panel 508 are included in or represent the reclamation treatment option(s) 118 of FIG. 1. The '“Estimated Reclaimed Fluid" is the amount of fluid estimated to be reclaimed after the treatment process. The “Estimated Fluid Replacement” is the amount of new fluid estimated to be required to replace the degraded fluid. The “Estimated Low Boilers Removed” is the volume of estimated low boilers removed during reclamation. The “Estimated High Boiling Residue Removed” refers to the predicted volume of high boilers removed. “Remaining Used Fluid in System” refers to the predicted amount of original EOL fluid chemistry remaining in the system after Reclamation. The “Estimated High Boilers / Low Boilers” is the predicted post-reclamation percentages of high boilers and low boilers. The “System Reclaimed Percent” is the predicted percentage of the total system volume reclaimed. The “Top-Off %” is the predicted percentage of the system volume that needs to be topped off with new fluid. “Low Boiler Generation / High Boiler Generation” refers to the predicted rates at which low boilers and high boilers are generated per time period, such as a year. The “Time Before Venting” parameter refers to the predicted time before venting is necessary to remove volatile components. The “Time Before Replacement (HB Limit / CR Limit)” refers to the estimated time before the fluid needs replacement due to high boiler or carbon residue limits. The “Replacements in Time Interval” is the predicted number of replacements required within a specific time interval, based on HB and CR limits. The “Leakage Makeup / Avg. Vent Makeup / Avg. Fluid Use” is the amount of fluid predicted to be lost due to leakage and venting, and the average fluid use per year.

[0148] In some embodiments, the values as indicated in the right panel 508 are predicted via the “HTF Reclamation Prediction Model” (e.g., the model(s) 108 of FIG.1) based on the values indicated in the left panel 502. For example, Estimated Reclaimed Fluid (MT) is calculated as follows in some embodiments:Estimated Reclaimed Fluid = Fill Volume x System Reclaimed Percent.For example, if the Fill Volume is 100 MT (from the left panel) and the System Reclaimed Percent, 69.2%, the Estimated Reclaimed Fluid=100 MT*0.692=69.2 MT.

[0149] In some embodiments, the Estimated Fluid Replacement (MT) is calculated as follows:Estimated Fluid Replacement = Fill Volume * ( 1 —For example, if the Fill Volume is 100 MT and the Estimated Reclaimed Fluid is 59.5 MT, the Estimated Fluid Replacement = 100 MT-59.5 MT=40.5 MT.

[0150] In some embodiments, the Estimated Low Boilers Removed (MT) is calculated as follows: Estimated Low Boilers Removed = Fill Volume x Current Low Boilers xSystem Recla‘™edPercentexample, if the Fill Volume is 100MT, Current Low Boilers: 5%, System Reclaimed Percent: 69.2%. the Estimated Low Boilers Removed is 100 MT *0.05 *0.692=3.46 MT. Tn some embodiments, the estimated High Boiling Residue Removed (MT) is similarly calculated as follows: Estimated High Boiling Residue Removed = Fill Volume *T T. , > .. System Reclaimed PercentCurrent Hig °h Boilers * - 100 .

[0151] In some embodiments, Estimated High Boilers (Post-Reclamation prediction, %) is calculated as follows:Estimated High Boilers = Current High Boilers - High Boilers Removed PercentageFor example, if the current High Boilers is 10%, the High Boilers Removed Percentage is based on the removal process effectiveness - Estimated High Boilers = 10% - 8.5% = 1.5%. In some embodiment, the estimated Low Boilers (Post-Reclamation predict, %) is similarly calculated as: Estimated Low Boilers = Current Low Boilers -Low Boilers Removed Percentage

[0152] In some embodiments, the System Reclaimed Percent (%) is derived from the "Estimated Reclaimed Fluid" and the "Fill Volume." For example, this is represented as follows in some embodiments:Svstem Reclaimed Percent = (Estimated ReclaimedFimd\^QQV Fill Volume / In some embodiments, “Top-off’ is calculated as follows:Top-off 100

[0153] In some embodiments, “Low Boiler Generation / High Boiler Generation (Annual Rates)” are calculated based on historical data and system operating conditions (e.g., operating temperature, severity factor). In some embodiments, “Time Before Venting / Replacement” is calculated Based on degradation rates and critical limits (e.g., low boilers, high boilers, TAN). In some embodiments, “Leakage Makeup / Avg. Vent Makeup / Avg. Fluid Use” is calculated based on known leakage and venting rates and operational data (e.g., leakage rate, vent rate, on-stream time). The values in the HTF Reclamation Prediction Model panel are predicted using these equations, incorporating data from the system's current state and historical degradation patterns. These calculations allow for informed decisions on maintenance and reclamation strategies, ensuring optimal system performance and fluid quality.

[0154] FIG. 6 depicts a diagram of one or more example neural network(s) 605 (referred to as a “neural network”) that is trained to generate multiple decision statistics indicative of various predictions, according to some embodiments. In some embodiments, the neural network 605 represents the reclamation model(s) 108 of FIG.1, and / or any of the predictions described in relation to FIG. 2, FIG. 3, FIG. 4, or FIG.5. In some embodiments, the neural network 605 represents any suitable model functionality, such as supervised learning (e.g., using logistic regression, using back propagation neural networks, using random forests, decision trees, etc.), unsupervised learning (e.g., using an Apnori algorithm, using K-means clustering), semi-supervised learning, reinforcement learning (e.g., using a Q-leaming algorithm, using temporal difference learning), a regression algorithm (e.g., ordinary least squares, logistic regression, stepwise regression, multivariate adaptive regression splines, locally estimated scatterplot smoothing, etc.), an instance-based method (e.g., k-nearest neighbor, learning vector quantization, self-organizing map, etc.), a regularization method (e.g., ridge regression, least absolute shrinkage and selection operator, elastic net, etc.), a decision tree learning method (e.g., classification and regression tree, iterative dichotomiser 3. C4.5, chi-squared automatic interaction detection, decision stump, random forest, multivariate adaptive regression splines, gradient boosting machines, etc.), a Bayesian method (e.g., naive Bayes, averaged one-dependenceestimators, Bayesian belief network, etc.), a kernel method (e.g., a support vector machine, a radial basis function, a linear discriminate analysis, etc ), a clustering method (e.g.. k-means clustering, expectation maximization, etc.), an associated rule learning algorithm (e.g., an Apriori algorithm, an Eclat algorithm, etc.), an artificial neural network model (e.g., a Perceptron method, a back-propagation method, a Hopfield network method, a self-organizing map method, a learning vector quantization method, etc.), a deep learning algorithm (e.g., a restricted Boltzmann machine, a deep belief network method, a convolution network method, a stacked auto-encoder method, etc.), a dimensionality reduction method (e.g., principal component analysis, partial lest squares regression, Sammon mapping, multidimensional scaling, projection pursuit, etc.), an ensemble method (e.g., boosting, bootstrapped aggregation, AdaBoost, stacked generalization, gradient boosting machine method, random forest method, etc.), and / or any suitable form of machine learning algorithm.

[0155] The neural network 605 is modeled as a data flow graph (DFG), where each node (e.g., 621) in the DFG is an operator with an input and output tensor, such as 620 and 622. A "‘tensor” (e.g., a vector) is a data structure that contains values representing the input, output, and / or transformations processed by the operator. Each edge of the DFG depicts the dependency between the operators. Neural network 605 includes an input layer, an output layer and one or more hidden layers. An Input layer is the first layer of the neural network 605. The input layer receives pre-processed (e.g., via the pre-processing 604 or 616) input data represented by 603 and 615. The Output layer (e.g., a classification layer) is the last layer of neural network 605. The output layer generates the predictions, which is represented by the inference and predictions 609 and 607. Neural network 605 may include any number of hidden layers. Hidden layers are intermediate layers in neural network 605 that perform various operations.

[0156] Each node in FIG. 6, such as node 621 , is associated with or includes an activation tensor, such as input tensor 620, output tensor 622, and / or intermediate tensors. An “activation tensor” is a tensor that is an input, intermediate, and / or output to at least one neural network layer (e.g., as modeled going from left to right), as illustrated by the flow of data from input tensor 620 to output tensor 622. This is different than a weight tensor, such as 624, where weight tensors are modeled asflowing upward (not being actual inputs or outputs). In other words activation tensors represent some form of the neural network inputs 603 and 615. For example, the input tensor 620 or node 621 can represent specific data points, such as the presence or absence of particular parameter values (e.g., low boiler values, viscosity, etc.), whereas a weight tensor represents the weight values indicating node activation / inhibition values indicating significance of the particular data point for the overall prediction at 609 or 607.

[0157] Each node in the network 605 is also be associated with or include and / or a weight tensor (e.g., 624), which include weight values. A “weight” in the context of machine learning may represent the importance or significance of a feature or feature value for prediction. For example, each feature (e.g., parameter value, treatment option, etc.) may be associated with an integer or other real number where the higher the real number, the more significant the feature is for its prediction. In some aspects, a weight in a neural network represents the strength of a connection between nodes or neurons from one layer (an input) to the next layer (a hidden or output layer). A weight of 0 may mean that the input (e.g., the input tensor 620) will not change the output (e.g., the output tensor 622), whereas a weight higher than 0 changes the output. The higher the value of the input or the closer the value is to 1, the more the output will change or increase. Likewise, there can be negative weights. Negative weights may proportionately reduce the value of the output. For instance, the more the value of the input increases, the more the value of the output decreases. Negative weights may contribute to negative scores. For example, a particular low boiler value may be highly correlated with a specific recommended reclamation treatment option and so neural network layers or nodes representing the particular low boiler value may be weighted higher so that that this data is activated or taken into account when making a final prediction score.

[0158] Each node of the neural network 605 may additionally perform a function using the activation tensors and weight tensors, such as activation functions, matrix multiplication, normalization, or the like. In some examples, the nodes in the neural network 605 are fully connected or partially connected. Continuing with FIG. 6, each node may process an input in 603 and 615 (or portion thereof) using activationtensors and weight tensors. In some examples, in response to receiving the deployment input(s) 603 and the training data input(s) 315, the neural network 605 first performs pre-processing 604 or 616, such as encoding or converting such input into machine-readable indicia representing the entire input (e.g., a tensor representing all of the deployment input(s) 603). Responsively, the node may then receive an input tensor, which may, for example, represent whether a feature (e.g., a particular parameter, such as density value) are present in the input. In some examples, the input tensor is an Tridimensional tensor, where N can be greater than or equal to one. In some examples, an input tensor 620 represents the input data of neural network 605 if the node is in the input layer. In some examples, the input tensor 620 is also the output of another node in the preceding layer. In some examples after a node, such as the node 621, performs an operation using the input tensor 620, it generates an output tensor 622, which is then passed to the other neurons in the hidden layer and / or output layer. The output tensor 622 represents the output processed by the node 621. For example, the output tensor 622 may be a matrix representing the product of matrix multiplication or a matrix indicating whether particular parameter values were present. In various aspects, the output tensor 622 represents an input of another node in the succeeding layer (i.e., the output layer).

[0159] In some examples, node 621 applies a weight tensor 624 to the input tensor 620 via a linear operation (e.g., matrix multiplication, addition, scaling, biasing, or convolution). All other nodes in the neural network may perform identical functionality. In some examples, the result of the linear operation is processed by a nonlinear activation, such as a step function, a sigmoid function, a hyperbolic tangent function (tan h), and rectified linear unit functions (ReLU) or the like. The result of the activation or other operation is an output tensor 622 that is sent to a subsequent connected node that is in the next layer of neural network 605. The subsequent node uses the output tensor 622 as the input activation tensor to another node.

[0160] Each of the functions in the neural netw ork 605 may be associated with different coefficients (e.g.. weights and kernel coefficients) that are adjustable during training. For example, after preprocessing 616 (e.g., normalization, feature scaling, and extraction) in various aspects, the neural netw ork 605 is trained using a data set of thepreprocessed training data inputs 615 in order to make acceptable loss training predictions at the appropriate weights to set the weight tensors. This will help later at deployment time to make a correct inference 609. In some aspects, learning or training includes minimizing a loss function between the target variable (for example, a correct reclamation treatment option recommendation) and the actual predicted variable (for example, an incorrect reclamation treatment option recommendation). Based on the loss determined by a loss function (for example, Mean Squared Error Loss (MSEL), crossentropy loss, etc.), the loss function leams to reduce the error in prediction over multiple epochs or training sessions so that the neural network 605 leams which features and weights are indicative of the correct inferences, given the inputs. Accordingly, it is desirable to arrive as close to 100% confidence in a particular classification or inference as much as possible so as to reduce the prediction error.

[0161] Subsequent to a first round / epoch of training, the neural network 605 makes predictions with a particular weight value, which may or may not be at acceptable loss function levels. For example, the neural network 605 may process the pre-processed additional training data inputs 615 a second time to make another pass of predictions. This process may then be repeated over multiple iterations or epochs until the weight values in the weight tensors are learned for optimal predicted values and / or the loss function reduces the error in prediction to acceptable levels of confidence.

[0162] Continuing with FIG. 6, in some examples, the neural network 605 is trained in a supervised manner using annotations or labels. For example, in some examples, training includes (or is preceded by) annotating / labeling training data 615 so that the neural network 605 learns associations between the features or weights and corresponding labels, which is used to change the weights / neural node connections for future predictions. As illustrated, for example, in the training data input(s) 615, they include several annotated pairs - parameter - treatment option pairs, treatment option - FL pairs, treatment option - cost pairs, treatment option - new fluid pairs, treatment option - % of system reclaimed pairs, and treatment option - waste removal pairs.

[0163] Each of these input-output pairs in 615 are used to teach a model how to map inputs to the desired outputs. This process is essential in supervised learning, wherethe goal is to learn a function f that maps inputs x to outputs y. The neural network 605 leams by adjusting its internal parameters to minimize the difference between the predicted outputs and the actual outputs (ground truth labels). A loss function measures the discrepancy between the model's predictions and the actual outputs. For example, in regression tasks, mean squared error (MSE) is commonly used, while in classification, cross-entropy loss is typical. The neural network 605 uses the loss to update its parameters during training, aiming to reduce the loss over time. During training, the model processes each input (of the pairs), predicts an output (of the pairs), and then calculates the loss using the actual output (of the pairs). It then adjusts its weights using optimization algorithms like gradient descent to minimize the loss.

[0164] In certain contexts, positive and negative samples are used in the example input-output pairs. Positive samples are examples where the output label corresponds to the target class or condition the neural network 605 should recognize. The neural network 605 leams to associate specific input features with positive outcomes. Negative samples are examples where the output label corresponds to the opposite or non-target class, (i.e., the incorrect prediction).

[0165] The “parameter-treatment option pairs” in 615 are used to predict the reclamation treatment option(s) in 607. In other words, a dataset of various heat transfer fluid samples and their parameter values (e.g., viscosity, high boilers, fill volume, etc.) are labeled with a particular treatment option that was successfully performed given those parameters. The “treatment option-FL pairs” are used to predict the projected FLE in 607. In other words, a dataset of particular reclamation treatment options (and / or other parameters) are labeled with the fluid life that the given heat transfer fluid experience when exposed to the particular reclamation treatment option. The “treatment option - cost pairs” are used to predict the “estimated cost” (e.g., corresponding to the cost in 210 of FIG. 2) associated with implementing a particular treatment option. In other words, a dataset that includes a particular treatment option (and / or other parameters) is labeled with a ground truth cost to actually implement such particular treatment option.

[0166] The “treatment option - new fluid pairs” are used to predict the “quantity and type of new fluid required” (e.g., as illustrated in 406 of FIG. 4) as indicated in 607.In other words, a dataset of a particular treatment option (and / or other parameters) are labeled with the actual quantity and type of heat transfer fluid used for a given reclamation process, for example. The “treatment option - % of system reclaimed pairs” are used to predict the “% of system needing to be reclaimed” (e.g., 410 of FIG. 4) in 607. In other words, a dataset including a particular reclamation treatment option is labeled with a ground truth indicating the actual percentage of a heat transfer fluid system that was reclaimed when such option was implemented. The “treatment option - waste removal pairs” are used to predict the “predicted waste removal” (e.g., 418 of FIG. 4) as indicated in 607. In other words, a dataset that includes a particular treatment option is labeled with a ground truth indication of how much waste was actually removed when the treatment option was implemented.

[0167] Subsequent to the neural network 605 training, the neural network 605 (for example, in a deployed state) receives the pre-processed deployment input(s) 603 (e.g., the heat transfer fluid sample data 102 of FIG. 1). When a machine learning model is deployed, it has been trained, tested, and packaged so that it can process data it has never processed. Responsively, in some aspects, the deployment input(s) 603 are fed to the neural network 605, which then uses the same weight tensors (e.g., 624) that were learned via training so that the neural network 605 can produce the correct inference predictions 609. For example, the input tensor 620 can include new values (e.g., parameter values corresponding to new data) which is then multiplied or otherwise combined with the weight tensor 624, representing the same weight values learned at training, in order to make the inference prediction(s) 609.

[0168] FIG. 7 is a flow diagram of an example process for generating a model prediction of at least one reclamation treatment option for a given heat transfer fluid sample, according to some embodiments. The process 700 (and / or any of the functionality described herein, such as 800) may be performed by processing logic that comprises hardware (e.g., an Al hardware accelerator, circuitry', dedicated logic, programmable logic, microcode, etc.), software (e.g., instructions run on a processor to perform hardware simulation), firmware, or a combination thereof. Although particular blocks described in this disclosure are referenced in a particular order at a particular quantity, it is understood that any block may occur substantially parallel with or beforeor after any other block. Further, more (or fewer) blocks may exist than illustrated. Added blocks may include blocks that embody any functionality described herein (e.g., as described with respect to FIG. 1 through FIG. 10). The computer-implemented method, the system (that includes at least one computing device having at least one processor and at least one computer readable storage medium), and / or the computer readable medium as described herein may perform or be caused to perform the process 700 or any other functionality described herein.

[0169] Per block 702, some embodiments receive first data representing one or more parameters of a sample of a heat transfer fluid (or simply one or more parameters of a heat transfer fluid). For example, such parameters can be any of those described with respect to the system information 504 and / or the mobile reclamation information 506 of the left panel 502 in FIG. 5. Alternatively or additionally, parameters include a high boilers value, a low boilers value, a moisture content value, an acid number value, an insoluble solids value, a viscosity value, a carbon residue value, a flash point value, a non-evaporable content value, a contamination value, and a mixture value. Each of these parameters are described in more detail below.

[0170] Various parameters extracted from heat transfer fluid samples are important or indicative of particular reclamation treatment options, quality / condition scores, or states when they are at certain levels or ranges. Provided for convenience below are descriptions of several parameters as they relate to heat transfer fluids. In some embodiments, there are six key parameters used to determine a fluid quality or fluid life — viscosity, moisture content, flash point, acidity, insoluble solids, and composition / degradation. These can indicate developing problems, enabling potential corrective actions to protect fluid performance / life.

[0171] Viscosity can be readily measured in a fluids lab. e.g., by the ASTM-445 test method, or similar technique. In the ASTM D-445 method, a fluid sample can be held at a precisely controlled temperature while the time for a known volume of the fluid to pass through a calibrated tube is measured. From the elapsed time, the viscosity is calculated.

[0172] The fluid s viscosity is a measure of its resistance to flow. Fluids of greater viscosity7will require higher-pumping horsepower requirements and willadversely affect the degree of turbulence at heat exchange surfaces which can lower heat transfer coefficients. Not only can elevated viscosity reduce heat transfer performance at high temperatures, it can also affect the ability to pump the heat transfer fluid during cold weather start-up conditions. Viscosity is related to the molecular weight of fluid components. Generally, lower molecular weight components decrease viscosity and higher molecular weight components increase viscosity of the heat transfer fluid. Contamination from leaked process streams, incorrect material added to the heat transfer fluid system, and solvents from system cleaning, as well as thermal stressing and oxidation, may be the source of materials that increase or decrease viscosity. The typical corrective action (e.g., as determined via the maintenance recommendation module 108) to address too low a viscosity would be the removal of low boiling components by circulating the heated fluid through the expansion tank with inert gas purge of the vapor space while venting to a safe location. Condensation and collection for proper disposal of the removed low-boiling organics is recommended unless vented to a properly designed flare. Correcting for high viscosity requires either aged fluid removal and replacement or dilution with unused heat transfer fluid.

[0173] Inert gas blanketing is an effective method of minimizing fluid oxidation by blanketing the expansion tank with an inert gas such as nitrogen, carbon dioxide, or natural gas. The purpose of inert gas blanketing is to maintain a nonreactive atmosphere in the vapor space of the expansion tank, preventing the entrance of air and moisture which can adversely affect fluid life. An uninterrupted supply of inert gas, usually nitrogen, controlled by pressure regulators for both inlet and outlet flow may be necessary' to obtain this protection. Pressures used should be kept as low as possible inside the expansion tank to minimize inert gas usage. Maintaining a positive pressure slightly over atmospheric barometric pressure is all that may be necessary to prevent air and moisture from entering the tank. A manual vent valve also should be installed to facilitate purging of the expansion tank’s vapor space if it becomes necessary.

[0174] Moisture content may be analyzed by the Karl Fischer titration technique. Moisture content should be kept quite low when operating at high temperatures to avoid issues related to its flashing into water vapor. Inability7tomaintain low moisture content is an indicator of either an aqueous leak into the system, or perhaps the addition of “wet” make up fluid.

[0175] When operating at very high temperatures, excess moisture content can prevent the ability to circulate the heat transfer fluid due to its flashing into vapor at the circulation pump intake, creating cavitation. Extended operation with cavitation can lead to excessive heat transfer fluid degradation in heater coils due to lower mass flow rates delivered from the pump. Also corrosion may be induced by elevated concentrations of system moisture. In cooling systems, a high moisture content of the fluid will increase the risk of formation of ice crystals on chiller surfaces. This can decrease the efficiency of heat transfer and deteriorate the overall system performance. On new system start-ups, operators may remove residual moisture from the system (from hydrostatic testing) to enable the heat transfer fluid to heat fully to the desired operating temperature. For systems in operation, increasing moisture content may be caused by in-leakage of water from aqueous process steams or from steam systems, or by moisture intake via an expansion tank open to atmosphere. Excess moisture can typically be vented from the expansion tank using the low-boiler venting method. To achieve low ppm moisture levels required for the cooling operation, molecular sieves can be placed in side stream operation.

[0176] The flash point of a high-temperature fluid is commonly measured by the Cleveland Open Cup (COC) method, ASTM D-92. A closed cup technique is also useful in classifying fluids and is run per ASTM D-93. The flash point is the lowest temperature of the fluid under the test conditions where ignition of the vapors above the liquid can occur, yet evaporation rate is too low to sustain combustion. Flash points are important in electrical classification and hazard analysis.

[0177] While many heat transfer fluids have relatively high flash points, they often are not classified as fire resistant. However, heat transfer fluid systems are typically closed systems. Therefore, a release of fluid may only occur in case of accidents or malfunctions and it is typically safe to operate such well-designed and maintained systems and fluids even at temperatures well above the fluid’s flash point. Flash point is a property to be considered in the hazard evaluation of operating systems with combustible fluids. A significantly depressed flash point of the in-service heattransfer fluid may not only increase the fire hazard in case of leakages and the presence of an effective ignition source, it may also affect the area electrical classification of the system in extreme cases. Typically, routine venting of low-boiling thermal degradation products from the expansion tank to a safe location will maintain the fluid’s open-cup flash point to within 25°C (45°F) of the flash point of unused fluid.

[0178] Acidity of fluids is commonly measured by ASTM D-664, which is a potentiometric titration. Fluid oxidation results in accumulations of carboxylic acids which lower the apparent pH or raise the total acid number (TAN). Typically, unused organic heat transfer fluids will have a near zero acid number.

[0179] High acid numbers could indicate severe fluid oxidation, which is most often a result of hot fluid exposure to air in the expansion tank. But they may also indicate possible contamination from improper material added to the heat transfer fluid system inadvertently or fluid leaked from the process side of heat exchangers. If the acidity becomes excessive, the machine components could corrode and fail. Oxidation and corrosion products can form sludge and deposits that can also decrease heat transfer rates by fouling. A condition of this nature is typically best corrected by removing the material and replacing it with new fluid, with serious consideration given to a system flush to remove residual acidity. If the high acidity was caused by oxidation, inerting the vapor space in the expansion tank should be considered. System inerting is a highly effective means of protecting against unwanted increases in fluid acidity and oxidative degradation.

[0180] Insoluble solids content is essentially a measure of the concentration of solids in the fluid at room temperature. Organic solids result from exceeding their solubility limit in the fluid. Other solids can include carbon, small portions of gasket materials and metal shavings, and some rust.

[0181] The presence of solvent (e.g., acetone or pentane) insoluble solids generally indicates contamination from dirt, corrosion products, or severe oxidative or thermal stressing. This condition may cause fouling of heat transfer surfaces which would deteriorate heat transfer performance. Also, plugging of small diameter lines or narrow heat transfer passages could occur. Finally, large amounts of insoluble solids may contribute to wear and plugging of mechanical seals and valves resulting inequipment failure, operational problems, and increased maintenance requirements. If these problems occur, side stream fdtration can usually provide ongoing protection against solids-related deposits and their potential consequences. If solids contamination is extremely high, fluid may need to be removed for external filtration and the system may need to be cleaned. Specialized flushing fluids, designed by heat transfer fluid suppliers, can be effective in removing fouling deposits from most synthetic and mineral oil fluid systems. Modest solids content may require filtering with successively finer rated filter element sizes to get the situation under control. A suggested filter rating generally is 10 to 25 micron for ongoing fluid maintenance.

[0182] Gas chromatography allows the quantification of compounds which have boiling points lower than the initial boiling point (low boilers [LBs]) and higher than the final boiling point (high boilers [HBs]) of the unstressed heat transfer fluid. This analysis provides a measure of the degree of fluid degradation experienced and can provide an indicator of organic contamination. Gas chromatography typically does not directly provide a measure of inorganic contamination of organic fluids.

[0183] Thermal cracking of the heat transfer fluid will result in components which are lower in molecular weight and commonly are known as low boilers. High boilers also can be generated when some compounds recombine to produce higher molecular weight materials. Both low- and high-boiling degradation products can create an unfavorable environment for efficient heat transfer system operation.

[0184] Low-boiling components can affect system operation in several ways. First, when present in significant quantities, low boilers can lead to pump cavitation. Severe cases may cause damage to pump seals and, if allowed to continue uncorrected, can damage impellers. Second, when low boilers are present in excessive concentrations, the heat transfer fluid flashpoint and viscosity may be lowered. Third, the increased fluid vapor pressure resulting from the presence of low-boiling components can cause premature and unexpected pressure relief and venting. Finally, excessively rapid formation of low boilers will result in unacceptably high fluid make up costs as the low boilers removed from the system are replaced with fresh fluid. Removal of low boilers is typically accomplished by venting from the expansion tank to a safe location.

[0185] Since an expansion tank is usually installed at a high point in the system, it also can serve as the main venting point of the system for excess levels of low boilers and moisture which may accumulate in the heat transfer fluid. To properly vent a heat transfer fluid system, the expansion tank should be capable of accommodating the circulating flow of hot heat transfer fluid. To remove low boilers, the temperature in the expansion tank should be increased and the tank pressure may be lowered while venting. As they flash into the vapor space, the excess low boilers and moisture can be more effectively removed by sweeping out the expansion tank through the vent line to a safe area (preferably via a cooled condenser). Modest pressure decreases help minimize the loss of good heat transfer fluid in the vent stream.

[0186] The presence of high boilers can increase heat transfer fluid viscosity, which will affect the fluid’s pump-ability at low temperatures and the system’s heat transfer efficiency. Unlike low boilers, high-boiling compounds are typically not removed from the system easily once they are formed. Hence, high boilers continue to accumulate until the maximum recommended concentrations are reached, thereby signaling the end of the recommended fluid life. If high-boiler concentrations are allowed to accumulate beyond that point, sludge and tar deposits can form as the solubility limits for the higher molecular weight compounds are exceeded. Added costs of operation as a result of these sludge deposits include downtime, repairs, clean-out, and lost production. Corrective action would be either a replacement of the fluid or a major dilution with virgin fluid to maintain fluid properties within normal range.

[0187] In some embodiments, the sample of the heat transfer fluid at block 702 represents a most recent or latest sample of heat transfer fluid taken from a system that is currently in operation. Having the most recent or latest sample of heat transfer fluid (HTF) is useful for accurately assessing the current condition of the fluid. The properties of HTF, such as viscosity, acid number (TAN), moisture content, low boilers, high boilers, and insoluble solids, can change rapidly under operational conditions due to factors like thermal degradation, oxidation, and contamination. A recent sample provides up-to-date information, which is useful for, determining the fluid's degradation state, identifying immediate maintenance needs, and deciding whether a reclamation treatment is necessary. In some instances, the accuracy of the reclamation model'spredictions for fluid quality, reclamation treatment options, and system performance improvements depends heavily on the quality and timeliness of the input data. Using the most recent sample ensures that predictions reflect the current state of the system, leading to more accurate and reliable forecasts. Treatment recommendations may be based on the most relevant data, avoiding outdated or irrelevant information that might lead to suboptimal decisions. This effectively allows for adjustments based on real-time conditions, optimizing maintenance schedules and resource allocation.

[0188] Per block 704, some embodiments encode the one or more parameters into first machine-readable characters. For example, block 704 in some embodiments represents or includes the functionality7described with respect to the encoder 106 of FIG. 1. For example, encoding can include vectorizing, normalizing, and / or scaling the parameter(s). Per block 706, some embodiments access, from a data store, in computer storage, second data, where the second data includes at least one of historical customer data (e.g., historical customer data 114 of FIG. 1), a degradation formula (e.g., degradation formula(s) 110 of FIG. 1), existing reclamation service provider data (e.g., reclamation service provider data 116 of FIG. 1), or a laboratory analysis (e.g., laboratory analysis data 112 of FIG. 1) of a reclaimed fluid, where the second data is encoded into second machine-readable characters (e.g., via the encoder 106 of FIG. 1). In some embodiments, the historical customer data includes data records from past instances where users have used a respective heat transfer fluid and undergone a reclamation process. For example, as illustrated in FIG. 1, the reclamation model(s) 108 access or receive any of the datasets in 110, 112, 114, and / or 116.

[0189] Per block 708, some embodiments provide the first machine-readable characters and the second machine-readable characters as input into a model (e.g., the reclamation model(s) 108 as illustrated in FIG. 1), where the model generates a first decision statistic indicative of a prediction of at least one reclamation treatment option recommendation for the heat transfer fluid based at least in part on the parameter(s) and the second data. For example, as illustrated in FIG. 2, the prediction can be of each of the reclamation treatment options 202, 212, 222, and 232. Additionally or alternatively, the reclamation treatment option recommendation(s) can be those described with respect to the inference 609 or training predict! on(s) 607 of FIG. 7. Additionally oralternatively, such reclamation treatment options can be any of those described with respect to 406, 410, and / or 418 of FIG. 4. In some embodiments, block 708 alternatively represents the more active step of a model, such as generating, via a model, the first decision statistic based at least in part on the one or more parameters.

[0190] In some embodiments, the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid at block 708 includes a least one of: estimating a quantity an type of new heat transfer fluid that needs to be purchased (as illustrated and described in 406 of FIG. 4), estimating a proportion of a customer’s system needs to be reclaimed or has been reclaimed (as described in 410 of FIG. 4), or estimating a duration of the at least one treatment option (as described in FIG. 4). In some embodiments, the model used at block 708 model is a machine learning model that is trained on historical data, where the historical data is labeled with input-output pairs (e.g., as described with respect to the training data input(s) 615 of FIG. 6).

[0191] In some embodiments, the generation of the first decision statistic at block 708 is based on system information (e.g., 504 of FIG. 5) associated with a system that uses the heat transfer fluid and reclamation information (e.g., 506 of FIG. 5). In some embodiments, the system information includes at least one of, fill volume of the system, density of the heat transfer fluid, on-stream time of the system, severity' factor of the system, operating temperature of the system, leakage rate of the system, a max low boiler of the system, a max high boiler of the system, a max CR percentage limit of the system, and a time interval of the system, as illustrated in 506 of FIG. 5. In some embodiments, the reclamation information includes at least one of, a heat transfer fluid type of the heat transfer fluid, a current low boilers value of the heat transfer fluid, a current high boilers value of the heat transfer fluid, a total PHT and TP of the heat transfer fluid, a current carbon residue of the heat transfer fluid, or a fluid lifetime associated with the heat transfer fluid, as illustrated under the right panel 508 of FIG.5.

[0192] Per block 710, in response to (or based at least in part on) the generating of the first decision statistic at block 708, some embodiments cause presentation of a first user interface element that indicates the at least one reclamation treatment optionfor the heat transfer fluid. Examples of this are described wi th respect to any of the user interface elements illustrated in FIG. 2, FIG. 3, or FIG. 4.

[0193] In some embodiments, some embodiments generate, via the model at block 708 or a second model, a second decision statistic indicative of predicting a posttreatment fluid quality7(e.g., the "Post-Reclamation Fluid” 304 of FIG. 3) and performance of the one or more parameters (e.g., the parameters 306 of FIG. 3) of the heat transfer fluid according to the at least one reclamation treatment option based at least in part on the model generating the first decision statistic indicative of the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid. And in response to the generating of the second decision statistic, some embodiments cause presentation of a second user interface element that indicates the post-treatment fluid quality or performance of the one or more parameters when the at least one reclamation option is performed. Examples of this are described with respect to FIG. 3 where the model generates a Post-Reclamation Fluid score indicating what the predicted fluid condition would be of a heat transfer fluid assuming that a particular reclamation treatment option is selected and then responsively causing presentation of the dial 304, indicating “Good” fluid condition.

[0194] Some embodiments generate, via the model at block 708 or a second model, a second decision statistic indicative of predicting a life expectancy (e.g., the “Fluid Life Expectancy” FLE of FIG. 2) of the heat transfer fluid based at least in part on the model generating the first decision statistic indicative of the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid. And in response to the generating of the second decision statistic, some embodiments cause presentation of a second user interface element (e.g., the dial 204 of FIG. 2) that indicates the life expectancy of the heat transfer fluid when the at least one reclamation treatment option performed for the heat transfer fluid. Examples of this are described in FIG. 2 where the FLE- and the “Projected End of Life” are calculated.

[0195] In some embodiments, the second data includes as indicated at block 706 includes the existing reclamation service provider data, and the existing reclamation service provider data includes monetary costs to perform reclamation services for a particular heat transfer fluid (e.g., representing the “cost” in the“treatment option -cost pairs” in 615). Some embodiments then compute a monetary cost to implement the at least one treatment option based at least in part on the existing reclamation service provider data and cause presentation of a second user interface element indicating the monetary cost to implement the at least one treatment option. Examples of this are described with respect to the “estimated cost” at inference 609 or training prediction(s) 607, and the “Estimated Option Cost” as indicated in 210 of FIG.2.

[0196] Some embodiments generate, via the model at block 708 or a second model, a second decision statistic indicative of predicting a quantity of waste that needs to be disposed based at least in part on the model generating the first decision statistic indicative of the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid. Examples of this are described with respect to the prediction 418 of FIG. 4 (predicting 31 MT waste will be removed for a given reclamation treatment option) and waste disposal 124 of FIG. 1.

[0197] FIG. 8 is a flow diagram of an example process 800 for training a machine learning model to predict a reclamation treatment option, according to some embodiments. In some embodiments, the process 800 represents how the neural network(s) 605 of FIG. 6 is trained. Per block 803 some embodiments collect historical data on heat transfer fluid (HTF) systems. For example, in some aspects, such historical data includes initial HTF fluid conditions (viscosity, TAN, moisture content, low boilers, high boilers, insoluble solids), operational parameters, and outcomes of previous reclamation treatments.

[0198] Per block 805, some embodiments then label the data (e.g., based on the success of the reclamation treatments). For example, various embodiments label a dataset with particular fluid conditions (e.g.. particular viscosity values) with a particular treatment option that was successful and / or not successful. For instance, positive samples are instances where the treatment was successful, leading to significant improvements in fluid quality or system performance. Negative samples are instances where the treatment was unsuccessful or less optimal, resulting in minimal improvements or negative outcomes.

[0199] Per block 807, some embodiments then extract one or more features, which is indicative of feature engineering. For example, some embodiments extract relevant features from the data, such as fluid condition parameters, treatment types, and operational settings that could influence the outcome of the reclamation process. Per block 809, some embodiments initialize a model’s weights randomly. Initializing the model's weights randomly means assigning initial values to the parameters (weights) of the machine learning model in a way that is not predetermined. These initial values are crucial because they serve as the starting point for the model's learning process.

[0200] Per block 811, particular embodiments initiate a forward pass by computing a decision statistic indicative of a predicted reclamation treatment option for each input sample (e.g., of an input-output pair) using the current weights. Per block 813, particular embodiments then calculate, via a loss function, the loss by computing the prediction against the ground truth output (e.g., of the input-output pair). The loss function measures the difference between the model's predictions and the actual outcomes (ground truth). It guides the model in adjusting its parameters to improve accuracy. For classification, for instance, Binary Cross-Entropy Loss is used when the output is a binary label (e.g., recommend or not recommend) in some embodiments. For regression, Mean Squared Error (MSE) is used in some embodiments when predicting continuous outcomes, like the degree of improvement.

[0201] Per block 815, some embodiments then engage in a backward pass by computing the gradients of the loss with respect to each weight. These gradients indicate how much and in which direction each weight should be adjusted to reduce the loss. For example, some embodiments use Gradient Descent to compute the gradients.

[0202] Per block 817, some embodiments then update the weights using the gradients and a learning rate alpha a. The learning rate controls the step size of the update. Gradients are partial derivatives of the loss function with respect to each weight. They represent the slope of the loss function and indicate how much the loss will change with a small change in the weight. The gradient for a weight informs of the direction in which embodiments should adjust that weight to reduce the loss. The learning rate alpha a is a hyperparameter that controls the size of the steps the algorithm takes in thedirection of the gradient. It essentially scales the gradient, determining how much the weights are adjusted in each iteration.

[0203] Per block 819, it is determined whether the loss converges to a minimum value. This indicates that the training process 800 continues until the loss function reaches its lowest possible value (e.g., a global or local minimum), or at least a point where further decreases in the loss are negligible. The goal of training is to minimize this loss function, as a lower loss indicates better model performance. Convergence refers to the process by which the training of the model stabilizes, meaning that further iterations do not result in significant improvements (i.e., reductions) in the loss function. Convergence is achieved when the model's parameters (weights) have been optimized to the point where the loss function reaches the minimum value, or at least an acceptably low value. If the loss does not converge to the minimum value, then embodiments repeat the forw ard pass, loss computation, backward pass, and parameter update steps (i.e., blocks 811, 813, 815, and 817) for a set number of iterations or until the loss converges to the minimum value, at which point the process stops.

[0204] The invention may be described in the general context of computer code or machine-useable instructions, including computer-executable instructions such as program modules, being executed by a computer (or one or more processors) or other machine, such as a personal data assistant or other handheld device. Generally, program modules including routines, programs, objects, components, data structures, etc., refer to code that perform particular tasks or implement particular abstract data types. The invention may be practiced in a variety of system configurations, including hand-held devices, consumer electronics, general-purpose computers, more specialty7computing devices, etc. The invention may also be practiced in distributed computing environments where tasks are performed by remote-processing devices that are linked through a communications network.

[0205] FIG. 9 is a block diagram of a computing environment 900 in which aspects of the present disclosure are employed in, according to certain embodiments. Although the environment 900 illustrates specific components at a specific quantity, it is recognized that more or less components may be included in the computing environment 900. For example, in some embodiments, there are multiple user devices902 and multiple servers 904, such as nodes in a cloud or distributing computing environment. In some embodiments, some, or each of the components of the system 100 of FIG. 1 are hosted in the one or more servers 904. In some embodiments, the user device(s) 902 and / or the server(s) 904 are embodied in any physical hardware, such as the computing device 1000 of FIG. 10.

[0206] The one or more user devices 902 are communicatively coupled to the server(s) 904 via the one or more networks 910. In practice, the connection may be any viable data transport network, such as, for example, a LAN or WAN. Network(s) 910 can be for example, a local area network (LAN), a wide area network (WAN) such as the Internet, or a combination of the two, and include wired, wireless, or fiber optic connections. In general, network(s) 910 can be any combination of connections and protocols that will support communications between the control server(s) 904 and the user devices 902.

[0207] In some embodiments, a user issues a query on the one or more user devices 902, after which the user device(s) 902 communicate, via the network(s) 910, to the one or more servers 904 and the one or more servers 904 executes the query (e.g., via one or more components of FIG. 1) and causes or provides for display information back to the user device(s) 902. For example, the user may issue a query at the user device 902 that is indicative of a request to provide a reclamation treatment option recommendation for a given heat transfer fluid sample. Responsively, the server(s) 904 can perform functionality necessary to generate the scores.

[0208] The invention may be described in the general context of computer code or machine-useable instructions, including computer-executable instructions such as program modules, being executed by a computer (or one or more processors) or other machine, such as a personal data assistant or other handheld device. Generally, program modules including routines, programs, objects, components, data structures, etc., refer to code that perform particular tasks or implement particular abstract data types. The invention may be practiced in a variety of system configurations, including hand-held devices, consumer electronics, general-purpose computers, more specialty computing devices, etc. The invention may also be practiced in distributed computingenvironments where tasks are performed by remote-processing devices that are linked through a communications network.

[0209] With reference to FIG. 10, computing device 1000 includes bus 10 that directly or indirectly couples the following devices: memory 12, one or more processors 14, one or more presentation components 16, input / output (I / O) ports 18, input / output / output components 20, and illustrative power supply 22. Bus 10 represents what may be one or more busses (such as an address bus, data bus, or combination thereof). Although the various blocks of FIG. 22 are shown with lines for the sake of clarity, in reality, delineating various components is not so clear, and metaphorically, the lines would more accurately be grey and fuzzy7. For example, one may consider a presentation component such as a display device to be an I / O component. Also, processors have memory. The inventors recognize that such is the nature of the art, and reiterate that this diagram is merely illustrative of an exemplary computing device that can be used in connection with one or more embodiments of the present invention. Distinction is not made between such categories as “workstation,” “server,” “laptop,” “hand-held device.” etc., as all are contemplated within the scope of FIG. 10 and reference to “computing device.”

[0210] In some embodiments, the computing device 1000 represents the physical embodiments of one or more systems and / or components described above. For example, the computing device 1000 can represent: the one or more user devices 902, and / or the server(s) 904 of FIG. 9. The computing device 1000 can also perform some or each of the blocks in the process 700, 800, and / or any functionality described herein with respect to FIGs 1-10. It is understood that the computing device 1000 is not to be construed necessarily as a generic computer that performs generic functions. Rather, the computing device 1000 in some embodiments is a particular machine or specialpurpose computer. For example, in some embodiments, the computing device 1000 is or includes: a hardware accelerator (e.g., an Al hardware accelerator), a multi-user mainframe computer system, one or more cloud computing nodes, a single-user system, or a server computer or similar device that has little or no direct user interface, but receives requests from other computer systems (clients), a desktop computer, portable computer, laptop or notebook computer, tablet computer, pocket computer, telephone,smart phone, smart watch, or any other suitable ty pe of electronic device. An Al hardware accelerator is a specialized computing device designed to speed up artificial intelligence (Al) tasks, particularly those involving machine learning and deep learning. These accelerators optimize the performance of algorithms that require intensive mathematical computations, such as matrix multiplications and convolutions, which are common in neural networks. Unlike general-purpose processors (CPUs), Al hardware accelerators are tailored for specific operations used in Al workloads. This specialization allows them to handle tasks more efficiently and with lower power consumption. Many Al tasks, like training deep neural networks, involve a large number of operations that can be processed in parallel. Accelerators are designed with a high degree of parallelism, enabling them to execute multiple operations simultaneously. Al accelerators are capable of processing large volumes of data quickly. Examples of Al accelerators include Graphics Processing Units (GPU), Tensor Processing Units (TPU), Field Programmable Gate Arrays (FPGA), and Applicationspecific Integrated Circuits (ASIC).

[0211] Computing device 1000 typically includes a variety of computer-readable media. Computer-readable media can be any available media that can be accessed by computing device 1000 and includes both volatile and nonvolatile media, removable and non-removable media. By way of example, and not limitation, computer-readable media may comprise computer storage media and communication media. Computer storage media includes both volatile and nonvolatile, removable, and non-removable media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology. CD-ROM. digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can be accessed by computing device 1000. Computer storage media does not comprise signals per se. Communication media typically embodies computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transportmechanism and includes any information delivery media. The term “modulated data signal” means a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal. By way of example, and not limitation, communication media includes wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, RF, infrared, and other wireless media. Combinations of any of the above should also be included within the scope of computer-readable media.

[0212] Memory 12 includes computer storage media in the form of volatile and / or nonvolatile memory. The memory may be removable, non -removable, or a combination thereof. Exemplary hardware devices include solid-state memory, hard drives, optical-disc drives, etc. Computing device 1000 includes one or more processors 14 that read data from various entities such as memory 12 or I / O components 20. Presentation component(s) 16 present data indications to a user or other device. Exemplary presentation components include a display device, speaker, printing component, vibrating component, etc.

[0213] I / O ports 18 allow computing device 1000 to be logically coupled to other devices including I / O components 20, some of which may be built in. Illustrative components include a microphone, joystick, game pad, satellite dish, scanner, printer, wireless device, etc. The I / O components 20 may provide a natural user interface (NUI) that processes air gestures, voice, or other physiological inputs generated by a user. In some instance, inputs may be transmitted to an appropriate network element for further processing. A NUI may implement any combination of speech recognition, touch and stylus recognition, facial recognition, biometric recognition, gesture recognition both on screen and adjacent to the screen, air gestures, head and eye-tracking, and touch recognition associated with displays on the computing device 1000. The computing device 1000 may be equipped with depth cameras, such as, stereoscopic camera systems, infrared camera systems, RGB camera systems, and combinations of these for gesture detection and recognition. Additionally, the computing device 1000 may be equipped with accelerometers or gyroscopes that enable detection of motion.DEFINITIONS

[0214] “And / or’" is the inclusive disjunction, also known as the logical disjunction and commonly known as the “inclusive or.” For example, the phrase “A, B, and / or C,” means that at least one of A or B or C is true; and "‘A, B, and / or C” is only false if each of A and B and C is false.

[0215] A “set of’ items means there exists one or more items; there must exist at least one item, but there can also be two, three, or more items. A “subset of’ items means there exists one or more items within a grouping of items that contain a common characteristic.

[0216] A “plurality of" items means there exists more than one item; there must exist at least two items, but there can also be three, four, or more items.

[0217] “Includes” and any variants (e.g., including, include, etc.) means, unless explicitly noted otherwise, “includes, but is not necessarily limited to.”

[0218] A “user” or a “subscriber” includes, but is not necessarily limited to: (i) a single individual human; (ii) an artificial intelligence entity with sufficient intelligence to act in the place of a single individual human or more than one human; (iii) a business entity for which actions are being taken by a single individual human or more than one human; and / or (iv) a combination of any one or more related “users” or “subscribers’" acting as a single “user” or “subscriber.”

[0219] The terms “receive.” “provide,” "‘send,” “input,” “output,” and “report” should not be taken to indicate or imply, unless otherwise explicitly specified: (i) any particular degree of directness with respect to the relationship between an object and a subject; and / or (ii) a presence or absence of a set of intermediate components, intermediate actions, and / or things interposed between an object and a subject.

[0220] A “module” or “component” is any set of hardware, firmware, and / or software that operatively works to do a function, without regard to whether the module is: (i) in a single local proximity; (ii) distributed over a wide area; (iii) in a single proximity within a larger piece of software code; (iv) located within a single piece of software code; (v) located in a single storage device, memory, or medium; (vi) mechanically connected; (vii) electrically connected; and / or (viii) connected in data communication. A “sub-module” is a “module” within a “module.”

[0221] The terms first (e.g., first cache), second (e.g., second cache), etc. are not to be construed as denoting or implying order or time sequences unless expressly indicated otherwise. Rather, they are to be construed as distinguishing two or more elements. In some embodiments, the two or more elements, although distinguishable, have the same makeup. For example, a first memory' and a second memory may indeed be two separate memories but they both may be RAM devices that have the same storage capacity (e.g., 4 GB).

[0222] The term '‘causing’’ or '‘cause” means that one or more systems (e.g., computing devices) and / or components (e.g., processors) may in in isolation or in combination with other systems and / or components bring about or help bring about a particular result or effect. For example, a server computing device may “cause” a message to be displayed to a user device (e.g., via transmitting a message to the user device) and / or the same user device may “cause” the same message to be displayed (e.g., via a processor that executes instructions and data in a display memory' of the user device). Accordingly, one or both systems may in isolation or together “cause” the effect of displaying a message.EXAMPLE LITERAL SUPPORT

[0223] One or more embodiments described below may be combined with one or more other embodiments. In an example embodiment, a computer-implemented method, comprises: receiving first data representing one or more parameters of a sample of a heat transfer fluid; encoding the one or more parameters into first machine-readable characters; accessing, from a data store in computer storage, second data, the second data includes at least one of, historical customer data, a degradation formula, existing reclamation service provider data, or a laboratory analysis of a reclaimed fluid, the historical customer data including data records from past instances where users have used a respective heat transfer fluid and undergone a reclamation processes, the second data being encoded into second machine-readable characters; providing the first machine-readable characters and the second machine-readable characters as input into a model, wherein the model generates a first decision statistic indicative of a prediction of at least one reclamation treatment option recommendation for the heat transfer fluidbased at least in part on the one or more parameters and the second data; and in response to the generating of the first decision statistic, causing presentation of a first user interface element that indicates the at least one reclamation treatment option recommendation for the heat transfer fluid.

[0224] In some embodiments, the one or more parameters include at least one parameter value selected from a group of parameter values consisting of: a high boilers value, a low boilers value, a moisture content value, an acid number value, an insoluble solids value, a viscosity value, a carbon residue value, a flash point value, a non-evaporable content value, a contamination value, and a mixture value.

[0225] In some embodiments, the method further comprises: generating, via the model or a second model, a second decision statistic indicative of predicting a posttreatment fluid quality and performance of the one or more parameters of the heat transfer fluid according to the at least one reclamation treatment option based at least in part on the model generating the first decision statistic indicative of the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid; and in response to the generating of the second decision statistic, causing presentation of a second user interface element that indicates the post-treatment fluid quality or performance of the one or more parameters when the at least one reclamation option is performed.

[0226] In some embodiments, the sample of the heat transfer fluid represents a most recent or latest sample of heat transfer fluid taken from a system that is currently in operation.

[0227] In some embodiments, the method further comprises : generating, via the model or a second model, a second decision statistic indicative of predicting a life expectancy of the heat transfer fluid based at least in part on the model generating the first decision statistic indicative of the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid; and in response to the generating of the second decision statistic, causing presentation of a second user interface element that indicates the life expectancy of the heat transfer fluid when the at least one reclamation treatment option performed for the heat transfer fluid.

[0228] In some embodiments, the second data includes the existing reclamation service provider data, and wherein the existing reclamation service provider data includes monetary costs to perform reclamation services for a particular heat transfer fluid, and wherein the method further comprising: computing a monetary cost to implement the at least one treatment option based at least in part on the existing reclamation service provider data; and causing presentation of a second user interface element indicating the monetary cost to implement the at least one treatment option.

[0229] In some embodiments, the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid includes at least one of: estimating a quantity and type, or estimating a duration of the at least one treatment option.

[0230] In some embodiments, the method further comprises : generating, via the model or a second model, a second decision statistic indicative of predicting a quantity of waste that needs to be disposed based at least in part on the model generating the first decision statistic indicative of the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid.

[0231] In some embodiments, the model is a machine learning model that is trained on historical customer data, and wherein the customer historical data is labeled with sample-reclamation treatment pairs.

[0232] In some embodiments, the generation of the first decision statistic is based on system information associated with a system that uses the heat transfer fluid and reclamation information , the system information includes at least one of, fill volume of the system, density of the heat transfer fluid, on-stream time of the system, severity factor of the system, operating temperature of the system, leakage rate of the system, a max low boiler of the system, a max high boiler of the system, a max CR percentage limit of the system, and a time interval of the system, the reclamation information including at least one of, a heat transfer fluid type of the heat transfer fluid, a current low boilers value of the heat transfer fluid, a current high boilers value of the heat transfer fluid, a total PHT and TP of the heat transfer fluid, a current carbon residue of the heat transfer fluid, or a fluid lifetime associated with the heat transfer fluid.

[0233] In one embodiments, a computerized system comprises: one or more processors; and computer storage memory having computer-executable instructions stored thereon which, when executed by the one or more processors, implement a method comprising: receiving first data representing one or more parameters of a heat transfer fluid; encoding the one or more parameters into first machine-readable characters; providing the first machine-readable characters as input into a model, wherein the model generates a first decision statistic indicative of a prediction of at least one reclamation treatment option recommendation for the heat transfer fluid based at least in part on the one or more parameters and second data, the second data includes at least one of, historical customer data, a degradation formula, existing reclamation service provider data, or a laboratory analysis of a reclaimed fluid; and based at least in part on the generating of the first decision statistic, causing presentation of a first user interface element that indicates the at least one reclamation treatment option recommendation for the heat transfer fluid.

[0234] In some embodiments of the system, the method further comprises: generating, via the model or a second model, a second decision statistic indicative of predicting a post-treatment fluid quality and performance of the one or more parameters of the heat transfer fluid according to the at least one reclamation treatment option based at least in part on the model generating the first decision statistic indicative of the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid; and in response to the generating of the second decision statistic, causing presentation of a second user interface element that indicates the post-treatment fluid quality or performance of the one or more parameters when the at least one reclamation option is performed.

[0235] In some embodiments of the system, the method further comprises: generating, via the model or a second model, a second decision statistic indicative of predicting a life expectancy of the heat transfer fluid based at least in part on the model generating the first decision statistic indicative of the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid; and in response to the generating of the second decision statistic, causing presentation of asecond user interface element that indicates the life expectancy of the heat transfer fluid when the at least one reclamation treatment option performed for the heat transfer fluid.

[0236] In some embodiments of the system, the second data includes the existing reclamation service provider data, and wherein the existing reclamation service provider data includes monetary7costs to perform reclamation sendees for a particular heat transfer fluid, and wherein the method further comprising: computing a monetary cost to implement the at least one treatment option based at least in part on the existing reclamation service provider data; and causing presentation of a second user interface element indicating the monetary cost to implement the at least one treatment option.

[0237] In some embodiments of the system, the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid includes a least one of: estimating a quantity and type of new heat transfer fluid that needs to be purchased, estimating a proportion of a customer’s system that can be returned as reclaimed, or estimating a duration of the at least one treatment option.

[0238] In some embodiments of the system, the method further comprises: generating, via the model or a second model, a second decision statistic indicative of predicting a quantity of waste that needs to be disposed based at least in part on the model generating the first decision statistic indicative of the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid.

[0239] In some embodiments of the system, the model is a machine learning model that is trained on historical customer data, and wherein the historical customer data is labeled with sample-reclamation treatment pairs.

[0240] In some embodiments of the system, the generation of the first decision statistic is based on system information associated with a system that uses the heat transfer fluid and reclamation information , the system information includes at least one of, fill volume of the system, density of the heat transfer fluid, on-stream time of the system, severity factor of the system, operating temperature of the system, leakage rate of the system, a max low boiler of the system, a max high boiler of the system, a max CR percentage limit of the system, and a time interval of the system, the reclamation information including at least one of, a heat transfer fluid type of the heat transfer fluid, a current low boilers value of the heat transfer fluid, a current high boilers value of theheat transfer fluid, a total PHT and TP of the heat transfer fluid, a current carbon residue of the heat transfer fluid, or a fluid lifetime associated with the heat transfer fluid.

[0241] In one embodiments, one or more computer storage media has computer-executable instructions embodied thereon that, when executed, by one or more processors, cause the one or more processors to perform a method, the method comprising: receiving first data representing one or more parameters of a heat transfer fluid; based at least in part on the one or more parameters, generating, via a model, a first decision statistic indicative of a prediction of at least one reclamation treatment option recommendation for the heat transfer fluid based at least in part on the one or more parameters; and based at least in part on the generating of the first decision statistic, causing presentation of a first user interface element that indicates the at least one reclamation treatment option recommendation for the heat transfer fluid.

[0242] In some embodiments of the one or more computer storage media, the one or more parameters include system information associated with a system that uses the heat transfer fluid and reclamation information, the system information includes at least one of, fill volume of the system, density of the heat transfer fluid, on-stream time of the system, severity factor of the system, operating temperature of the system, leakage rate of the system, a max low boiler of the system, a max high boiler of the system, a max CR percentage limit of the system, and a time interval of the system, the reclamation information including at least one of, a heat transfer fluid type of the heat transfer fluid, a current low boilers value of the heat transfer fluid, a current high boilers value of the heat transfer fluid, a total PHT and TP of the heat transfer fluid, a current carbon residue of the heat transfer fluid, or a fluid lifetime associated with the heat transfer fluid.

Claims

1. CLAIMSWhat is claimed is:

1. A computer-implemented method, comprising:receiving first data representing one or more parameters of a sample of a heat transfer fluid;encoding the one or more parameters into first machine-readable characters; accessing, from a data store in computer storage, second data, the second data includes at least one of, historical customer data, a degradation formula, existing reclamation service provider data, or a laboratory analysis of a reclaimed fluid, the historical customer data including data records from past instances where users have used a respective heat transfer fluid and undergone a reclamation processes, the second data being encoded into second machine-readable characters;providing the first machine-readable characters and the second machine-readable characters as input into a model, wherein the model generates a first decision statistic indicative of a prediction of at least one reclamation treatment option recommendation for the heat transfer fluid based at least in part on the one or more parameters and the second data; andin response to the generating of the first decision statistic, causing presentation of a first user interface element that indicates at least one reclamation treatment option recommendation for the heat transfer fluid.

2. The method of claim 1, wherein the one or more parameters include at least one parameter value selected from a group of parameter values consisting of: a high boilers value, a low boilers value, a moisture content value, an acid number value, an insoluble solids value, a viscosity value, a carbon residue value, a flash point value, a non-evaporable content value, a contamination value, and a mixture value.

3. The method of claim 1, further comprising:generating, via the model or a second model, a second decision statistic indicative of predicting a post-treatment fluid quality and performance of the one or more parameters of the heat transfer fluid according to the at least one reclamation treatment option based at least in part on the model generating the first decision statistic indicative of the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid; andin response to the generating of the second decision statistic, causing presentation of a second user interface element that indicates the post-treatment fluid quality or performance of the one or more parameters when the at least one reclamation option is performed.

4. The method of claim 1, wherein the sample of the heat transfer fluid represents a most recent or latest sample of heat transfer fluid taken from a system that is currently in operation.

5. The method of claim 1, further comprising:generating, via the model or a second model, a second decision statistic indicative of predicting a life expectancy of the heat transfer fluid based at least in part on the model generating the first decision statistic indicative of the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid; and in response to the generating of the second decision statistic, causing presentation of a second user interface element that indicates the life expectancy of the heat transfer fluid when the at least one reclamation treatment option performed for the heat transfer fluid.

6. The method of claim 1 , wherein the second data includes the existing reclamation service provider data, and wherein the existing reclamation service provider data includes monetary costs to perform reclamation sen-ices for a particular heat transfer fluid, and wherein the method further comprising:computing a monetary cost to implement the at least one treatment option based at least in part on the existing reclamation service provider data; andcausing presentation of a second user interface element indicating the monetary cost to implement at least one treatment option.

7. The method of claim 1, wherein the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid includes at least one of: estimating a quantity and type, or estimating a duration of the at least one treatment option.

8. The method of claim 1, further comprising:generating, via the model or a second model, a second decision statistic indicative of predicting a quantity of waste that needs to be disposed based at least in part on the model generating the first decision statistic indicative of the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid.

9. The method of claim 1, wherein the model is a machine learning model that is trained on historical customer data, and wherein the customer historical data is labeled with sample-reclamation treatment pairs.

10. The method of claim 1, wherein the generation of the first decision statistic is based on system information associated with a system that uses the heat transfer fluid and reclamation information , the system information includes at least one of, fill volume of the system, density of the heat transfer fluid, on-stream time of the system, severity factor of the system, operating temperature of the system, leakage rate of the system, a max low boiler of the system, a max high boiler of the system, a max CR percentage limit of the system, and a time interval of the system, the reclamation information including at least one of, a heat transfer fluid type of the heat transfer fluid, a current low boilers value of the heat transfer fluid, a current high boilers value of the heat transfer fluid, a total PHT and TP of the heat transfer fluid, a current carbon residue of the heat transfer fluid, or a fluid lifetime associated with the heat transfer fluid.

11. A computerized system, comprising:one or more processors; andcomputer storage memory having computer-executable instructions stored thereon which, when executed by the one or more processors, implement a method comprising:receiving first data representing one or more parameters of a heat transfer fluid; encoding the one or more parameters into first machine-readable characters; providing the first machine-readable characters as input into a model, wherein the model generates a first decision statistic indicative of a prediction of at least one reclamation treatment option recommendation for the heat transfer fluid based at least in part on the one or more parameters and second data, the second data includes at least one of, historical customer data, a degradation formula, existing reclamation service provider data, or a laboratory analysis of a reclaimed fluid; andbased at least in part on the generating of the first decision statistic, causing presentation of a first user interface element that indicates the at least one reclamation treatment option recommendation for the heat transfer fluid.

12. The system of claim 11, wherein the method further comprising: generating, via the model or a second model, a second decision statistic indicative of predicting a post-treatment fluid quality and performance of the one or more parameters of the heat transfer fluid according to the at least one reclamation treatment option based at least in part on the model generating the first decision statistic indicative of the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid; andin response to the generating of the second decision statistic, causing presentation of a second user interface element that indicates the post-treatment fluid quality or performance of the one or more parameters when the at least one reclamation option is performed.

13. The system of claim 11, wherein the method further comprising:generating, via the model or a second model, a second decision statistic indicative of predicting a life expectancy of the heat transfer fluid based at least in part on the model generating the first decision statistic indicative of the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid; and in response to the generating of the second decision statistic, causing presentation of a second user interface element that indicates the life expectancy of the heat transfer fluid when the at least one reclamation treatment option performed for the heat transfer fluid.

14. The sy stem of claim 11, wherein the second data includes the existing reclamation service provider data, and wherein the existing reclamation service provider data includes monetary costs to perform reclamation services for a particular heat transfer fluid, and wherein the method further comprising:computing a monetary cost to implement the at least one treatment option based at least in part on the existing reclamation service provider data; and causing presentation of a second user interface element indicating the monetary cost to implement at least one treatment option.

15. The system of claim 11, wherein the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid includes a least one of: estimating a quantity and type of new heat transfer fluid that needs to be purchased, estimating a proportion of a customer’s system that can be returned as reclaimed, or estimating a duration of the at least one treatment option.

16. The system of claim 11. wherein the method further comprising: generating, via the model or a second model, a second decision statistic indicative of predicting a quantity of waste that needs to be disposed based at least in part on the model generating the first decision statistic indicative of the prediction of at least one reclamation treatment option recommendation for the heat transfer fluid.

17. The system of claim 11 , wherein the model is a machine learning model that is trained on historical customer data, and wherein the historical customer data is labeled with sample-reclamation treatment pairs.

18. The system of claim 11, wherein the generation of the first decision statistic is based on system information associated with a system that uses the heat transfer fluid and reclamation information , the system information includes at least one of, fill volume of the system, density of the heat transfer fluid, on-stream time of the system, severity factor of the system, operating temperature of the system, leakage rate of the system, a max low boiler of the system, a max high boiler of the system, a max CR percentage limit of the system, and a time interval of the system, the reclamation information including at least one of, a heat transfer fluid type of the heat transfer fluid, a current low boilers value of the heat transfer fluid, a current high boilers value of the heat transfer fluid, a total PHT and TP of the heat transfer fluid, a current carbon residue of the heat transfer fluid, or a fluid lifetime associated with the heat transfer fluid.

19. One or more computer storage media having computerexecutable instructions embodied thereon that, when executed, by one or more processors, cause the one or more processors to perform a method, the method comprising:receiving first data representing one or more parameters of a heat transfer fluid; based at least in part on the one or more parameters, generating, via a model, a first decision statistic indicative of a prediction of at least one reclamation treatment option recommendation for the heat transfer fluid based at least in part on the one or more parameters; andbased at least in part on the generating of the first decision statistic, causing presentation of a first user interface element that indicates the at least one reclamation treatment option recommendation for the heat transfer fluid.

20. The system of claim 11, wherein the one or more parameters include system information associated with a system that uses the heat transfer fluidand reclamation information, the system information includes at least one of, fill volume of the system, density of the heat transfer fluid, on-stream time of the system, severity factor of the system, operating temperature of the system, leakage rate of the system, a max low boiler of the system, a max high boiler of the system, a max CR percentage limit of the system, and a time interval of the system, the reclamation information including at least one of, a heat transfer fluid type of the heat transfer fluid, a current low boilers value of the heat transfer fluid, a current high boilers value of the heat transfer fluid, a total PHT and TP of the heat transfer fluid, a current carbon residue of the heat transfer fluid, or a fluid lifetime associated with the heat transfer fluid.

Images