System, apparatus and method for managing multiple assets

Abstract

A system, apparatus and method (400) for managing a plurality of assets (105, 302A-N) in a technical facility are disclosed. The method (400) includes receiving, by a processing unit (135), a set of requirements for managing a plurality of assets (105, 302A-N); selecting one or more assets (105, 302A-N) from among the assets (105, 302A-N) based on the received set of requirements; mapping the one or more assets (105, 302A-N) to a corresponding sensing unit (125); extracting information associated with the selected one or more assets (105, 302A-N) and the received set of requirements; determining a performance metric based on information extracted from the knowledge base (122) and the mapped sensing units (125); defining a workflow to be executed based on the determined at least one performance indicator; selecting a configured machine learning model from a machine learning model set based on the defined workflow; and determining a result of the selected machine learning model based on the received set of requirements.

Description

[0001] The present invention relates to the field of predictive systems and more particularly to the management and predictive maintenance of one or more assets in a technical facility.

[0002] In an industrial environment, a large number of assets are used to perform different operations. Assets may include mechanical systems, electromechanical systems, electronic systems, building equipment, etc. Each asset may have different attributes such as availability, performance, behavior, efficiency, maintainability, reliability, serviceability, etc. The attributes of each asset in the equipment may in turn affect the overall performance of the assets in the industrial environment. In order to predict several parameters and determine the overall performance of the assets, predictive analytics techniques are used. Existing predictive analytics solutions use a very general approach to determine the predicted parameters or detect anomalies, where a specific solution is designed once for each requirement.

[0003] There are several shortcomings in the current approaches to managing assets in industrial environments. One major problem with existing solutions is that they do not meet the specific requirements of the target sectors, such as continuous processing (oil and gas industry, pharmaceutical industry, etc.), discrete industries (automotive, FMCG, etc.), building spaces, manufacturing plant settings, etc. Therefore, the existing solutions have low accuracy and low reliability as the solutions are not customized as per the specific requirements. Another major problem with existing solutions is the inability to automatically determine the sensors associated with the assets required for various tasks such as prediction or management of assets. Yet another problem is the lack of integration of new assets into the system in a seamless manner, making the overall workflow complex and time consuming.

[0004] Another limitation of existing solutions is the lack of ability to scale the solution to the number “n” of assets that can be added to an industrial environment. Furthermore, other limitations of current systems are the lack of integration of domain knowledge into the system in order to address the specific requirements of assets of a particular industrial environment. Furthermore, current solutions also lack the ability to automatically train and retrain machine learning models using available domain knowledge and sensor data.

[0005] For example, in cases where assets like heat exchangers, pumps, compressors, chillers need to be managed in order to improve their efficiency and availability through condition monitoring and predictive maintenance, current solutions require manual steps to configure each individual asset, including their required workflows. Moreover, domain experts must additionally give their input to select the right sensors and choose the features of importance for the AI ​​predictive model for each device. This approach is time-consuming and expensive when the task needs to be applied to multiple assets in an industrial environment.

[0006] In view of the above, there exists a need to use the disclosed predictive maintenance techniques to manage one or more assets in a technical facility.

[0007] Therefore, an object of the present invention is to provide a system, an apparatus and a method for managing one or more assets in a technical facility.

[0008] The objects of the present invention are achieved by a method for managing multiple assets in a technical facility. The term "asset" refers to any equipment, system, instrument or machine manufactured or used in industry that can be used to perform an operation. Examples of assets include any machine in a technical system or technical facility / equipment, such as motors, gears, bearings, shafts, switchgear, rotors, circuit breakers, protection devices, remote terminal units, transformers, reactors, disconnectors, gear transmissions, gradient coils, magnets, radio frequency coils, etc. Example technical facilities include turbines, large drives, manufacturing units, factory settings, buildings, etc. The method includes receiving a set of requirements for managing multiple assets.

[0009] Throughout this disclosure, the term "requirements set" may refer to any information related to an expected result or analysis requested by an operator (hereinafter, sometimes referred to as a "user"). Advantageously, the requirements set may be information about operation, performance, service life, anomalies, failures, or any other information that an operator may request from a processing unit.

[0010] In one embodiment, the method also includes generating user profiles for one or more users. In this article, each user profile is associated with a set of functions accessible to the user. Advantageously, the user profile enables users to select roles according to their expertise, which in turn can enhance user experience.

[0011] The method includes selecting one or more assets from the plurality of assets based on the received set of requirements. Advantageously, the present invention is able to select relevant one or more assets from the plurality of assets 105 based on the set of requirements. This ensures that the workflow is initiated for a target number of assets, thereby increasing the processing speed of the processing unit.

[0012] The method includes mapping one or more assets to corresponding sensing units. In one embodiment, the method of mapping one or more assets to corresponding at least one sensing unit includes determining a relationship between the asset and the sensing unit using engineering data. In addition, the method includes mapping the asset to the sensing unit based on the determined relationship. According to one embodiment, the method of mapping the asset and the sensing unit also includes automatically updating the mapping between the asset and the sensing unit in real time when the sensing unit is online.

[0013] The method includes extracting information associated with the selected one or more assets and the received set of requirements. In this context, the information includes domain knowledge stored in a knowledge base. Advantageously, information from the domain knowledge and historical data is also considered for managing multiple assets in a technical facility.

[0014] The method includes determining at least one performance indicator based on information extracted from a knowledge base and a mapped sensing unit. Throughout this disclosure, the term "performance indicator" (also referred to herein as a key performance indicator or KPI) as used herein may refer to a parameter that affects the performance of an asset in a technical facility. Advantageously, the present invention is able to automatically determine a performance indicator based on a set of requirements from a user to initiate a workflow. At the same time, the present invention also implements the ability to calculate performance indicators for new assets and new asset classes so that new assets can be seamlessly integrated into the system.

[0015] According to one embodiment, a method for determining at least one performance indicator comprises selecting at least one sensing unit required for calculating at least one performance indicator. The method comprises obtaining data required for determining at least one performance indicator from the selected at least one sensing unit. The method comprises calculating at least one performance indicator based on the obtained data.

[0016] According to one embodiment, a method for calculating at least one performance indicator when a new asset class is added to a technical facility includes mapping the new asset class to a sensing unit associated with the new asset class. Advantageously, the present invention is capable of automatically recognizing that a new asset class has been added to the technical facility and ultimately mapping the new asset class to the sensing unit. In addition, the method includes obtaining information related to the new asset class from a database. In addition, the method includes updating a knowledge base with information related to the new asset class. In addition, the method includes calculating a set of formulas, wherein the set of formulas is a function of data received from the sensing unit and known performance indicators. In addition, the method includes determining at least one performance indicator of the new asset class based on the calculated set of formulas. Advantageously, the system is capable of automatically calculating at least one performance indicator of the new asset class and thus integrating the new asset class into the technical facility in an efficient and accurate manner. In addition, the method includes re-updating the knowledge base with the determined performance indicator. Advantageously, the new asset class can be added to the asset database without any changes to the backend code.

[0017] The method includes defining a workflow to be performed based on the determined at least one performance indicator. Throughout this disclosure, the term "workflow" refers to a set of instructions or tasks that need to be performed in order to determine a requested result. In one embodiment, the task is at least one of: anomaly detection in an asset, root cause analysis of an asset, remaining useful life estimation of an asset, prediction of a parameter associated with an asset, performance optimization of an asset, and energy optimization of an asset.

[0018] The method includes selecting a configured machine learning model from a set of machine learning models based on a defined workflow, wherein the set of machine learning models includes one or more machine learning models configured to determine an outcome for a specific task. Throughout this disclosure, the term "machine learning model" as used herein refers to a model used to determine a desired outcome given a set of training data sets. As used herein, a "configured machine learning model" refers to any machine learning model that is configured or trained to determine an outcome for a specific task, such as anomaly detection in an asset, root cause analysis of an asset, remaining useful life estimation of an asset, performance optimization of an asset, and energy optimization of an asset.

[0019] The method includes determining an outcome of a selected machine learning model based on the received set of requirements. In this document, the outcome of the selected machine learning model is a function of at least one determined performance indicator of one or more assets.

[0020] According to one embodiment, a method of determining an outcome of a selected machine learning model includes adjusting one or more coefficients of the selected machine learning model using data received from at least one sensing unit and information extracted. The method includes determining an accuracy of the adjusted machine learning model based on a received set of requirements. The method includes determining an outcome of the adjusted machine learning model if the accuracy of the adjusted machine learning model is above a first predefined threshold.

[0021] According to one embodiment, a method for determining the results of a selected machine learning model when a new asset is added to a technical facility includes mapping the new asset to a sensing unit associated with the new asset. The method includes selecting the new sensing unit required for determining at least one performance indicator. The method includes re-adjusting the coefficients of the selected machine learning model based on data obtained from the new sensing unit. The method includes determining the results of the re-adjusted machine learning model when the new asset is added.

[0022] According to one embodiment, a method for determining an outcome of a machine learning model for a new asset class includes determining a new workflow for the new asset class to be executed based on at least one determined performance indicator. The method includes selecting a configured machine learning model from a set of preconfigured machine learning models based on the newly determined workflow. The method includes adjusting one or more coefficients of the configured machine learning model using data received from at least one sensing unit and a knowledge base. The method includes determining an accuracy of the machine learning model adjusted for the new asset class based on a set of requirements. The method includes determining an outcome of the machine learning model adjusted for the new asset class if the accuracy of the adjusted machine learning model is above a second predefined threshold.

[0023] According to one embodiment, a method of adding a new machine learning model to a collection of machine learning models includes obtaining data associated with an asset for which the new machine learning model is to be added. The method includes training the new machine learning model using the data obtained from a database. The method includes determining an accuracy of the new machine learning model. The method includes adding the new machine learning model to the collection of machine learning models associated with the asset for a specific task if the accuracy of the new machine learning model is above a third predefined threshold.

[0024] According to one embodiment, the method further comprises presenting the determined results on a client device for managing a plurality of assets in a technical facility. In further embodiments, the method comprises an updated set of requirements from a user, and determining an updated result from the selected machine learning model based on the received updated set of requirements. Furthermore, the method comprises presenting the updated results on a client device associated with the user.

[0025] The object of the present invention is also achieved by a device for managing one or more assets in a technical facility. The device comprises one or more processing units and a memory unit communicatively coupled to the one or more processing units. The memory unit comprises an asset management module stored in the form of machine-readable instructions executable by the one or more processing units, wherein the asset management module is configured to perform the method steps as described above.

[0026] The object of the present invention is also achieved by a system for managing multiple assets in a technical facility. The system includes: one or more devices configured to provide a set of requirements for predictive maintenance of multiple assets; one or more sensing units for providing data associated with multiple assets; a knowledge base including domain knowledge related to multiple assets; and a device communicatively coupled to the one or more devices. The device is configured to manage multiple assets of a technical facility according to the above method.

[0027] The object of the present invention is also achieved by a computer program product having machine-readable instructions stored therein, which, when executed by one or more processing units, cause the processing units to perform the above method.

[0028] The above-mentioned properties, features and advantages of the present invention and the ways to achieve them will become more obvious and understandable (clear) by using the following description of the embodiments of the present invention in conjunction with the corresponding drawings. The illustrated embodiments are intended to illustrate, but not limit, the present invention.

[0029] Figure 1A A block diagram of a system for predictive maintenance of multiple assets in a technical facility according to an embodiment of the present invention is shown;

[0030] Figure 1BA block diagram showing an apparatus for predictive maintenance of multiple assets in a technical facility according to an embodiment of the present invention;

[0031] Figure 2 A block diagram illustrating the architecture of the system of FIG. 1 according to an embodiment of the present disclosure;

[0032] Figure 3 A block diagram illustrating the architecture of the system of FIG. 1 according to another embodiment of the present disclosure;

[0033] Figure 4 depicts a flow chart of a method for predictive maintenance of a plurality of assets in a technical facility according to an embodiment of the present invention;

[0034] Figure 5 depicts a flow chart of a method for determining at least one performance indicator according to an embodiment of the present invention;

[0035] Figure 6 depicts a flow chart of a method for determining the results of a selected machine learning model according to an embodiment of the present invention;

[0036] Figure 7 depicts a flow chart of a method for determining the results of a selected machine learning model when a new asset is added to a technical facility in accordance with an embodiment of the present invention;

[0037] Figure 8 : calculating at least one performance indicator when a new asset class is added to a technical facility according to an embodiment of the present invention;

[0038] Fig. 9 A flowchart depicting a method for determining results of a machine learning model for a new asset class according to an embodiment of the present invention; and

[0039] Fig.10 A flowchart of a method for adding a new machine learning model to a collection of machine learning models according to an embodiment of the present invention is depicted.

[0040] Hereinafter, embodiments for carrying out the present invention are described in detail. Various embodiments are described with reference to the accompanying drawings, wherein similar reference numerals are used to refer to similar elements throughout. In the following description, for the purpose of explanation, many specific details are set forth in order to provide a thorough understanding of one or more embodiments. It may be apparent that such embodiments may be practiced without these specific details.

[0041] Figure 1AA block diagram of a system 100 for managing multiple assets 105 in a technical facility according to an embodiment of the present invention is shown. Examples of technical facilities may be complex industrial or non-industrial settings with multiple assets 105, such as power plants, wind farms, power grids, manufacturing equipment, processing plants, buildings (residential or non-residential), etc. Throughout this disclosure, the term "asset" 105 refers to any equipment, system, instrument, or machine that is manufactured or used in an industry and can be used to perform an operation. In some cases, assets 105 may also include any equipment or instrument deployed or operated in a non-industrial environment such as a building. Examples of assets 105 include any machine in a technical system or technical facility / equipment, such as a motor, gear, bearing, shaft, switchgear, rotor, circuit breaker, protection device, remote terminal unit, transformer, reactor, disconnector, gear transmission, gradient coil, magnet, radio frequency coil, electrical appliance, electronic equipment, chiller, pump, heat exchanger, cooling tower, air compressor, boiler, fluidized bed dryer, coating machine, carbonizing tower, etc.

[0042] Throughout this disclosure, the term "requirements set" may refer to information related to an expected result or analysis requested by an operator. Hereinafter, throughout this disclosure, the terms "operator" and "user" have been used interchangeably without limitation. The requirements set may be input to the client device 130 by an operator. In one example, the requirements set may include information related to one or more parameters for predicting the condition of the asset 105. In another example, the requirements set may include conditions for estimating the remaining useful life of the asset 105. In yet another example, the requirements set may include information for continuously monitoring the asset 105. In yet another example, the requirements set may include information for calculating efficiency or performance metrics for multiple assets. Advantageously, the requirements set may be information about operation, performance, useful life, anomalies, failures, or any other information that an operator may request from the device 110. Throughout this disclosure, the term "updated requirements set" may refer to information related to an expected result or analysis requested by an operator after the processing unit has executed the previous workflow.

[0043] Throughout this disclosure, the term "sensing unit" 125 may refer to one or more electronic devices for acquiring and transmitting data from a plurality of assets 105. The sensing unit 125 includes, but is not limited to, a position sensor, a rotary encoder, a force gauge, a proximity sensor, a current sensor, an accelerometer, a temperature sensor, a sound sensor, a voltage sensor. The sensing unit 125 may acquire real-time condition data indicating one or more operating conditions of the asset 105 in real time.

[0044] Throughout this disclosure, the term "knowledge base" 122 may refer to a database including data and information related to a technical facility in the form of a knowledge graph including a plurality of nodes linked to each other. The knowledge base 122 may include domain knowledge related to assets 105 in a technical facility.

[0045] The set of requirements may be stored in a memory of the edge device 115, or may be input to the edge device 115 by an operator. For example, the edge device 115 may be communicatively coupled to a client device 130. Non-limiting examples of client devices 130 include personal computers, workstations, personal digital assistants, and human-machine interfaces. The client device 130 may enable a user to input one or more requirements through a web-based interface. Upon receiving the one or more requirements from the user, the edge device 115 transmits a request to the apparatus 110 for managing the plurality of assets 105.

[0046] In this embodiment, the device 110 is deployed in a cloud computing environment. As used herein, a "cloud computing environment" refers to a processing environment that includes configurable computing physical and logical resources (e.g., networks, servers, storage, applications, services, etc.) and data distributed over a network 120 (e.g., the Internet). The cloud computing environment provides on-demand network access to a shared pool of configurable computing physical and logical resources. The device 110 may include an asset management module ( Figure 1B ), which is configured to manage, monitor, detect anomalies and recommend actions based on predictions. Additionally, the apparatus 110 may include a network interface (not shown) for communicating with one or more edge devices 115 via the network 120.

[0047] In another embodiment, the apparatus 110 may be an edge computing device. As used herein, "edge computing" refers to a computing environment that can be executed on an edge device (e.g., one end connected to one or more sensing units in an industrial setting and the other end connected to a (one or more) remote server such as for (one or more) computing servers or (one or more) cloud computing servers), which may be a dense computing device with a small form factor and resource constraints in terms of computing power. A network of edge computing devices may also be used to implement the apparatus. Such a network of edge computing devices is referred to as a fog network.

[0048] like Figure 1BAs shown in , the device 110 includes a processing unit 135, a memory unit 140, a storage unit 145, an input unit 155, an output unit 160, and a standard interface or bus 195. The device 110 can be a computer, a workstation, a virtual machine running on host hardware, a microcontroller, or an integrated circuit. Alternatively, the device 110 can be a real or virtual computer group (the technical term for a real computer group is a "cluster", and the technical term for a virtual computer group is a "cloud").

[0049] As used herein, the term "processing unit" 135 means any type of computing circuit, such as, but not limited to, a microprocessor, a microcontroller, a complex instruction set computing microprocessor, a reduced instruction set computing microprocessor, a very long instruction word microprocessor, an explicitly parallel instruction computing microprocessor, a graphics processor, a digital signal processor, or any other type of processing circuit. The processing unit 135 may also include an embedded controller, such as a general or programmable logic device or array, an application specific integrated circuit, a single chip computer, etc. In general, the processing unit 135 may include hardware elements and software elements. The processing unit 135 can be configured for multithreading, that is, the processing unit 135 can host different computing processes simultaneously, thereby executing active and passive computing processes in parallel or switching between them.

[0050] The processing unit 135 is configured to receive a set of requirements for managing a plurality of assets 105. Furthermore, the processing unit 135 is configured to select one or more assets 105 based on the received set of requirements. Furthermore, the processing unit 135 is configured to map the one or more assets 105 to corresponding sensing units 125. Furthermore, the processing unit 135 is configured to extract information associated with the selected one or more assets 105 and one or more parameters associated with the assets 105, such as voltage, current, temperature, humidity, etc. The information includes domain knowledge stored in the knowledge base 122. Furthermore, the processing unit 135 is configured to determine at least one performance indicator based on the information extracted from the knowledge base 122 and the mapped sensing units 125. Furthermore, the processing unit 135 is configured to define a workflow to be executed based on the determined at least one performance indicator. Furthermore, the processing unit 135 is configured to determine a configured machine learning model from a set of machine learning models based on the defined workflow. The set of machine learning models includes one or more machine learning models configured to determine a result of a specific task. Furthermore, the processing unit 135 is configured to determine a result of the selected machine learning model based on the received set of requirements. The results of the selected machine learning model are a function of the determined at least one performance metric of one or more assets 105.

[0051] The memory unit 140 may be a volatile memory and a non-volatile memory. The memory unit 140 may be coupled for communication with the processing unit 135. The processing unit 135 may execute instructions and / or codes stored in the memory unit 140. Various computer-readable storage media may be stored in the memory unit 140 and may be accessed from the memory unit 140. The memory unit 140 may include any suitable element for storing data and machine-readable instructions, such as a read-only memory, a random access memory, an erasable programmable read-only memory, an electrically erasable programmable read-only memory, a hard disk drive, a removable media drive for handling compact disks, digital video disks, magnetic disks, tape cartridges, memory cards, and the like.

[0052] The memory unit 140 includes an asset management module 165 in the form of machine-readable instructions on any of the above-mentioned storage media, and can communicate with and be executed by the processing unit 135. The asset management module 165 includes a data acquisition module 170, an asset mapping module 175, an information extraction module 180, a performance indicator determination module 182, a workflow definition module 185, a machine learning model selection module 190, and an output module 192.

[0053] The data acquisition module 170 is configured to obtain a set of requirements for managing the plurality of assets 105. The data acquisition module 170 is also configured to acquire data from the sensing unit 125 and the knowledge base 122. In addition, the data acquisition module 170 is configured to pre-process the data and information received in a format that can be processed by another module. Pre-processing may include data standardization, data translation, data conversion, missing value checking, data cleaning, noise filtering, and the like.

[0054] The asset mapping module 175 is configured to select one or more assets 105 from the plurality of assets 105 based on the received set of requirements. The asset mapping module 175 is configured to map the one or more assets 105 to corresponding sensing units 125. In addition, the asset mapping module 175 is configured to determine the relationship between the assets 105 and the sensing units 125, and thereby map the assets 105 to the corresponding sensing units 125 based on the determined relationship.

[0055] The information extraction module 180 is configured to extract information associated with the selected one or more assets 105 and the received set of requirements. The information extraction module 180 is also configured to extract information from the knowledge base 122, which includes domain knowledge stored in the form of a knowledge graph.

[0056] The performance indicator determination module 182 is configured to determine at least one performance indicator based on the information extracted from the knowledge base 122 and the mapped sensing units 125. The performance indicator determination module 182 is also configured to select at least one sensing unit 125 required to calculate the at least one performance indicator. In addition, the performance indicator determination module 182 is configured to obtain data required to determine the at least one performance indicator from the selected at least one sensing unit 125, and then calculate the at least one performance indicator.

[0057] The workflow definition module 185 is configured to define a workflow to be executed based on at least one performance indicator.

[0058] The machine learning model selection module 190 is configured to configure a machine learning model from a set of machine learning models based on the determined task set.

[0059] The output module 192 is configured to determine the results of the selected machine learning model based on the received set of requirements. In this article, the results of the selected machine learning model are a function of the determined at least one performance indicator of one or more assets 105. The output module 192 is also configured to present the determined results on the client device 130. In addition, the output module 192 is configured to determine updated results from the selected machine learning model based on the received updated set of requirements, and present the updated results to the client device 130.

[0060] The storage unit 145 includes a non-volatile memory that stores domain knowledge and data such as received from the sensing unit. The storage unit 145 includes a database 150, which includes a knowledge graph, historical data (such as previously detected anomalies, previous suggestions, etc.). The database 150 may also include engineering data, trained machine learning models, synthetic data, etc. The bus 195 serves as an interconnect between the processing unit 135, the memory unit 140, the storage unit 145, the input unit 155, and the output unit 160. The input unit 155 enables a user to input one or more requirements or make requests for managing the asset 105. The output unit 160 is configured to present the status of the asset 105, the detected anomalies, the predicted parameters, the generated suggestions, etc.

[0061] One of ordinary skill in the art will appreciate that Figure 1A and Figure 1BThe hardware depicted in the drawings may vary for different implementations. For example, other peripheral devices such as optical disk drives, local area network (LAN) / wide area network (WAN) / wireless (e.g., Wi-Fi) adapters, graphics adapters, disk controllers, input / output (I / O) adapters, network connection devices may also be used in addition to or in place of the depicted hardware. The depicted examples are provided for purposes of explanation only and are not meant to imply architectural limitations with respect to the present disclosure.

[0062] Figure 2 A block diagram 200 of the architecture of the system of FIG. 1 according to an embodiment of the present disclosure is shown. The block diagram 200 shows a layered software architecture view of the system 100. As shown, the architecture of the system includes a predictive analytics engine user interface layer 202, a predictive analytics engine backend layer 204, a data processing layer 206, a data storage layer 208, a communication layer 210, and a data collection layer 212. It should be understood that the layers and components disclosed herein are for illustrative purposes only and should not be construed as limited to the layers and components as described herein.

[0063] The predictive analysis engine user interface layer 202 includes an AI model library module 214, an AI model trainer module 216, an AI model simulator module 218, an AI workflow manager module 220, and a system management module 222. The predictive analysis engine user interface layer 202 is configured to select a set of analysis tasks or workflows and / or specific technical facilities (or assets / equipment) to be performed. In one example, the selected workflow can be a process industry, a building technology, a discrete factory, a brownfield, a greenfield, or an asset (such as a heat exchanger, a chiller, or a pump). This decision can be made by user selection or automatically made via the predictive analysis engine user interface layer 202 based on the user's requirement set. The predictive analysis engine user interface layer 202 is configured to determine the required performance indicators (KPIs), the fault descriptions of different assets as a function of available sensors, and the integration of domain-based techniques in the front end.

[0064] The predictive analysis engine user interface layer 202 is also configured to provide network-based access, select network checks, and different user profiles. In addition, the predictive analysis engine user interface layer 202 is configured to transmit domain information to the predictive analysis engine backend layer 204 for further processing. In addition, the predictive analysis engine user interface layer 202 is configured to start AI model configuration, for example, to select different AI model workflows, which can be pre-configured for different assets (such as pumps or heat exchangers). For example, there may be an exhaustive list of fault monitoring AI workflows provided for pumps. In this case, the predictive analysis engine user interface layer 202 is configured to select an appropriate AI workflow based on the available sensing unit 125. In addition, the predictive analysis engine user interface layer 202 is configured to start real-time data ingestion services, configure the sensing unit 125 to access relevant data, select a path to a data source (e.g., a file server, an OPC-UA server, a historian), and so on. In addition, the predictive analysis engine user interface layer 202 is configured to start AI model training, AI model retraining, AI model testing, or AI model deployment workflows. The predictive analysis engine user interface layer 202 is also configured as a debugging system.

[0065] The predictive analysis engine backend layer 204 includes an AI model library module 224, an AI model trainer module 226, an AI model simulator module 228, an AI workflow manager module 230, and a system management module 232. The predictive analysis engine backend layer 204 is configured to interact with the predictive analysis engine user interface layer 202. The predictive analysis engine backend layer 204 is configured to authenticate users, for example, using LDAP, SSO, and other technologies. In addition, the predictive analysis engine backend layer 204 is configured to configure a network check for a user profile. In addition, the predictive analysis engine backend layer 204 is configured to configure an analysis workflow to be executed, for example, a process industry, a building technology, a discrete factory, or for a specific asset 105. In one example, the predictive analysis engine backend layer 204 is also configured to allow a user to select a workflow with multiple assets that are bundled together for configurations such as overall optimization and virtual sensing, which may require information from different assets. Furthermore, the predictive analytics engine backend layer 204 is configured to train the selected AI model using both offline and online training, autonomous data pre-processing, optimization workflows, and the like.

[0066] In addition, the predictive analysis engine backend layer 204 is configured to deploy AI models, manage AI model libraries, interact with other visualization or model building tools, provide debugging data in the form of workflow logs, and the like. In addition, the predictive analysis engine backend layer 204 also provides templated features for new assets or new asset categories in technical facilities. The predictive analysis engine backend layer 204 provides the ability to seamlessly add any new asset to the system by adding all relevant metadata, charts, sensors, KPIs, and any other available information. In addition, the predictive analysis engine backend layer 204 is configured to use various DOEs (experimental designs) to generate synthetic data for the purpose of proof of concept of several workflows. In addition, the simulation model provides the ability to merge new assets into the AI ​​model library. The predictive analysis engine backend layer 204 is configured to use synthetic data / real data to train offline AI models and simulations. Similarly, the predictive analysis engine backend layer 204 is configured to train online AI models and use real-time data acquired from assets 105 and sensing units 125. In addition, the predictive analytics engine backend layer 204 is configured to use test data to test the accuracy of the AI ​​model, the ability to perform blind testing, and provide insights into output quality (multivariate input data range checking).

[0067] The data processing layer 206 includes a KPI calculator module 234, a data cleansing module 236, a data structuring module 238, and a data normalization module 240. The data processing layer 206 is configured to pre-process the data received from the data storage layer 208.

[0068] The data storage layer 208 includes a knowledge base 242, a mongoDB 244, and a time series database 246. The data storage layer 208 is configured to store data sensor data, knowledge domain information, AI models, etc. received from the data collection layer 212 through the data communication layer 210. The communication layer 210 includes an OPC UA module 248, a TCP / IP module 250, and an MQTT module 252. The communication layer 210 acts as an interface between the data collection layer 212 and the data processing layer 208, and is configured to transmit the collected data from the data collection layer 212 to the data processing layer 208.

[0069] The data collection layer 212 includes one or more assets, including chillers 254, pumps 256, HVAC 258, PLCs 260, exhaust fans 262. It should be understood that these are merely examples and are not limited to the specific assets mentioned herein. The data collection layer 212 is configured to interact directly with the sensing units 125 associated with the corresponding assets to acquire data from the assets in real time.

[0070] Figure 3A block diagram 300 of the architecture of the system of FIG. 1 according to another embodiment of the present disclosure is shown. As can be seen in the block diagram 300, the architecture includes a plurality of assets 302, such as pumps 302A, wells 302B, heat exchangers 302C, HVAC 302D, chillers 302E, motors 302N. In addition, the architecture includes a server 304 communicatively coupled to the plurality of assets 302A-N via a network. The server 304 includes a user interface (UI) layer 306, a backend layer 308, and a business connector layer 310.

[0071] The UI layer 306 also includes a navigation application 312 , an application framework 314 (which includes configuration, dashboards, workflow, asset monitoring, etc.). In addition, the UI layer includes a graphics framework 316 and an angular framework 318 . The UI layer 306 communicates with the backend layer 308 .

[0072] The backend layer 308 also includes a programming language-specific kit 320 and a programming language-specific editor 322. In one example, the programming language used herein may be Java or Python. Therefore, the Java / Python kit may include an AI model training module 324, an AI model simulation module 326, an AI model prediction module 328, a workflow manager module 330, and a knowledge graph module 332. In addition, a programming language-specific editor 322 (e.g., node.js) may include a management module 334 and a navigation module 336.

[0073] The application programming interface (API) of the business connector layer 310 also includes a database access API 338, a security API 340, other APIs 342, and a communication protocol access API 344 (such as Kafka). The database access API 338 is communicatively coupled to a Mongo database 346 and an influx database 348. In addition, the Mongo database 346 and the influx database 348 are configured to receive data from a communication module 350, which includes an OPC / UA module 352, a TCP / IP module 354, and an MQTT module 356. In addition, the access API 344 is connected to a communication protocol module 358. It should be understood that the layers and components disclosed herein are for illustrative purposes only and should not be construed as being limited to the layers and components as described herein.

[0074] Figure 4A flow chart of a method 400 for managing multiple assets in a technical facility according to an embodiment of the present invention is depicted. At step 402, a set of requirements for maintaining multiple assets 105 is received. The set of requirements can be input by an operator using a client device 130. In one example, the set of requirements can include information related to one or more parameters for predicting the condition of the asset 105. In another example, the set of requirements can include conditions for estimating the remaining useful life of the asset 105. In yet another example, the set of requirements can include information for continuously monitoring the asset 105. In yet another example, the set of requirements can include information for calculating efficiency or performance metrics for multiple assets 105. Advantageously, the set of requirements can be information about operation, performance, useful life, anomalies, failures, or any other information that an operator can request from the device 110.

[0075] One or more assets 105 may include, but are not limited to, any machine in a technical system or technical facility / equipment, such as motors, gears, bearings, shafts, switchgear, rotors, circuit breakers, protection devices, remote terminal units, transformers, reactors, disconnectors, gear transmissions, gradient coils, magnets, radio frequency coils, electrical appliances, electronic equipment, chillers, pumps, heat exchangers, cooling towers, air compressors, boilers, fluidized bed dryers, coaters, carbonizing towers, etc.

[0076] Throughout this disclosure, the term “asset” 105 may refer to any asset 105 that already exists in a technical facility.

[0077] Throughout this disclosure, the term “new asset” 105 may refer to any asset 105 that is not currently integrated into a technical facility but is similar in specification and operation to one or more assets 105 that already exist in the technical facility.

[0078] Throughout this disclosure, the term "new asset class" may refer to any asset that is not currently integrated into a technical facility and is not similar in specification and operation to any asset 105 already existing in the technical facility.

[0079] It should be understood that management of assets 105 may include effectively managing the operation of one or more assets 105, prediction of one or more parameters of assets 105, determination of performance and efficiency of assets 105, energy optimization of assets 105, prediction of various parameters (such as load prediction of assets 105), calculation of remaining useful life of assets 105, anomaly detection, root cause analysis of faults detected in assets 105, training of artificial intelligence (AI) models for desired outputs, adding new assets to the system, and the like.

[0080] According to one embodiment, the method also includes generating a user profile for one or more users associated with the client device 130. In this article, each user profile is associated with a set of functions accessible to the user. In one example, the user profile can be defined by the processing unit 135 based on one or more qualifications and expertise of a specific user. In another example, the user profile can also be defined based on the user's designation or role in the technical setting. In yet another example, the user profile can be manually assigned to the user by the administrator of the system. The label of the user profile can also be defined by the administrator of the system. As an example, a user employed to manage equipment that is only responsible for viewing results, executing workflows, and monitoring the operation of different assets 105 can be assigned an "operator user profile". Users with "operator user profiles" are completely isolated from functions in the back end (such as configuring AI models, adding new workflows, adding new AI models, training AI models, etc.). In contrast, users with expertise in AI model training, configuration, and coding may be assigned a "developer user profile". A user with a "developer user profile" is allowed to perform functions such as configuring AI models, adding new workflows, adding new AI models, training AI models, etc., as well as accessing other functions. Advantageously, the user profile enables users to select roles based on their expertise, which in turn enhances the user experience. It will be appreciated that the term "user" as used herein may also include a natural person, a robot, an artificial intelligence machine, without any limitation.

[0081] In step 404, one or more assets 105 are selected from a plurality of assets 105 based on the received set of requirements. The selection of one or more assets 105 or the entire technical facility may be based on a set of requirements such as received from an operator. In one example, when the set of requirements is "predict the efficiency of pumps in the system", the pumps in the technical facility are selected to determine the desired output, i.e., the efficiency of the pumps. In another example, when the set of requirements is "provide a maintenance schedule for chillers", the chillers installed in the outlet or factory will be selected to start the workflow. In yet another example, when the requirement is "predict the energy consumption in the building", the entire building including a plurality of assets is selected to determine such a result. It will be appreciated that the disclosed system is capable of selecting the relevant one or more assets 105 from a plurality of assets 105 based on the set of requirements. This ensures that the workflow is started for a target number of assets, thereby increasing the processing speed of the system.

[0082] At step 406, one or more assets 105 are mapped to corresponding sensing units 125. Correctly mapping the selected assets 105 to the sensing units 125 to obtain relevant data is important for the accuracy of the system. According to one embodiment, the mapping of one or more assets 105 to the corresponding at least one sensing unit 125 includes determining the relationship between the assets 105 and the sensing units 125 using engineering drawings. Throughout this disclosure, the term "engineering drawing" as used herein may refer to a diagram used to convey technical information related to a plurality of assets 105. Engineering drawings include diagrams such as circuit diagrams, schematic diagrams, technical manuals, asset specification data, wiring diagrams, floor plans, asset data sheets, piping and instrumentation diagrams, etc. In an exemplary embodiment, a piping and instrumentation diagram (P&ID) is used to determine the relationship between the assets 105 and the sensing units 125. In this article, a piping and instrumentation diagram (P&ID) represents a detailed diagram in the process industry, which shows piping and process equipment as well as instrumentation and control equipment. As known in the art, a P&ID is initially drawn during the design phase based on a combination of process flow chart data, mechanical process equipment design, and instrumentation engineering design. During the design phase, the diagram also provides the basis for the development of the system control scheme, allowing further safety and operational investigations, such as hazard and operability studies (HAZOPs). The P&ID therefore shows the physical sequence of assets and sensing units in a technical facility, and how these assets are connected. The P&ID also plays a significant role in the maintenance and modification of processes after initial construction.

[0083] In one example, the P&ID may contain information related to assets 105, such as pressure vessels, columns, tanks, pumps, compressors, heat exchangers, furnaces, fans, cooling towers, screens and filters, orifices, steam traps, moisture traps, flares and vents, silencers, etc. In another example, the P&ID may contain information related to process piping, sizing and identification, such as piping in piping class and line number, flow direction, interconnect references, start-up lines, flush lines, bypass lines, flow lines, pipes, blinds, insulation and heat tracing, etc. In yet another example, the P&ID may also include process control instruments and designations, such as names, numbers, unique tag identifiers, valves and their types and identifications, control inputs and outputs, vents, drains, special fittings, sampling lines, reducers, and other miscellaneous information.

[0084] According to another embodiment, the assets 105 may be mapped to the sensing units 125 based on the information stored in the knowledge base 122. According to another embodiment, the user / operator may also be prompted to map the relevant assets 105 to the sensing units 125 using the web-based interface in the client device 130. According to another embodiment, the method of mapping the assets 105 to the sensing units 125 also includes automatically updating the mapping between the assets 105 and the sensing units 125 in real time when the sensing units 125 are online. It will be appreciated that the sensing units 125 and the assets 105 may be mapped at the start of any workflow and at any time during the execution of the workflow. Let us consider the case when a critical sensing unit 125 is offline or unavailable during the mapping phase. In this case, the specific sensing unit 125 can be mapped to the sensor later, and the workflow can be updated in real time, thereby increasing the efficiency and accuracy of the system.

[0085] These relationships can be determined by analyzing the information stored in the engineering drawing. Thereafter, based on the determined relationships, the selected asset 105 is mapped to at least one sensing unit 125. In addition, once the asset 105 is mapped to the sensing unit 125, data from the sensing unit 125 is acquired for the requested further analysis. According to one embodiment, a method of obtaining data from the sensing unit 125 includes establishing a connection between a processing unit 135 and a data source associated with the sensing unit 125. In this document, the source is the address of the sensing unit 125. In addition, data is obtained from the sensing unit in real time via the established connection. In an exemplary embodiment, the processing unit 135 runs a data ingestion service in a backend server to obtain real-time data from an OPC-UA / WinCC server. In addition, data can be directly acquired from a SCADA system associated with the asset 105.

[0086] According to another embodiment, the method for obtaining data from the sensing unit 125 includes establishing a connection between the processing unit 135 and a data source associated with the sensing unit 125. In this article, the source is a database. In one example, the data from the sensing unit 125 can be acquired and stored in the database on a periodic basis. In another example, the data can be uploaded to the database by an operator for analysis. In addition, once the connection is established, data is obtained from the database. In an exemplary embodiment, the processing unit 135 runs a data loader service to obtain a specific instance of a database (e.g., Influx, SQlite, etc.), and obtains data from the specific instance according to a request.

[0087] At step 408, information associated with the selected one or more assets 105 and the received set of requirements is extracted from the knowledge base 122. In this article, the information includes domain knowledge stored in the knowledge base 122. The knowledge base 122 may include information in the form of a knowledge graph. The knowledge graph may contain one or more link nodes that capture domain knowledge and indicate parameters and corresponding responses related to one or more components of the asset 105, failure history of the asset 105, historical data related to parameter values ​​of one or more components of the asset 105, failure predictions and corresponding failure causes, etc. It should be noted that each link node is connected using one or more paths or edges. The knowledge graph includes asset-specific information, data related to the conceptualization of the asset 105, data related to the design of the asset 105, data related to the manufacture of the asset 105, data related to the assembly of the asset 105, data related to the manufacture of the asset 105, process-specific information, general process information, and any other information related to the business requirements can be linked to each other in a defined order. Additionally, the knowledge graph includes failures and failure causes in a linked manner. In one example, a fault in a motor may be linked to reduced lubrication in a rotor, a rotor fault may be linked to a defect in a bearing, etc. Additionally, the knowledge graph also includes standard procedure data that may be specific to a process in the asset 105. Standard procedure data includes predefined rule sets, standard operating procedures, predefined activities or processes, etc. In addition, the knowledge graph also includes physics-based models of specific processes, such as manufacturing, assembly, testing, and operation of the asset 105. In addition, the knowledge graph also includes data related to previously detected anomalies or faults and actions initiated to resolve previously predicted faults.

[0088] At step 410, at least one performance indicator is determined based on the information extracted from the knowledge base 122 and the mapped sensing unit 125. Throughout this disclosure, the term "performance indicator" as used herein may refer to a parameter that affects the performance of an asset 105 in a technical facility. The term "performance" as used herein refers to a measure that affects the efficiency, availability, or reliability of an asset 105. In this disclosure, a performance indicator may be the overall efficiency, overall availability, overall reliability of an asset 105, or a performance indicator may also define a specific indicator for a specific asset (such as a chiller, pump, heat exchanger, etc.). This information related to the association between the sensing unit 125 and the asset 105 and the performance indicator may be extracted from the knowledge base 122 and may then be used for further analysis. It will be appreciated that the knowledge base 122 may be used to directly map the performance indicators of existing assets, as well as to calculate the performance indicators for new assets or new asset classes, and then update the performance indicators in the knowledge base 122. Advantageously, the present invention is able to automatically determine the performance indicators based on a set of requirements from a user to initiate a workflow. At the same time, the present invention also implements the ability to calculate performance indicators for new assets and new asset classes, so that new assets can be seamlessly integrated into the system.

[0089] exist Figure 5 The method for determining performance indicators of existing assets 105 is further explained in. Figure 5 A flow chart of a method 500 for determining at least one performance indicator according to an embodiment of the present invention is depicted. At step 502, at least one sensing unit 125 required for calculating at least one performance indicator is selected. It should be understood that all data from the mapped sensing units 125 may not be required to determine a particular performance indicator. Therefore, it is important to intelligently select the sensing units 125, which is crucial for the calculation of the determined performance indicator. At step 504, the data required for determining at least one performance indicator is obtained from the selected at least one sensing unit 125. At step 506, at least one performance indicator is calculated based on the obtained data.

[0090] In one example, let us consider a heat exchanger as an existing asset in a technical facility. In this case, domain knowledge and performance indicator definitions are already available in the knowledge base 122. The requirement set may be "determine leakage" and "determine fouling". In this context, the processing unit 135 first determines the sensing units required to determine the performance indicators for the requested analysis, and then determines the performance indicators that will be required for the analysis. Table 1 depicts the determined relationship between the requirement set, the performance indicators, and the required sensing units 125, as follows:

[0091]

[0092] Table 1

[0093] It will be appreciated that new assets belonging to the same asset class can be seamlessly integrated into the technical facility because the approach is scalable. Consider the above example of a heat exchanger. Since all the information, data, metadata, domain knowledge, sensor information, performance indicator information, the relationship between the heat exchanger and the required sensors for different requirements are already known and stored in the knowledge base 122, any limited number of new heat exchangers that can be added to the technical facility can be easily integrated into the system. Therefore, when, for example, 50 new heat exchangers are added, there may be no challenges in the integration of the new heat exchangers.

[0094] exist Figure 8 The methodology used to determine the performance metrics for new asset classes is further explained in . Figure 8 A flow chart of a method 800 according to an embodiment of the present invention is depicted, the method 800 being used to calculate at least one performance indicator when a new asset class is added to a technical facility. It should be understood that performance indicators may not be available for the new asset class being added to the technical facility, and therefore performance indicators are to be calculated for such new asset classes. In step 802, the new asset class is mapped to a sensing unit 125 associated with the new asset class. In this context, the new asset class may be a new asset type that may not exist in the technical facility at a particular time, or there is no a priori information about the new asset in the knowledge base. In one example, the new asset class may be a pump, a heat exchanger, or any other asset that was not previously available in the technical facility. Advantageously, the present invention is able to automatically recognize that a new asset class has been added to the technical facility, and ultimately map the new asset class to the sensing unit 125. The mapping of the new asset class to the sensing unit 125 may be performed in a manner similar to Figure 4 In one example, the new asset category is mapped to the corresponding sensing unit by determining the relationship between the new asset category and the sensing unit using an engineering diagram such as obtained from a database. In another example, the operator is prompted to manually map the new asset category to the sensing unit 125 using a graphical user interface on the client device 130.

[0095] In step 804, information related to the new asset class is obtained from a database. Information such as new specifications, asset-related data, parameter information, domain-specific information, or any other relevant metadata can be obtained from an associated database. In one embodiment, such information can be manually input through a web-based interface. In addition, data from a sensing unit 125 associated with the new asset class is also obtained. In one embodiment, a method for obtaining data from a sensing unit 125 associated with a new asset class includes establishing a connection between a processing unit 135 and a data source associated with the sensing unit 125. In this article, the source is the address of the sensing unit 125. In addition, data is obtained from the sensing unit 125 in real time via the established connection. According to another embodiment, a method for obtaining data from a sensing unit 125 associated with a new asset class includes establishing a connection between a processing unit 135 and a data source associated with the sensing unit 125. In this article, the source is a database. In one example, data from the sensing unit 125 can be periodically acquired and stored in a database. In another example, the data can be uploaded to the database by an operator for analysis. In addition, once the connection is established, data is obtained from the database.

[0096] At step 806, the knowledge base 122 is updated with information related to the new asset class. As new information is discovered, the knowledge base 122 is updated by adding nodes and edges to the knowledge graph. Advantageously, updating the knowledge base 122 with new information acquired for the new asset class allows the knowledge base 122 to continually learn and establish new nodes and connections.

[0097] At step 808, a set of formulas is calculated, where the set of formulas is a function of the data received from the sensing unit 125 and known performance indicators. In this context, the set of formulas refers to mathematical statements derived from the data received from the sensing unit 125 and the performance indicators for other assets stored in the knowledge base 122. In one example, the set of formulas is defined using mathematical operators, such as addition, subtraction, multiplication, and division of parameter values ​​obtained from the sensing unit 125. In another example, the set of formulas is defined using already known performance indicators (KPIs) that are known and stored in the knowledge base 122.

[0098] For example:

[0099] KPI: Delta Pressure = Pressure_out - Pressure_in is a function of the inlet and outlet pressure sensors only.

[0100] KPI: Efficiency = fxn (Delta Pressure, Flow, speed) is a function of the calculated KPI (Delta Pressure) and sensor variables flow and speed, and fxn can be represented by a machine learning model (such as a neural network).

[0101] At least one performance indicator is determined for the new asset class based on the calculated set of formulas at step 810. Advantageously, the system is capable of automatically calculating at least one performance indicator for the new asset class and thereby integrating the new asset class into the technical facility in an efficient and accurate manner.

[0102] At step 812, the knowledge base 122 is re-updated with the determined performance indicators. In an exemplary embodiment, let us consider an oil and gas plant, where a distillation column is a new asset class that needs to be integrated into the oil and gas plant. Advantageously, the distillation column can be added to the asset library without any changes to the backend code. In this case, model templates and libraries for calculating various performance indicators are not available because this is a new requirement. Therefore, KPIs need to be added according to the domain skills of the operator, especially the user with the developer profile. As a first step, available sensors are added according to the P&ID and equipment data sheet and / or SCADA diagram. In addition, an "Add_KPI" configuration for the distillation column is added in the front end to initialize the workflow, which adds the KPI and also provides a template in the back end to configure the required formula set for the distillation column. The required independent variables in terms of available sensors can also be added in the back end. In addition, the system determines the performance indicators of the distillation column by calculating the formula set. The performance indicator can be a function of the sensor value or it can also be a function of the KPI that has been selected. Such a formula set can then be integrated within the source code of the distillation column. This also allows the distillation column to be easily integrated as part of the model library. Once the performance metric is determined, specify the "Add_Analytics" workflow with the target KPI and input features. The input features can be sensors around the distillation column, or other KPIs that have been defined. Subsequently, add the model training workflow in the backend, where the input features and target variable will already be available.

[0103] In step 412, a workflow to be executed is defined based on the determined at least one performance indicator and the set of requirements received from the operator. In this article, the term "workflow" refers to a set of instructions or tasks that need to be executed in order to determine the requested result. It is worth noting that every action in the system is modeled as a workflow. The workflow can be understood as an entry point to the system backend. Define and start a workflow for an asset 105. The workflow may include one or more tasks for each of a plurality of assets 105. In one embodiment, the task is at least one of the following: anomaly detection in an asset 105, root cause analysis of an asset 105, remaining useful life estimation of an asset 105, performance optimization of an asset 105, prediction of one or more parameters associated with an asset 105, and energy optimization of an asset 105. One or more tasks may vary for different assets and different technical facilities based on requirements. In one example, when the technical facility is a building, the task may be to predict load, predict temperature, predict humidity, and the like. Once the workflow is defined as required, the workflow is submitted for execution. Examples of workflows may include, but are not limited to, "train only", "deploy only", "train and deploy", and "test". Advantageously, defining workflows allows models to be seamlessly deployed on edge devices (where the model was trained elsewhere), or directly on internal servers where the model can be trained at the customer site itself. Similarly, the same workflow can be executed on the cloud. This allows flexibility in deployment and model management.

[0104] In step 414, a configured machine learning model is selected from a set of machine learning models based on a defined workflow. In this article, a set of machine learning models includes one or more machine learning models configured to determine the results of a specific task. The term "machine learning model" as used herein refers to a predictive model given a set of training data sets. Typically, each individual sample of training data is a pair containing a data set (e.g., operating conditions of asset 105 and corresponding parameter values ​​of asset 105) and an expected output value or data set (e.g., parameter values ​​of asset 105). The machine learning model analyzes the training data and generates a prediction function. Once the prediction function is derived through training, it is able to reasonably predict or estimate the correct output value or data set. Exemplary machine learning models may include artificial intelligence models, such as deep neural networks, convolutional neural networks, LSTM models, and the like.

[0105] As used herein, a “configured machine learning model” refers to any machine learning model that is configured or trained to determine an outcome for a specific task, such as anomaly detection in an asset, root cause analysis of an asset 105, remaining useful life estimation of an asset 105, performance optimization of an asset 105, parameter prediction associated with an asset 105, and energy optimization of an asset. Notably, the collection of machine learning models is stored in a database in the server backend. Advantageously, the system is able to automatically determine the configured machine learning model from the collection of machine learning models in order to determine the requested outcome. Furthermore, it will be appreciated that the need for a user to understand the underlying source code or require coding skills to execute a workflow and determine an outcome from a selected machine learning model is minimal. In one embodiment, the system intelligently selects a configured machine learning model and determines an outcome. In another embodiment, the system allows users with a developer profile to train a machine learning model or configure / reconfigure a machine learning model based on their knowledge and expertise. This may be the case when a new machine learning model is to be added to the system. In Fig.10 This approach for adding a new machine learning model to an ensemble of machine learning models is further explained in .

[0106] Fig.10 A flow chart of a method 1000 for adding a new machine learning model to a collection of machine learning models is depicted according to an embodiment of the present invention. In step 1002, data related to an asset 105 to which a new machine learning model is to be added is obtained. In this document, the data is stored in a database. The asset 105 can be an existing asset or a new asset to which a new machine learning model is to be added. The data can be conditional data, operational data, specification data, parameter data, historical data, etc. that may be required for training a new machine learning model. In one example, the data is synthetic data designed using experimental design (DOE) techniques. In the event that data for training a new machine learning model is not available, synthetic data may be required for a new asset. In another example, the data is historical data collected for an asset over a period of time. Historical data can be used for existing assets.

[0107] At step 1004, the new machine learning model is trained using the data obtained from the database. In one example, in the case of a new asset, the new machine learning model is trained on synthetic data. In another example, in the case of an existing asset, the new machine learning model is trained on historical data. The training of the new machine learning model can be accomplished using techniques used in the art. It will be appreciated that the new machine learning model can be configured by adding, reducing, or modifying the architecture of the new machine learning model (such as the number of layers or nodes per layer) based on the requirements of the task and the requested workflow.

[0108] In addition, the new machine learning model can be configured to adjust the weights of the nodes in the new machine learning model in order to achieve the requested task. For example, if a user has requested to calculate the remaining useful life (RUL) of a chiller that did not previously exist in the system, a new machine learning model for calculating the RUL of the chiller is configured and trained for RUL estimation of the chiller. Advantageously, the system allows users with developer profiles to make new machine learning models on request.

[0109] At step 1006, the accuracy of the new machine learning model is determined. In this article, the accuracy of the machine learning embodiment can be determined by testing the new machine learning model using a test data set. In one embodiment, the accuracy of the machine learning model can be determined by executing the new machine learning model in a simulation environment. The simulation environment can include a digital twin of the asset 105, and then the accuracy of the new machine learning algorithm is determined based on the results of the simulation instance. In this article, the term "digital twin" as used herein refers to a dynamic virtual replica based on one or more of the following: a physics-based model, a computer-aided design (CAD) model, a computer-aided engineering (CAE) model, a one-dimensional (1D) model, a two-dimensional (2D) model, a three-dimensional (3D) model, a finite element (FE) model, a descriptive model, a meta-model, a stochastic model, a parametric model, a reduced-order model, a statistical model, a heuristic model, a predictive model, an aging model, a machine learning model, an artificial intelligence model, a deep learning model, a system model, a knowledge graph, etc. The digital twin interacts with the knowledge base 122 to generate simulation instances and determine the accuracy of the new machine learning model.

[0110] At step 1008, if the accuracy of the new machine learning model is above a third predefined threshold, the new machine learning model is added to the set of machine learning models associated with the asset 105 used for a specific task. The third predefined threshold may be a quantifiable threshold set by the operator depending on the task for which the new machine learning model will be configured. When the accuracy of the new machine learning model is derived from the test data or the results of the simulation environment, it is compared with the third predefined threshold. If the accuracy of the new machine learning model is above the third predefined threshold, the new machine learning is added to the set of machine learning models, such as in an AI model library. If the accuracy of the new machine learning model is below the third predefined threshold, the new machine learning model is retrained until the desired accuracy is achieved.

[0111] Back to reference Figure 4, at step 416, the result of the selected machine learning model is determined based on the received set of requirements. In this article, the result of the selected machine learning model is a function of at least one performance indicator determined for one or more assets 105. The result of the selected machine learning model can be the result of the specific task requested. In one embodiment, the result can be related to a specific task, such as anomaly detection in assets 105, root cause analysis of assets 105, remaining useful life estimation of assets 105, performance optimization of assets 105, prediction of one or more parameters associated with assets 105, and energy optimization of assets. The determined result is a function of at least one performance indicator determined for one or more assets 105. In one example, the efficiency of the motor will be determined based on the performance indicators of the motor (such as current value, voltage value, torque, power, load, etc.).

[0112] In one embodiment, the results of the selected machine learning model are presented on the client device 130 for managing multiple assets 105 in the technical facility. In one embodiment, the method also includes determining updated results from the selected machine learning model based on the updated set of requirements received. In addition, the method includes presenting the updated results on the client device 130. It will be appreciated that the set of requirements input to the system can be varied by the user to visualize different results on the client device 130 based on the received varying input parameters. Notably, this technique enables the system to perform sensitivity analysis on a particular workflow. Advantageously, the sensitivity analysis helps a user (such as a plant operator) understand the impact of various parameter variations on the performance of one or more assets 105. In one example, the performance of an asset 105 (such as efficiency (%)) is captured by a performance indicator (such as input liquid carbon dioxide, input liquid ammonia, seawater flow, etc.). Therefore, by varying the input features of the performance indicator, the user is able to understand how changes in a particular input feature (e.g., a performance indicator) affect a target feature (e.g., the efficiency of an asset). Advantageously, the ability to analyze and visualize the sensitivity of input parameters on asset 105 ensures that a user can make informed decisions regarding one or more parameter values ​​that will ensure optimal values ​​for the various performance indicators and, therefore, the performance of the overall asset or technical facility.

[0113] In an exemplary embodiment, the asset 105 is a carbonization tower. The carbonization tower may have the following variables as input features, such as input liquid carbon dioxide, input liquid ammonia, seawater flow, chilled water flow, bottom gas flow, intermediate gas flow, seawater inlet temperature, chilled water inlet temperature, and output liquid flow, etc. The system is capable of predicting the efficiency of the carbonization tower. The user can enter several parameter value combinations as input features to observe the sensitivity of the features to the efficiency of the carbonization tower. For this purpose, a graphical user interface (GUI) is provided to the user to enter parameter values ​​and visualize the changes in real time on the GUI. The user may be provided with an input interface, such as a slider, a knob, a manual input box, etc., to select the value of each parameter within an operating range. Once any of these parameters is changed, a sensitivity curve graph with input features versus output features is presented on the client device 130.

[0114] Later in Figure 6 , Figure 7 and Fig. 9 The method for determining the results of the selected machine learning model for existing assets 105, new assets 105, and new asset classes 105 is explained in detail.

[0115] Figure 6 A flow chart of a method 600 for determining the results of a selected machine learning model according to an embodiment of the present invention is depicted. At step 602, one or more coefficients of the selected machine learning model are adjusted using data received and information extracted from at least one sensing unit. In particular, the weights of the nodes in different layers of the machine learning model are modified so as to adjust the results of the machine learning model as desired. It will be appreciated that a selected machine learning model that is preconfigured for a specific task (such as predicting anomalies in an asset 105) is not used as a fixed solution for all assets. Beneficially, the predictive model is adjusted to improve the accuracy of the machine learning model selected for the asset 105. In one example, the selected machine learning model can be further trained on data received from at least one sensing unit 105 associated with the asset 105, which sensing unit 105 is required to determine the performance indicator.

[0116] At step 604, the accuracy of the adjusted machine learning model is determined based on the received set of requirements. The operator can set the sensitivity of the accuracy of the result based on the requirements. For example, the accuracy of manufacturing or processing hazardous materials in a factory will be higher than the accuracy of manufacturing clothes in a factory. Therefore, the accuracy of the machine learning model can be set by the operator or the system itself. As previously explained, the accuracy of the adjusted machine learning model can be determined by testing the adjusted machine learning with a test data set or by executing the adjusted machine learning model in a simulation environment. At step 606, if the accuracy of the adjusted machine learning model is higher than a first predefined threshold, the result of the adjusted machine learning model is determined. The first predefined threshold may depend on the sensitivity and criticality of the requirements. The first predefined threshold may be defined by the system or manually defined based on the set of requirements. In the case where the accuracy of the adjusted machine learning model is higher than the first predefined threshold, the adjusted machine learning model is run and the results, such as predictions of different parameter values, asset efficiency, maintenance schedules, detected faults, RCA of faults, etc., are displayed on the display device 130 according to the request.

[0117] Figure 7 A flow chart of a method 700 for determining the results of a selected machine learning model when a new asset 105 is added to a technical facility according to an embodiment of the present invention is depicted. In step 702, the new asset 105 is mapped to a sensing unit 125 associated with the new asset 105. The mapping of the new asset 105 to the sensing unit 125 may be performed in the form of Figure 4 In a similar manner as mentioned in step 406 of . In one example, the new asset 105 is mapped to the corresponding sensing unit by determining the relationship between the new asset and the sensing unit using an engineering drawing such as obtained from a database. In another example, the operator is prompted to manually map the new asset 105 to the sensing unit 125 using a graphical user interface on the client device 130. In step 704, a new sensing unit 125 required for determining at least one performance indicator is selected. It should be understood that all data from the mapped sensing unit 125 may not be required to determine a particular performance indicator. Therefore, it is important to intelligently select a new sensing unit 125, which may be essential for determining the determined performance indicator. The performance indicator for the new asset 105 can be obtained from the knowledge base 122 based on the similarity between the new asset and assets of the same category that have been integrated into the system in the knowledge base 122.

[0118] In step 706, the coefficients of the selected machine learning model are readjusted based on the data obtained from the new sensing unit. The data may be condition data, operational data, specification data, parameter data, historical data, etc., which may be required to retrain the selected machine learning model. In one example, the data is historical data collected for the asset 105 over a period of time, which is used to retrain the machine learning model for the new asset 105. In step 708, when a new asset 105 is added, the results of the readjusted machine learning model are determined. Advantageously, the present invention is able to extend the results of the machine learning model to a large number of assets (e.g., assets of the same category) to be integrated into the system without having to redetermine the entire workflow. The machine learning model is readjusted based on the data received from the new sensing unit 125, and thus the readjusted machine learning model is also able to determine the results for the newly added asset 105.

[0119] Fig. 9 A flow chart of a method 900 for determining results of a machine learning model for a new asset class according to an embodiment of the present invention is depicted. At step 902, a new workflow is determined for the new asset class to be executed based on the determined at least one performance indicator. Figure 8 A method of determining at least one performance indicator for a new asset class is explained in . Since a workflow can be understood as an entry point to the system backend, a new workflow will be defined and started for a new asset class. In this context, a new asset class may be a new asset type that may not exist in the technical facility at a particular time, or there is no a priori information about the new asset class 105 in the knowledge base 122. In one example, the new asset class may be a pump, a heat exchanger, or any other asset that was not previously available in the technical facility. Based on the example Figure 8 Based on the set of requirements and performance indicators determined in the workflow, it is necessary to define a new workflow for the new asset class. The workflow may include one or more tasks for each of the plurality of assets 105. In one embodiment, the task is at least one of the following: anomaly detection in the asset 105, root cause analysis of the asset 105, remaining useful life estimation of the asset 105, performance optimization of the asset 105, prediction of one or more parameters associated with the asset 105, and energy optimization of the asset 105. Once the workflow is defined as per the requirements, the workflow is submitted for execution. Examples of workflows may include, but are not limited to, "train only", "deploy only", "train and deploy", and "test".

[0120] At step 904, a configured machine learning model is selected from the set of machine learning models based on the newly determined workflow. As used herein, a "configured machine learning model" refers to any machine learning model that is configured or trained to determine an outcome for a particular task, such as anomaly detection in an asset 105, root cause analysis of an asset 105, remaining useful life estimation of an asset 105, performance optimization of an asset 105, predicting one or more parameters associated with an asset 105, and energy optimization of an asset 105. Advantageously, the system is capable of automatically determining a configured machine learning model from the set of machine learning models in order to determine the requested outcome.

[0121] At step 906, one or more coefficients of the configured machine learning model are adjusted using the data received from the at least one sensing unit and the data from the knowledge base 122. In particular, the weights of the nodes in the different layers of the machine learning model are modified so as to adjust the results of the machine learning model as desired in the new workflow.

[0122] At step 908, the accuracy of the adjusted machine learning model is determined for the new asset class based on the set of requirements. The operator can set the sensitivity of the accuracy of the results based on the requirements. For example, the accuracy of manufacturing or processing hazardous materials in a factory will be higher than the accuracy of manufacturing clothes in a factory. Therefore, the accuracy of the machine learning model can be set by the operator or the system itself. As previously explained, the accuracy of the adjusted machine learning model can be determined by testing the adjusted machine learning with a test data set, or by executing the adjusted machine learning model in a simulated environment.

[0123] At step 910, if the accuracy of the adjusted machine learning model is above a second predefined threshold, the results of the adjusted machine learning model for the new asset class are determined. The second predefined threshold may depend on the sensitivity and criticality of the requirements. The second predefined threshold may be defined by the system or manually defined based on the set of requirements. In the event that the accuracy of the adjusted machine learning model for the new asset class is above the second predefined threshold, the adjusted machine learning model is run and the results such as predictions of different parameter values, efficiency of the asset, maintenance schedule, detected faults, root cause analysis (RCA) of detected faults, etc. are displayed on the display device 130 upon request.

[0124] The present invention aims to integrate domain knowledge, artificial intelligence expertise, predictive analytics in a flexible, reliable and scalable manner. The present invention enables users to implement high-quality predictive analytics solutions in a time-efficient manner. Beneficially, the present invention allows users to seamlessly configure assets, select related assets, define performance indicators, define workflows, select machine learning models, configure and train machine learning models, and determine the desired results based on the user's request. It will be appreciated that the present invention integrates domain knowledge and historical information into the system by generating a knowledge graph and continuously updating the knowledge graph. In other words, the front-end configurator is embedded with the knowledge of the assets being configured. This allows the complete asset to be configured in one go without going back and forth between the front-end and the back-end, thus significantly reducing the overall development effort.

[0125] In addition, the present invention executes workflows and runs tasks such as anomaly detection, parameter prediction, training AI models, storing AI models, updating knowledge bases, etc. in a time-efficient manner with minimal intervention from software code / AI / ML experts. The present invention allows training and retraining of AI models. Simulation models or AI model simulators endowed with synthetic data can be used to generate and test new assets, thereby providing users with access to the final solution in the shortest possible time. Beneficially, the present invention allows users to create new workflows based on available template workflows or modify existing workflows designed by the system. It will be appreciated that embodiments of the present invention enable users to define entire new workflows based on the user's requirements, or to expand existing workflows based on the requirements in an efficient manner. In other words, the present invention enables existing assets and new assets to be integrated into the system in a seamless manner. Advantageously, the present invention reduces costs, increases accuracy, and is time-efficient.

[0126] The present invention is not limited to a specific computer system platform, processing unit, operating system or network. One or more aspects of the present invention may be distributed in one or more computer systems, for example, a server configured to provide one or more services to one or more client computers or to perform a complete task in a distributed system. For example, according to various embodiments, one or more aspects of the present invention may be performed on a client-server system including components distributed in one or more server systems that perform multiple functions. These components include, for example, executable code, intermediate code or interpreted code that communicates over a network using a communication protocol. The present invention is not limited to being executable on any particular system or group of systems, and is not limited to any particular distributed architecture, network or communication protocol.

[0127] Although the present invention has been described in detail with reference to the preferred embodiments, the present invention is not limited to the disclosed examples. Other variations may be derived by those skilled in the art without departing from the scope of protection of the present invention.

[0128] List of reference numerals:

[0129] 100 Systems for managing multiple assets

[0130] 105 Assets

[0131] 110 Devices

[0132] 115 Edge Devices

[0133] 120 Network

[0134] 122 Knowledge Base

[0135] 125 Sensing unit

[0136] 130 Client Device / User Device

[0137] 135 Processing Units

[0138] 140 memory cells

[0139] 145 Storage Unit

[0140] 150 Database

[0141] 155 Input unit

[0142] 160 Output Units

[0143] 165 Asset Management Module

[0144] 170 Data Acquisition Module

[0145] 175 Asset Mapping Module

[0146] 180 Asset Mapping Module

[0147] 182 Performance Index Determination Module

[0148] 185 Workflow Definition Module

[0149] 190 Machine Learning Model Selection Module

[0150] 192 Output Module

[0151] 195 Bus

[0152] 200 Block diagram of system architecture

[0153] 202 Predictive Analysis Engine User Interface Layer

[0154] 204 Predictive Analysis Engine Backend Layer

[0155] 206 Predictive Analysis Engine Backend Layer

[0156] 208 Data Storage Layer

[0157] 210 Communication Layer

[0158] 212 Data Collection Layer

[0159] 214 AI Model Library Module

[0160] 216 AI model trainer module

[0161] 218 AI Model Simulator Module

[0162] 220 AI Workflow Manager Module

[0163] 222 System Management Module

[0164] 224 AI Model Library Module

[0165] 226 AI model trainer module

[0166] 228 AI Model Simulator Module

[0167] 230 AI Workflow Manager Module

[0168] 232 System Management Module

[0169] 234 KPI calculator module

[0170] 236 Data Cleaning Module

[0171] 238 Data Structuring Module

[0172] 240 Data Standardization Module

[0173] 242 Knowledge Base

[0174] 244 mongoDB

[0175] 246 Time Series Database

[0176] 248 OPC UA Module

[0177] 250 TCP / IP module

[0178] 252 MQTT module

[0179] 254 Cooler

[0180] 256 Pump

[0181] 258 HVAC

[0182] 260 PLC

[0183] 262 Exhaust Fan

[0184] 300 System Architecture Block Diagram

[0185] 302+ assets

[0186] 302A Pump

[0187] 302B Oil Well

[0188] 302C Heat Exchanger

[0189] 302D HVAC

[0190] 302E Cooler

[0191] 302N Motor

[0192] 304 Server

[0193] 306 User Interface (UI) Layer

[0194] 308 Backend Layer

[0195] 310 Business Connector Layer

[0196] 312 Navigation Applications

[0197] 314 Application Framework

[0198] 316 Graphics Frame

[0199] 318 Angle Frame

[0200] 320 Programming Language Specific Kits

[0201] 322 Programming language specific editors

[0202] 324 AI model training module

[0203] 326 AI model simulation module

[0204] 328 AI model prediction module

[0205] 330 Workflow Manager Module

[0206] 332 Knowledge Graph Module

[0207] 334 Management Module

[0208] 336 Navigation Module

[0209] 338 Database Access API

[0210] 340 Security API

[0211] 342 Other APIs

[0212] 344 Communication Protocol Access API

[0213] 346 Mongo Database

[0214] 348 Inflow Database

[0215] 350 Communication Module

[0216] 352 OPC / UA Module

[0217] 354 TCP / IP module 354

[0218] 356 MQTT module

[0219] 358 Communication protocol module

[0220] 400 Methods for managing multiple assets in a technical facility

[0221] 500 Method for determining at least one performance indicator

[0222] 600 Methods for determining the results of the selected machine learning model

[0223] 700 Method for determining the results of a selected machine learning model when a new asset is added to a technical facility

[0225] 800 Method for calculating at least one performance indicator when a new asset class is added to a technical facility

[0227] 900 Methods for determining the results of machine learning models for new asset classes

[0228] 1000 Methods for adding new machine learning models to a collection of machine learning models

Claims

1. A method (400) for managing a plurality of assets (105, 302A-N) in a technical facility, the method comprising: Receiving, by a processing unit (135), a set of requirements for managing a plurality of assets (105, 302A-N); selecting, by a processing unit (135), one or more assets (105, 302A-N) from a plurality of assets (105, 302A-N) based on the received set of requirements; Mapping, by a processing unit (135), one or more assets (105, 302A-N) to corresponding sensing units (125); extracting, by a processing unit (135), information associated with the selected one or more assets (105, 302A-N) and the received set of requirements, wherein the information includes domain knowledge stored in a knowledge base (122); determining, by a processing unit (135), at least one performance indicator based on information extracted from a knowledge base (122) and a mapped sensing unit (125); defining, by a processing unit (135), a workflow to be executed based on the determined at least one performance indicator; Selecting, by a processing unit (135), a configured machine learning model from a set of machine learning models based on a defined workflow, wherein the set of machine learning models includes one or more machine learning models configured to determine an outcome of a specific task; An outcome of a selected machine learning model is determined by a processing unit (135) based on the received set of requirements, wherein the outcome of the selected machine learning model is a function of at least one determined performance metric of the selected one or more assets (105, 302A-N).

2. The method (400) of claim 1, further comprising presenting the determined results on a client device (130) for use in managing a plurality of assets (105, 302A-N) in a technical facility.

3. The method (400) of claim 1, further comprising: Receiving, by a processing unit (135), an updated set of requirements; determining updated results from the selected machine learning model based on the received updated set of requirements; The updated results are presented on the client device.

4. The method (500) of claim 1, wherein determining the at least one performance indicator comprises: Selecting, by a processing unit (135), at least one sensing unit (125) required for calculating the at least one performance indicator; Obtaining, by the processing unit (135), data required for determining the at least one performance indicator from the selected at least one sensing unit (125); The at least one performance indicator is calculated based on the obtained data.

5. The method (400) of claim 1, further comprising generating user profiles for one or more users associated with the client device (130), wherein each user profile is associated with a set of functions accessible to the user.

6. The method (400) of claim 1, wherein mapping one or more assets (105, 302A-N) to corresponding at least one sensing unit (125) comprises: determining a relationship between an asset (105, 302A-N) and a sensing unit (125) using the engineering data; and Assets (105, 302A-N) are mapped to sensing units (125) based on the determined relationships.

7. The method (600) of claim 1, wherein determining the results of the selected machine learning model comprises: Using data received from at least one sensing unit (125) and information extracted to adjust one or more coefficients of a selected machine learning model; determining an accuracy of the adjusted machine learning model based on the received set of requirements; If the accuracy of the adjusted machine learning model is above a first predefined threshold, a result of the adjusted machine learning model is determined.

8. The method (700) of claims 1 and 6, further comprising determining a result of the selected machine learning model when a new asset (105) is added to the technical facility, wherein determining a result of the selected machine learning model comprises: mapping the new asset (105) to a sensing unit (125) associated with the new asset (105); selecting a new sensing unit (125) required for determining the at least one performance indicator; re-adjusting coefficients of the selected machine learning model based on data obtained from the new sensing unit (125); and When new assets (105) are added to the technology facility, the results of the re-tuned machine learning model are determined.

9. The method (800) of claim 1, further comprising calculating at least one performance indicator when a new asset class (105) is added to the technical facility, wherein calculating at least one performance indicator comprises: mapping the new asset class (105) to a sensing unit (125) associated with the new asset class (105); obtaining information related to the new asset class (105) from a database (150); updating the knowledge base (122) with information related to the new asset class (105); calculating a set of formulas, wherein the set of formulas is a function of data received from the sensing unit (125) and known performance indicators; determining at least one performance indicator for the new asset class (105) based on the calculated set of formulas; The knowledge base is updated using the determined performance indicators (122).

10. The method (900) of claim 7, further comprising determining a result of the machine learning model for the new asset class (105), wherein determining the result comprises: determining a new workflow for a new asset class (105) to be executed based on the determined at least one performance indicator; Based on the newly determined workflow, select a configured machine learning model from a collection of pre-configured machine learning models; adjusting one or more coefficients of a configured machine learning model using data received from at least one sensing unit (125) and domain knowledge from a knowledge base (122); determining the accuracy of the machine learning model adjusted for the new asset class (105) based on the set of requirements; If the accuracy of the adjusted machine learning model is above a second predefined threshold, then determining the results of the machine learning model adjusted for the new asset class (105).

11. The method (1000) of claim 1, further comprising adding a new machine learning model to the set of machine learning models, wherein adding the new machine learning model comprises: obtaining data related to an asset (105, 302A-N) for which a new machine learning model is to be added, wherein the data is stored in a database (150); training a new machine learning model using data obtained from the database (150); Determine the accuracy of new machine learning models; If the accuracy of the new machine learning model is above a third predefined threshold, the new machine learning model is added to the set of machine learning models associated with the asset (105, 302A-N) used for the specific task.

12. A method (400) according to any of the preceding claims, wherein the task is at least one of: anomaly detection in an asset (105, 302A-N), root cause analysis of an asset (105, 302A-N), remaining useful life estimation of an asset (105, 302A-N), prediction of parameters associated with an asset (105, 302A-N), performance optimization of an asset (105, 302A-N), and energy optimization of an asset (105, 302A-N).

13. The method of claim 1, further comprising automatically updating a mapping between the asset (105, 302A-N) and the sensing unit (125) in real time when the sensing unit (125) is online.

14. An apparatus (110) for managing a plurality of assets (105, 302A-N) in a technical facility, the apparatus (110) comprising: one or more processing units (135); and A memory unit (140) communicatively coupled to one or more processing units (135), wherein the memory unit (140) includes an asset management module (165) stored in the form of machine-readable instructions executable by the one or more processing units (135), wherein the asset management module (165) is configured to perform the method steps according to any one of claims 1 to 13.

15. A system (100) for managing a plurality of assets (105, 302A-N) in a technical facility, the system comprising: one or more devices (130) configured to provide a set of requirements for managing a plurality of assets (105, 302A-N); one or more sensing units (125) for providing data associated with a plurality of assets (105, 302A-N); a knowledge base (122) comprising domain knowledge related to a plurality of assets (105, 302A-N); and The apparatus (110) of claim 14, being communicatively coupled to one or more devices (130), wherein the apparatus (110) is configured to manage a plurality of assets (105, 302A-N) based on a set of requirements according to any of method claims 1 to 13.

16. A computer program product having machine-readable instructions stored therein, which, when executed by one or more processing units (135), cause the processing units (135) to perform the method according to any one of claims 1 to 13.

Images