Monitoring a complex process with at least two process steps.

Abstract

A computer implementation method for monitoring a complex chemical process having at least two process steps, comprising: a step of providing monitoring data by at least one monitoring process for each of the at least two process steps; a step of providing aggregated data by an aggregation layer by summarizing and / or combining all of the provided monitoring data; and a step of providing instruction data based on the aggregated data.

Description

Technical Field

[0001] The present disclosure relates to a computer-implemented method for monitoring a complex process having at least two process steps, an apparatus for monitoring a complex process having at least two process steps, the use of control data for controlling or adapting a complex process and / or at least one process step of a complex process, and computer program elements.

[0002] Background Art A general background of the present disclosure is the monitoring of complex processes such as chemical processes having at least two process steps.

[0003] In general technical practice, the monitoring of complex processes is involved and very complex based on the interaction of multiple different, especially fundamentally different processes or sub-processes. Therefore, in general technical practice, the monitoring of complex processes has to be split into multiple sub-monitoring processes that each monitor different sub-processes of the complex process as a whole.

[0004] It has been found that there is a need for an alternative, smarter way of monitoring complex processes. Also, by including monitoring processes and aggregation layers in the monitoring of complex processes, multiple sub-monitoring processes can be rendered ineffective, making the monitoring of complex processes easier and less complex. Furthermore, based on this, resources such as power consumption and hardware requirements can be conserved and reduced.

[0005] Summary of the Invention In one aspect of the present disclosure, there is provided a computer-implemented method for monitoring a complex process having at least two process steps, comprising: providing monitoring data by at least one monitoring process for each of the at least two process steps; The steps include: providing aggregated data by an aggregation layer by summarizing and / or combining all the provided monitoring data; Steps to provide instruction data based on aggregated data and Computer implementation methods, including the above, are presented.

[0006] In a further aspect of this disclosure, there is an apparatus for monitoring a complex process having at least two process steps, One or more computing nodes and one or more computer-readable media, which, when executed by one or more computing nodes, the device, A step of providing monitoring data by at least one monitoring process for each of at least two process steps, The steps include: providing aggregated data by an aggregation layer by summarizing and / or combining all the provided monitoring data; Steps to provide instruction data based on aggregated data and One or more computer-readable media having computer-executable instructions structured to perform the following actions A device comprising the above is presented.

[0007] A system for monitoring a complex process having at least two process steps, optionally A first providing unit for providing monitoring data by at least one monitoring process for each of at least two process steps, A second providing unit provides aggregated data by an aggregation layer, which summarizes and / or combines all of the provided monitoring data, A third supply unit that provides instruction data based on aggregated data and A system is presented that includes the following features.

[0008] In a further embodiment, the use of control data is presented to control or adapt a complex process and / or at least one process step of a complex process.

[0009] In a further embodiment, a computer element, in particular a computer program product or computer-readable medium, is presented having instructions configured to perform any step of the method disclosed herein in the apparatus disclosed herein when executed on one or more computing devices.

[0010] Any disclosures and embodiments described herein relate to the methods, apparatus, systems, and computer program elements outlined above, and vice versa. Advantageously, any advantages provided by any embodiment and example apply equally to all other embodiments and examples, and vice versa.

[0011] As used herein, “determining” also includes “initiating or causing a decision to be made, estimating, calculating, or modeling,” and “providing” also includes “initiating or causing a decision, generation, selection, measurement, delivery, or receipt.”

[0012] The methods, apparatus, and computer elements disclosed herein provide efficient, sustainable, and robust methods for monitoring complex processes. In particular, efficient, sustainable, and robust methods for monitoring complex processes are based at least on including multiple monitoring processes and aggregation layers in the monitoring of the complex process. Therefore, monitoring complex processes can be made easier and less complex, as multiple sub-monitoring processes become inactive and unnecessary. Furthermore, resources such as power consumption and hardware requirements can be saved and reduced.

[0013] The object of the present invention is to provide an efficient, sustainable, and robust method for monitoring complex processes. The above and other objects will become clear from the following description and are resolved by the subject matter of the independent claims. Dependent claims refer to preferred embodiments of the present invention.

[0014] The term "complex process" should be understood broadly in this context to represent any process involving at least two process steps, and in particular multiple process steps. For example, a complex process may be, but is not limited to, a chemical process, a continuous process, a continuous chemical process, a manufacturing process, a band reactor process, a sandwich panel manufacturing process, a foam manufacturing process, a duroplastic foam manufacturing process, and / or a thermoplastic foam manufacturing process, a polyurethane manufacturing process, and / or a polyisocyanurate manufacturing process, and / or a polystyrene manufacturing process, and / or a melamine resin manufacturing process. A complex process may also include multiple subprocesses.

[0015] The term "monitoring process" should be understood broadly in this context to represent any execution of an algorithm or model configured for monitoring, tracking, measuring, inspecting, and / or observing a state, function, process, and / or manufacturing process. Furthermore, a monitoring process is configured to provide monitoring data that shows the results of monitoring the state, function, process, and / or manufacturing process. The monitoring data may be timestamped with the time of provision / measurement. For example, a monitoring process may be, but is not limited to, a machine learning algorithm, a neural network, a measurement, a region-based convolutional neural network, AI, i.e., artificial intelligence, a model, a machine learning network with a residual neural network, a machine learning model, a classification model, and / or a sum of squared deviations algorithm. Multiple monitoring processes may be identical or different from one another; that is, they may monitor the same parameters or variables, or they may monitor different parameters or variables. A monitoring process may be executed for only one of at least two process steps of a complex process, or it may be executed separately for multiple process steps of at least two process steps of a complex process; that is, each process step may be monitored by its own monitoring process.

[0016] The aggregation layer provides aggregated data by summarizing and / or combining all the provided monitoring data. In particular, the aggregation layer may summarize and / or combine all the provided monitoring data and / or the provided intermediate aggregated data. That is, monitoring data is summarized and / or combined for each of at least two process steps in order to provide aggregated data. More specifically, the term aggregation layer should be understood broadly in this case to represent any algorithm and / or model for summarizing, combining and / or comparing all the provided monitoring data and / or the provided intermediate aggregated data with reference data. In other words, the aggregation layer is the final aggregation layer that receives all the provided data, i.e., monitoring data and / or intermediate aggregated data. For example, the aggregation layer may be an algorithm or a machine learning algorithm, but is not limited to these. The aggregation layer may receive monitoring data and / or provided intermediate aggregated data from all, i.e., multiple, identical or different monitoring processes and / or intermediate aggregation layers. The results of the aggregation layer are provided as data, i.e., so-called aggregated data. The reference data may be any predetermined, pre-configured, and / or pre-defined data that indicates an optimal operating state. Reference datasets can be extracted during normal steady-state operation, as long as the machines and equipment are operating within their optimal parameter windows. It is worth noting that the data is progressively transformed from abstract measurement data to specific process control datasets with each integration layer.

[0017] The terms "compare" or "comparison" should be understood broadly in this context to refer to any process of matching or correlating provided monitoring data and / or provided intermediate aggregated data with reference data.

[0018] The term "indicator data" should be understood broadly in this context and refers to any data that indicates the state of a complex process. For example, indicator data may include alarms, interruptions in a complex process, error notifications in a complex process, and / or recommendations on how to deal with interruptions, alarms, and / or errors. Indicator data is based on aggregated data.

[0019] The term "intermediate aggregation layer" should be understood broadly in this context to represent any algorithm and / or model for summarizing, combining, and / or comparing one or more provided monitoring data sets with reference data. For example, the intermediate aggregation layer may be, but is not limited to, an algorithm or a machine learning algorithm. The intermediate aggregation layer may receive monitoring data provided from one or more identical or different monitoring sources. The results of the intermediate aggregation layer are provided as data, i.e., so-called intermediate aggregation data. The intermediate aggregation layer sends the provided intermediate aggregation data to the aggregation layer for further processing. The reference data may be any predetermined, pre-configured, and / or pre-defined data that represents an optimal operating state. Thus, the intermediate aggregation layer behaves similarly to the aggregation layer described above, but breaks down the aggregation task into smaller modular steps.

[0020] As used herein, the term "provide" should be understood broadly in this context to represent any method for receiving, measuring, determining, generating, selecting, sending, or receiving parameters or data. For example, data can be provided and received via the Internet, or provided or modified / adapted by manual input by a user via a user interface.

[0021] The control data used in this specification should be understood broadly in this case and relates to any data configured to control or adapt a complex and continuous chemical process and / or at least one process step of a complex and continuous chemical process based on the instruction data. The control data can be provided by a control unit and configured to control one or more technical means of a process / manufacturing device, but is not limited thereto. The control data may be one-dimensional, two-dimensional or three-dimensional data. The control data is based on the provided instruction data.

[0022] In a method for monitoring a complex process having at least two process steps of one embodiment, the aggregated data is provided by an aggregation layer by comparing the provided monitoring data with reference data. The aggregation layer enables splitting the complex monitoring tasks of the complex process into smaller monitoring modules. Each of the smaller monitoring modules is modular and can be developed and maintained independently. In addition, these smaller monitoring modules can be reused and reconfigured with the help of the aggregation layer to monitor different complex processes.

[0023] In a method for monitoring a complex process having at least two process steps of a further embodiment, the monitoring data for each of the at least two process steps is provided by a separate monitoring process of at least one monitoring process.

[0024] In a method for monitoring a complex process having at least two process steps of a further embodiment, the monitoring data for at least two of the at least two process steps is provided by the same monitoring process of at least one monitoring process.

[0025] In a method for monitoring a complex process having at least two process steps of a further embodiment, the provision of aggregated data by an aggregation layer is provided by an algorithm, particularly a machine learning algorithm. The machine learning algorithm can have much higher generality with respect to adoption for different complex processes or process steps. Further, the machine learning algorithm is generally more robust against non-process-related environmental conditions such as illumination, shadowing, color balance, or noise.

[0026] In a method for monitoring a complex process having at least two process steps of a further embodiment, the method further includes the step of providing intermediate aggregated data by an intermediate aggregation layer based on a subset of the provided monitoring data, and the provision of aggregated data by the aggregation layer is based on the provided intermediate aggregated data. In particular, the provision of aggregated data by the aggregation layer is based on all of the provided intermediate aggregated data.

[0027] In a method for monitoring a complex process having at least two process steps of a further embodiment, the provision of intermediate aggregated data by the aggregation layer is provided by an algorithm, particularly a machine learning algorithm.

[0028] In a method for monitoring a complex process having at least two process steps of a further embodiment, the monitoring process is at least one of a group consisting of a machine learning algorithm, a neural network, a measurement, a region-based convolutional neural network, an AI model, a machine learning network having a residual neural network, a machine learning model, a classification model, and a sum of squared deviations algorithm.

[0029] In a method for monitoring a complex process having at least two process steps of a further embodiment, if at least one monitoring process is multiple monitoring processes, the monitoring processes are different from each other. Alternatively, if at least one monitoring process is multiple monitoring processes, the monitoring processes are identical from each other. By using different monitoring processes, if at least one monitoring process is multiple monitoring processes, monitoring of multiple different process steps can be performed in a specific manner.

[0030] In a method for monitoring a complex process having at least two process steps of a further embodiment, the instruction data includes at least one of a group consisting of alarms, interruptions in the complex process, error notifications in the complex process, and recommendations on how to deal with interruptions, alarms, and errors, respectively. This instruction data enables a feedback loop. One possible feedback loop is a direct, manual, or automated interaction with a process control system to keep the complex process within an optimal process window, or to shift process parameters to move the complex process back into its optimal window if it has moved out of or begun to move out of the optimal process window. Furthermore, this instruction data can also be used to automatically loop back and retrain a monitoring algorithm. Finally, the instruction data can even be used to automatically find a more optimized manufacturing time by an unattended, computer-driven, for example, brute-force feedback trial-and-error loop.

[0031] In a method for monitoring a complex process having at least two process steps of a further embodiment, the method further includes the step of providing control data to control or adapt at least one process step of the complex process based on instruction data. Alternatively or additionally, the provision of control data may also control or adapt an entire complex, sequential chemical process based on instruction data. Automation can be provided by providing control data.

[0032] A method for monitoring a complex process having at least two process steps in a further embodiment, wherein the complex process is at least one of the group consisting of a chemical process, a continuous process, a complex continuous chemical process, a manufacturing process, a band reactor process, a sandwich panel manufacturing process, a foam manufacturing process, a polyurethane manufacturing process, and / or a polyisocyanurate manufacturing process.

[0033] The following further explanation of this disclosure will be provided with reference to the attached figures. [Brief explanation of the drawing]

[0034] [Figure 1] This diagram shows a computer implementation method for monitoring a complex process that has at least two process steps. [Figure 2] This outlines common methods for monitoring complex processes that have at least two process steps, with ML standing for Machine Learning. [Figure 3] This diagram shows an exemplary process flow chart for manufacturing thermal insulation panels using a computer implementation method for monitoring complex processes. [Figure 4] A box diagram of a system for monitoring a complex process having at least two process steps is shown.

[0035] Modes for carrying out the invention The following embodiments are merely examples of how to implement the methods and systems disclosed herein and should not be considered limiting.

[0036] Figure 1 shows a flowchart of a computer implementation method for monitoring complex processes having at least two process steps, namely, chemical processes, continuous processes, manufacturing processes, band reactors, sandwich panel manufacturing processes, foam manufacturing processes, duroplastic foam manufacturing processes, thermoplastic foam manufacturing processes, polyurethane manufacturing processes, polyisocyanurate manufacturing processes, polystyrene manufacturing processes, and melamine resin manufacturing processes. The following describes an exemplary sequence of steps according to this disclosure. However, the provided sequence is not mandatory; that is, all or some steps may be performed in a different order or simultaneously.

[0037] The method steps shown in Figure 1 can also be performed by a system.

[0038] In the first step, monitoring data is provided by at least one monitoring process for each of at least two process steps. The monitoring processes are machine learning algorithms, neural networks, measurements, region-based convolutional neural networks, AI models, machine learning networks with residual neural networks, machine learning models, classification models, and / or sum-of-squares-deviations algorithms. The monitoring data represents the results of the monitoring processes. Monitoring data for at least one of the at least two process steps is provided by the same monitoring process for at least one of the monitoring processes.

[0039] In the second step, aggregated data is provided by the aggregation layer by summarizing and / or combining all of the provided monitoring data, and in particular, the aggregated data is further provided by comparing the provided monitoring data with reference data. The provision of aggregated data by the aggregation layer is provided by a machine learning algorithm.

[0040] In the third step, instructional data is provided based on aggregated data. This instructional data includes alarms, interruptions in complex processes, errors in complex processes, and / or recommendations on how to address interruptions, alarms, and / or errors.

[0041] Optionally, the method further includes the step of providing intermediate aggregated data by an intermediate aggregate layer based on a subset of the provided monitoring data. The provision of aggregated data by the aggregate layer is based on all of the provided intermediate aggregated data. The provision of intermediate aggregated data is provided by the aggregate layer by a machine learning algorithm.

[0042] Optionally, the method further includes a step of providing control data for controlling or adapting a complex process and / or at least one process step of a complex process.

[0043] Figure 2 outlines a common method for monitoring a complex process that has at least two process steps.

[0044] A complex process involves multiple process steps, particularly substeps. Each process step is monitored by another monitoring process and / or the same monitoring process. The monitoring processes are machine learning, ML algorithms, or other measurements, such as those provided by sensors. The machine learning algorithms may be the same or different. The other measurements may be the same or different. The results of the monitoring processes are monitoring data. The monitoring data is provided to at least one intermediate aggregation layer. The intermediate aggregation layer is a machine learning algorithm. The intermediate aggregation layer provides intermediate aggregated data and provides this provided intermediate aggregated data to the aggregation layer. The aggregation layer also includes machine learning algorithms for providing aggregated data. Based on the aggregated data, monitoring, control, etc., can be provided.

[0045] Figure 3 shows an exemplary process flow diagram for manufacturing an insulating panel using a computer implementation method for monitoring a complex process. This process provides process and quality control for a complex industrial chemical process through multiple (n≧2), parallel, independently operating simple neuron networks.

[0046] This process included the following steps for manufacturing an insulating panel using a solid foam core. In particular, the sandwich panel manufacturing process: an insulating foam core with layered surfaces on both sides sandwiched between covered metal covers, i.e., the top and bottom surfaces are covered with metal sheets. The foam is manufactured in-line directly by polyurethane or polyisocyanurate reaction.

[0047] In Step 1, the coated metal cover (coil) is pre-treated. During pre-treatment, sub-steps are performed including profiling of the metal coating (surface area and edge connection profile), cleaning of the metal surface (mechanical, corona, etc.), and heating to prepare for ideal PU reaction conditions. During pre-treatment, monitoring detects defects, voids, scratches, impurities, and cracks within or on the surface of the (coated) metal sheet. Ideally, this surface should be homogeneous and completely clean. To detect defects, voids, scratches, impurities, and cracks within or on the surface of the (coated) metal sheet, an optical camera is used to cover the entire width of the metal sheet, and video / image data from multiple cameras is provided and stitched together to obtain a full image of the entire width of the metal sheet. Furthermore, a Faster R-CNN (Region-based Convolutional Neural Network) network is trained to detect defects, voids, scratches, impurities, and cracks within or on the surface of the (coated) metal sheet that differ from the clean (coated) metal sheet. The ideal output of Faster R-CNN is 0, meaning nothing is detected. An output of Faster R-CNN other than 0 indicates a potential problem. Faster R-CNN, i.e., the AI ​​model, is trained with OK data (human-monitored) from a working solution. Defective data was artificially generated by intentionally adding scratches, liquid and solid impurities to metal surfaces during manufacturing shutdowns. For another complex process, the inventors use the YOLO object detection algorithm to detect the aforementioned defects. This algorithm enables a much faster detection speed, at the expense of some accuracy, compared to Faster R-CNN, Single-Shot MultiBox Detector (SSD), and Retina-Net detection models, which the inventors also tested.

[0048] Furthermore, monitoring detects erroneous or heterogeneous temperature distributions on the metal sheet surface non-contact during manufacturing. To detect erroneous or heterogeneous temperature distributions on the metal sheet surface, a spectral imaging device, i.e., an IR thermography camera, is used to monitor the accurate and homogeneous temperature of the metal sheet. To detect this, a simple linear regression model with two output layers is used, where the first output layer represents absolute temperature and the second output layer represents temperature homogeneity from 0 (heterogeneous) to 1 (perfectly homogeneous / uniform). OK data was generated during manufacturing by parallel monitoring with a handheld portable thermal camera. Non-OK data was recorded on sample metal sheets treated with ice cubes and a blowtorch to create temperature heterogeneity.

[0049] In Step 2, the adhesion promoter is applied. The adhesion promoter, which is a reactive (1K or 2K) liquid, is applied to the inside of one or both sides of the metal cover sheet, and there are many different application methods for the promoter, such as moving nozzles, spray nozzles, drip beams, and rotating washer discs. In this step, monitoring detects complete or partial absence of the adhesion promoter, as well as non-uniform distribution on one or more metal sheets. To detect this, an optical camera is used to cover the entire width of the metal sheet covered with the adhesion promoter. Video / image data from multiple cameras are stitched together to obtain a full image of the entire width of the metal sheet. This detection is provided by a machine learning ML network with a Resnet (Residual Neural Network) architecture configured to detect deviations of the metal sheet from the ideal adhesion promoter coating. The model was trained on OK data from normal manufacturing runs. Within these runs, the distribution of the adhesion promoter was monitored and controlled by human visual inspection, and the amount of adhesion promoter was controlled by performing offline weight control following flow measurement. Defective data, or non-OK data, was recorded during the regular start and stop procedures of the process. This automatically generated highly uneven and flawed videos and images.

[0050] In step 3, one or more liquid raw materials for the insulating foam core are applied. The foam raw materials are pre-treated and pre-mixed before application. The reactive mixture is applied across the entire width of the lower metal sheet, and there are many different application methods with different outlet shapes, such as nozzles, multiple nozzles, multiple nozzles on a bar ("poker"), and pre-foaming pipes. The application device can be fixed or it can move the metal sheet from side to side, i.e., vibrate. All application methods aim to cover the entire width of the metal sheet with the reactive raw material. In this step, monitoring is performed to detect (premature, partial, or complete) blockages of the raw material outlet device that result in defects or non-uniformity in the raw material distribution, which can lead to defects in the foam core that is ultimately produced. For this detection, in the case of a fixed outlet device, an optical camera is used to cover the entire width of the device and the full size of the outgoing raw material flow until it hits the lower metal sheet. In the case of a mobile dispensing device, the camera can be fixedly mounted to cover the entire width of the vibrating area, or more preferably mounted to move with the dispensing device. The detection method employs two machine learning (ML) models. In the first step, a very shallow and fast Mobilenet R-CNN clone is used to detect the actual location (region of interest (ROI)) of the ejection device. In the second step, an autoencoder model (encoder + decoder) is used to reconstruct an image from the input image using a weighted sum of squared pixel differences. If the difference exceeds a threshold, an ejection blockage (premature, partial, or complete) is detected. Both networks are trained with OK images during normal manufacturing operations monitored by humans. Defective images are not required for this method.

[0051] In step 4, the raw material is formed into the final foam core. The liquid raw material begins leveling and foaming (chemical or physical foaming). The foam expands until it is trapped by hitting the upper metal sheet (the side regions are trapped by moving temporary blocks). Simultaneously, chemical reactions initiate crosslinking and hardening of the foam structure. In this step, monitoring detects heterogeneity in foam leveling and defects in the growing foam, such as bubbles, blow-offs, and ripples. For detection, video / image data is provided by an optical camera to cover the entire width of the raw material initiating foaming on the lower metal sheet. Detection is provided by a high-speed classification model, such as VisionTransformer, which is perfectly suited to this task. OK data is generated during normal operation. Defect data regarding bubbles and blow-offs is collected during process startup. Defect data regarding foam leveling was created by intentionally using incorrect (higher) temperatures for the metal sheet and raw material. In this step, monitoring is performed on foaming rate, i.e., foam expansion; gelation time, i.e., chemical reaction rate; and contact time and contact location, i.e., when the foam contacts the upper panel. Monitoring is provided by optical measuring devices such as laser rangefinders, and laser profilers (2D, 3D) and 3D industrial cameras may also function. Due to the limited space and constraints of the ATEX zone, the inventors use a 3D laser profiler. Monitoring is provided by an LSTM (Long Short-Term Memory) model to analyze the time series of foaming rate, gelation time, contact time, and contact location. Based on actual and historical (time-series) data, future development of these values ​​is predicted. Network training is performed during normal operation, with the operator changing the manufacturing parameters each time and then keeping them constant for more than one day of manufacturing. Training data was collected during manufacturing time exceeding four weeks. In this method, the inventors obtain training start measurement data results for almost all relevant manufacturing parameters.

[0052] Step 5 provides pressing and curing of the sandwich while the foam is curing.

[0053] Step 6 provides instructions for cutting / sewing the endless sandwich into panel sizes.

[0054] In step 7, the cooling of the panel storage area is provided. The panel with the reacted foam core is stored until it cools to room temperature. Before, during, and after this step, the panel is moved very slowly, allowing it to find an ideal position for inspecting the final panel quality. In this step, monitoring detects dimensions and shape (buckling, expansion, convex or concave bending). For this purpose, two 2D laser profilometers are used at the entrance to the cooling area. One profilometer is from the top surface and the other from the bottom surface of the panel, and the panel moves through / between the profilometers, allowing both flat surfaces to be scanned by the profilometers. By combining the upper 2D profile (measured from above) with the lower 2D profile (measured from below), complete 3D shape data can be extracted. The combination can be done, for example, by subtracting a given calibrated distance between the upper 2D profilometer and the lower 2D profilometer from these profiles. When a panel is continuously moving between profilers, the 2D data recorded over time can be converted to 3D by calculating the missing length parameter z in the direction of movement using z = velocity·time. In particular, absolute thickness and thickness homogeneity, as well as any bending, can be easily tracked from the 3D shape data. Detection is provided by a simple sum of squared deviations algorithm to compare the measured 3D shape (2D profile over time) with a given ideal geometric shape. A simple threshold provides feedback on whether the panel is within or outside of specifications.

[0055] Step 8 provides the final preparation, milling, and cleaning of the panel.

[0056] Step 9 involves packaging and storage / preparation for transport.

[0057] Figure 4 shows a box diagram of a system for monitoring a complex process having at least two process steps. System 30 for monitoring a complex process having at least two process steps comprises a first providing unit 31 for providing monitoring data by at least one monitoring process for each of the at least two process steps. The first providing unit 31 includes machine learning algorithms, neural networks, measurements, domain-based convolutional neural networks, AI models, machine learning networks with residual neural networks, machine learning models, classification models and / or sum-of-squares-deviation algorithms. System 30 further comprises a second providing unit 32 for providing aggregated data. The second providing unit 32 includes an aggregation layer, which provides aggregated data by summarizing and / or combining all of the provided monitoring data, in particular, by comparing the provided monitoring data with reference data. The aggregation layer is a machine learning algorithm. System 30 further comprises a third providing unit 33 for providing instructional data based on the aggregated data. The instructional data includes alarms, interruptions in the complex process, errors in the complex process, and / or recommendations on how to deal with interruptions, alarms, and / or errors.

[0058] Optionally, the system 30 further comprises a fourth providing unit 34 for providing intermediate aggregated data by an intermediate aggregation layer based on a subset of the provided monitoring data. The fourth providing unit 34 includes an intermediate aggregation layer. The provision of intermediate aggregated data is provided by the aggregation layer by a machine learning algorithm.

[0059] Optionally, the system 30 further comprises a fifth supply unit 35 for providing control data to control or adapt a complex process and / or at least one process step of a complex process.

[0060] This disclosure is described in conjunction with preferred embodiments as examples. However, those skilled in the art will be able to understand and implement other variations by carrying out the claimed invention from consideration of the drawings, this disclosure, and the claims. In particular, any of the presented steps can be carried out in any order; that is, the present invention is not limited to a particular order of these steps. Furthermore, the different steps do not need to be carried out at a specific location or one node in a distributed system; that is, each step may be carried out at a different node using different equipment / data processing units.

[0061] In this specification and in the claims, the phrase "comprising" does not exclude other elements or steps, and the indefinite article "a" or "a" does not exclude the plural. A single element or other unit may perform the function of several entities or items described in the claims. The mere fact that certain means are described in different dependent claims does not indicate that combinations of these means cannot be used in advantageous embodiments.

Claims

1. A computer implementation method for monitoring a complex process having at least two process steps, A step of providing monitoring data by at least one monitoring process for each of the at least two process steps, The steps include providing aggregated data by an aggregation layer by summarizing and / or combining all of the aforementioned provided monitoring data, A step of providing instruction data based on the aggregated data. Computer implementation methods, including those mentioned above.

2. The aggregated data is provided by the aggregation layer by comparing the provided monitoring data with the reference data. The computer implementation method according to claim 1.

3. The monitoring data for each of the at least two process steps is provided by a separate monitoring process for the at least one monitoring process. The computer implementation method according to claim 1 or 2.

4. The monitoring data for at least two of the two process steps is provided by the same monitoring process of at least one of the monitoring processes. The computer implementation method according to claim 1 or 2.

5. The provision of aggregated data by the aforementioned aggregation layer is provided by an algorithm, particularly a machine learning algorithm. The computer implementation method according to any one of claims 1 to 4.

6. The intermediate aggregation layer provides intermediate aggregated data based on a subset of the aforementioned monitoring data. It further includes, The provision of aggregated data by the aggregation layer is based on the provided intermediate aggregated data. The computer implementation method according to any one of claims 1 to 5.

7. The provision of intermediate aggregated data by the aforementioned intermediate aggregate layer is provided by an algorithm, particularly a machine learning algorithm. The computer implementation method according to claim 6.

8. The monitoring process is at least one of the group consisting of a machine learning algorithm, a neural network, measurement, a region-based convolutional neural network, an AI model, a machine learning network having a residual neural network, a machine learning model, a classification model, and a sum of squared deviations algorithm. The computer implementation method according to any one of claims 1 to 7.

9. If the aforementioned at least one monitoring process is a plurality of monitoring processes, the monitoring processes are different from each other. The computer implementation method according to any one of claims 1 to 8.

10. The instruction data includes at least one of the following: an alarm, an interruption of the complex process, an error notification in the complex process, and recommendations on how to deal with the interruption, alarm, and error, respectively. The computer implementation method according to any one of claims 1 to 9.

11. A step of providing control data for controlling or adapting at least one process step of the complex process based on the instruction data. A computer implementation method according to any one of claims 1 to 10, further comprising:

12. The complex process is at least one of the group consisting of chemical processes, continuous processes, manufacturing processes, band reactor processes, sandwich panel manufacturing processes, foam manufacturing processes, duroplastic foam manufacturing processes, thermoplastic foam manufacturing processes, polyurethane manufacturing processes, polyisocyanurate manufacturing processes, polystyrene manufacturing processes, and melamine resin manufacturing processes. The computer implementation method according to any one of claims 1 to 11.

13. An apparatus for monitoring a complex process having at least two process steps, wherein the apparatus comprises one or more computing nodes and one or more computer-readable media, which, when executed by the one or more computing nodes, the apparatus provides A step of providing monitoring data by at least one monitoring process for each of the at least two process steps, The steps include providing aggregated data by an aggregation layer by summarizing and / or combining all of the aforementioned provided monitoring data, A step of providing instruction data based on the aggregated data. A device comprising one or more computer-readable media having computer-executable instructions structured to perform a certain action.

14. Use of control data generated by the computer implementation method according to claim 11 for controlling or adapting a complex process and / or at least one process step of the complex process.

15. A computer program element having instructions configured to perform a step of the method according to any one of claims 1 to 12 in the apparatus according to claim 13 when executed on a computing device of a computing environment.

Images