A modeling evaluation method and system for complex equipment operation

Through the multi-level multi-modeling integrated design method based on machine learning, the problems of low generalization and low accuracy in device modeling analysis are solved, real-time status evaluation and fault diagnosis of complex equipment are realized, and equipment safety production is improved and intelligent operation and maintenance management capabilities.

CN113919230BActive Publication Date: 2024-11-12CHINA ORDNANCE EQUIP GRP AUTOMATION RES INST CO LTD
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
CN202111230046.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-10-20
Publication Date
2024-11-12
Estimated Expiration
2041-10-20

AI Technical Summary

Technical Problem

The existing technology has problems of low generalization and low accuracy in equipment modeling and analysis, and cannot effectively identify the operating status of complex equipment, and traditional algorithm models cannot adaptively adjust, which increases the risk of equipment production safety.

Method used

Using a multi-level multi-modeling integrated design method based on machine learning, through object-oriented modeling, combining basic device resources, evaluation strategies, algorithm model library optimization, evaluation and decision-making, a multi-level algorithm model library is formed, and multi-threading technology is used. Carry out model retraining and optimization to realize real-time evaluation and fault diagnosis of equipment operating status.

Benefits of technology

It improves the accuracy and generalization of equipment operating status evaluation, enhances the flexibility and efficiency of modeling and analysis, reduces the model adjustment time after the equipment is put into production, and reduces the later production cost.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113919230B_ABST
    Figure CN113919230B_ABST
Patent Text Reader

Abstract

In order to solve the problems of low generalization and low accuracy in existing modeling methods, the present invention proposes a modeling and evaluation method for the operation of complex equipment based on machine learning. The purpose is to propose a modeling and evaluation method for the operation of complex equipment in view of the shortcomings of existing analysis model technology, to integrate a variety of commonly used algorithms of machine learning in a universal and integrated manner, to form a basic algorithm model library for equipment operation status evaluation, and to design a management UI for the algorithm model library, evaluation strategy, etc. The algorithm library takes the production parameters, production mode and environmental parameters of complex equipment as input to provide data input for the model. The algorithm library votes on the analysis results based on the results of each algorithm, and finally analyzes and evaluates the operating status of the complex equipment. At the same time, the model of the algorithm model library is evaluated and optimized. Enterprises can use the algorithm library to accurately evaluate the operating status of complex equipment and effectively manage the equipment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent manufacturing, and in particular relates to a modeling evaluation method and system for complex equipment operation. Background Art

[0002] Facing a new wave of intelligent manufacturing in the world, in order to improve product R&D efficiency, reduce R&D costs and respond quickly to the market, domestic high-end manufacturing fields such as aviation and aerospace have taken intelligent manufacturing and digital twins as the core of manufacturing transformation. Modeling and intelligent analysis of equipment, workshops and factories are hot spots in digital twins. Modeling and analysis provide enterprises with production strategies and intelligent management as the core in the entire life cycle of the production process. In the field of power automation technology, a clustering-based equipment operation status evaluation method is adopted to establish an operation status evaluation model of power equipment. Based on the historical data of power equipment, the establishment of the power equipment analysis model is completed. Through model evaluation, quantitative analysis, demonstration and comparison of the operation status of power automation equipment, the operation status evaluation of power automation equipment is realized; in the field of shipbuilding, a marine engine status evaluation method and system are established based on digital twins. According to the parameter transfer relationship between the working process model, fault diagnosis model and structural strength finite element model, a corresponding digital twin model is established. The input of the digital twin model is the operation parameters and environmental parameters. The model finally outputs the corresponding evaluation parameters to finally determine the operation status of the marine engine. In the field of automobile manufacturing, a digital twin-driven intelligent drive-by-wire chassis system and its fault diagnosis method were proposed. A digital twin model system of the drive-by-wire chassis device was established through digital twin technology. Based on data drive, the virtual model and the physical entity operated consistently, realizing real-time data interaction and fault diagnosis between the drive-by-wire chassis device and the analysis model.

[0003] As an important direction for the transformation and development of the manufacturing industry today, smart manufacturing and digital twins are expected to solve the intelligent management problems of the actual production process through workshop, production line and equipment modeling and simulation analysis, and assist managers in making decisions through the results of massive data modeling and analysis in the production process. In the digital twin system, the real equipment is mapped to the virtual modeling based on data drive for real-time drive, and the real-time interaction of data is realized. However, when the model is analyzed for the equipment, a large amount of invalid data is often generated in the initial construction, which will lead to the low accuracy of the actual application model trained by the model, and it is impossible to accurately evaluate the real-time status of the equipment; secondly, the data analysis modeling for a device is often established by feature mining to establish a fixed algorithm model, or a model based on artificial experience knowledge. This model established in the early stage of production is difficult to accurately identify the operation status of the equipment after the enterprise's later production tasks and modes are adjusted, which will affect the cost control of the enterprise; in addition, after the production mode of the equipment is adjusted, its fault type or operation status often increases, and the traditional customized algorithm analysis model cannot be adjusted adaptively, which increases the risk of equipment safety production.

[0004] Before the equipment is actually put into production, modeling and analysis is an important part of the enterprise's intelligent production line to realize its digital twin, and it is a necessary link to provide a more ideal production plan to increase the core competitiveness of the enterprise. In recent years, industrial equipment such as industry and metallurgy have shown a trend towards large-scale and complex development. These complex equipment cannot completely rely on traditional methods to establish accurate physical models for management and monitoring. In addition, the equipment production process generates a large amount of reaction operation mechanism and real-time equipment status data. Based on production optimization, technology and other factors, in order to make better use of them, statistical analysis models and qualitative empirical knowledge analysis methods have been formed. However, these analysis methods all have the following problems: 1) The data characteristics of external factors are not combined with the model, resulting in low model generalization; 2) A large amount of historical data is required to establish an analysis model, which cannot meet the data volume requirements for model training before the equipment is produced and operated, resulting in low model accuracy; 3) The algorithm model used is fixed, and after the equipment is online, there is a situation where it is not applicable to actual production. Summary of the invention

[0005] In order to solve the problems of low generalization and low accuracy in existing modeling methods, the present invention proposes a modeling and evaluation method for the operation of complex equipment based on machine learning. The purpose is to propose a modeling and evaluation method for the operation of complex equipment in view of the shortcomings of existing analysis model technology, to integrate a variety of commonly used algorithms of machine learning in a universal and integrated manner, to form a basic algorithm model library for equipment operation status evaluation, and to design a management UI for the algorithm model library, evaluation strategy, etc. The algorithm library takes the production parameters, production mode and environmental parameters of complex equipment as input to provide data input for the model. The algorithm library votes on the analysis results based on the results of each algorithm, and finally analyzes and evaluates the operating status of the complex equipment. At the same time, the model of the algorithm model library is evaluated and optimized. Enterprises can use the algorithm library to accurately evaluate the operating status of complex equipment and effectively manage the equipment.

[0006] The present invention is achieved through the following technical solutions: a modeling and evaluation method for the operation of complex equipment is provided, and an object-oriented modeling method is adopted for the common processing mode of multiple subsystems of complex equipment in digital intelligent workshop production lines. The multi-modeling evaluation method needs to be mainly based on four parts: basic equipment resources, evaluation strategy, algorithm model library optimization, evaluation and decision-making.

[0007] After the above four parts are constructed, combined with the complex equipment production and operation process, the specific process of the method implementation can be summarized as follows: The basic equipment resources first read the equipment production parameters and external environment parameter information, and then through data cleaning, select the corresponding multiple models for each subsystem of the complex equipment from the model library resources, call the algorithm evaluation strategy parameters from the evaluation strategy, and evaluate the accuracy and generalization of the algorithm model of each subsystem. When the administrator who uses the development model needs to optimize the algorithm model settings or reset the evaluation strategy, a management UI interface for the algorithm model library and evaluation strategy is provided in the model. By setting parameters such as strategies on the management UI, the model can be quickly retrained, optimized, and self-adapted through multi-threaded execution, and the state evaluation algorithm model of each subsystem is optimally bound, thereby re-forming the real-time state evaluation service of the entire device.

[0008] Basic equipment resources: mainly composed of equipment process parameters and environmental process parameters. They mainly realize the management of equipment process parameters, equipment control program parameters, external environment parameters, management UI and strategy parameters. Equipment process parameters are based on material information, equipment process information and resource information in complex systems. They are composed of unit basic models, algorithm library evaluation strategies, algorithm optimization strategies, evaluation and analysis basic model units and external information interfaces to form multi-subsystems, multi-algorithm models, multi-evaluation strategies and other multi-level reconfigurable models. The internal part adopts modular encapsulation to form a new multi-level algorithm model library for complex equipment. Its input is the reconstructed feature data, which is directly and quickly mapped to the basic model resources, and the new model parameters are used as the input for production modeling for status evaluation.

[0009] Algorithm model library optimization: mainly includes optimization of new faults, working status and accuracy. The new fault and working status type strategies are optimized by re-training the original model or combining multiple algorithms. It consists of four modules: fault label classification, historical training data cleaning, algorithm model training interface and algorithm model accuracy verification. Each module calls the corresponding interface method internally and is associated with each basic unit control. It is encapsulated in the model library resources and collaborates with the basic model resources as the input for modeling analysis and evaluation. The accuracy optimization of the model is achieved by cross-validating the cleaned training data to improve generalization.

[0010] Algorithm evaluation strategy: Based on the original system, combined with basic equipment resource management and algorithm model optimization UI, the calling interface method is adopted to input information such as strategy parameters on the management UI into the multi-level algorithm model library of multiple subsystems, and configure the integrated optimization system framework. According to the optimization needs of different users, optimization parameters can be set, optimization strategy settings can be set, evaluation strategy settings can be set, the optimal solution can be viewed, and the original system detailed settings can be called to call the integrated UI interface. Through multi-threading technology, the optimal algorithm model and evaluation strategy for each subsystem can be quickly calculated in the final equipment model.

[0011] Status assessment and decision-making: After the modeling of complex equipment is completed, the process feasibility, system rationality, and algorithm model applicability are comprehensively considered, and the algorithm model library of complex equipment is comprehensively trained as a model for assessment and decision-making. After the training is completed, the accuracy of the analysis results of each algorithm model of each subsystem is verified by cross-validation method through the input data in the basic model resources. The evaluation results of the operating status of each subsystem, such as fault, idle, running, overload, etc., are provided to the integrated learning algorithm interface for equipment evaluation, which is used for real-time operating status evaluation and fault diagnosis of complex equipment. If it does not meet the actual situation, the target algorithm model needs to be optimized until the accuracy of the evaluation results meets the production requirements, and then automatic deployment can be carried out, and the evaluation report will be provided to the user.

[0012] In summary, compared with the prior art, the present invention has the following advantages and beneficial effects: The present invention is a multi-level multi-modeling integrated design method based on machine learning, and its key points and points to be protected are: First, the basic algorithm model of machine learning, algorithm optimization strategy, evaluation optimization strategy, etc. are combined to form a multi-level multi-model algorithm model library, with the basic model resource and control strategy split as the basic idea, introducing basic data resources, and selecting matching modeling objects from the model library and evaluation strategy to achieve multi-level one-to-one mapping association, and efficiently completing the generalization modification of the analysis model due to changes in the equipment. Second, the model is developed in a modular form, interacts with the integrated UI through an interface, is integrated through evaluation strategies, and combined with multi-model optimization algorithms, so that multi-level multi-model optimization training can be quickly set up and implemented on the integrated UI framework.

[0013] According to the design planning requirements of different complex equipment, one-to-one mapping can be performed based on the model library, evaluation strategy, etc. At the same time, the algorithm model has strong generalization and expansion capabilities, realizing the multi-level and multi-model analysis and evaluation required by enterprise users, and enhancing the complexity and high accuracy of modeling analysis. The modular form and integrated UI interface development avoids the errors and inefficiency problems in the original model library method, comprehensively solves the forward-looking problems before the equipment is put into production, effectively controls the analysis model adjustment problems caused by the later production, and greatly saves the model adjustment time after the equipment is put into production.

[0014] In summary, the present invention effectively overcomes the problems of low flexibility and accuracy in the existing modeling and analysis process, and performs reasonable, fast and efficient model resource configuration training on the models of each subsystem of the implementation equipment, and optimizes the operation through multi-threading technology, which has high significance and practical use value. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention.

[0016] Figure 1 It is a flow chart of the modeling integrated design method of the embodiment.

[0017] Figure 2 It is a schematic diagram of the basic data resource principle of the embodiment.

[0018] Figure 3 It is a schematic diagram of the principle of the algorithm model optimization strategy of the embodiment.

[0019] Figure 4 It is a schematic diagram of the principle of the device evaluation strategy of the embodiment.

[0020] Figure 5 Principle of the embodiment Figure 1.

[0021] Figure 6 Principle of the embodiment Figure 2 . DETAILED DESCRIPTION

[0022] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0023] In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is apparent to one of ordinary skill in the art that these specific details are not necessarily employed to practice the present invention. In other instances, well-known structures, circuits, materials, or methods are not specifically described in order to avoid obscuring the present invention.

[0024] Throughout the specification, references to "one embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment of the present invention. Therefore, the phrases "one embodiment," "an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily all refer to the same embodiment or example. In addition, particular features, structures, or characteristics may be combined in one or more embodiments or examples in any suitable combination and / or sub-combination. In addition, it will be appreciated by those of ordinary skill in the art that the figures provided herein are for illustrative purposes and that the figures are not necessarily drawn to scale. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0025] In the description of the present invention, it should be understood that the terms "front", "rear", "left", "right", "up", "down", "vertical", "horizontal", "high", "low", "inside", "outside", etc., indicating directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the scope of protection of the present invention.

[0026] The embodiment provides a modeling and evaluation method for the operation of complex equipment. Aiming at the common processing mode of complex equipment with multiple subsystems in digital intelligent workshop production lines, an object-oriented modeling method is adopted. The multi-modeling evaluation method needs to be mainly based on four parts: basic equipment resources, evaluation strategy, algorithm model library optimization, evaluation and decision-making, such as Figure 1 shown.

[0027] After the above four parts are constructed, combined with the complex equipment production and operation process, the specific process of the method implementation can be summarized as follows: The basic equipment resources first read the equipment production parameters and external environment parameter information, and then through data cleaning, select the corresponding multiple models for each subsystem of the complex equipment from the model library resources, call the algorithm evaluation strategy parameters from the evaluation strategy, and evaluate the accuracy and generalization of the algorithm model of each subsystem. When the administrator who uses the development model needs to optimize the algorithm model settings or reset the evaluation strategy, a management UI interface for the algorithm model library and evaluation strategy is provided in the model. By setting parameters such as strategies on the management UI, the model can be quickly retrained, optimized, and self-adapted through multi-threaded execution, and the state evaluation algorithm model of each subsystem is optimally bound, thereby re-forming the real-time state evaluation service of the entire device.

[0028] This embodiment proposes a complex equipment real-time status recognition and evaluation modeling method based on machine learning. Based on the real-time production parameters of field equipment, environmental temperature and humidity factors, workshop energy and other data, starting from the various subsystem modules of complex equipment, multiple production modes and key control points of energy control, different subsystems of complex equipment are modeled separately, and evaluated according to different production modes to achieve equipment fault information prediction and operation status recognition, so as to minimize equipment failures, ensure the safe production of equipment and improve intelligent operation and maintenance management capabilities, such as Figure 5-Figure 6 shown.

[0029] Basic data resources: Basic data resources are an important source of data for the algorithm model library of complex equipment, providing data support for the establishment of the entire analysis model. They mainly include four parts: basic data tables, evaluation strategies, algorithm model optimization strategies, and external data interface quantities. The schematic diagram of the basic data resource principle is shown below: Figure 2 As shown. By configuring ODBC to connect to the database, the program remotely reads various types of information in the database, reads the operating parameters of all equipment subsystem algorithm models, reads the historical production parameters and external environment parameters in the database, and combines multiple machine learning algorithms for model training; classifies the various working states of the equipment and associates them with the multi-level and multi-models in the model library resources; edits the equipment evaluation strategy configuration, configures the initialization, reset and simulation end API methods, and can perform real-time data evaluation through the optimal evaluation model of each subsystem in the algorithm model library before providing real-time status evaluation services. Initialization calls are used to import and bind basic model resources; after the algorithm analysis model training is completed, the corresponding statistical results can be collected and viewed through the statistical table in the data table.

[0030] Algorithm model optimization strategy: mainly includes optimization for newly added working state modes, newly added equipment fault types and algorithm model accuracy. Among them, the newly added working state mode of the equipment can be re-annotated by combining the original historical data and the new working state mode, and then the algorithm models of the subsystems of the equipment can be retrained, or the working state types that are not recognized by the model can be reserved at the beginning of the training mode, and the data under the new working state of the equipment can be added to the training data to optimize the existing model. For example, a new working mode is added on the equipment model library management UI interface, and the data under the new mode is imported to re-acquire the features of the mode for training. Secondly, for the newly added equipment fault type, this new fault type can generally directly locate which subsystem has failed. The fault data and historical data of the subsystem can be imported for annotation and optimization training to improve the accuracy of the subsystem in judging the fault type. Finally, the algorithm model accuracy optimization can be set through the equipment model library management UI to accurately optimize the configuration information, in which the equipment status recognition program can be set to optimize the training, timed training, and quantitative training when it is idle, so that the model can maintain a high accuracy in identifying the equipment status. In addition, the accuracy of the algorithm model of a subsystem can be improved by the joint verification of the model to improve the accuracy of the subsystem's operating status and fault diagnosis. After the algorithm model library management UI is set up, the strategy is loaded into the background of the device status recognition program and the multi-threading technology is used to automatically and quickly complete each model optimization task. The schematic diagram of the algorithm model optimization strategy principle is as follows Figure 3 shown.

[0031] Algorithm evaluation strategy: On the device library model management UI interface, combined with the basic data resource management interface, the webapi interface method is used to write a complex device multi-model modeling evaluation program based on machine learning in C# language, which is encapsulated into a type object. The WPF is used to develop the framework of the configuration model library configuration management system. The framework is mainly composed of two parts: the UI module of the management interface and the algorithm library model optimization module. Figure 4 As shown. The model library management UI module includes the configuration process of initializing the equipment status recognition and fault diagnosis algorithm library. After initialization, the analysis model is loaded and run through the model accuracy setting. If the accuracy requirements in the evaluation strategy are met, the equipment status analysis results of the optimal model can be directly output; at the same time, the evaluation can be set as an integrated algorithm comprehensive evaluation strategy. At this time, the model library of the equipment integrates the evaluation results and accuracy information of each subsystem to conduct a comprehensive evaluation of the equipment status and faults, forming an API interface for the evaluation results, which is provided to the evaluation and decision-making modules.

[0032] The optimization and evaluation strategies can also be encapsulated with the model library objects and can be embedded in the entity model as user customization. However, when faced with different complex equipment, the workload of modification and adjustment due to multiple models increases and it does not have good flexibility.

[0033] Status evaluation and decision-making: Start running the real-time equipment status recognition program through model pre-training. Users use the management control interface developed by WPF to build an evaluation and decision-making integrated UI control framework, use controls and model optimization interfaces to bind each other, and write model optimization programs in the interface methods. When viewing controls, indirectly call the data source in the statistical tool for graphical output display, analyze the rationality and manufacturability of statistical data such as equipment status recognition, fault diagnosis, and safe production under the current model parameter settings, and provide a status recognition evaluation report for equipment model analysis. Users can call the internal interface to view detailed attribute information in the report, and can compare the identification status information, fault information, etc. of the equipment after the model program is run with the optimized equipment safe production. The evaluation results will serve as an important reference for the new multi-level and multi-model. If the actual equipment status information is not met, select the target model for re-optimization training based on the decision evaluation.

[0034] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A modeling and evaluation method for complex equipment operation, characterized in that: Construct an equipment operation status assessment model that includes basic equipment resources, equipment algorithm model library, algorithm evaluation, algorithm model library optimization, status assessment and decision-making; Based on this evaluation model, the basic equipment resources first read the equipment production parameters and external environment parameter information, and then clean the data; Select multiple corresponding algorithm models for each subsystem of the complex equipment from the equipment algorithm model library; Call algorithm evaluation parameters from the algorithm evaluation strategy to evaluate the accuracy and generalization of the algorithm model of each subsystem; When the algorithm model library needs to be optimized or the algorithm evaluation parameters need to be reset, the device algorithm model library provides a management UI for algorithm model library optimization and algorithm evaluation parameters. The algorithm model retraining optimization and adaptation are implemented through the management UI, and the state evaluation algorithm model of each subsystem is optimally bound. This then constitutes a real-time status assessment service for the entire device.

2. The modeling and evaluation method for complex equipment operation according to claim 1, characterized in that: In basic equipment resources, equipment production parameters include equipment process parameters and equipment control program parameters, while equipment process parameters include material information, equipment process information, and resource information in complex systems; The characteristic data after reconstruction of the basic equipment resources is input and mapped to the equipment algorithm model library; if the algorithm model library is optimized, the state evaluation is performed according to the optimized algorithm model as input.

3. The modeling and evaluation method for complex equipment operation according to claim 2, characterized in that: The basic data resources of the evaluation model include basic equipment resources, algorithm evaluation strategies, and algorithm model library optimization strategies. They are connected to the equipment algorithm model library through a data interface, and connected to the basic data resource database by configuring ODBC; Remotely read the operating parameters of the algorithm models of each subsystem of the equipment, remotely read the historical production parameters and external environment parameters, and combine multiple machine learning algorithms to train the algorithm models; classify the various working modes of the equipment, and perform multi-level and multi-model association with the algorithm models in the equipment algorithm model library; Before providing the real-time status assessment service for equipment, edit the algorithm assessment strategy configuration, configure the initialization, reset and simulation end API methods, perform real-time data assessment based on the optimal assessment model for each subsystem in the algorithm model library, and initialize the call to import and bind the basic algorithm model resources; After the algorithm model training is completed, the corresponding statistical results are collected and viewed through the statistical table in the basic equipment resources.

4. The modeling and evaluation method for complex equipment operation according to claim 3, characterized in that: The algorithm model library optimization includes optimization of new fault types, new working modes and accuracy; The optimization strategy for new fault types and new working modes is to optimize by retraining the original algorithm model or combining multiple algorithms. The fault label classification, historical training data cleaning, algorithm model training interface and algorithm model accuracy verification modules are encapsulated in the algorithm model library resources by internally calling the corresponding interface methods and associating them with the basic unit controls, and then coordinated with the basic model resources as the input for modeling analysis and evaluation. The accuracy of the algorithm model is optimized by cross-validating the cleaned training data to improve generalization.

5. The modeling and evaluation method for complex equipment operation according to claim 4, characterized in that: According to the algorithm model library optimization strategy for the newly added working mode, after re-annotating the data by combining the original historical data and the new working mode, the algorithm models of each equipment subsystem are retrained; Alternatively, when the algorithm model training mode starts, the working mode type that is not recognized by the model is reserved, and the data of the new working mode of the device is added to the training data to optimize the existing algorithm model; That is, add a new working mode on the device algorithm model library management UI interface, import the data under the new working mode, and re-acquire the features of the working mode for training; For new fault types of equipment, directly locate the subsystem where the new fault type occurs, import the fault data and historical data of the subsystem, and perform optimization training after annotation; To optimize the accuracy of the algorithm model, accurately optimize the configuration information through the device model library management UI, set the device status recognition program to optimize training, timed training, and quantitative training when it is idle, and jointly verify the accuracy of the subsystem's algorithm model through the algorithm model.

6. The modeling and evaluation method for complex equipment operation according to claim 5, characterized in that: The algorithm evaluation strategy is based on the original algorithm model, combined with basic equipment resources and algorithm model optimization UI, and uses the method of calling interfaces to input information such as strategy parameters on the management UI into the multi-subsystem multi-level algorithm model library to configure an integrated optimization evaluation model framework; According to the optimization needs of different users, optimization parameters are set, optimization strategies are set, evaluation strategies are set, the optimal solution is viewed, and the original evaluation model is set in detail to call the integrated UI interface. The multi-threaded technology is used to calculate the optimal algorithm model and evaluation strategy for each subsystem in the final evaluation model.

7. The modeling and evaluation method for complex equipment operation according to claim 6, characterized in that: The algorithm evaluation strategy is based on the device library model management UI, combined with the basic data resource management interface, and adopts the webapi interface method to write a complex device multi-algorithm model modeling and evaluation program based on machine learning in C# language; The framework of the configuration algorithm model library configuration management system is developed using WPF. The framework consists of the UI module of the management interface and the algorithm model library optimization module. The management UI module includes the initialization configuration process of the device status recognition and fault diagnosis algorithm library, etc. After initialization, the algorithm model is loaded and run after the accuracy of the algorithm model is set; If the accuracy requirements in the evaluation strategy are met, the equipment status evaluation results of the optimal algorithm model are directly output; at the same time, the evaluation method is set to an integrated multi-level algorithm comprehensive evaluation strategy. At this time, the algorithm model library of the equipment integrates the evaluation results and accuracy information of each subsystem, conducts a comprehensive evaluation of the equipment status and faults, and has an API interface for the evaluation results, which is provided to the evaluation and decision-making modules for use.

8. The modeling and evaluation method for complex equipment operation according to claim 7, characterized in that: State assessment and decision-making comprehensively considers process feasibility, system rationality, and algorithm model applicability, and conducts comprehensive training of the algorithm model library of complex equipment as a model for state assessment and decision-making; After the training is completed, the accuracy of the analysis results of each algorithm model of each subsystem is verified by cross-validation method through the input data of the basic model resources; Combine the evaluation results of each subsystem's operating status, including fault, idle, running, and overload, and provide them to the integrated algorithm model interface for equipment evaluation, which is used for real-time operating status evaluation and fault diagnosis of complex equipment; If the actual situation is not met, the target algorithm model needs to be optimized until the accuracy of the evaluation results meets the requirements of production, then automatic deployment can be carried out and the evaluation report can be provided to the user.

9. The modeling and evaluation method for complex equipment operation according to claim 8, characterized in that: State assessment and decision-making starts running the real-time equipment state recognition program of the assessment model through algorithm model pre-training, uses the management control interface developed by WPF to build the assessment and decision-making integrated UI control framework, and uses the control and model optimization interface to bind each other; And write the algorithm model optimization program in the interface method, and when providing viewing controls, indirectly call the data source in the statistical tool for graphical output display; Analyze the rationality and manufacturability of equipment status identification, fault diagnosis, and safety production statistics under the algorithm model parameter settings, and provide an equipment status identification evaluation report; By calling the internal interface to view the attribute information of the report, that is, by evaluating the identification status information and fault information of the equipment after the model program is run, the safety production of the equipment after optimization is compared, and the evaluation results are used as a reference for the new algorithm model; If the actual status information of the device is not satisfied, the target algorithm model is selected according to the decision evaluation for re-optimization training.

10. A modeling and evaluation system for complex equipment operation, characterized in that: It includes basic equipment resource module, equipment algorithm model library module, evaluation strategy module, algorithm model library optimization module, and state evaluation and decision-making module; Basic equipment resource module, used to read equipment production parameters and external environment parameter information, and perform data cleaning; The equipment algorithm model library module is used to select multiple corresponding algorithm models for each subsystem of complex equipment; The evaluation strategy module is used to call the algorithm evaluation parameters and evaluate the accuracy and generalization of the algorithm model of each subsystem; The algorithm model library optimization module is used to provide algorithm model library optimization and algorithm evaluation parameter management UI through the device algorithm model library when it is necessary to optimize the algorithm model library or reset the algorithm evaluation parameters, realize algorithm model retraining optimization and self-adaptation, and optimize the binding of the state evaluation algorithm model of each subsystem; The status assessment and decision-making module is used to provide assessment services for the real-time status of the entire equipment.

Citation Information

Patent Citations

  • Distribution network device status evaluating system

    CN102509178A

  • Complex device performance evaluating and predicting method of multi-source no-label data machine learning

    CN107545112A

  • Operation and maintenance method, device and system

    CN108173671A

  • Enterprise production safety risk assessment and management system

    CN109508848A

  • Simulation model parameterization integration system and method based on multi-level reconstruction

    CN111144025A