Complex system fault diagnosis method based on hybrid simulation model dynamic evaluation
By constructing a hybrid simulation model and an adaptive threshold optimization method, the problems of insufficient accuracy and reliability in fault diagnosis of complex systems are solved, enabling timely and accurate fault diagnosis of complex systems and improving the reliability and operating efficiency of the system.
Patent Information
- Application Number
- CN202511427349.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-30
- Publication Date
- 2026-01-16
AI Technical Summary
Existing technologies for fault diagnosis of complex systems suffer from the "black box" problem of data-driven methods and the problem of experience solidification in expert systems, making it difficult to adapt to changing operating conditions and multi-field fault coupling scenarios, resulting in insufficient diagnostic accuracy and reliability.
A hybrid simulation model combining mechanistic and data models is constructed, and OPC UA is used to achieve real-time data exchange. An anomaly monitoring module is constructed by combining a long short-term memory neural network. Residual evaluation and fault diagnosis are performed through an adaptive threshold optimization method, and the threshold is dynamically adjusted to adapt to changes in operating conditions.
It enables timely and accurate diagnosis of faults in complex systems, improves system reliability and operating efficiency, and meets the fault diagnosis needs of complex equipment in scenarios with varying operating conditions and multiple coupled faults.
Smart Images
Figure CN121349039A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of health management technology for complex systems, and more specifically, to a method for fault diagnosis of complex systems based on dynamic evaluation using a hybrid simulation model. Background Technology
[0002] Against the backdrop of continuous global industrialization, various industrial equipment and mechanical systems are becoming increasingly complex. Key equipment such as engines and pumps have become core assets in enterprise production and operation. The stable operation of these complex systems directly affects production efficiency, product quality, and enterprise economic benefits. Therefore, effective health management of complex systems and timely and accurate diagnosis of potential faults are of paramount practical importance for ensuring reliable and stable system operation, and are gradually becoming one of the key requirements in the industrial sector.
[0003] Currently, there are two main technological approaches in the field of complex system health management. The first is a data-driven approach, which relies on large amounts of historical data to discover latent patterns and achieve fault diagnosis, such as the application of machine learning algorithms like artificial neural networks and support vector machines. The second is a system based on expert knowledge, which relies on the experience and knowledge of domain experts to build a rule base for fault reasoning and judgment. However, data-driven approaches often suffer from a "black box" problem, meaning the model's decision-making process is difficult to interpret; while expert systems face numerous shortcomings, including the rigidity of experience, difficulty adapting to new fault modes, and instability caused by differences in expert knowledge. Both approaches struggle to effectively address the complex and ever-changing actual operating conditions of complex equipment and multi-field fault coupling scenarios, significantly reducing the accuracy and reliability of fault diagnosis and failing to meet the high requirements of modern industrial production for complex system fault diagnosis. Summary of the Invention
[0004] In response to at least one defect or improvement requirement in the prior art, this application provides a method for diagnosing faults in complex systems based on dynamic evaluation using a hybrid simulation model, which can solve at least one of the problems existing in the background art.
[0005] To achieve the above objectives, according to the first aspect of this application, a method for fault diagnosis of complex systems based on dynamic evaluation using a hybrid simulation model is provided, characterized by comprising the following steps: A hybrid simulation model combining a mechanistic model and a data model is constructed. The mechanistic model is a parameterized model of an electromechanical-hydraulic coupling system built using modeling and simulation languages and exported in a specified format. The data model is an anomaly monitoring module built using a long short-term memory neural network method and encapsulated in the specified format using tools. Real-time data exchange between the mechanistic model and the data model is achieved based on OPC UA, thus constructing the hybrid simulation model. The input of the real system is connected to the input terminal of the hybrid simulation model to simulate the operation of the real system. The outputs of the real system and the hybrid simulation model are respectively subtracted to obtain residual signals. The time-domain feature vector, frequency-domain feature vector and spatial correlation feature vector are extracted from them. The feature vectors are weighted and fused to obtain normalized feature weights. An adaptive threshold optimization method is used for residual assessment and fault diagnosis. An initial threshold is set, and the threshold is dynamically adjusted by a sensitivity coefficient based on the deviation between the current operating condition parameter vector and the rated operating condition parameter vector. The residual characteristics are then analyzed to obtain the fault diagnosis results.
[0006] Furthermore, in the aforementioned method for fault diagnosis of complex systems based on dynamic evaluation using a hybrid simulation model, the mechanism model construction specifically includes: Parametric modeling of each component of the electromechanical-hydraulic coupling system is performed based on modeling and simulation languages; Export the constructed mechanism model in the specified format.
[0007] Furthermore, in the aforementioned method for fault diagnosis of complex systems based on dynamic evaluation using a hybrid simulation model, the data model construction specifically includes: A long short-term memory neural network is used to learn and train on historical system data to construct an anomaly detection module; Use tools to encapsulate the data model into a specified format.
[0008] Furthermore, the aforementioned method for fault diagnosis of complex systems based on dynamic evaluation using a hybrid simulation model specifically includes the following steps when extracting residual features: The residual signal is calculated by real-time monitoring and comparison of the input and output data of the real system and the hybrid simulation model. The residual signal is feature extracted from three dimensions: time domain, frequency domain, and spatial correlation. The time domain feature vector, frequency domain feature vector, and spatial correlation feature vector are extracted respectively.
[0009] Furthermore, in the aforementioned method for fault diagnosis of complex systems based on dynamic evaluation using a hybrid simulation model, the weighted fusion of feature vectors specifically includes: Each feature vector is standardized to eliminate dimensional differences and bring all features to the same scale. The weights of each feature vector are determined based on its inverse variance. The smaller the measurement error, the greater the feature weight, thus obtaining the normalized feature weights.
[0010] Furthermore, in the aforementioned complex system fault diagnosis method based on dynamic evaluation using a hybrid simulation model, the adaptive threshold optimization method specifically includes: The initial threshold is set to
[0011] in The historical residual mean This represents the standard deviation of the residuals.
[0012] Furthermore, in the aforementioned method for fault diagnosis of complex systems based on dynamic evaluation using a hybrid simulation model, the dynamic adjustment of the threshold specifically includes: Based on the deviation between the current operating condition parameter vector and the rated operating condition parameter vector
[0013] in, This is the parameter vector for the current operating condition. This is the vector of rated operating parameters. This is the sensitivity coefficient; Using sensitivity coefficient The influence of changes in control conditions on the threshold is controlled. When the control conditions deviate from the rated control conditions, the threshold ratio is relaxed, and the residual threshold is dynamically optimized adaptively.
[0014] According to a second aspect of this application, a fault diagnosis device for complex systems based on dynamic evaluation using a hybrid simulation model is also provided, characterized in that it comprises: The model building module is used to build a hybrid simulation model of mechanism model and data model. The mechanism model is constructed using modeling and simulation languages to build a parameterized model of electromechanical-hydraulic coupling system and exported to a specified format. The data model is constructed using a long short-term memory neural network method to build an anomaly monitoring module and encapsulated into the specified format through tools. Real-time data exchange between mechanism model and data model is realized based on OPCUA to build a hybrid simulation model. The feature extraction module is used to connect the input of the real system to the input terminal of the hybrid simulation model to simulate the operation of the real system, calculate the difference between the output of the real system and the output of the hybrid simulation model to obtain the residual signal, extract the time domain feature vector, frequency domain feature vector and spatial correlation feature vector, and perform weighted fusion on the feature vector to obtain normalized feature weighting; The fault diagnosis module is used to perform residual assessment fault diagnosis using an adaptive threshold optimization method. It sets an initial threshold and dynamically adjusts the threshold based on the deviation between the current operating condition parameter vector and the rated operating condition parameter vector through a sensitivity coefficient. It then analyzes the residual characteristics to obtain fault diagnosis results.
[0015] According to a third aspect of this application, a fault diagnosis device for complex systems based on dynamic evaluation using a hybrid simulation model is also provided, comprising at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program that, when executed by the processing unit, causes the processing unit to perform the steps of any of the methods described above.
[0016] According to a fourth aspect of this application, a storage medium is also provided, which stores a computer program executable by a complex system fault diagnosis device based on a hybrid simulation model for dynamic evaluation. When the computer program is run on the complex system fault diagnosis device based on a hybrid simulation model for dynamic evaluation, the complex system fault diagnosis device based on a hybrid simulation model for dynamic evaluation performs the steps of any of the methods described above.
[0017] In summary, compared with the prior art, the above-described technical solutions conceived in this application can achieve the following beneficial effects: This application constructs a hybrid simulation model combining a mechanistic model and a data model, and combines it with OPC UA to achieve real-time data exchange. The input of the real system is connected to the input end of the hybrid simulation model to simulate the operation of the real system and obtain the residual signal by subtraction. An adaptive threshold optimization method is used for residual evaluation and fault diagnosis. The threshold is dynamically adjusted according to the deviation of the operating condition parameter vector. This can accurately simulate the actual operating state of complex systems and adapt to the fault diagnosis needs of complex equipment in scenarios with changing operating conditions and multi-field fault coupling. This enables timely and accurate diagnosis of faults in complex systems and improves the reliability and operating efficiency of the system. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 This is a flowchart illustrating a method for diagnosing faults in complex systems based on dynamic evaluation using a hybrid simulation model, as provided in an embodiment of this application. Detailed Implementation
[0020] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. Furthermore, the technical features involved in the various embodiments described below can be combined with each other as long as they do not conflict with each other.
[0021] The terms "first," "second," "third," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.
[0022] Figure 1 A flowchart illustrating a complex system fault diagnosis method based on dynamic evaluation using a hybrid simulation model, as provided in this application embodiment, is shown below. Figure 1 As shown in the embodiments of this application, a method for fault diagnosis of complex systems based on dynamic evaluation using a hybrid simulation model includes the following steps: A hybrid simulation model combining a mechanistic model and a data model is constructed. The mechanistic model is a parameterized model of an electromechanical-hydraulic coupling system built using modeling and simulation languages and exported in a specified format. The data model is an anomaly monitoring module built using a long short-term memory neural network method and encapsulated in the specified format using tools. Real-time data exchange between the mechanistic model and the data model is achieved based on OPC UA, thus constructing the hybrid simulation model. The input of the real system is connected to the input terminal of the hybrid simulation model to simulate the operation of the real system. The outputs of the real system and the hybrid simulation model are respectively subtracted to obtain residual signals. The time-domain feature vector, frequency-domain feature vector and spatial correlation feature vector are extracted from them. The feature vectors are weighted and fused to obtain normalized feature weights. An adaptive threshold optimization method is used for residual assessment and fault diagnosis. An initial threshold is set, and the threshold is dynamically adjusted by a sensitivity coefficient based on the deviation between the current operating condition parameter vector and the rated operating condition parameter vector. The residual characteristics are then analyzed to obtain the fault diagnosis results.
[0023] Specifically, in one embodiment provided in this application, the modeling and simulation language can be Modelica to parametrically model the various components of the electromechanical-hydraulic coupling system. Modelica is an open, object-oriented, equation-based computer language mainly used for modeling complex physical systems across multiple domains. Its core function is to describe the dynamic behavior of multi-domain physical systems through mathematical equations, supporting multi-disciplinary collaborative simulation. For example, in an engine system, parametric models are established for the engine's mechanical structure, electrical system, and hydraulic system respectively. By describing the physical characteristics of each component and their interconnections using Modelica, a parametric model of the electromechanical-hydraulic coupling system that accurately reflects the system's physical characteristics is constructed. The constructed mechanism model is then exported to a specified format. In one embodiment provided in this application, the specified format is FMU (Functional Mock-up Unit) format.
[0024] Historical system operation data is collected, including normal operation data and fault operation data. An LSTM (Long Short-Term Memory) neural network method is used to learn and train on this data to build an anomaly detection module. This module can determine whether there are anomalies in the system based on the characteristics of the input data. The data model is encapsulated into FMU format using specific tools (such as Python FMU).
[0025] Real-time data exchange between the mechanistic model and the data model is achieved based on OPC UA (Open Platform Communications Unified Architecture). OPC UA is a communication protocol and architecture used to realize data exchange and communication between devices in industrial automation systems. It provides a standardized, secure, and scalable mechanism that enables seamless data exchange and communication between different devices and systems. OPC UA provides a secure and reliable data transmission mechanism, ensuring real-time data interaction between the two models. By constructing an FMI (Functional Mock-up Interface) co-simulation framework, the mechanistic model and the data model are integrated to form a hybrid simulation model.
[0026] The input signals of the real system (such as engine speed and load) are connected to the input of the constructed hybrid simulation model to simulate the operation of the real system. During operation, the output signals of the real system and the output signals of the hybrid simulation model are compared, and the difference is obtained to obtain the residual signal. For example, in pump operation monitoring, the actual flow output of the pump and the flow output of the hybrid simulation model are acquired in real time, and the residual signal between the two is calculated. This signal reflects the difference between the actual operating state of the system and the simulation model.
[0027] Feature extraction is performed on the residual signal, extracting time-domain feature vectors, frequency-domain feature vectors, and spatial correlation feature vectors. In time-domain feature extraction, the mean, variance, peak value, and other characteristic values of the residual signal are calculated. In frequency-domain feature extraction, a Fast Fourier Transform (FFT) is performed on the residual signal to obtain its spectral characteristics, extracting characteristic values such as the dominant frequency and band energy. In spatial correlation feature extraction, the correlation between the residual signal and different sensor locations is analyzed, and spatial correlation feature vectors are extracted.
[0028] The extracted time-domain feature vectors, frequency-domain feature vectors, and spatially correlated feature vectors are weighted and fused. First, each feature vector is standardized to eliminate dimensional differences, ensuring all features are on the same scale. Then, the weights are determined based on the inverse variance of each feature; features with smaller measurement errors have larger weights. The normalized feature weighted sum F is obtained through weighted fusion calculation. fusion This enables the effective integration and weighted expression of different features, highlighting the contribution of important features to fault diagnosis.
[0029] The historical residual mean and standard deviation are calculated based on historical residual data. The historical residual mean reflects the baseline value of the system under normal conditions, while the residual standard deviation characterizes the range of data fluctuation. Initial threshold settings are set based on the historical residual mean and standard deviation, providing a basic reference for subsequent dynamic adjustments.
[0030] The system acquires the current operating condition parameter vector (such as speed and load) and the rated operating condition parameter vector in real time. It calculates the degree of deviation between the two and dynamically adjusts the threshold based on this deviation using a sensitivity coefficient. When the operating condition deviates from the rated operating condition, the threshold ratio is appropriately relaxed to avoid false alarms caused by changes in operating conditions. For example, during system startup or speed change phases, dynamically adjusting the threshold allows the system to adapt to changes in operating conditions.
[0031] Weighted sum of normalized features and F fusion Compare and analyze the results with the dynamically adjusted threshold. If F fusionIf the threshold is exceeded, the system is judged to have malfunctioned, and the type and location of the malfunction are determined based on the analysis results of the feature vector. For example, in the fault diagnosis of industrial equipment, this method can promptly and accurately diagnose equipment faults, providing maintenance personnel with accurate fault information so that timely measures can be taken for repair and maintenance, thereby improving the reliability and operating efficiency of the system.
[0032] This application constructs a hybrid simulation model combining a mechanistic model and a data model, and combines it with OPC UA to achieve real-time data exchange. The input of the real system is connected to the input end of the hybrid simulation model to simulate the operation of the real system and obtain the residual signal by subtraction. An adaptive threshold optimization method is used for residual evaluation and fault diagnosis. The threshold is dynamically adjusted according to the deviation of the operating condition parameter vector. This can accurately simulate the actual operating state of complex systems and adapt to the fault diagnosis needs of complex equipment in scenarios with changing operating conditions and multi-field fault coupling. This enables timely and accurate diagnosis of faults in complex systems and improves the reliability and operating efficiency of the system.
[0033] Optionally, the fault diagnosis method for complex systems based on dynamic evaluation using a hybrid simulation model provided in this application embodiment specifically includes the following steps during the construction of the mechanism model: Parametric modeling of each component of the electromechanical-hydraulic coupling system is performed based on modeling and simulation languages; Export the constructed mechanism model in the specified format.
[0034] Specifically, parametric modeling of various components of the electromechanical-hydraulic coupling system is performed using the Modelica language. Taking an industrial pump system as an example, the mechanical components of the pump, such as the impeller and pump shaft, are first parametrically described according to their geometric dimensions, material properties, and motion characteristics. Using components from Modelica's mechanical library, a rotating model of the impeller is constructed, setting parameters such as the number of blades, their angles, and moment of inertia. For the pump shaft, parameters such as its length, diameter, and torsional stiffness are defined to simulate its rotation and deformation under torque.
[0035] Next, the hydraulic components of the pump are modeled. Using Modelica's hydraulic library, models of the pump's suction chamber, discharge chamber, and flow channels are constructed. Fluid parameters such as density and viscosity are set, and based on the pump's working principle, flow equations for the fluid between various components are defined, including the continuity equation and Bernoulli's equation, to simulate pressure changes and flow distribution within the pump. Simultaneously, the pump's electrical components, such as the motor, are parametrically modeled. Using Modelica's electrical library, parameters such as motor voltage, current, resistance, and inductance are set, and circuit and electromagnetic coupling models of the motor are constructed to simulate the motor's start-up, operation, and speed regulation processes.
[0036] After completing the parametric modeling of each component of the electromechanical-hydraulic coupling system, the entire mechanistic model is integrated and verified. Using the simulation capabilities of the Modelica development environment, a preliminary simulation is run on the constructed mechanistic model to check for syntax errors, inconsistencies in equations, and other issues, ensuring the model functions correctly and outputs the expected results. The verified mechanistic model is then exported to FMU format using the export function provided by Modelica. During the export process, appropriate FMU export options are selected according to actual needs, and the exported FMU file is packaged and organized for subsequent integration and co-simulation with the data model, enabling research and application in fault diagnosis of complex systems.
[0037] Optionally, the complex system fault diagnosis method based on dynamic evaluation using a hybrid simulation model provided in this application embodiment specifically includes the following steps during data model construction: A long short-term memory neural network is used to learn and train on historical system data to construct an anomaly detection module; Use tools to encapsulate the data model into a specified format.
[0038] Specifically, an LSTM neural network is used to learn and train on historical system data to construct an anomaly monitoring module. Taking the fault diagnosis of a certain type of industrial engine as an example, a large amount of historical data on the engine under normal and fault conditions is collected. The data includes multiple variables such as speed, temperature, pressure, and vibration signals. The collected historical data is preprocessed to construct an LSTM neural network model. Through training, the LSTM neural network learns the data characteristics and patterns of the system under normal and fault conditions, thereby constructing a module capable of monitoring system anomalies.
[0039] After training, the LSTM data model is packaged into FMU format using the Python FMU tool. Utilizing relevant functions in the Python FMU library, the trained LSTM model and its associated weight parameters, input / output definitions, and other information are packaged and encapsulated to obtain a pre-packaged data model FMU file. This file contains an anomaly detection module based on the LSTM neural network, which can be used for subsequent joint simulation research on fault diagnosis of complex systems.
[0040] Optionally, the fault diagnosis method for complex systems based on dynamic evaluation using a hybrid simulation model provided in this application embodiment specifically includes the following steps when extracting residual features: The residual signal is calculated by real-time monitoring and comparison of the input and output data of the real system and the hybrid simulation model. The residual signal is feature extracted from three dimensions: time domain, frequency domain, and spatial correlation. The time domain feature vector, frequency domain feature vector, and spatial correlation feature vector are extracted respectively.
[0041] Specifically, in the process of fault diagnosis of complex systems, it is first necessary to monitor and compare the input and output data of the real system and the hybrid simulation model in real time in order to calculate the residual signal. Taking a motor-driven hydraulic pump system as an example, in actual operation, the input and output data such as the motor speed, current, hydraulic pump outlet pressure and flow rate are collected in real time by sensors, while the corresponding simulated output data is obtained from the hybrid simulation model.
[0042] Specifically, the input signals of the real system (such as the motor's power supply voltage and load commands) are connected to the input of the hybrid simulation model to simulate the system's operating state under actual working conditions. During the simulation, the output results of the simulation model are recorded in real time, including the simulated motor's speed, the simulated hydraulic pump's outlet pressure, and flow rate. Simultaneously, corresponding output data is collected from the real system. The output data of the real system and the output data of the simulation model are then subtracted to obtain residual signals. For example, for motor speed, the residual signal is the difference between the real motor speed and the simulated motor speed; for hydraulic pump outlet pressure, the residual signal is the difference between the real outlet pressure and the simulated outlet pressure, and so on.
[0043] After acquiring the residual signal, features are extracted from the residual signal from three dimensions: time domain, frequency domain, and spatial correlation, in order to obtain comprehensive feature information.
[0044] In time-domain feature extraction, the mean, variance, peak value, and kurtosis of the residual signal are calculated. These features reflect the statistical characteristics of the signal over time, helping to identify the characteristics of the system under different operating conditions. For example, the mean reflects the average magnitude of the signal, the variance reflects the degree of signal fluctuation, the peak value represents the maximum deviation of the signal, and the kurtosis reflects the sharpness of the signal distribution. By calculating these features, a time-domain feature vector can be obtained.
[0045] A Fast Fourier Transform (FFT) is performed on the residual signal to transform it into the frequency domain, obtaining the signal's spectral characteristics. In the frequency domain, feature values such as the dominant frequency, band energy, and spectral entropy are extracted. The dominant frequency represents the most energetic frequency component in the signal, the band energy reflects the energy distribution of the signal within a specific frequency band, and the spectral entropy describes the complexity of the signal's spectrum. By analyzing these frequency domain feature values, a frequency domain eigenvector can be obtained.
[0046] Analyzing the correlation between residual signals at different sensor locations allows for the extraction of spatial correlation feature vectors. In complex systems, multiple sensors are typically distributed across different locations to monitor various parts of the system. By calculating the correlation coefficient, mutual information, and other metrics between the residual signals of different sensors, the degree of spatial correlation between the signals can be understood. For example, in a motor-driven hydraulic pump system, analyzing the correlation between the motor current residual signal and the hydraulic pump pressure residual signal allows for the extraction of spatial correlation feature vectors.
[0047] Through the above steps, the residual signal is comprehensively extracted from three dimensions: time domain, frequency domain, and spatial correlation, providing rich feature information for subsequent feature fusion and fault diagnosis.
[0048] Optionally, the fault diagnosis method for complex systems based on dynamic evaluation using a hybrid simulation model provided in this application embodiment specifically includes the following during the feature vector weighted fusion process: Each feature vector is standardized to eliminate dimensional differences and bring all features to the same scale. The weights of each feature vector are determined based on its inverse variance. The smaller the measurement error, the greater the feature weight, thus obtaining the normalized feature weights.
[0049] Specifically, before weighted fusion of feature vectors, each feature vector is first standardized to eliminate dimensional differences and bring all features to the same scale.
[0050] After standardization, the weights of each feature vector are determined based on its inverse variance. A larger inverse variance indicates a smaller measurement error for that feature, and therefore a larger weight. The standardized feature vectors are then weighted and fused to obtain the normalized feature weights. The feature vectors are weighted and fused to obtain a normalized feature weighted sum. F fusion
[0051] in, This indicates that for each feature vector F k Standardize, This represents the mean of the k-th eigenvalue. This represents the standard deviation of the k-th vector. Its function is to eliminate dimensional differences and bring all features to the same scale.
[0052] Weight Determined by the inverse variance of each feature. , This represents the measurement error of the k-th feature. The smaller the error (the higher the accuracy), the greater the feature weight.
[0053] This normalized feature weighted vector reflects the combined contribution of each feature to fault diagnosis. This combined feature vector can be used as the final feature representation for subsequent fault diagnosis analysis.
[0054] Optionally, the adaptive threshold optimization method in the complex system fault diagnosis method based on dynamic evaluation using a hybrid simulation model provided in this application embodiment specifically includes: The initial threshold is set to
[0055] in The historical residual mean This represents the standard deviation of the residuals.
[0056] Specifically, in this embodiment of the application, the mean of the historical residuals is calculated by collecting and analyzing the residual data of the pump system during its historical operation. and standard deviation This is used to set the initial threshold.
[0057] The collected residual data undergoes initial screening to remove obvious outliers and erroneous data. Statistical tools are then used to calculate the mean of the filtered residual data. and standard deviation Calculate the initial threshold:
[0058] Through the above steps, the embodiments of this application can provide a reasonable initial threshold setting method for fault diagnosis of complex systems, laying the foundation for subsequent fault diagnosis and system optimization.
[0059] Optionally, the fault diagnosis method for complex systems based on dynamic evaluation using a hybrid simulation model provided in this application embodiment, specifically includes the following when dynamically adjusting the threshold: Based on the deviation between the current operating condition parameter vector and the rated operating condition parameter vector
[0060] in, This is the parameter vector for the current operating condition. This is the vector of rated operating parameters. This is the sensitivity coefficient; Using sensitivity coefficient The influence of changes in control conditions on the threshold is controlled. When the control conditions deviate from the rated control conditions, the threshold ratio is relaxed, and the residual threshold is dynamically optimized adaptively.
[0061] Specifically, In this embodiment, the threshold is dynamically adjusted in the following manner.
[0062] in, This is the current operating condition parameter vector (speed, load, etc.). This is the vector of rated operating parameters. The sensitivity coefficient represents the intensity of the influence of changes in control conditions on the threshold.
[0063] When the operating conditions deviate from the rated operating conditions, the threshold ratio is relaxed to avoid false alarms (such as during startup / speed change). The residual threshold is adaptively and dynamically optimized. The residual features are used as input to the dynamic threshold optimization module, and the fault diagnosis results are obtained through comparative analysis.
[0064] The dynamically adjusted threshold is applied to the fault diagnosis process of industrial motor systems. Combined with other diagnostic tools and methods, residual signals exceeding the threshold are further analyzed to determine whether a fault exists in the system, and if so, what type and location it is. Through these steps, the embodiments of this application can achieve dynamic threshold adjustment for fault diagnosis of complex systems, enabling the system to better adapt to changes in operating conditions and improving the accuracy and reliability of fault diagnosis.
[0065] Optionally, embodiments of this application also provide a fault diagnosis device for complex systems based on dynamic evaluation using a hybrid simulation model, comprising: The model building module is used to build a hybrid simulation model of mechanism model and data model. The mechanism model is constructed using modeling and simulation languages to build a parameterized model of electromechanical-hydraulic coupling system and exported to a specified format. The data model is constructed using a long short-term memory neural network method to build an anomaly monitoring module and encapsulated into the specified format through tools. Real-time data exchange between mechanism model and data model is realized based on OPCUA to build a hybrid simulation model. The feature extraction module is used to connect the input of the real system to the input terminal of the hybrid simulation model to simulate the operation of the real system, calculate the difference between the output of the real system and the output of the hybrid simulation model to obtain the residual signal, extract the time domain feature vector, frequency domain feature vector and spatial correlation feature vector, and perform weighted fusion on the feature vector to obtain normalized feature weighting; The fault diagnosis module is used to perform residual assessment fault diagnosis using an adaptive threshold optimization method. It sets an initial threshold and dynamically adjusts the threshold based on the deviation between the current operating condition parameter vector and the rated operating condition parameter vector through a sensitivity coefficient. It then analyzes the residual characteristics to obtain fault diagnosis results.
[0066] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method. The computer-readable storage medium may include, but is not limited to, any type of disk, including floppy disks, optical disks, DVDs, CD-ROMs, microdrives, as well as magneto-optical disks, ROMs, RAMs, EPROMs, EEPROMs, DRAMs, VRAMs, flash memory devices, magnetic cards or optical cards, nanosystems (including molecular memory ICs), or any type of medium or device suitable for storing instructions and / or data.
[0067] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.
[0068] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.
[0069] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between devices or units may be electrical or other forms.
[0070] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0071] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0072] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.
[0073] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.
[0074] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of embodiments of this disclosure upon considering the specification and practicing the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described herein. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.
[0075] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0076] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.
Claims
1. A complex system fault diagnosis method based on hybrid simulation model dynamic evaluation, characterized in that, The method comprises the following steps: a hybrid simulation model is constructed, the mechanism model is constructed by using a modeling and simulation language to construct a parameterized model of the mechatronic-hydraulic coupling system and is exported in a specified format, the data model is constructed by using a long short-term memory neural network method to construct an anomaly monitoring module and is packaged in the specified format by using a tool, real-time data exchange of the mechanism model and the data model is realized based on OPC UA, and the hybrid simulation model is constructed; real system inputs are connected to input ends of the hybrid simulation model to simulate real system operation, residual error signals are obtained by respectively subtracting outputs of the real system and the hybrid simulation model, time domain feature vectors, frequency domain feature vectors and spatial correlation feature vectors are extracted from the residual error signals, feature vectors are weighted and fused to obtain normalized feature weights; an adaptive threshold optimization method is used for residual error evaluation and fault diagnosis, an initial threshold is set, the threshold is dynamically adjusted by using a sensitivity coefficient according to a deviation between a current working condition parameter vector and a rated working condition parameter vector, and residual error features are analyzed to obtain a fault diagnosis result.
2. The method for complex system fault diagnosis based on hybrid simulation model dynamic evaluation according to claim 1, wherein, When the mechanism model is constructed, the following steps are specifically included: each component of the mechatronic-hydraulic coupling system is parameterized modeled based on the modeling and simulation language; the constructed mechanism model is exported in the specified format.
3. The method of claim 1, wherein the method further comprises: When the data model is constructed, the following steps are specifically included: the long short-term memory neural network is used to learn and train historical system data to construct the anomaly monitoring module; the data model is packaged in the specified format by using the tool.
4. The method for complex system fault diagnosis based on hybrid simulation model dynamic evaluation according to claim 1, wherein, When the residual error features are extracted, the following steps are specifically included: real-time monitoring and comparison are performed on input and output data of the real system and the hybrid simulation model, and residual error signals are calculated; time domain feature vectors, frequency domain feature vectors and spatial correlation feature vectors are extracted from the residual error signals in three dimensions of time domain, frequency domain and spatial correlation.
5. The method for complex system fault diagnosis based on hybrid simulation model dynamic evaluation according to claim 1, wherein, When the feature vectors are weighted and fused, the following steps are specifically included: each feature vector is normalized to eliminate dimensional differences and make all features in the same scale; weights of the feature vectors are determined based on inverse variances, a feature with a smaller measurement error has a larger weight, and normalized feature weights are obtained.
6. The method for complex system fault diagnosis based on hybrid simulation model dynamic evaluation according to claim 1, characterized in that, In the adaptive threshold optimization method, the following steps are specifically included: the initial threshold is set as wherein is the historical residual mean, is the residual standard deviation.
7. The method for complex system fault diagnosis based on hybrid simulation model dynamic evaluation according to claim 6, characterized in that, When the threshold is dynamically adjusted, the following steps are specifically included: a deviation between a current working condition parameter vector and a rated working condition parameter vector is calculated wherein, is the current operating parameter vector, is the nominal operating parameter vector, is the sensitivity coefficient; Utilizing sensitivity coefficients The influence intensity of the control working condition change on the threshold value is controlled, and when the working condition deviates from the rated working condition, the threshold value is relaxed, and the residual threshold value is adaptively and dynamically optimized.
8. A complex system fault diagnosis device based on hybrid simulation model dynamic evaluation, characterized in that, The method comprises the following steps: a model construction module is configured to construct a hybrid simulation model of a mechanism model and a data model, the mechanism model is constructed by using a modeling and simulation language to construct a parameterized model of a mechatronic-hydraulic coupling system and is exported in a specified format, the data model is constructed by using a long short-term memory neural network method to construct an anomaly monitoring module and is packaged in the specified format by using a tool, real-time data exchange of the mechanism model and the data model is realized based on OPC UA, and the hybrid simulation model is constructed; a feature extraction module is configured to connect real system inputs to input ends of the hybrid simulation model to simulate real system operation, obtain residual error signals by respectively subtracting outputs of the real system and the hybrid simulation model, extract time domain feature vectors, frequency domain feature vectors and spatial correlation feature vectors from the residual error signals, and weight and fuse feature vectors to obtain normalized feature weights; and The fault diagnosis module is used for residual error evaluation fault diagnosis by using an adaptive threshold optimization method, setting an initial threshold, dynamically adjusting the threshold according to the deviation of the current working condition parameter vector and the rated working condition parameter vector, and analyzing the residual error characteristics to obtain a fault diagnosis result.
9. A complex system fault diagnosis device based on hybrid simulation model dynamic evaluation, characterized in that, The computer program product comprises at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, and when the computer program is executed by the processing unit, the processing unit executes the steps of the method in any one of claims 1-7.
10. A storage medium, characterized by The computer program product comprises at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, and when the computer program is executed by the processing unit, the processing unit executes the steps of the method in any one of claims 1-7.