An evaluation method of an operating mechanism and related device

Abstract

Embodiments of the present application provide an evaluation method of an operating mechanism and a related device, applied to the technical field of data processing, input the first score and the second score of each state evaluation index into an equilibrium game theory model, and obtain the optimal score of each state evaluation index output by the equilibrium game theory model. Since the first score represents the score of each to-be-evaluated index based on expert experience, and the second score represents the score of each to-be-evaluated index based on an artificial intelligence model, the optimal score obtained by fusing the first score and the second score through game theory improves the evaluation accuracy of the to-be-evaluated index. Moreover, the equilibrium game theory model is a game theory model constructed based on a dynamic game function, the weight of the first score is a first dynamic weight coefficient, and the weight of the second score is a second dynamic weight coefficient. It can be seen that the accuracy of the optimal score is further improved through the game theory with variable weights.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to an evaluation method and related apparatus for an operating mechanism. Background Technology

[0002] Operating mechanisms in power systems are used to perform various switching operations, such as closing and opening circuit breakers. Since operating mechanisms are usually mechanical devices, they are susceptible to failure due to mechanical wear, corrosion, and other factors under outdoor conditions. This results in low reliability of the operating mechanisms, which will affect the execution of various switching operations in the power system and further affect the safety, stability, and reliability of the power system operation.

[0003] As the automation level and reliability requirements of power distribution networks increase, the reliability requirements for operating mechanisms also continue to rise. Accurately assessing the health status of operating mechanisms is an important basis for improving their reliability. Summary of the Invention

[0004] This application provides a method and related apparatus for evaluating operating mechanisms, with the aim of improving the accuracy of evaluation results for operating mechanisms, as follows:

[0005] The first aspect of this application provides a method for evaluating an operating mechanism, comprising:

[0006] Collect operating data of the operating mechanism to be evaluated. The operating data includes a sequence of index values ​​for multiple status evaluation indicators.

[0007] The operation data is input into the expert system to obtain the expert score set of each state evaluation index output by the expert system. The expert system includes multiple expert clients, each expert client is bound to an expert user, and the expert score set includes the expert scores uploaded by each expert user through the expert client.

[0008] Based on the expert score set of each of the aforementioned state assessment indicators, the first score of each of the aforementioned state assessment indicators is obtained;

[0009] The manipulation data is input into a pre-built hybrid deep learning model to obtain the second score of each of the state evaluation metrics output by the hybrid deep learning model;

[0010] The first and second scores of each of the aforementioned state evaluation indicators are input into the equilibrium game theory model to obtain the optimal scores of each of the aforementioned state evaluation indicators output by the equilibrium game theory model. The equilibrium game theory model is a game theory model constructed based on a dynamic game function. The dynamic game function is used to indicate that the optimal score of each of the aforementioned state evaluation indicators is equal to the weighted sum of the first score and the second score, wherein the weight of the first score is the first dynamic weight coefficient, and the weight of the second score is the second dynamic weight coefficient.

[0011] In one possible implementation, operating data of the actuator to be evaluated is collected, including:

[0012] Obtain a set of candidate indicators, which are divided into energy storage unit indicator set, electric drive unit indicator set, motor unit indicator set and transmission unit indicator set according to the organizational unit;

[0013] Based on the set of candidate indicators, an indicator relationship network is obtained, wherein each candidate indicator is a node, and the directed edges between nodes point from the upper-level indicator to the lower-level indicator.

[0014] Based on the aforementioned indicator relationship network, the highest-level indicator set is obtained as the first candidate indicator set.

[0015] Calculate the Pearson correlation coefficient between each candidate indicator and the reference candidate indicator, and select the set of candidate indicators that meet the preset correlation conditions as the second candidate indicator set. The reference candidate indicators include click angle, and the correlation conditions include a correlation coefficient greater than the preset threshold.

[0016] Obtain the intersection of the first candidate indicator set and the second candidate indicator set as the evaluation indicator set, which includes multiple indicators to be evaluated.

[0017] In one possible implementation, based on the set of expert scores for each of the state assessment indicators, a first score for each of the state assessment indicators is obtained, including:

[0018] Each time, a first preset number of expert scores are selected from the expert score set to form an expert score array, resulting in multiple expert score arrays, wherein each expert score array is not completely identical, and the first preset number is less than the number of expert clients;

[0019] Calculate the variance of each expert score array;

[0020] The average of the expert scores in the expert score array with the smallest variance is selected as the comprehensive expert score of the state assessment index.

[0021] The comprehensive expert scores of each of the aforementioned state assessment indicators are normalized to obtain the first score of each of the aforementioned state assessment indicators.

[0022] In one possible implementation, the hybrid deep learning model is built upon a convolutional neural network (CNN), a gated recurrent unit (GRU), and a self-attention mechanism (Attention network).

[0023] In one possible implementation, the process of assigning second scores to the various state evaluation metrics output by the hybrid deep learning model includes:

[0024] The CNN performs convolution operations on the input manipulation data to extract local features of the manipulation data; and performs pooling operations on the local features of the manipulation data to obtain the pooled features of the manipulation data.

[0025] The GRU performs cyclic processing on the pooling features of the manipulation data to obtain the cyclic features of the manipulation data;

[0026] The Attention network performs weighted processing on the cyclic features of the manipulation data based on a self-attention mechanism, generates a second score for each of the state evaluation indicators, and outputs it.

[0027] In one possible implementation, the first and second scores of each of the aforementioned state evaluation indicators are input into an equilibrium game theory model to obtain the optimal scores of each of the aforementioned state evaluation indicators output by the equilibrium game theory model, including:

[0028] Based on the idea of ​​minimizing deviation in game theory and the differential property of matrices, the first and second fixed-weight coefficients of each state evaluation index are calculated.

[0029] The first dynamic weight coefficient of the state evaluation index is calculated based on the first score of the state evaluation index, the preset equilibrium function correction coefficient, and the first fixed weight coefficient.

[0030] The second dynamic weight coefficient of the state evaluation index is calculated based on the second score of the state evaluation index, the adjustment coefficient of the equilibrium function, and the second fixed weight coefficient.

[0031] The optimal score for each of the state evaluation indicators is calculated using the dynamic game function.

[0032] A second aspect of this application provides an evaluation device for an operating mechanism, comprising:

[0033] The data acquisition unit is used to collect the operating data of the operating mechanism to be evaluated. The operating data includes a sequence of index values ​​for multiple status evaluation indicators.

[0034] An expert scoring unit is used to input the operation data into an expert system, obtain a set of expert scores for each status evaluation indicator output by the expert system, the expert system includes multiple expert clients, each expert client is bound to an expert user, and the set of expert scores includes the expert scores uploaded by each expert user through the expert client; based on the set of expert scores for each status evaluation indicator, a first score for each status evaluation indicator is obtained.

[0035] An artificial intelligence scoring unit is used to input the manipulation data into a pre-built hybrid deep learning model and obtain the second score of each of the state evaluation indicators output by the hybrid deep learning model.

[0036] The game unit is used to input the first and second scores of each of the state evaluation indicators into the equilibrium game theory model to obtain the optimal scores of each of the state evaluation indicators output by the equilibrium game theory model. The equilibrium game theory model is a game theory model constructed based on a dynamic game function. The dynamic game function is used to indicate that the optimal score of each of the state evaluation indicators is equal to the weighted sum of the first score and the second score, wherein the weight of the first score is a first dynamic weight coefficient, and the weight of the second score is a second dynamic weight coefficient.

[0037] A third aspect of this application provides an evaluation device for an operating mechanism, comprising: a memory and a processor;

[0038] The memory is used to store programs;

[0039] The processor is used to execute the program to implement the various steps of the evaluation method for the operating mechanism as described above.

[0040] A fourth aspect of this application provides a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the various steps of the evaluation method for the operating mechanism as described above.

[0041] The fifth aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the various steps of the evaluation method for the operating mechanism as described above.

[0042] As can be seen from the above technical solution, the evaluation method and related apparatus for the operating mechanism provided in this application collect operating data of the operating mechanism to be evaluated. The operating data includes a sequence of index values ​​for multiple state evaluation indicators. The operating data is input into an expert system to obtain a set of expert scores for each state evaluation indicator output by the expert system. Based on the set of expert scores for each state evaluation indicator, a first score for each state evaluation indicator is obtained. The operating data is input into a pre-built hybrid deep learning model to obtain a second score for each state evaluation indicator output by the hybrid deep learning model. The first and second scores for each state evaluation indicator are input into an equilibrium game theory model to obtain the optimal score for each state evaluation indicator output by the equilibrium game theory model. Since the expert system includes multiple expert clients, each expert client is bound to an expert user, and the set of expert scores includes the expert scores uploaded by each expert user through the expert client, the first score represents the score for each indicator to be evaluated based on expert experience, and the second score represents the score for each indicator to be evaluated based on an artificial intelligence model. The optimal score obtained by fusing the first and second scores through game theory can improve the evaluation accuracy of the indicator to be evaluated. Furthermore, since the equilibrium game theory model is a game theory model based on a dynamic game function, the dynamic game function represents that the optimal score of each state evaluation index is equal to the weighted sum of the first score and the second score, where the weight of the first score is the first dynamic weight coefficient, and the weight of the second score is the second dynamic weight coefficient. Therefore, this application further improves the accuracy of the optimal score through game theory with variable weights. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art 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.

[0044] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0045] Figure 1 A schematic diagram of the structure of an evaluation system for an operating mechanism provided in an embodiment of this application;

[0046] Figure 2 A flowchart illustrating a specific implementation of an evaluation method for an operating mechanism provided in this application embodiment;

[0047] Figure 3 A flowchart illustrating an evaluation method for an operating mechanism provided in an embodiment of this application;

[0048] Figure 4 A flowchart illustrating a method for obtaining state assessment indicators provided in an embodiment of this application;

[0049] Figure 5 This application provides a schematic diagram of the structure of an index relationship network.

[0050] Figure 6 A schematic diagram of a correlation matrix provided in an embodiment of this application;

[0051] Figure 7 A schematic diagram illustrating the specific structure of a hybrid deep learning model provided in this application embodiment;

[0052] Figure 8 A schematic diagram of an operation based on an attention mechanism provided for an embodiment of this application;

[0053] Figure 9 This is a schematic diagram illustrating the results of an evaluation effect test provided in an embodiment of this application;

[0054] Figure 10 A schematic diagram of the structure of an evaluation device for an operating mechanism provided in an embodiment of this application;

[0055] Figure 11 This is a schematic diagram of the structure of an evaluation device for an operating mechanism provided in an embodiment of this application. Detailed Implementation

[0056] Embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure.

[0057] It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure. The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the description below. The modifications of "a" and "a plurality" mentioned in this disclosure are illustrative and not restrictive, and those skilled in the art should understand that unless explicitly indicated otherwise in the context, they should be understood as "one or more".

[0058] This application provides an evaluation method for an operating mechanism, which is applied to... Figure 1 The evaluation system shown is as follows: Figure 1 As shown, the evaluation system includes an operational data acquisition system, an expert system, a hybrid deep learning model, and a game theory model.

[0059] In this embodiment, the operation data acquisition system is used to collect operation data of the operating mechanism to be evaluated, and send the operation data to the expert system and the hybrid deep learning model respectively. The operation data includes multiple index values ​​of the operating mechanism. Optionally, the operating mechanism to be evaluated is a direct-drive motor operating mechanism. Figure 2 An example is a schematic diagram of a direct-drive operating mechanism for an electric motor, such as... Figure 2 As shown, the direct-drive operating mechanism includes components such as a coupling, connecting flange, torsion bar, crank arm, overtravel spring, and insulating tie rod. The switching action is driven by a permanent magnet synchronous motor connected to the coupling. Viewed from the direction away from the coupling, clockwise is the opening direction, and counterclockwise is the closing direction.

[0060] In this embodiment, the expert system is used to obtain expert evaluation results based on the manipulation data and send the expert evaluation results to the game theory model. The hybrid deep learning model is used to obtain model evaluation results based on the manipulation data and send the model evaluation results to the game theory model. The game theory model is used to perform a fusion operation on the expert evaluation results and the model evaluation results based on game theory to obtain the game evaluation result of the manipulation mechanism to be evaluated.

[0061] Figure 3 This is a flowchart illustrating an evaluation method for an operating mechanism provided in an embodiment of this application. The evaluation method for the operating mechanism can be applied to, for example... Figure 1 The evaluation system shown is as follows: Figure 3 As shown, this method includes:

[0062] S301. Collect the operating data of the operating mechanism to be evaluated.

[0063] In this embodiment, the operation data includes a sequence of index values ​​for multiple state evaluation indicators. The index value sequence includes multiple index values ​​arranged in sequence. Optionally, the state evaluation indicators include d-axis current, q-axis current, motor speed, motor angle, capacitor voltage, and capacitor current.

[0064] Specifically, the operation data acquisition system collects the index values ​​of multiple status evaluation indicators through a preset acquisition method and a preset acquisition frequency, and obtains the index value sequence of each status evaluation indicator. For specific acquisition methods, please refer to existing technologies.

[0065] It should be noted that the status assessment indicators are pre-configured based on the indicator selection method. For the specific indicator selection method, please refer to the following example.

[0066] S302. Input the operation data into the expert system and obtain the set of expert scores for each status evaluation index output by the expert system.

[0067] In this embodiment, the expert score set includes multiple expert scores. Each expert score is output by an expert client in the expert system, and the expert score is used to indicate the human score of the operating mechanism's status. It should be noted that the expert system includes multiple expert clients, and each expert client is bound to an expert user. Optionally, the operating data is distributed to each expert client in the expert system, and the expert scores of each status evaluation index in the operating data output by each expert client are obtained.

[0068] Taking the manipulation data as a sequence of index values ​​for n state evaluation indicators and the expert system as having S expert clients as an example, the set of expert scores for the i-th (i∈[1,n]) state evaluation indicator Xi is obtained. ,in, For the s-th ( The expert score of the status assessment index Xi output by the expert client.

[0069] S303. Based on the expert score set of each state assessment indicator, obtain the first score of each state assessment indicator.

[0070] In this embodiment, for any state assessment index, the specific method for obtaining the first score of the state assessment index based on the expert score set of the state assessment index includes:

[0071] A1. Each time, a first preset number of expert scores are selected from the expert score set to form an expert score array, resulting in multiple expert score arrays. Each expert score array is not exactly the same, and the first preset number is less than the number of expert clients.

[0072] A2. Calculate the variance of each expert score array.

[0073] A3. Select the average of all expert scores in the expert score array with the smallest variance as the comprehensive expert score of the state assessment index.

[0074] The set of expert scores for the state assessment index Xi includes Taking the first preset quantity as an example, where the quantity is equal to S-2, for... Perform permutations and combinations, each time starting from... Select S-2 expert scores to form an expert score array, and obtain Given several expert score arrays, calculate the variance of each array and select the array with the smallest variance as the result array. Then, average the scores in the result array to obtain the comprehensive expert score for the state assessment index Xi. .

[0075] A4. Normalize the comprehensive expert scores of each status assessment indicator to obtain the first score of each status assessment indicator.

[0076] In this embodiment, the specific method for normalizing the comprehensive expert scores of each state evaluation index includes:

[0077] B1. For any state assessment indicator, divide the overall expert score of the state assessment indicator by the sum of the overall expert scores of all state assessment indicators to obtain the percentage of expert scores.

[0078] In this embodiment, the overall expert score is determined by the percentage of expert scores, making it a value between 0 and 1. Continuing from the previous example, the percentage of expert scores for the state assessment index Xi is used. The calculation formula is:

[0079] ;

[0080] in, .

[0081] B2. For any state assessment indicator, divide the percentage of expert scores for the state assessment indicator by the sum of the percentages of expert scores for all state assessment indicators to obtain the normalized comprehensive expert score, which is also the first score.

[0082] In this embodiment, standardization is used to make the sum of the scores of all experts equal to 1.

[0083] Continuing from the previous example, the first score of the i-th state evaluation index Xi The calculation formula is:

[0084] .

[0085] S304. Input the manipulation data into the hybrid deep learning model and obtain the second score of each state evaluation index output by the hybrid deep learning model.

[0086] In this embodiment, the hybrid deep learning model is constructed from CNN (Convolutional Neural Network), GRU (Gate Recurrent Unit), and the self-attention mechanism Attention network.

[0087] Specifically, the manipulation data consists of a data sequence including various state evaluation metrics. This manipulation data is input into a hybrid deep learning model. A CNN performs convolution operations on the input manipulation data to extract local features. Pooling is then applied to these local features to obtain pooled features, reducing the dimensionality of the local features and thus lowering the model's computational complexity. A GRU iteratively processes these pooled features to obtain cyclic features, capturing sequence dependencies. Finally, an Attention network weights these cyclic features to generate and output a second score for each state evaluation metric.

[0088] It should be noted that due to the characteristics of the protective equipment in direct-drive motor operating mechanisms, such as the limited number of actions throughout their entire lifecycle and the continuous recording of waveforms during a single action, the operational data of direct-drive motor operating mechanisms is a set of data with a clear sequence, limited data volume, and relatively obvious local characteristics. Therefore, this step uses a hybrid deep learning model to analyze the operational data. CNN excels at capturing local features, while RNN excels at capturing sequence dependencies. The attention mechanism can weight the input sequence according to its importance, thereby improving the performance of the hybrid deep learning model and thus improving the accuracy of the output results.

[0089] It should be noted that the specific structure of the hybrid deep learning model and the method for obtaining the second score of each state evaluation index based on the hybrid deep learning model can be found in the following embodiments.

[0090] S305. Input the first and second scores of each state evaluation index into the equilibrium game theory model to obtain the optimal scores of each state evaluation index output by the equilibrium game theory model.

[0091] In this embodiment, the equilibrium game theory model is a game theory model constructed based on a dynamic game function. The dynamic game function is used to represent that the optimal score of each state evaluation index is equal to the weighted sum of the first score and the second score, where the weight of the first score is the first dynamic weight coefficient, and the weight of the second score is the second dynamic weight coefficient.

[0092] In this embodiment, the first score of the state evaluation index Xi is The corresponding dynamic weighting coefficient is denoted as The second score of the state assessment index Xi is The corresponding dynamic weighting coefficient is denoted as .

[0093] In this embodiment, the process by which the equilibrium game theory model calculates the optimal score for each state evaluation index based on the first and second scores of each state evaluation index includes:

[0094] C1. Obtain the first score vector and the second score vector.

[0095] In this embodiment, the first score vector is composed of the first scores of each state evaluation indicator arranged in sequence, and the second score vector is composed of the second scores of each state evaluation indicator arranged in sequence. For example, the first score vector... =[ ],in, ( The first score is the value of the state assessment index Xi. The second score vector is... =[ ],in, ( ) is the second score of the state assessment index Xi.

[0096] C2. Based on the idea of ​​minimizing deviation in game theory and the differential property of matrices, calculate the first and second fixed-weight coefficients of each state evaluation index.

[0097] Specifically, based on the idea of ​​minimizing deviations in game theory, the optimization formula is as follows:

[0098] ;

[0099] Where G=2, k=1~2, that is, it includes two types of scores: the first score and the second score. The optimal score for the state assessment index Xi is... This represents the minimum deviation.

[0100] Differential property formulas based on matrices The intermediate formula is as follows:

[0101] ;

[0102] in, include and , The first fixed weighting coefficients for each state evaluation index The vector formed The second fixed-weight coefficients are derived from the various state evaluation indicators. The constructed vector, based on the intermediate formula, is calculated. and That is, the first fixed weighting coefficient of each state evaluation index. Second fixed weight coefficient .

[0103] C3. The first dynamic weight coefficient of the state evaluation index Xi is calculated based on the first score of the state evaluation index Xi, the preset equilibrium function correction coefficient, and the first fixed weight coefficient. The second dynamic weight coefficient of the state evaluation index Xi is calculated based on the second score of the state evaluation index Xi, the preset equilibrium function correction coefficient, and the second fixed weight coefficient.

[0104] In this embodiment, the method for calculating the dynamic weight coefficient based on the variable weight formula containing the equilibrium function is as follows:

[0105] ;

[0106] Where rt is the equilibrium function correction coefficient.

[0107] It should be noted that in weight determination methods, weights are generally divided into constant weights and variable weights. When the constant weights deviate from their values ​​during the evaluation process, they cannot accurately reflect the true state of the evaluated object. Therefore, game theory-based combinatorial weighting is proposed to improve the variable weight theory, using a variable weight formula containing an equilibrium function to calculate the first dynamic weight coefficient of the state evaluation index Xi. Second dynamic weight coefficient .

[0108] C4. Use dynamic game functions to calculate the optimal scores for each state evaluation index.

[0109] In this embodiment, the optimal score of each state evaluation index is equal to the weighted sum of the first score and the second score, where the weight of the first score is the first dynamic weight coefficient and the weight of the second score is the second dynamic weight coefficient.

[0110] Specifically, the optimal weight of the state assessment index Xi Equal to the first dynamic weighting coefficient Second dynamic weight coefficient The weighting coefficient is used for the first score. Second score We perform a weighted summation to obtain the weighted sum result, and the corresponding dynamic game function is as follows:

[0111] .

[0112] As can be seen from the above technical solutions, the evaluation method for an operating mechanism provided in this application adopts expert group evaluation and multi-dimensional evaluation indicators to improve the accuracy of expert evaluation methods. A self-attention mechanism is introduced into the CNN-GRU network, and the output results of the GRU layer are updated by reasonably allocating attention weights, thereby reducing potential biases. Game theory is used to integrate and optimize the results of the expert group evaluation method and the CNN-GRU network method based on the attention mechanism. In the game theory weight determination method, a game theory combination weighting is proposed to improve the contingency theory. Expert systems are suitable for situations relying on professional knowledge, while game theory methods can consider the interaction between different factors. Choosing game theory to optimize and analyze the results of both methods to obtain the final state evaluation result results in high accuracy and strong targeting.

[0113] It should be noted that, Figure 3 This application provides only one specific implementation of an evaluation method for an operating mechanism. Other implementations are also possible. For example, in step S303, when obtaining the first score of each state evaluation indicator based on the expert score set of each state evaluation indicator, the average of the expert scores in the expert score set can be directly used as the first score of the state evaluation indicator. Alternatively, this application may also obtain the operating state of the operating mechanism based on the optimal score of each state evaluation indicator output by an equilibrium game theory model. Optionally, the operating state of the operating mechanism can be divided into multiple categories, such as A (normal), B (slight jamming), C (moderate jamming), and D (locked, sampling abnormality, undervoltage, etc.). A mapping relationship between the scores of each state evaluation indicator and the operating state is preset, and the optimal score of each state evaluation indicator is mapped to the operating state using a lookup table method.

[0114] Figure 4 A flowchart illustrating a method for obtaining state assessment indicators provided in this application embodiment is shown below. Figure 4 As shown, this method includes:

[0115] S401. Obtain the set of candidate indicators.

[0116] In this embodiment, the set of candidate indicators is divided into the energy storage unit indicator set, the electric drive unit indicator set, the motor unit indicator set, and the transmission unit indicator set according to the mechanical unit.

[0117] In this embodiment, the energy storage unit index set includes multiple energy storage unit indices, the electric drive unit index set includes multiple electric drive unit indices, the motor unit index set includes multiple motor unit indices, and the transmission unit index set includes multiple transmission unit indices.

[0118] Specifically, the candidate indicators for each indicator set are shown in Table 1 below.

[0119] Table 1. Schematic diagram of unit indicator set

[0120]

[0121] S402. Based on the set of candidate indicators, obtain the indicator relationship network.

[0122] In this embodiment, the indicator relationship network uses each candidate indicator as a node, and directed edges between nodes point from higher-level indicators to lower-level indicators. The higher-level indicators of the target indicator include indicators that influence the target indicator; that is, lower-level indicators change along with higher-level indicators.

[0123] In this embodiment, the method for obtaining the indicator relationship network based on the candidate indicator set includes:

[0124] 1. Based on the hardware structure and motor operating principle, the logical relationships between the various candidate indicators are analyzed. Here, the logical relationship refers to the influence relationship between variables.

[0125] Specifically, the overall structure of the transmission unit adopts a four-bar linkage. Geometric analysis can reveal the relationship between the rotational angle of the motor drive shaft and the linear transmission stroke at the end of the transmission unit. In three-phase linkage, the deformation of the transmission shaft affects the synchronization of the three phases. The deformation of the transmission shaft is related to its stiffness and the opening and closing speeds. Especially under high acceleration, insufficient transmission shaft stiffness leads to a higher degree of deformation the further away from the motor, ultimately resulting in poor synchronization of the three phases and affecting the equipment's service function.

[0126] The main parameters involved in the motor unit, such as stator resistance, d-axis and q-axis inductance, permanent magnet flux linkage, and stator and rotor temperatures, can be obtained through offline or online methods. Furthermore, stator temperature monitoring can be achieved based on stator resistance identification, and demagnetization fault detection and rotor temperature monitoring can be achieved through permanent magnet flux linkage identification. Optionally, using the three-phase current, speed, and d-axis and q-axis current signals from the electric drive unit, a genetic algorithm can be employed to simultaneously identify the four parameters: stator resistance, d-axis and q-axis inductance, and permanent magnet flux linkage.

[0127] As the energy source for the direct-drive operating mechanism of the motor, the energy storage unit experiences short-term high-current discharges during operation. The voltage and current data of the energy storage capacitor can help determine its operating status and health. Optionally, the least squares method can be used to identify the capacitor value. This method is simple in structure, requiring only the voltage and current signals of the energy storage unit, as well as the injection of a specific signal, to identify the capacitor value. The capacitance calculation error using the least squares method is less than 1%.

[0128] 2. Establish logical relationships between the various candidate indicators and construct an indicator relationship network.

[0129] Figure 5 This is a schematic diagram of the structure of an index relationship network provided in an embodiment of this application. For example... Figure 5 As shown, specifically, the candidate indicators in the electric drive unit indicator set are the superior indicators of the candidate indicators in the motor unit indicator set; that is, the candidate indicators in the motor unit indicator set can be calculated from the candidate indicators in the electric drive unit indicator set. The motor speed and motor angle in the electric drive unit indicator set are the superior indicators of the candidate indicators in the transmission unit. In the motor unit indicator set, stator resistance is the superior indicator of stator temperature, and permanent magnet flux linkage is the superior indicator of rotor temperature. In the energy storage unit indicator set, capacitor voltage and capacitor current are the superior indicators of capacitor capacitance.

[0130] S403. Based on the indicator relationship network, obtain the highest-level indicator set as the first candidate indicator set.

[0131] In this embodiment, if there is no candidate indicator in the first indicator set that is a lower-level indicator of the candidate indicator in other indicator sets, then the first indicator set is taken as the highest-level indicator set.

[0132] like Figure 5 As shown, based on the index relationship network, the electric drive unit and energy storage unit are considered as the first-level units, and the motor unit and transmission unit are considered as the second-level units. Therefore, the highest-level index set includes the electric drive unit index set and the energy storage unit index set. That is, the first candidate index set includes all candidate indicators from the electric drive unit index set and all candidate indicators from the energy storage unit index set.

[0133] S404. Calculate the Pearson correlation coefficient between each candidate index and the motor angle, and select the set of candidate indices that meet the preset correlation conditions as the second candidate index set.

[0134] In this embodiment, the correlation condition includes a Pearson correlation coefficient greater than a preset correlation coefficient threshold.

[0135] In this embodiment, the Pearson correlation coefficient between the target candidate indicator and the motor angle is used to measure the linear relationship between the target candidate indicator and the motor angle. The larger the Pearson correlation coefficient, the stronger the correlation between the target candidate indicator and the motor angle. Since the motor angle indicates the actual feedback position of the motor, this step uses the motor angle as the key metric, selects candidate indicators whose Pearson correlation coefficient with the motor angle is greater than the correlation coefficient threshold, and uses this candidate indicator set as the second candidate indicator set.

[0136] In this embodiment, the correlation coefficient is used to verify the association between the candidate features. The correlation coefficient is a statistical indicator used to measure the strength and direction of the linear relationship between two variables. Optionally, the correlation coefficient includes the Pearson correlation coefficient, which measures the linear relationship between two continuous variables. The Pearson correlation coefficient ranges from -1 to 1, where 1 indicates a perfect positive correlation, -1 indicates a perfect negative correlation, and 0 indicates no correlation.

[0137] The Pearson correlation coefficient between candidate feature x and candidate feature y For example, let the sequence of candidate feature x be { }, Let represent the i-th value of the candidate feature x, and let the sequence of candidate features y be { }, This represents the i-th value of the candidate feature y, where i ∈ [1, n]. The calculation method is shown in the following formula:

[0138] ;

[0139] in, , and are the means of the candidate features x and y, respectively.

[0140] S405. Obtain the intersection of the first candidate indicator set and the second candidate indicator set as the evaluation indicator set.

[0141] As can be seen from the above technical solutions, the embodiments of this application determine the state evaluation index based on the logical relationship and correlation between the candidate indicators. Since the candidate indicators in the evaluation index set meet the conditions that there are no superior indicators and the Pearson correlation coefficient with the motor angle is greater than the correlation coefficient threshold, the candidate indicators in the evaluation index set are used as state evaluation indicators. When the actual feedback position of the motor is the most important measure, the state evaluation index has a high representativeness for the state evaluation of the operating system, thus improving the evaluation effect.

[0142] It should be noted that, Figure 4 This is merely one specific implementation of a method for obtaining state assessment indicators provided in this application embodiment. In other optional implementations, the method for obtaining state assessment indicators can be implemented through various optional specific implementation processes. For example, in S405, one optional implementation method is to obtain the Pearson correlation coefficient between each candidate indicator and the motor angle in real time, and select the candidate indicator set consisting of candidate indicators with a correlation coefficient threshold as the second candidate indicator set. Another optional implementation method is to perform correlation analysis on each candidate indicator in advance, calculate the Pearson correlation coefficient between every two candidate indicators, and obtain a correlation matrix. Figure 6 An example of a correlation matrix diagram is shown, such as... Figure 6 As shown, the columns and rows of the correlation matrix correspond to the candidate indicators, and the numbers in the cells represent the Pearson correlation coefficients between the two candidate indicators represented in the corresponding row and column. This step can be started from... Figure 6 Extract the Pearson correlation coefficients between any number of candidate indicators from the correlation matrix shown. That is, extract the Pearson correlation coefficients between each candidate indicator and the motor angle from the correlation matrix, and select the candidate indicator set composed of candidate indicators that are greater than the correlation coefficient threshold as the second candidate indicator set.

[0143] like Figure 6 As shown, if the correlation coefficient threshold is equal to K, then the candidate indicators whose Pearson correlation coefficient with the motor angle is greater than the correlation coefficient threshold include all candidate indicators in the electric drive unit indicator set and all candidate indicators in the energy storage unit indicator set. Therefore, the second candidate indicator set includes all candidate indicators in the electric drive unit indicator set (motor angle, motor speed, three-phase current, d-axis current, and q-axis current) and all candidate indicators in the energy storage unit indicator set (voltage, current, and capacitance value).

[0144] Figure 6 The definitions of each indicator to be evaluated are as follows:

[0145] a: Phase A current;

[0146] b: Phase B current;

[0147] c: C-phase current;

[0148] Idref: Current reference value along the d-axis in the rotating coordinate system;

[0149] Idfb: Current feedback value along the d-axis in the rotating coordinate system;

[0150] Iqref: Current reference value along the q-axis in the rotating coordinate system;

[0151] Iqfb: Current feedback value along the q-axis in the rotating coordinate system;

[0152] Wref: Rotational speed reference value;

[0153] Wfb: Speed ​​feedback value;

[0154] Thetaref - Motor angle reference value;

[0155] Thetafd: Motor angle feedback value;

[0156] v - Capacitor voltage;

[0157] i: Capacitor current;

[0158] r - stator resistance;

[0159] dl: Direct-axis inductance;

[0160] ql: quadrature axis inductance;

[0161] phi-magnetic link.

[0162] Figure 7 A schematic diagram illustrating the specific structure of a hybrid deep learning model provided in this application embodiment is shown below. Figure 7 As shown, the hybrid deep learning model includes CNN, RNN, and Attention network, where the RNN is a GRU. The CNN consists of convolutional layers, pooling layers, and fully connected layers connected sequentially, while the RRU consists of multiple reset gates and update gates. For specific structures, please refer to existing technologies.

[0163] Combination Figure 7 One possible method for outputting a second score using a hybrid deep learning model includes:

[0164] 1. The CNN part is designed to effectively reduce the complexity of feature extraction and data reconstruction. It features characteristics such as pooling, local connections, and weight sharing, making it excellent for feature extraction from time-series data. The sequence of values ​​for each evaluation metric in the manifold data is directly input into the CNN layer as time-series data. Convolutional layers extract features (i.e., features of the evaluation metrics) from the time-series data. Subsequently, pooling layers further capture the most important features in the time dimension, and fully connected layers are used to abstract and combine these time-series features. Specifically, the mathematical expression of the network is as follows:

[0165] ;

[0166] in, This represents the input time-series data. Represents the operations of the convolutional layer. Represents the operations of the pooling layer. This represents the operation of the fully connected layer. This hybrid structure allows the network to extract important features from raw time-series data and to model time-series information in an efficient manner.

[0167] It should be noted that the structure of CNNs and the specific methods for processing manipulated data can be found in existing technologies.

[0168] 2. The output of the CNN is used as input to the GRU, which completes the model's training and prediction tasks. The GRU simplifies individual storage units and integrates gate functions, making the network's learning from training data more efficient and improving the model's ability to prevent overfitting. It should be noted that the structure of the GRU and the specific methods for predicting the CNN output can be found in existing technologies.

[0169] 3. The output of the GRU is input into the Attention network. The attention mechanism module filters out high-value information from a large amount of data based on the attention mechanism. The output of the GRU layer is updated by reasonably allocating attention weights, and evaluation result data is output to reduce possible bias. The evaluation result data includes the second score of each evaluation indicator.

[0170] Figure 8 This is a schematic diagram of an operation based on the attention mechanism, such as... Figure 8 As shown, this method includes: multiplying the output of GRU, i.e., the feature Xi, by the model parameters Wl, Wm, and Wn respectively to obtain three representation matrices Li, Mi, and Ni:

[0171] .

[0172] To calculate the global alignment weights for feature Xi, we need to calculate the weights of Xi with respect to all other Xi except itself:

[0173] .

[0174] Finally, the calculated global alignment weights are summed to obtain...

[0175] .

[0176] It should be noted that the structure of the Attention network and the specific methods for adjusting the output of GRU can be found in existing technologies.

[0177] This application embodiment also provides an evaluation effect testing method, which includes: dividing the operating state of the motor direct drive operating mechanism into four states: A (normal), B (slight jamming), C (moderate jamming), and D (locked rotor, sampling abnormality, undervoltage, etc.) according to actual operating conditions; performing 100 actions for each of the four operating states (A, B, C, and D); and obtaining the evaluation result for each operating state using different evaluation methods for each action. Optionally, an expert system is used to obtain the first evaluation result, a hybrid deep learning network (i.e., CNN-GRU-Attention) is used to obtain the second evaluation result, a game theory model (using fixed weight coefficients as the weights of the first and second scores) is used to play against the first and second evaluation results to obtain the third evaluation result, and an equilibrium game theory model (using dynamic weight coefficients as the weights of the first and second scores) is used to play against the first and second evaluation results to obtain the fourth evaluation result. The accuracy of each evaluation result in each operating state is calculated respectively. Taking the first evaluation result in state A as an example, obtain the first evaluation result for each action in state A. If the first evaluation result is not state A, it means that the first evaluation result is incorrect. If the first evaluation result is state A, it means that the first evaluation result is accurate. Thus, obtain the ratio of the number of times the first evaluation result is accurate to the total number of times 100, which is taken as the accuracy rate of the first evaluation result in state A.

[0178] It should be noted that the method for obtaining the running status through the scores of each evaluation indicator may include weighted summation, model fusion, or mapping table methods, etc., and this embodiment does not limit it.

[0179] Figure 9 The diagram illustrates the results of the evaluation test, showing a comparison of the accuracy of each evaluation result under different operating conditions. Figure 9 It can be seen that the evaluation results obtained by the state index evaluation method based on equilibrium game theory provided in this application embodiment, namely the fourth evaluation result, have the highest accuracy in all operating states.

[0180] Figure 10 This application provides a schematic diagram of the structure of an evaluation device for an operating mechanism, as shown in the embodiment of the present application. Figure 10 As shown, the evaluation device 1000 for the operating mechanism may include:

[0181] The data acquisition unit 1001 is used to collect the operation data of the operating mechanism to be evaluated. The operation data includes a sequence of index values ​​for multiple status evaluation indicators.

[0182] The expert scoring unit 1002 is used to input the operation data into the expert system, obtain the expert score set of each state evaluation indicator output by the expert system, the expert system includes multiple expert clients, each expert client is bound to an expert user, and the expert score set includes the expert scores uploaded by each expert user through the expert client; based on the expert score set of each state evaluation indicator, the first score of each state evaluation indicator is obtained.

[0183] The artificial intelligence scoring unit 1003 is used to input the operation data into a pre-built hybrid deep learning model and obtain the second score of each of the state evaluation indicators output by the hybrid deep learning model.

[0184] Game unit 1004 is used to input the first and second scores of each of the state evaluation indicators into the equilibrium game theory model to obtain the optimal scores of each of the state evaluation indicators output by the equilibrium game theory model. The equilibrium game theory model is a game theory model constructed based on a dynamic game function. The dynamic game function is used to indicate that the optimal score of each of the state evaluation indicators is equal to the weighted sum of the first score and the second score, wherein the weight of the first score is the first dynamic weight coefficient, and the weight of the second score is the second dynamic weight coefficient.

[0185] In one possible implementation, when the data acquisition unit is used to acquire the operating data of the operating mechanism to be evaluated, it is specifically used for:

[0186] Obtain a set of candidate indicators, which are divided into energy storage unit indicator set, electric drive unit indicator set, motor unit indicator set and transmission unit indicator set according to the organizational unit;

[0187] Based on the set of candidate indicators, an indicator relationship network is obtained, wherein each candidate indicator is a node, and the directed edges between nodes point from the upper-level indicator to the lower-level indicator.

[0188] Based on the aforementioned indicator relationship network, the highest-level indicator set is obtained as the first candidate indicator set.

[0189] Calculate the Pearson correlation coefficient between each candidate indicator and the reference candidate indicator, and select the set of candidate indicators that meet the preset correlation conditions as the second candidate indicator set. The reference candidate indicators include click angle, and the correlation conditions include a correlation coefficient greater than the preset threshold.

[0190] Obtain the intersection of the first candidate indicator set and the second candidate indicator set as the evaluation indicator set, which includes multiple indicators to be evaluated.

[0191] In one possible implementation, when the expert scoring unit obtains the first score for each of the state assessment indicators based on the set of expert scores for each of the state assessment indicators, it specifically performs the following:

[0192] Each time, a first preset number of expert scores are selected from the expert score set to form an expert score array, resulting in multiple expert score arrays, wherein each expert score array is not completely identical, and the first preset number is less than the number of expert clients;

[0193] Calculate the variance of each expert score array;

[0194] The average of the expert scores in the expert score array with the smallest variance is selected as the comprehensive expert score of the state assessment index.

[0195] The comprehensive expert scores of each of the aforementioned state assessment indicators are normalized to obtain the first score of each of the aforementioned state assessment indicators.

[0196] In one possible implementation, the hybrid deep learning model is built upon a convolutional neural network (CNN), a gated recurrent unit (GRU), and a self-attention mechanism (Attention network).

[0197] In one possible implementation, the process of assigning second scores to the various state evaluation metrics output by the hybrid deep learning model includes:

[0198] The CNN performs convolution operations on the input manipulation data to extract local features of the manipulation data; and performs pooling operations on the local features of the manipulation data to obtain the pooled features of the manipulation data.

[0199] The GRU performs cyclic processing on the pooling features of the manipulation data to obtain the cyclic features of the manipulation data;

[0200] The Attention network performs weighted processing on the cyclic features of the manipulation data based on a self-attention mechanism, generates a second score for each of the state evaluation indicators, and outputs it.

[0201] In one possible implementation, when the game unit inputs the first and second scores of each of the state evaluation indicators into the equilibrium game theory model to obtain the optimal scores of each of the state evaluation indicators output by the equilibrium game theory model, it has the following functions:

[0202] Based on the idea of ​​minimizing deviation in game theory and the differential property of matrices, the first and second fixed-weight coefficients of each state evaluation index are calculated.

[0203] The first dynamic weight coefficient of the state evaluation index is calculated based on the first score of the state evaluation index, the preset equilibrium function correction coefficient, and the first fixed weight coefficient.

[0204] The second dynamic weight coefficient of the state evaluation index is calculated based on the second score of the state evaluation index, the adjustment coefficient of the equilibrium function, and the second fixed weight coefficient.

[0205] The optimal score for each of the state evaluation indicators is calculated using the dynamic game function.

[0206] This application also provides an electronic device in its embodiments. (See reference...) Figure 11 The diagram illustrates a structural schematic suitable for implementing the electronic device in the embodiments of this application. The electronic device in the embodiments of this application may include, but is not limited to, fixed terminals such as mobile phones, laptops, PDAs (personal digital assistants), PADs (tablet computers), desktop computers, etc. Figure 11 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0207] like Figure 11 As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 1101, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1102 or a program loaded from a storage device 1108 into a random access memory (RAM) 1103. When the electronic device is powered on, the RAM 1103 also stores various programs and data required for the operation of the electronic device. The processing unit 1101, ROM 1102, and RAM 1103 are interconnected via a bus 1104. An input / output (I / O) interface 1105 is also connected to the bus 1104.

[0208] Typically, the following devices can be connected to I / O interface 1105: input devices 1106 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 1107 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1108 including, for example, memory cards, hard drives, etc.; and communication devices 1109. Communication device 1109 allows electronic devices to exchange data via wireless or wired communication with other devices. Although Figure 6 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.

[0209] This application also provides a computer program product including computer-readable instructions, which, when executed on an electronic device, cause the electronic device to implement any of the operating mechanism evaluation methods provided in this application.

[0210] This application also provides a computer-readable storage medium carrying one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any of the operating mechanism evaluation methods provided in this application.

[0211] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and 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 modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the device embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0212] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0213] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.

[0214] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

Claims

1. A method for evaluating an operating mechanism, characterized in that, include: Collect operating data of the operating mechanism to be evaluated. The operating data includes a sequence of index values ​​for multiple status evaluation indicators. The operation data is input into the expert system to obtain the expert score set of each state evaluation index output by the expert system. The expert system includes multiple expert clients, each expert client is bound to an expert user, and the expert score set includes the expert scores uploaded by each expert user through the expert client. Based on the expert score set of each of the aforementioned state assessment indicators, a first score for each of the aforementioned state assessment indicators is obtained; wherein, obtaining the first score for each of the aforementioned state assessment indicators based on the expert score set of each of the aforementioned state assessment indicators includes: selecting a first preset number of expert scores from the expert score set each time to form an expert score array, resulting in multiple expert score arrays, wherein each expert score array is not completely identical, and the first preset number is less than the number of expert clients; calculating the variance of each expert score array; selecting the average of the expert scores in the expert score array with the smallest variance as the comprehensive expert score of the aforementioned state assessment indicator; and normalizing the comprehensive expert scores of each of the aforementioned state assessment indicators to obtain the first score for each of the aforementioned state assessment indicators. The manipulation data is input into a pre-built hybrid deep learning model to obtain the second score of each of the state evaluation metrics output by the hybrid deep learning model; The first and second scores of each of the aforementioned state evaluation indicators are input into the equilibrium game theory model to obtain the optimal scores of each of the aforementioned state evaluation indicators output by the equilibrium game theory model. The equilibrium game theory model is a game theory model constructed based on a dynamic game function. The dynamic game function is used to indicate that the optimal score of each of the aforementioned state evaluation indicators is equal to the weighted sum of the first score and the second score, wherein the weight of the first score is the first dynamic weight coefficient, and the weight of the second score is the second dynamic weight coefficient. The step of inputting the first and second scores of each of the state evaluation indicators into an equilibrium game theory model to obtain the optimal scores of each of the state evaluation indicators output by the equilibrium game theory model includes: calculating the first and second fixed-weight coefficients of each of the state evaluation indicators based on the deviation minimization idea of ​​game theory and the differential property of matrices; calculating the first dynamic weight coefficient of the state evaluation indicator based on the first score of the state evaluation indicator, a preset equilibrium function correction coefficient, and the first fixed-weight coefficient; calculating the second dynamic weight coefficient of the state evaluation indicator based on the second score of the state evaluation indicator, the equilibrium function correction coefficient, and the second fixed-weight coefficient; and using the dynamic game function to calculate the optimal scores of each of the state evaluation indicators.

2. The evaluation method for the operating mechanism according to claim 1, characterized in that, The collection of operating data from the operating mechanism to be evaluated includes: Obtain a set of candidate indicators, which are divided into energy storage unit indicator set, electric drive unit indicator set, motor unit indicator set and transmission unit indicator set according to the organizational unit; Based on the set of candidate indicators, an indicator relationship network is obtained, wherein each candidate indicator is a node, and the directed edges between nodes point from the upper-level indicator to the lower-level indicator. Based on the aforementioned indicator relationship network, the highest-level indicator set is obtained as the first candidate indicator set. Calculate the Pearson correlation coefficient between each candidate indicator and the reference candidate indicator, and select the set of candidate indicators that meet the preset correlation conditions as the second candidate indicator set. The reference candidate indicators include click angle, and the correlation conditions include a correlation coefficient greater than the preset threshold. Obtain the intersection of the first candidate indicator set and the second candidate indicator set as the evaluation indicator set, which includes multiple indicators to be evaluated.

3. The evaluation method for the operating mechanism according to claim 1, characterized in that, The hybrid deep learning model is constructed based on convolutional neural networks (CNN), gated recurrent units (GRU), and attention networks with self-attention mechanisms.

4. The evaluation method for the operating mechanism according to claim 3, characterized in that, The process of assigning the second score to each state evaluation metric output by the hybrid deep learning model includes: The CNN performs convolution operations on the input manipulation data to extract local features of the manipulation data; and performs pooling operations on the local features of the manipulation data to obtain the pooled features of the manipulation data. The GRU performs cyclic processing on the pooling features of the manipulation data to obtain the cyclic features of the manipulation data; The Attention network performs weighted processing on the cyclic features of the manipulation data based on a self-attention mechanism, generates a second score for each of the state evaluation indicators, and outputs it.

5. An evaluation device for an operating mechanism, characterized in that, include: The data acquisition unit is used to collect the operating data of the operating mechanism to be evaluated. The operating data includes a sequence of index values ​​for multiple status evaluation indicators. An expert scoring unit is used to input the operation data into an expert system, obtain a set of expert scores for each state evaluation indicator output by the expert system, wherein the expert system includes multiple expert clients, each expert client is bound to an expert user, and the set of expert scores includes the expert scores uploaded by each expert user through the expert client; based on the set of expert scores for each state evaluation indicator, a first score for each state evaluation indicator is obtained; wherein, when the expert scoring unit obtains the first score for each state evaluation indicator based on the set of expert scores for each state evaluation indicator, it specifically performs the following: each time, it selects a first preset number of expert scores from the set of expert scores to form an expert score array, resulting in multiple expert score arrays, wherein each expert score array is not completely identical, and the first preset number is less than the number of expert clients; calculates the variance of each expert score array; selects the average of the expert scores in the expert score array with the smallest variance as the comprehensive expert score for the state evaluation indicator; and normalizes the comprehensive expert scores for each state evaluation indicator to obtain the first score for each state evaluation indicator. An artificial intelligence scoring unit is used to input the manipulation data into a pre-built hybrid deep learning model and obtain the second score of each of the state evaluation indicators output by the hybrid deep learning model. The game unit is used to input the first and second scores of each of the state evaluation indicators into the equilibrium game theory model to obtain the optimal scores of each of the state evaluation indicators output by the equilibrium game theory model. The equilibrium game theory model is a game theory model constructed based on a dynamic game function. The dynamic game function is used to indicate that the optimal score of each of the state evaluation indicators is equal to the weighted sum of the first score and the second score, wherein the weight of the first score is a first dynamic weight coefficient, and the weight of the second score is a second dynamic weight coefficient. The game theory unit, used to input the first and second scores of each of the state evaluation indicators into the equilibrium game theory model, specifically obtains the optimal scores of each of the state evaluation indicators output by the equilibrium game theory model by: calculating the first and second fixed-weight coefficients of each of the state evaluation indicators based on the deviation minimization idea of ​​game theory and the differential property of matrices; calculating the first dynamic weight coefficient of the state evaluation indicator based on the first score of the state evaluation indicator, the preset equilibrium function correction coefficient, and the first fixed-weight coefficient; calculating the second dynamic weight coefficient of the state evaluation indicator based on the second score of the state evaluation indicator, the equilibrium function correction coefficient, and the second fixed-weight coefficient; and using the dynamic game function to calculate the optimal scores of each of the state evaluation indicators.

6. An evaluation device for an operating mechanism, characterized in that, include: Memory and processor; The memory is used to store programs; The processor is configured to execute the program to implement the various steps of the evaluation method for the operating mechanism as described in any one of claims 1 to 4.

7. A readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements each step of the evaluation method for the operating mechanism as described in any one of claims 1 to 4.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements each step of the evaluation method for the operating mechanism as described in any one of claims 1 to 4.

Images