An engine coupling operation parameter optimization control system

Abstract

The application belongs to the technical field of intelligent manufacturing, and discloses an engine coupling operation parameter optimization control system; the system comprises: a parameter acquisition module, which is used for acquiring coupling operation parameters and coupling process parameters; a parameter classification module, which is used for dividing the coupling process parameters into a groups and matching each group with different parameters in the coupling operation parameters; a parameter identification module, which is used for performing abnormal analysis on the coupling operation parameters, identifying risk parameters in the coupling operation parameters, acquiring coupling process parameters corresponding to the risk parameters, and marking the coupling process parameters as detection parameters; a parameter detection module, which is used for performing abnormal detection on the detection parameters and identifying to-be-adjusted parameters in the detection parameters; and a parameter optimization module, which is used for formulating a process optimization strategy and optimizing the to-be-adjusted parameters; the application can effectively improve the overall performance and production efficiency of the coupling, and realize higher fuel economy and durability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent manufacturing, more specifically, the present application relates to an engine coupling operation parameter optimization control system. BACKGROUND

[0002] In the automotive powertrain system, the engine coupling is a key transmission component between the engine and the transmission, its performance directly affects the power output, fuel economy and durability of the vehicle; the running state of the coupling will be affected by many factors such as material properties, manufacturing process, assembly accuracy and vehicle operating conditions, therefore, in the development and production process of the coupling, it is essential to optimize its process parameters; currently, the traditional coupling manufacturing process mainly relies on experience and standardized parameter setting, and the analysis of test data in the actual running process is insufficient, resulting in some products may show low efficiency, high noise, short fatigue life and other problems under actual working conditions; in addition, the production of the coupling involves multiple processes, including material selection, heat treatment, precision machining and assembly, and small deviations in process parameters will affect the quality of the final product; therefore, an intelligent parameter control system is urgently needed to improve the overall performance and production efficiency of the coupling.

[0003] In recent years, the rise of intelligent manufacturing has brought new opportunities for the production of automotive parts; by using advanced sensor technology, data acquisition and analysis technology and automatic control system, real-time monitoring and optimization of parameters can be achieved; not only can improve the accuracy and efficiency of the production process, but also can significantly shorten the product development cycle and reduce production costs; for example, the patent with publication number CN118795853A discloses a processing process optimization system for automobile shock pad mold; again, the patent with publication number CN116165988B discloses a production quality control method and system for automobile center console; both realize the optimization control of process parameters in the production process of automobile parts.

[0004] However, the existing technology mainly relies on preset process parameters, quality standards or recommended models for process parameter optimization, and does not guide process adjustment through actual product running parameters, resulting in the optimization results not completely matching the actual use, thereby reducing the adaptability of process parameter optimization; at the same time, the existing technology relies on fixed optimization strategies, such as quality score analysis, fixed processing parameter recommendation, etc., and fails to dynamically adjust the optimization scheme combined with actual running parameters, resulting in reducing the accuracy and flexibility of the process parameter optimization process.

[0005] In view of this, the present application proposes an engine coupling operation parameter optimization control system to solve the above problems. SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art, in order to achieve the above-mentioned purposes, the present application provides the following technical scheme: an engine coupling operation parameter optimization control system, comprising:

[0007] A parameter acquisition module is configured to acquire coupling operation parameters and coupling process parameters.

[0008] A parameter classification module is configured to divide the coupling process parameters into a groups, and match each group with different parameters in the coupling operation parameters.

[0009] A parameter identification module is configured to perform abnormal analysis on the coupling operation parameters, calculate a parameter abnormality coefficient, identify a risk parameter in the coupling operation parameters based on the parameter abnormality coefficient, and acquire a coupling process parameter corresponding to the risk parameter according to a group corresponding to the risk parameter, and mark the coupling process parameter as a detection parameter.

[0010] A parameter detection module is configured to acquire historical parameters, construct an abnormality detection architecture based on the historical parameters, perform abnormality detection on the detection parameter according to the abnormality detection architecture, and identify a to-be-adjusted parameter in the detection parameter.

[0011] A parameter optimization module is configured to formulate a process optimization strategy, generate a corresponding optimization value for each to-be-adjusted parameter using the process optimization strategy, and optimize the to-be-adjusted parameter according to the optimization value.

[0012] Further, the coupling operation parameters are dynamic characteristics involved in the working process of the coupling, and the coupling process parameters are operation parameters involved in the manufacturing process of the coupling.

[0013] The step of dividing the coupling process parameters into a groups comprises:

[0014] Step S101: using a pre-trained word embedding model, converting each parameter in the coupling process parameters into a corresponding text vector;

[0015] Step S102: regarding each text vector as a node;

[0016] Step S103: selecting a nodes as center points, and regarding nodes that are not center points as sample points, a being the number of parameters in the coupling operation parameters;

[0017] Step S104: constructing a groups corresponding to the a center points, and calculating sample distances from each sample point to each center point in turn;

[0018] Step S105: preset a distance threshold, compare each sample distance corresponding to each sample point with the distance threshold respectively, mark the center point corresponding to the sample distance with a value less than the distance threshold as a classification point, and do not mark the sample distance with a value greater than or equal to the distance threshold, and sequentially divide each sample point into a group corresponding to the classification point;

[0019] Step S106: calculate the new center of each group, and replace the center point of each group with the corresponding new center;

[0020] Step S107: repeat steps S104-S106 until the new center of each group calculated in step S106 is consistent with the corresponding new center calculated in the last cycle, and the cycle is ended, and a group is obtained.

[0021] Further, in step S103, the step of screening a node as a center point comprises:

[0022] Step S201: randomly select a node as a center point;

[0023] Step S202: calculate the sample distance from each sample point to each center point, compare the sample distances corresponding to the same sample points, and take the smallest sample distance as the minimum distance of the corresponding sample point;

[0024] Step S203: square the minimum distance of each sample point to obtain the distance square;

[0025] Step S204: add the distance square of each sample point to obtain the distance square sum, and divide the distance square of each sample point by the distance square sum to obtain the screening probability corresponding to each sample point;

[0026] Step S205: select a node as a center point according to the screening probability;

[0027] Step S206: repeat steps S202-S205 until a node is selected, and the cycle is ended;

[0028] In step S104, the method for calculating the sample distance is: calculating the cosine similarity between the sample point and the center point, and taking the reciprocal of the cosine similarity as the sample distance;

[0029] In step S106, the method for calculating the new center comprises:

[0030] Count the number of nodes in each group and mark it as the number of nodes; add the nodes in each group in turn, and then divide by the corresponding number of nodes to obtain the average point corresponding to each group; calculate the sample distance of each node in each group to the corresponding average point and mark it as the average distance; compare the average distances of the corresponding groups, and take the node corresponding to the smallest average distance as the new centroid of the corresponding group.

[0031] Further, the method of screening nodes according to the screening probability in step S205 includes:

[0032] Sort each node according to the corresponding screening probability from high to low to generate a node sorting table; according to the node sorting table in ascending order, add the screening probability of each node to the screening probability of the corresponding previous node in turn to obtain the cumulative probability of each node; calculate the probability range of each node according to the cumulative probability of each node; generate a random number from the interval [0, 1], and mark the probability range where the random number falls as the screening range to screen the node corresponding to the screening range; wherein, mark a node as the current node, and the previous node of the current node is all the nodes in front of the current node in the node sorting table; the maximum value of the probability range corresponding to the current node is, the cumulative probability corresponding to the current node, and the minimum value of the probability range corresponding to the current node is the cumulative probability corresponding to the node in front of the current node in the node sorting table;

[0033] The method of matching each group with different parameters in the coupling operating parameter includes:

[0034] Each parameter in the coupling operating parameter is converted into a corresponding text vector using a pre-trained word embedding model, and is marked as an operating vector; the sample distance of each operating vector to the center point of each group is calculated and marked as operating distance; sort each operating distance corresponding to each operating vector from small to large to generate a distance sorting table corresponding to each operating vector; mark each operating distance in the first position of each distance sorting table as a matching distance, and match each operating vector with the group corresponding to the matching distance, and match each group with the parameter corresponding to the operating vector.

[0035] Further, the method of calculating the parameter abnormality coefficient includes:

[0036] The preset parameter threshold includes a threshold corresponding to each parameter in the coupling operation parameter; each parameter in the coupling operation parameter and the corresponding threshold are taken as a calculation set, and the calculation set corresponds to the parameter in the coupling operation parameter one by one; each calculation set is input into the trained coefficient calculation model in turn, and the corresponding parameter abnormality coefficient is calculated; wherein the parameter abnormality coefficient is a matrix of 1x a, the elements in the matrix correspond to the parameters in the coupling operation parameter one by one, and the coefficient calculation model is a deep neural network model; the training process of the coefficient calculation model includes:

[0037] A group of data sets are collected in advance, each data set includes a group of different calculation sets, the b groups of data sets are all different, the corresponding parameter abnormality coefficients are set for the b groups of data sets, b is an integer greater than 1, the data set and the corresponding parameter abnormality coefficient are converted into a corresponding group of feature vectors; each group of feature vectors is taken as the input of the coefficient calculation model, the coefficient calculation model takes a group of predicted parameter abnormality coefficients corresponding to each data set as the output, takes the actual parameter abnormality coefficient corresponding to each data set as the prediction target, and the actual parameter abnormality coefficient is the parameter abnormality coefficient corresponding to the data set set in advance; the sum of the prediction errors of all data sets is minimized as the training target; the coefficient calculation model is trained until the sum of the prediction errors converges, and the training is stopped;

[0038] The method for identifying the risk parameters in the coupling operation parameter includes:

[0039] An abnormal threshold is preset, each element in the parameter abnormality coefficient is compared with the abnormal threshold; the elements with a value greater than or equal to the abnormal threshold are marked as abnormal elements, and the elements with a value less than the abnormal threshold are not marked; the parameters corresponding to the abnormal elements in the coupling operation parameter are taken as the risk parameters.

[0040] Further, the historical parameters are normal parameters obtained at historical time, and the normal parameters are coupling process parameters in a normal state;

[0041] The abnormal detection architecture includes d detection sub-architectures, d is the number of parameters in the coupling process parameters, and the detection sub-architecture corresponds to the parameters in the coupling process parameters one by one; the detection sub-architecture corresponding to the detection parameter in the abnormal detection architecture is obtained and marked as a running sub-architecture; the abnormal detection step of the corresponding detection parameter according to one running sub-architecture includes:

[0042] Step S301: marking the detection parameter to be detected as a current parameter, screening the parameters corresponding to the current parameter from the historical parameters and marking them as reference parameters, and collectively marking the reference parameters and the current parameters as analysis parameters;

[0043] Step S302: Obtain the neighborhood parameters corresponding to each analysis parameter, and subtract each analysis parameter from the corresponding neighborhood parameter to obtain the parameter difference corresponding to each analysis parameter; the neighborhood parameter corresponding to the analysis parameter is the remaining r analysis parameters, and r is the number of reference parameters;

[0044] Step S303: Obtain the standard span of each analysis parameter according to the parameter difference;

[0045] Step S304: The neighborhood parameter corresponding to the parameter difference in the top y positions in the element sorting table is used as the standard parameter of the analysis parameter corresponding to the difference set, and 1 < y < r + 1;

[0046] Step S305: According to the parameter difference and the standard span, the adjacent span between each analysis parameter and the corresponding standard parameter is calculated in sequence, and the regional distribution degree corresponding to each analysis parameter is calculated.

[0047] Step S306: Based on the regional distribution degree, the outlier coefficient of the current parameter is calculated, and the outlier coefficient is compared with the preset identification coefficient. If the outlier coefficient is greater than the identification coefficient, the current parameter is marked as a parameter to be adjusted. If the outlier coefficient is less than or equal to the identification coefficient, the current parameter is not marked.

[0048] Further, in step S303, the method for obtaining the standard span includes:

[0049] The parameter difference of each analysis parameter is used as a set of difference sets, and the difference set corresponds to the analysis parameter one by one. The parameter differences in each difference set are sorted from small to large to generate a difference sorting table corresponding to each difference set. The parameter difference in the yth position in each difference sorting table is used as the standard span of the analysis parameter corresponding to the corresponding difference set.

[0050] In step S305, the method for calculating the adjacent span includes:

[0051] The parameter difference between each analysis parameter and the corresponding standard parameter is marked as an adjacent difference. One adjacent difference and one standard span corresponding to each analysis parameter are used as a set of span sets, that is, each set of span sets includes one adjacent difference and one standard span. The adjacent difference and the standard span in each span set are collectively referred to as an analysis span. Each analysis span in each span set is compared respectively, and the analysis span with the largest value is used as the adjacent span between the corresponding analysis parameter and the corresponding standard parameter.

[0052] The method for calculating the regional distribution degree includes:

[0053] Add the adjacent spans corresponding to each analysis parameter in sequence to obtain a comprehensive span corresponding to each analysis parameter; multiply the reciprocal of the comprehensive distance corresponding to each analysis parameter by y to obtain a regional distribution degree corresponding to each analysis parameter;

[0054] The method for calculating the outlier coefficient in step S306 includes:

[0055] Obtain the standard parameter corresponding to the real-time parameter and mark it as a calculation parameter; divide the regional distribution degree corresponding to each calculation parameter by the regional distribution degree corresponding to the real-time element to obtain a relative distribution degree corresponding to each calculation parameter; add the relative densities of each evaluation element in sequence and then divide by y to obtain the outlier coefficient corresponding to the real-time parameter.

[0056] Further, the step of generating a corresponding optimization value for each to-be-adjusted parameter includes:

[0057] Step S401: Obtain a parameter range, and construct m optimization sets according to the parameter range, each optimization set including adjustment values corresponding to each to-be-adjusted parameter;

[0058] Step S402: Define an initialization set, the initialization set including a candidate solution, an iteration threshold, a removal strategy set, a reconstruction strategy set, a strategy weight set, and an update factor, and set the iteration number to 0; wherein the candidate solution is one of the m optimization sets;

[0059] Step S403: Calculate a removal selection probability of each removal strategy in the removal strategy set and a reconstruction selection probability corresponding to each reconstruction strategy in the reconstruction strategy set according to the strategy weights in the strategy weight set;

[0060] Step S404: Select a removal strategy from the removal strategy set according to the removal selection probability, and obtain the strategy weight corresponding to the removal strategy from the strategy weight set and mark it as a removal weight;

[0061] Step S405: Perform a removal operation on the candidate solution using the removal strategy to obtain a partial solution;

[0062] Step S406: Select a reconstruction strategy from the reconstruction strategy set according to the reconstruction selection probability, and obtain the strategy weight corresponding to the reconstruction strategy from the strategy weight set and mark it as a reconstruction weight;

[0063] Step S407: Perform a reconstruction operation on the partial solution using the reconstruction strategy to obtain a reconstructed solution;

[0064] Step S408: Calculate the optimization effect of the reconstructed solution, calculate a corresponding update probability based on the optimization effect, and determine whether to update the candidate solution to the reconstructed solution according to the update probability;

[0065] Step S409: adjust the removal weight and the reconstruction weight respectively according to the update result;

[0066] Step S410: determine whether the iteration number is less than the iteration threshold, if yes, increase the iteration number by one and return to step S403, if not, enter step S411;

[0067] Step S411: obtain the optimization set corresponding to the candidate solution, and take each adjustment value in the optimization set as the optimization value of the corresponding to-be-adjusted parameter.

[0068] Further, in step S401, the parameter range includes a normal range corresponding to each to-be-adjusted parameter, the maximum value of the normal range corresponding to each to-be-adjusted parameter is the normal parameter with the largest value in the corresponding historical parameter, and the minimum value of the normal range corresponding to each to-be-adjusted parameter is the normal parameter with the smallest value in the corresponding historical parameter; a value is randomly selected from each normal range, and m sets of optimization sets are constructed, and the m sets of optimization sets are different from each other.

[0069] In step S403, the method for calculating the removal selection probability is: adding the strategy weights corresponding to each removal strategy in turn to obtain a removal total value; and dividing the strategy weights corresponding to each removal strategy by the removal total value to obtain the removal selection probability corresponding to each removal strategy.

[0070] The method for calculating the reconstruction selection probability is: adding the strategy weights corresponding to each reconstruction strategy in turn to obtain a reconstruction total value; and dividing the strategy weights corresponding to each reconstruction strategy by the reconstruction total value to obtain the reconstruction selection probability corresponding to each reconstruction strategy.

[0071] In step S404, the method for selecting one removal strategy from the removal strategy set is consistent with the method for selecting a node according to the screening probability in step S205.

[0072] In step S406, the method for selecting one reconstruction strategy from the reconstruction strategy set is consistent with the method for selecting a node according to the screening probability in step S205.

[0073] Further, in step S408, the method for calculating the optimization effect of the reconstruction solution includes:

[0074] An optimization set corresponding to the reconstruction solution is obtained and is marked as a current set; a value of each to-be-adjusted parameter in the coupling process parameter is replaced by a corresponding adjustment value in the current set, and the coupling process parameter after the replacement is marked as an optimization process parameter; the optimization process parameter, the coupling process parameter, and the coupling operation parameter are taken as prediction parameters, and the prediction parameters are input into the trained parameter prediction model to predict a corresponding optimization operation parameter; the training process of the parameter prediction model is consistent with the training process of the coefficient calculation model, and both are deep neural network models; the optimization operation parameter and the parameter threshold value are input into the coefficient calculation model to calculate a corresponding parameter anomaly coefficient and mark the parameter anomaly coefficient as an optimization anomaly coefficient; each element in the optimization anomaly coefficient is compared with the anomaly threshold value to identify an anomaly element in the optimization anomaly coefficient; a number of the anomaly elements in the optimization anomaly coefficient is counted and marked as a first anomaly number; a number of the anomaly elements in the parameter anomaly coefficient is counted and marked as a second anomaly number; the second anomaly number is subtracted by the first anomaly number to obtain an optimization effect.

[0075] The method for calculating the update probability is as follows: an iteration number corresponding to the current iteration process is obtained and is marked as a current number; the current number is multiplied by a preset influence factor and then multiplied by an update factor to obtain a current factor; an optimization effect of the candidate solution is calculated, the optimization effect of the candidate solution is marked as a candidate effect, and an optimization effect of the reconstruction solution is marked as a reconstruction effect; the candidate effect is compared with the reconstruction effect; if the candidate effect is less than or equal to the reconstruction effect, the corresponding update probability is 1; if the candidate effect is greater than the reconstruction effect, the reconstruction effect is subtracted by the candidate effect and then divided by the current factor to obtain an update coefficient; the natural constant is taken as a base number, and the update coefficient is taken as an index to perform an index operation to obtain the update probability.

[0076] The method for determining whether the candidate solution is updated to the reconstruction solution includes:

[0077] If the update probability is 1, the candidate solution is updated to the reconstruction solution; if the update probability is not 1, an update range [0, k] and a reserved range (k, 1] are set, k is the update probability; a judgment value is randomly generated from the interval [0, 1], if the judgment value is in the update range, the candidate solution is updated to the reconstruction solution, if the judgment value is in the reserved range, the candidate solution is not updated to the reconstruction solution.

[0078] In the step S409, the method for adjusting the removal weight and the reconstruction weight respectively includes:

[0079] The adjustment effect is obtained by subtracting the candidate effect from the reconstruction effect; the adjustment degree is obtained by multiplying the adjustment effect by the current factor; and the removal weight and the reconstruction weight are adjusted according to the adjustment degree.

[0080] The technical effects and advantages of the engine coupling operation parameter optimization control system of the present application are as follows:

[0081] By classifying the coupling process parameters and matching them with the coupling operation parameters, the association between the coupling process parameters and the coupling operation parameters is established. The deep technology is used for abnormal analysis and risk identification of the coupling operation parameters, so as to accurately find the potential problems existing in the coupling operation process. The abnormal detection architecture is constructed to perform real-time abnormal detection and identification on the detected parameters, so as to timely find the parameters that need to be optimized and adjusted, and improve the pertinence of the optimization effect. The dynamic process optimization strategy is used to generate corresponding optimization values for different to-be-adjusted parameters, so as to realize flexible optimization of the process parameters and effectively improve the overall performance and production efficiency of the coupling. Not only the dynamic performance and manufacturing precision of the coupling are improved, but also the development cycle is shortened, higher fuel economy and durability are realized, and the overall performance of the automobile power system is enhanced. BRIEF DESCRIPTION OF DRAWINGS

[0082] Figure 1 A flowchart of the engine coupling operation parameter optimization control system of the present application is shown in Figure 1.

[0083] Figure 2 A schematic diagram of the engine coupling operation parameter optimization control system of the present application is shown in Figure 2. DETAILED DESCRIPTION

[0084] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0085] Embodiment 1

[0086] Please refer to Figure 1 and Figure 2 , the engine coupling operation parameter optimization control system described in the present embodiment includes a parameter acquisition module, a parameter classification module, a parameter identification module, a parameter detection module, and a parameter optimization module. Each module is connected through wired and / or wireless means to realize data transmission between modules.

[0087] The parameter acquisition module is used to acquire the coupling operation parameters and the coupling process parameters.

[0088] The coupling operation parameter is a dynamic characteristic involved in the working process of the coupling, and is used to describe the performance of the coupling in the working state. The coupling operation parameter is, for example, the maximum rotating speed, the maximum torque transmission capacity, the maximum vibration, the maximum temperature, and the like. If the maximum rotating speed does not reach the rotating speed requirement, the process parameters such as the quenching temperature and the tempering time in the heat treatment stage, and the grinding accuracy in the finishing stage should be optimized and controlled to improve the hardness and the balance of the coupling. If the maximum torque transmission capacity does not reach the torque transmission capacity requirement, the process parameters such as the carburizing temperature and the carburizing depth in the carburizing heat treatment stage, and the shot pressure in the surface strengthening stage should be optimized and controlled to improve the load capacity and the fatigue strength of the coupling. If the maximum vibration exceeds the vibration limit, the process parameters such as the tool feed rate in the turning processing stage, and the dynamic balance level in the dynamic balance correction stage should be optimized and controlled to improve the coaxiality and the rotor mass distribution uniformity of the coupling. If the maximum temperature exceeds the temperature limit, the process parameters such as the carburizing temperature and the tempering temperature in the heat treatment stage, and the shaft hole interference amount in the surface treatment stage should be optimized and controlled to improve the high-temperature resistance and the fitting stability in the high-temperature environment of the coupling.

[0089] The coupling process parameter is an operation parameter involved in the manufacturing process of the coupling, and is used to describe the conditions that need to be controlled and optimized in the production and processing of the coupling, and directly affects the machining precision, the mechanical properties and the service life of the coupling. The coupling process parameter is, for example, the quenching temperature, the carburizing temperature, the shot pressure, the grinding accuracy, and the like.

[0090] The parameter classification module is used to divide the coupling process parameters into a groups, and match each group with a different parameter in the coupling operation parameters.

[0091] The step of dividing the coupling process parameters into a groups includes:

[0092] Step S101: using a pre-trained word embedding model (such as Word2Vec, GloVe, BERT, etc.), each parameter in the coupling process parameters is converted into a corresponding text vector;

[0093] Step S102: each text vector is taken as a node;

[0094] Step S103: a is selected as the center point, and the nodes that are not taken as the center point are taken as the sample points, and a is the number of parameters in the coupling operation parameters;

[0095] Step S104: a groups are constructed according to the a center points, and the sample distance of each sample point to each center point is calculated in turn;

[0096] Step S105: presetting a distance threshold, comparing each sample distance corresponding to each sample point with the distance threshold respectively, marking the center point corresponding to the sample distance with a value less than the distance threshold as a classification point, and not marking the sample distance with a value greater than or equal to the distance threshold, and sequentially dividing each sample point into a group corresponding to the classification point;

[0097] Step S106: calculating a new centroid corresponding to each group, and replacing the center point of each group with the corresponding new centroid;

[0098] Step S107: repeating steps S104-S106 until the new centroid of each group calculated in step S106 is consistent with the corresponding new centroid calculated in the last cycle, and the cycle ends, obtaining a groups.

[0099] In the above step S103, the step of screening a nodes as center points includes:

[0100] Step S201: randomly selecting a node as a center point;

[0101] Step S202: calculating the sample distance from each sample point to each center point, comparing the sample distances corresponding to the same sample points, and taking the smallest sample distance as the minimum distance of the corresponding sample point;

[0102] Step S203: sequentially squaring the minimum distance of each sample point to obtain the distance square;

[0103] Step S204: sequentially adding the distance squares of each sample point to obtain the distance square sum, and dividing the distance square of each sample point by the distance square sum to obtain the screening probability corresponding to each sample point;

[0104] Step S205: selecting a node as a center point according to the screening probability;

[0105] Step S206: repeating steps S202-S205 until a nodes are selected, and the cycle ends.

[0106] In the above step S104, the method for calculating the sample distance is: calculating the cosine similarity between the sample point and the center point, and taking the reciprocal of the cosine similarity as the sample distance; the calculation method of the cosine similarity is the prior art, which is not described in detail here.

[0107] In the above step S106, the method for calculating the new centroid includes:

[0108] Count the number of nodes in each group and mark it as the number of nodes; add the nodes in each group in turn, and then divide by the corresponding number of nodes to obtain the average point corresponding to each group; calculate the sample distance of each node in each group to the corresponding average point and mark it as the average distance; compare the average distances of the corresponding groups, and take the node corresponding to the smallest average distance as the new centroid of the corresponding group.

[0109] In the step S205, the method for screening nodes according to the screening probability comprises:

[0110] Sort each node according to the corresponding screening probability from high to low to generate a node sorting table; according to the node sorting table in ascending order, add the screening probability of each node to the screening probability of the corresponding previous node in turn to obtain the cumulative probability of each node; calculate the probability range of each node according to the cumulative probability of each node; generate a random number from the interval [0, 1], and mark the probability range where the random number falls as the screening range to screen the node corresponding to the screening range; wherein, mark a node as the current node, and the previous node of the current node is all nodes in front of the current node in the node sorting table; the maximum value of the probability range corresponding to the current node is, the cumulative probability corresponding to the current node is, and the minimum value of the probability range corresponding to the current node is, the cumulative probability corresponding to the node in front of the current node in the node sorting table.

[0111] The method for matching each group with different parameters in the coupling operating parameters comprises:

[0112] Each parameter in the coupling operating parameters is converted into a corresponding text vector using a pre-trained word embedding model, and is marked as an operating vector; the sample distance of each operating vector to each group center point is calculated and marked as operating distance; each operating vector corresponding operating distance is sorted from small to large to generate a distance sorting table corresponding to each operating vector; the operating distance in the first position of each distance sorting table is marked as a matching distance, and each operating vector is matched with the group corresponding to the matching distance, and each group is matched with the parameter corresponding to the operating vector.

[0113] The parameter identification module is used for abnormal analysis of the coupling operating parameters, calculation of the parameter abnormality coefficient, identification of the risk parameter in the coupling operating parameters based on the parameter abnormality coefficient, and acquisition of the coupling process parameter corresponding to the risk parameter according to the group corresponding to the risk parameter, and marking it as a detection parameter.

[0114] The method for calculating the parameter abnormality coefficient comprises:

[0115] The preset parameter threshold includes a threshold corresponding to each parameter in the coupling operation parameter, and the parameter threshold is preset by a person skilled in the art according to the performance requirement of the coupling; each parameter in the coupling operation parameter is taken as a calculation set corresponding to the threshold, and the calculation set corresponds to the parameter in the coupling operation parameter in one-to-one manner; each calculation set is sequentially input into the trained coefficient calculation model, and a corresponding parameter anomaly coefficient is calculated; the parameter anomaly coefficient is a 1×a matrix, the elements in the matrix correspond to the parameters in the coupling operation parameter in one-to-one manner, and the coefficient calculation model is a deep neural network model.

[0116] The training process of the coefficient calculation model includes:

[0117] B sets of data sets are collected in advance, each set of data sets includes a set of a different calculation set, the b sets of data sets are all different, the corresponding parameter anomaly coefficients are set for the b sets of data sets, b is an integer greater than 1, and the data set and the corresponding parameter anomaly coefficient are converted into a corresponding set of feature vectors; the parameter anomaly coefficient corresponding to the data set is collected by a person skilled in the art in the process of calculating the historical parameter anomaly coefficient, a set of calculation sets in each data set is analyzed in sequence according to the actual situation, the abnormal degree corresponding to each calculation set is evaluated, a corresponding matrix is constructed according to the abnormal degree of the a sets of calculation sets, the matrix is taken as the parameter anomaly coefficient of the corresponding data set, and the corresponding parameter anomaly coefficients are sequentially set for the b sets of data sets;

[0118] Each set of feature vectors is taken as the input of the coefficient calculation model, the coefficient calculation model takes a set of predicted parameter anomaly coefficients corresponding to each data set as the output, takes the actual parameter anomaly coefficient corresponding to each data set as the prediction target, and the actual parameter anomaly coefficient is the parameter anomaly coefficient corresponding to the data set set in advance; the training target is to minimize the sum of prediction errors of all data sets; wherein the calculation formula of the prediction error is η w =(θ w -ε w ) 2 , wherein η w is the prediction error, w is the group number of the feature vector corresponding to the data set, θ w is the predicted parameter anomaly coefficient corresponding to the wth data set, and ε w is the actual parameter anomaly coefficient corresponding to the wth data set; the coefficient calculation model is trained until the sum of prediction errors converges.

[0119] The method for identifying the risk parameter in the coupling operation parameter includes:

[0120] The preset abnormal threshold is preset by a person skilled in the art according to actual conditions; each element in the parameter abnormality coefficient is compared with the abnormal threshold; the element with a value greater than or equal to the abnormal threshold is marked as an abnormal element, and the element with a value less than the abnormal threshold is not marked; and the parameter corresponding to the abnormal element in the coupling running parameter is taken as a risk parameter.

[0121] The parameter detection module is configured to acquire historical parameters, construct an abnormality detection architecture based on the historical parameters, perform abnormality detection on the detection parameters according to the abnormality detection architecture, and identify the to-be-adjusted parameters in the detection parameters.

[0122] The historical parameters are normal parameters acquired at historical moments, and the normal parameters are coupling process parameters in a normal state; in the normal state, the coupling produced does not have a risk parameter in the coupling running parameter corresponding to the coupling.

[0123] The abnormality detection architecture includes d detection sub-architectures, d is the number of parameters in the coupling process parameters, and the detection sub-architectures correspond to the parameters in the coupling process parameters one by one; a detection sub-architecture corresponding to the detection parameter in the abnormality detection architecture is acquired and marked as a running sub-architecture; and the abnormality detection step performed on the corresponding detection parameter according to one running sub-architecture includes:

[0124] Step S301: marking the detection parameter to be subjected to abnormality detection as a current parameter, screening a parameter corresponding to the current parameter from the historical parameters and marking the parameter as a reference parameter, and collectively referring to the reference parameter and the current parameter as analysis parameters;

[0125] Step S302: acquiring a neighborhood parameter corresponding to each analysis parameter, subtracting each neighborhood parameter corresponding to each analysis parameter from each analysis parameter to acquire a parameter difference corresponding to each analysis parameter; the neighborhood parameter corresponding to the analysis parameter is the remaining r analysis parameters, and r is the number of reference parameters;

[0126] Step S303: acquiring a standard span of each analysis parameter according to the parameter difference;

[0127] Step S304: taking the neighborhood parameters corresponding to the parameter differences arranged in the first y positions in each element sorting table as standard parameters of the analysis parameters corresponding to the corresponding difference set, 1 < y < r + 1, and the specific value of y is set by a person skilled in the art according to actual conditions;

[0128] Step S305: sequentially calculating the adjacent spans between each analysis parameter and each standard parameter corresponding to the analysis parameter according to the parameter difference and the standard span, and calculating a regional distribution degree corresponding to each analysis parameter;

[0129] Step S306: Based on the regional distribution degree, the outlier coefficient of the current parameter is calculated, and the outlier coefficient is compared with the preset identification coefficient. If the outlier coefficient is greater than the identification coefficient, the current parameter is marked as a parameter to be adjusted. If the outlier coefficient is less than or equal to the identification coefficient, the current parameter is not marked. The identification coefficient is preset by a person skilled in the art according to the actual situation, and the value range of the identification coefficient is [1.5, 3].

[0130] In the above step S303, the method for obtaining the standard span includes:

[0131] The parameter difference of each analysis parameter is taken as a set of difference sets, and the difference set corresponds to the analysis parameter one by one. The parameter differences in each difference set are sorted from small to large to generate a difference sorting table corresponding to each difference set. The parameter difference ranked in the yth position in each difference sorting table is taken as the standard span of the analysis parameter corresponding to the corresponding difference set.

[0132] In the above step S305, the method for calculating the adjacent span includes:

[0133] The parameter difference between each analysis parameter and the corresponding standard parameter is marked as an adjacent difference. One adjacent difference and one standard span corresponding to each analysis parameter are taken as a set of spans, that is, each set of spans includes one adjacent difference and one standard span. The adjacent difference and the standard span in each span set are collectively referred to as analysis spans. Each analysis span in each span set is compared, and the analysis span with the largest value is taken as the adjacent span between the corresponding analysis parameter and the corresponding standard parameter.

[0134] The method for calculating the regional distribution degree includes:

[0135] The adjacent spans corresponding to each analysis parameter are sequentially added to obtain the comprehensive span corresponding to each analysis parameter. The reciprocal of the comprehensive distance corresponding to each analysis parameter is multiplied by y to obtain the regional distribution degree corresponding to each analysis parameter.

[0136] In the above step S306, the method for calculating the outlier coefficient includes:

[0137] The standard parameter corresponding to the real-time parameter is obtained and marked as a calculation parameter. The regional distribution degree corresponding to each calculation parameter is divided by the regional distribution degree corresponding to the real-time element to obtain the relative distribution degree corresponding to each calculation parameter. The relative densities of each evaluation element are sequentially added and then divided by y to obtain the outlier coefficient corresponding to the real-time parameter.

[0138] The parameter optimization module is configured to formulate a process optimization strategy, generate an optimization value corresponding to each parameter to be adjusted by using the process optimization strategy, and optimize the parameter to be adjusted according to the optimization value.

[0139] The step of generating a corresponding optimization value for each parameter to be adjusted comprises:

[0140] Step S401: Obtain a parameter range, and construct m sets of optimization sets according to the parameter range, each set of optimization sets including a tuning value corresponding to each parameter to be adjusted;

[0141] Step S402: Define an initialization set, the initialization set including a candidate solution, an iteration threshold, a removal strategy set, a reconstruction strategy set, a strategy weight set and an update factor, and set the iteration number to 0; wherein the candidate solution is one of the m sets of optimization sets;

[0142] Step S403: According to the strategy weight in the strategy weight set, calculate the removal selection probability of each removal strategy in the removal strategy set, and the reconstruction selection probability corresponding to each reconstruction strategy in the reconstruction strategy set;

[0143] Step S404: According to the removal selection probability, select a removal strategy from the removal strategy set, and obtain the strategy weight corresponding to the removal strategy from the strategy weight set, and mark it as a removal weight;

[0144] Step S405: Perform a removal operation on the candidate solution using the removal strategy to obtain a partial solution;

[0145] Step S406: According to the reconstruction selection probability, select a reconstruction strategy from the reconstruction strategy set, and obtain the strategy weight corresponding to the reconstruction strategy from the strategy weight set, and mark it as a reconstruction weight;

[0146] Step S407: Perform a reconstruction operation on the partial solution using the reconstruction strategy to obtain a reconstructed solution;

[0147] Step S408: Calculate the optimization effect of the reconstructed solution, calculate the corresponding update probability based on the optimization effect, and determine whether to update the candidate solution to the reconstructed solution according to the update probability;

[0148] Step S409: Adjust the removal weight and the reconstruction weight according to the update result respectively;

[0149] Step S410: Determine whether the iteration number is less than the iteration threshold, if yes, increase the iteration number by one, and return to step S403, if not, go to step S411;

[0150] Step S411: Obtain the optimization set corresponding to the candidate solution, and take each tuning value in the optimization set as the optimization value of the corresponding parameter to be adjusted.

[0151] In the step S401, the parameter range includes a normal range corresponding to each to-be-adjusted parameter, the maximum value of the normal range corresponding to each to-be-adjusted parameter is the normal parameter with the maximum value in the corresponding historical parameter, and the minimum value of the normal range corresponding to each to-be-adjusted parameter is the normal parameter with the minimum value in the corresponding historical parameter. A value is randomly selected from each normal range to construct an optimization set, and m optimization sets are constructed, and the m optimization sets are different.

[0152] In the step S402, the removal strategy set includes multiple removal strategies, such as random removal (that is, randomly removing a part of the adjustment values corresponding to the to-be-adjusted parameters from the candidate solution), cost-driven removal (that is, removing a part of the adjustment values corresponding to the to-be-adjusted parameters that have a greater impact on the optimization effect from the candidate solution), and the like. The reconstruction strategy set includes multiple reconstruction strategies, such as random value insertion (that is, inserting a new value randomly selected from the corresponding normal range into the candidate solution for the removed adjustment value), neighborhood search insertion (that is, adjusting the removed adjustment value by a small value and inserting the adjusted value into the candidate solution), and the like. The strategy weight set includes the strategy weight corresponding to each removal strategy and reconstruction strategy. The value range of the update factor is [0, 0.5]. The strategy weight set, the update factor, and the iteration threshold are all set in advance by the person skilled in the art according to the actual situation.

[0153] It should be understood that the purpose of using the removal strategy and the reconstruction strategy is to break the existing structure by removing a part of the candidate solution, to rebuild the removed part, to avoid the search from falling into a local optimum, to enhance the exploratory nature of the process optimization strategy, and to improve the global optimization capability.

[0154] In the step S403, the method for calculating the removal selection probability is that the strategy weights corresponding to each removal strategy are added in sequence to obtain a removal total value, and the strategy weights corresponding to each removal strategy are divided by the removal total value to obtain the removal selection probability corresponding to each removal strategy.

[0155] The method for calculating the reconstruction selection probability is that the strategy weights corresponding to each reconstruction strategy are added in sequence to obtain a reconstruction total value, and the strategy weights corresponding to each reconstruction strategy are divided by the reconstruction total value to obtain the reconstruction selection probability corresponding to each reconstruction strategy.

[0156] In the step S404, the method for selecting a removal strategy from the removal strategy set is consistent with the method for selecting a node according to the screening probability in the step S205.

[0157] In the step S406, the method for selecting a reconstruction strategy from the reconstruction strategy set is consistent with the method for selecting a node according to the screening probability in the step S205.

[0158] The method for calculating the optimization effect of the reconstructed solution in step S408 includes:

[0159] An optimization set corresponding to the reconstructed solution is obtained and marked as a current set. The value of each to-be-adjusted parameter in the coupling process parameter is replaced by the corresponding adjustment value in the current set, and the coupling process parameter after the replacement is marked as an optimization process parameter. The optimization process parameter, the coupling process parameter, and the coupling operation parameter are taken as prediction parameters, and the prediction parameters are input into the trained parameter prediction model to predict a corresponding optimization operation parameter. The optimization operation parameter is a coupling operation parameter corresponding to a coupling product obtained by producing the coupling according to the optimization process parameter. The training process of the parameter prediction model is consistent with the training process of the coefficient calculation model, and both are deep neural network models. The optimization operation parameter and the parameter threshold value are input into the coefficient calculation model to calculate a corresponding parameter abnormality coefficient and mark it as an optimization abnormality coefficient. Each element in the optimization abnormality coefficient is compared with the abnormality threshold value to identify abnormal elements in the optimization abnormality coefficient. The number of abnormal elements in the optimization abnormality coefficient is counted and marked as a first abnormality number. The number of abnormal elements in the parameter abnormality coefficient is counted and marked as a second abnormality number. The second abnormality number is subtracted from the first abnormality number to obtain the optimization effect.

[0160] The method for calculating the update probability includes:

[0161] The iteration number corresponding to the current iteration process is obtained and marked as a current number. The current number is multiplied by a preset influence factor and then multiplied by an update factor to obtain a current factor. The value range of the influence factor is [0, 1], and the influence factor is preset by a person skilled in the art according to the actual situation. The optimization effect of the candidate solution is calculated, and the optimization effect of the candidate solution is marked as a candidate effect and the optimization effect of the reconstructed solution is marked as a reconstructed effect. The candidate effect and the reconstructed effect are compared. If the candidate effect is less than or equal to the reconstructed effect, the corresponding update probability is 1. If the candidate effect is greater than the reconstructed effect, the reconstructed effect is subtracted from the candidate effect and then divided by the current factor to obtain an update coefficient. The natural constant is taken as the base number and the update coefficient is taken as the index to perform the index operation to obtain the update probability.

[0162] The method for determining whether to update the candidate solution to the reconstructed solution includes:

[0163] If the update probability is 1, the candidate solution is updated to the reconstructed solution. If the update probability is not 1, an update range [0, k] and a retention range (k, 1] are set, and k is the update probability. A judgment value is randomly generated from the interval [0, 1]. If the judgment value is in the update range, the candidate solution is updated to the reconstructed solution. If the judgment value is in the retention range, the candidate solution is not updated to the reconstructed solution.

[0164] It should be appreciated that the purpose of determining whether to update the candidate solution to the reconstructed solution according to the update probability is that if the candidate solution is updated according to the better reconstructed solution only, then the local optimum will be quickly reached, and if the candidate solution is updated based on the update probability, then the local optimum can be jumped out, the global search capability is enhanced, and thus the search efficiency is improved.

[0165] In the step S409, the method of adjusting the removal weight and the reconstruction weight respectively includes:

[0166] The adjustment effect is obtained by subtracting the candidate effect from the reconstruction effect; the adjustment degree is obtained by multiplying the adjustment effect by the current factor; and the removal weight and the reconstruction weight are adjusted respectively according to the adjustment degree.

[0167] The embodiment establishes the association between the process parameters and the operation parameters of the coupling by classifying the process parameters of the coupling and matching the process parameters with the operation parameters of the coupling; performs abnormal analysis and risk identification on the operation parameters of the coupling by using deep technology, and accurately finds potential problems existing in the operation process of the coupling; constructs an abnormal detection architecture to perform real-time abnormal detection and identification on the detected parameters, so that the parameters that need to be optimized and adjusted can be found in time, and the pertinence of the optimization effect is improved; generates corresponding optimization values for different to-be-adjusted parameters by using a dynamic process optimization strategy, realizes flexible optimization of the process parameters, and effectively improves the overall performance and production efficiency of the coupling; not only improves the dynamic performance and manufacturing precision of the coupling, but also shortens the development cycle, realizes higher fuel economy and durability, and enhances the overall performance of the automobile power system.

[0168] Embodiment 2

[0169] The application also provides an electronic device. The electronic device can include one or more processors and one or more memories. The memory stores computer-readable code which, when executed by the one or more processors, can perform an engine coupling operation parameter optimization control system as described above.

[0170] The method or system according to the embodiments of the application can also be implemented by means of the architecture of the electronic device shown in the application. The electronic device can include a bus, one or more CPUs, ROM, RAM, a communication port connected to a network, input / output, a hard disk, etc. The storage device in the electronic device, such as ROM or hard disk, can store an engine coupling operation parameter optimization control system provided by the application. Further, the electronic device can also include a user interface. Of course, the architecture shown in the application is only exemplary, and when different devices are implemented, one or more components of the electronic device shown in the application can be omitted according to actual needs.

[0171] Embodiment 3

[0172] One embodiment of the present application discloses a computer readable storage medium. The computer readable storage medium stores computer readable instructions. When the computer readable instructions are run by a processor, an engine coupling operating parameter optimization control system according to the embodiment of the present application described above with reference to the accompanying drawings can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM), cache memory, and the like. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, and the like.

[0173] In addition, according to the embodiments of the present application, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, the present application provides a non-transitory machine readable storage medium storing machine readable instructions executable by a processor to perform instructions corresponding to the method steps provided by the present application, for example: an engine coupling operating parameter optimization control system. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0174] The above describes only the specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0175] Finally: the above described only the preferred embodiments of the present application, and not for limiting the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application, should be included in the protection scope of the present application.

Claims

1. An engine coupling operating parameter optimization control system characterized by, The application relates to a coupling parameter optimization method and device. The application comprises: a parameter acquisition module for acquiring coupling operation parameters and coupling process parameters; a parameter classification module for dividing the coupling process parameters into a groups and matching each group with different parameters in the coupling operation parameters; a parameter identification module for performing abnormal analysis on the coupling operation parameters, calculating a parameter abnormality coefficient, identifying a risk parameter in the coupling operation parameters based on the parameter abnormality coefficient, and acquiring the coupling process parameters corresponding to the risk parameter according to the group corresponding to the risk parameter and marking the coupling process parameters as detection parameters; a parameter detection module for acquiring historical parameters, constructing an abnormality detection framework based on the historical parameters, performing abnormality detection on the detection parameters according to the abnormality detection framework, and identifying to-be-adjusted parameters in the detection parameters; 2. The engine coupling operating parameter optimization control system of claim 1, wherein, a parameter optimization module for formulating a process optimization strategy, generating corresponding optimization values for each to-be-adjusted parameter by using the process optimization strategy, and optimizing the to-be-adjusted parameters according to the optimization values. The coupling operation parameters are dynamic characteristics involved in the working process of the coupling; the coupling process parameters are operation parameters involved in the manufacturing process of the coupling. The step of dividing the coupling process parameters into a groups comprises: Step S101: a pre-trained word embedding model is used to convert each parameter in the coupling process parameters into a corresponding text vector; Step S102: each text vector is taken as a node; Step S103: a node is selected as a center point, and a node that is not taken as the center point is taken as a sample point, and a is the number of parameters in the coupling operation parameters; Step S104: a corresponding a group is constructed according to the a center points, and the sample distance from each sample point to each center point is calculated in sequence; Step S105: a distance threshold is preset, each sample distance corresponding to each sample point is compared with the distance threshold, the center point corresponding to the sample distance with a value less than the distance threshold is marked as a classification point, and the sample distance with a value greater than or equal to the distance threshold is not marked, and each sample point is sequentially divided into a group corresponding to the classification point; Step S106: a new center of each group is calculated, and the center point of each group is replaced by the corresponding new center; 3. An engine coupling operating parameter optimization control system according to claim 2, characterized by, Step S107: steps S104-S106 are cycled until the new center of each group calculated in step S106 is consistent with the corresponding new center calculated in the last cycle, and the cycle is ended, and the a groups are acquired. In step S103, the step of selecting a node as a center point comprises: Step S201: a node is randomly selected as a center point; Step S202: the sample distance from each sample point to each center point is calculated, the sample distances corresponding to the same sample points are compared, and the sample distance with the minimum value is taken as the minimum distance of the corresponding sample point; Step S203: the minimum distance of each sample point is sequentially squared to obtain a distance square; Step S204: the distance squares of each sample point are sequentially added to obtain a distance square sum, and the distance square of each sample point is divided by the distance square sum to obtain a screening probability corresponding to each sample point. Step S205: screening a node as a center point according to a screening probability; Step S206: repeating steps S202-S205 until a node is screened, and the loop ends; In step S104, the method for calculating the sample distance is: calculating the cosine similarity between the sample point and the center point, and taking the reciprocal of the cosine similarity as the sample distance; In step S106, the method for calculating the new centroid includes: Counting the number of nodes in each group and marking it as the node number; adding the nodes in each group in turn, and then dividing by the corresponding node number to obtain the average point corresponding to each group; calculating the sample distance of each node in each group to the corresponding average point, and marking it as the average distance; comparing the average distances corresponding to the same group, and taking the node corresponding to the smallest average distance as the new centroid of the corresponding group.

4. An engine coupling operating parameter optimization control system according to claim 3, characterized by, In step S205, the method for screening a node according to a screening probability includes: Sort each node according to the corresponding screening probability from high to low to generate a node sorting table; according to the node sorting table in ascending order, add the screening probability of each node to the screening probability of the corresponding previous node in turn to obtain the cumulative probability of each node; calculate the probability range of each node according to the cumulative probability of each node; generate a random number from the interval [0, 1], and mark the probability range where the random number falls as the screening range to screen the node corresponding to the screening range; wherein, mark a node as the current node, and the previous node of the current node is all nodes in front of the current node in the node sorting table; the maximum value of the probability range corresponding to the current node is, the cumulative probability corresponding to the current node, and the minimum value of the probability range corresponding to the current node is, the cumulative probability corresponding to the node in front of the current node in the node sorting table; The method for matching each group with different parameters in the coupling running parameter includes: Using a pre-trained word embedding model, each parameter in the coupling running parameter is converted into a corresponding text vector, and is marked as a running vector; calculate the sample distance of each running vector to the center point of each group, and mark it as the running distance; sort each running distance corresponding to each running vector from small to large to generate a distance sorting table corresponding to each running vector; mark each running distance in the first place in each distance sorting table as a matching distance, and match each running vector with the group corresponding to the matching distance, and match each group with the parameter corresponding to the corresponding running vector.

5. An engine coupling operating parameter optimization control system according to claim 4, characterized by, The method for calculating the parameter anomaly coefficient includes: The preset parameter threshold value includes a threshold value corresponding to each parameter in the coupling operation parameter; each parameter in the coupling operation parameter is respectively taken as a calculation set corresponding to the threshold value, and the calculation set corresponds to the parameter in the coupling operation parameter in one-to-one manner; each calculation set is sequentially input into the trained coefficient calculation model to calculate a corresponding parameter anomaly coefficient; the parameter anomaly coefficient is a 1x a matrix, the elements in the matrix correspond to the parameters in the coupling operation parameter in one-to-one manner, and the coefficient calculation model is a deep neural network model; the training process of the coefficient calculation model includes: b sets of data sets are collected in advance, each set of data sets includes a set of different calculation sets, the b sets of data sets are all different, a corresponding parameter anomaly coefficient is set for each set of data sets, b is an integer greater than 1, each data set and the corresponding parameter anomaly coefficient are converted into a corresponding set of feature vectors; each set of feature vectors is taken as the input of the coefficient calculation model, the coefficient calculation model takes a set of predicted parameter anomaly coefficients corresponding to each set of data sets as the output, takes the actual parameter anomaly coefficient corresponding to each set of data sets as the prediction target, and the actual parameter anomaly coefficient is the parameter anomaly coefficient corresponding to the data set set in advance; the sum of the prediction errors of all data sets is minimized as the training target; the coefficient calculation model is trained until the sum of the prediction errors converges, and the training is stopped; The method for identifying the risk parameter in the coupling operation parameter includes: An abnormal threshold value is preset, each element in the parameter anomaly coefficient is compared with the abnormal threshold value; the elements with a value greater than or equal to the abnormal threshold value are marked as abnormal elements, and the elements with a value less than the abnormal threshold value are not marked; the parameters corresponding to the abnormal elements in the coupling operation parameter are taken as the risk parameters.

6. An engine coupling operating parameter optimisation control system according to claim 5, wherein, The historical parameters are normal parameters obtained at a historical moment, and the normal parameters are coupling process parameters in a normal state; The abnormal detection architecture includes d detection sub-architectures, d is the number of parameters in the coupling process parameters, and the detection sub-architecture corresponds to the parameters in the coupling process parameters in one-to-one manner; a detection sub-architecture corresponding to the detection parameter in the abnormal detection architecture is obtained and marked as a running sub-architecture; The abnormal detection step of the corresponding detection parameter according to one running sub-architecture includes: Step S301: the detection parameter to be detected is marked as a current parameter, the parameters corresponding to the current parameter are selected from the historical parameters and marked as reference parameters, and the reference parameters and the current parameter are collectively referred to as analysis parameters; Step S302: the neighborhood parameters corresponding to each analysis parameter are obtained, each analysis parameter is subtracted by each corresponding neighborhood parameter to obtain the parameter difference corresponding to each analysis parameter; the neighborhood parameters corresponding to the analysis parameters are the remaining r analysis parameters, and r is the number of reference parameters; Step S303: the standard span of each analysis parameter is obtained according to the parameter difference; Step S304: the neighborhood parameters corresponding to the parameter differences arranged in the first y positions in each element sorting table are taken as the standard parameters of the analysis parameters corresponding to the corresponding difference set, and 1 < y < r + 1. Step S305: According to the parameter difference and the standard span, the adjacent span between each analysis parameter and each corresponding standard parameter is calculated in sequence, and the regional distribution degree corresponding to each analysis parameter is calculated. Step S306: Based on the regional distribution degree, the outlier coefficient of the current parameter is calculated, and the outlier coefficient is compared with the preset identification coefficient. If the outlier coefficient is greater than the identification coefficient, the current parameter is marked as a parameter to be adjusted. If the outlier coefficient is less than or equal to the identification coefficient, the current parameter is not marked.

7. An engine coupling operating parameter optimisation control system according to claim 6, wherein, In the step S303, the method for obtaining the standard span comprises: Taking the parameter difference of each analysis parameter as a set of difference amounts, the set of difference amounts corresponds to the analysis parameter one by one; sorting the parameter differences in each set of difference amounts from small to large to generate a difference sorting table corresponding to each set of difference amounts; taking the parameter difference ranked in the yth position in each difference sorting table as the standard span of the analysis parameter corresponding to the corresponding set of difference amounts; In the step S305, the method for calculating the adjacent span comprises: Taking the parameter difference between each analysis parameter and each corresponding standard parameter as an adjacent difference; taking one adjacent difference and one standard span corresponding to each analysis parameter as a span set, that is, each span set includes one adjacent difference and one standard span; collectively taking the adjacent difference and the standard span in each span set as an analysis span, comparing each analysis span in each span set respectively, and taking the analysis span with the maximum value as the adjacent span between the corresponding analysis parameter and the corresponding standard parameter; The method for calculating the regional distribution degree comprises: Adding the adjacent spans corresponding to each analysis parameter in sequence to obtain the comprehensive span corresponding to each analysis parameter; multiplying the reciprocal of the comprehensive distance corresponding to each analysis parameter by y to obtain the regional distribution degree corresponding to each analysis parameter; In the step S306, the method for calculating the outlier coefficient comprises: Obtaining the standard parameter corresponding to the real-time parameter and marking it as a calculation parameter; dividing the regional distribution degree corresponding to each calculation parameter by the regional distribution degree corresponding to the real-time element to obtain the relative distribution degree corresponding to each calculation parameter; adding the relative densities of each evaluation element in sequence and then dividing by y to obtain the outlier coefficient corresponding to the real-time parameter.

8. An engine coupling operating parameter optimisation control system according to claim 7, wherein, The step of generating the corresponding optimization value for each parameter to be adjusted comprises: Step S401: Obtain the parameter range, and construct m optimization sets according to the parameter range, each optimization set including the adjustment value corresponding to each parameter to be adjusted; Step S402: Define an initialization set, which includes a candidate solution, an iteration threshold, a removal strategy set, a reconstruction strategy set, a strategy weight set, and an update factor, and set the iteration number to 0; wherein the candidate solution is one of the m optimization sets; Step S403: According to the strategy weight in the strategy weight set, the removal selection probability of each removal strategy in the removal strategy set is calculated, and the reconstruction selection probability corresponding to each reconstruction strategy in the reconstruction strategy set is calculated; Step S404: According to the removal selection probability, a removal strategy is selected from the removal strategy set, and the strategy weight corresponding to the removal strategy is obtained from the strategy weight set, and is marked as a removal weight; Step S405: The removal strategy is used to remove the candidate solution to obtain a partial solution; Step S406: According to the reconstruction selection probability, a reconstruction strategy is selected from the reconstruction strategy set, and the strategy weight corresponding to the reconstruction strategy is obtained from the strategy weight set, and is marked as a reconstruction weight; Step S407: The reconstruction strategy is used to reconstruct the partial solution to obtain a reconstructed solution; Step S408: The optimization effect of the reconstructed solution is calculated, the corresponding update probability is calculated based on the optimization effect, and whether the candidate solution is updated to the reconstructed solution is determined according to the update probability; Step S409: The removal weight and the reconstruction weight are adjusted according to the update result respectively; Step S410: It is judged whether the iteration number is less than the iteration threshold, if yes, the iteration number is increased by one, and the step S403 is returned, if not, the step S411 is entered; Step S411: The optimization set corresponding to the candidate solution is obtained, and each adjustment value in the optimization set is used as an optimization value of the corresponding to-be-adjusted parameter.

9. An engine coupling operating parameter optimisation control system according to claim 8, wherein, In the step S401, the parameter range includes a normal range corresponding to each to-be-adjusted parameter, the maximum value of the normal range corresponding to each to-be-adjusted parameter is the normal parameter with the largest value in the corresponding historical parameter, and the minimum value of the normal range corresponding to each to-be-adjusted parameter is the normal parameter with the smallest value in the corresponding historical parameter; A value is randomly selected from each normal range, and m groups of optimization sets are constructed, and the m groups of optimization sets are different; In the step S403, the method for calculating the removal selection probability is that the strategy weights corresponding to each removal strategy are added in sequence to obtain a removal total value, and the strategy weights corresponding to each removal strategy are divided by the removal total value to obtain the removal selection probability corresponding to each removal strategy; The method for calculating the reconstruction selection probability is that the strategy weights corresponding to each reconstruction strategy are added in sequence to obtain a reconstruction total value, and the strategy weights corresponding to each reconstruction strategy are divided by the reconstruction total value to obtain the reconstruction selection probability corresponding to each reconstruction strategy; In the step S404, the method for selecting a removal strategy from the removal strategy set is consistent with the method for screening out a node according to the screening probability in the step S205; In the step S406, the method for selecting a reconstruction strategy from the reconstruction strategy set is consistent with the method for screening out a node according to the screening probability in the step S205.

10. An engine coupling operating parameter optimisation control system according to claim 9, wherein, In the step S408, the method for calculating the optimization effect of the reconstructed solution includes: An optimization set corresponding to the reconstruction solution is obtained and is marked as a current set; a value of each to-be-adjusted parameter in the coupling process parameter is replaced by a corresponding adjustment value in the current set, and the coupling process parameter after the replacement is marked as an optimization process parameter; the optimization process parameter, the coupling process parameter, and the coupling operation parameter are taken as prediction parameters, and the prediction parameters are input into the trained parameter prediction model to predict a corresponding optimization operation parameter; the training process of the parameter prediction model is consistent with the training process of the coefficient calculation model, and both are deep neural network models; the optimization operation parameter and the parameter threshold value are input into the coefficient calculation model to calculate a corresponding parameter anomaly coefficient and mark the parameter anomaly coefficient as an optimization anomaly coefficient; each element in the optimization anomaly coefficient is compared with an anomaly threshold value to identify an anomaly element in the optimization anomaly coefficient; a number of the anomaly elements in the optimization anomaly coefficient is counted and is marked as a first anomaly number; a number of the anomaly elements in the parameter anomaly coefficient is counted and is marked as a second anomaly number; the second anomaly number is subtracted from the first anomaly number to obtain an optimization effect; The method for calculating the update probability is as follows: an iteration number corresponding to the current iteration process is obtained and is marked as a current number; the current number is multiplied by a preset influence factor and then by an update factor to obtain a current factor; the optimization effect of the candidate solution is calculated, and the optimization effect of the candidate solution is marked as a candidate effect, and the optimization effect of the reconstruction solution is marked as a reconstruction effect; the candidate effect is compared with the reconstruction effect; if the candidate effect is less than or equal to the reconstruction effect, the corresponding update probability is 1; if the candidate effect is greater than the reconstruction effect, the reconstruction effect is subtracted from the candidate effect, and the result is divided by the current factor to obtain an update coefficient; the natural constant is taken as a base, and the update coefficient is taken as an index to perform an index operation to obtain the update probability; The method for determining whether the candidate solution is updated to the reconstruction solution includes the following steps: If the update probability is 1, the candidate solution is updated to the reconstruction solution; if the update probability is not 1, an update range [0, k] and a reserved range (k, 1] are set, k is the update probability; a judgment value is randomly generated from the interval [0, 1]; if the judgment value is in the update range, the candidate solution is updated to the reconstruction solution; if the judgment value is in the reserved range, the candidate solution is not updated to the reconstruction solution; In the step S409, the method for adjusting the removal weight and the reconstruction weight respectively includes the following steps: The adjustment effect is obtained by subtracting the candidate effect from the reconstruction effect; the adjustment degree is obtained by multiplying the adjustment effect by the current factor; and the removal weight and the reconstruction weight are adjusted according to the adjustment degree.

Images