Gearbox gear knocking evaluation method and system based on data analysis

Abstract

The invention discloses a gearbox gear knocking evaluation method and system based on data analysis. The gearbox gear knocking evaluation method comprises the steps that historical gear knocking monitoring information of a target gearbox in different operation states is acquired, and a historical knocking monitoring data set is constructed; carrying out evaluation index selection and gear knocking evaluation system construction, and carrying out performance evaluation on historical knocking monitoring examples in the historical knocking monitoring data set; establishing a gear knocking evaluation model by adopting a support vector machine, inputting real-time gear knocking monitoring information to carry out gear knocking evaluation on the target gearbox, and analyzing the real-time performance of the target gearbox; and analyzing whether the performance of the target gearbox accords with the current operation condition reference, if not, carrying out abnormal reason analysis, and generating a maintenance early warning report according to the abnormal reason analysis for pushing. The gearbox performance is prevented from being affected by gear problems, and therefore the use risk is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of gear knock evaluation, and in particular to a gearbox gear knock evaluation method and system based on data analysis. Background Art

[0002] Transmission gear knock is an important control content of automobile transmission NVH (noise, vibration and harshness) performance. With the widespread application of dual-clutch automatic transmission (DCT) in domestic passenger cars, its fast and smooth shifting process, good power, comfort and fuel economy have been widely recognized. However, gear knock, as a common problem of DCT, not only affects driving comfort, but may also have an adverse effect on the life and reliability of the transmission. Traditional transmission gear knock evaluation methods mainly rely on subjective evaluation and simple physical tests. However, these methods often have problems such as inconsistent evaluation standards and evaluation results that are easily affected by human factors.

[0003] Therefore, how to better understand the gearbox knock characteristics to avoid affecting the gearbox performance due to gear problems and thus reduce the risk of use is an urgent problem to be solved. Summary of the invention

[0004] The present invention overcomes the defects of the prior art and provides a transmission gear knock evaluation method and system based on data analysis, an important purpose of which is to improve the accuracy and objectivity of the transmission gear knock evaluation.

[0005] To achieve the above-mentioned purpose, the present invention provides a first aspect of a gearbox gear knock evaluation method based on data analysis, comprising: Obtain historical gear knock monitoring information of target gearboxes in different operating states and perform preprocessing, build a historical knock monitoring data set based on the preprocessing results, and use the historical knock monitoring data set to select evaluation indicators and build a gear knock evaluation system; According to the constructed gear knock evaluation system, the performance of historical knock monitoring instances in the historical knock monitoring data set is evaluated and a training sample set is constructed. The gear knock evaluation model is constructed using support vector machine. Performing gear knock monitoring on the target gearbox to obtain real-time gear knock monitoring information, using the gear knock evaluation model to perform gear knock evaluation on the target gearbox, analyzing the real-time performance of the target gearbox, and obtaining knock evaluation result information; Based on the knocking evaluation result information, it is analyzed whether the performance of the target gearbox meets the current operating condition benchmark. If not, an abnormality cause analysis is performed, and a maintenance warning report is generated and pushed based on the abnormality cause analysis.

[0006] In this solution, the historical gear knock monitoring information of the target gearbox in different operating states is obtained and preprocessed, and a historical knock monitoring data set is constructed according to the preprocessing results, specifically including: Acquiring historical gear knock monitoring information of a target gearbox in different operating states based on data retrieval, wherein the historical gear knock monitoring information includes vibration monitoring information of a gearbox housing, acoustic monitoring information when the gearbox knocks, and knock monitoring environment information; Preprocessing the acquired historical gear knock monitoring information, eliminating missing data and abnormal data in the historical gear knock monitoring data, and using the preprocessed historical gear knock monitoring data to form a first data set; Based on the first data set, the state parameters corresponding to each historical knock monitoring data are obtained, wherein the state parameters include operation state parameters and working condition state parameters, and a data label is constructed according to the obtained state parameters; Using a clustering algorithm to cluster the historical percussion monitoring data in the first data set with data labels as categories, dividing the data sets into several different categories and merging them to output several category subsets; The operating status attributes of each category subset are extracted, and the normal knocking category subset and the abnormal knocking category subset are divided according to the extracted health status attributes, and a historical knocking monitoring data set is constructed.

[0007] In this scheme, the selection of evaluation indicators and the construction of gear knock evaluation system using historical knock monitoring data sets specifically include: Acquire a historical knock monitoring data set, perform feature extraction on the historical knock monitoring data set, obtain knock monitoring features of each knock category subset, perform matching analysis based on the knock monitoring features, and obtain matching differences between each subset based on the matching analysis results; The matching difference is compared with a preset threshold, the knock monitoring feature whose matching difference is greater than the preset threshold is used as a first performance evaluation indicator, and the Spearman correlation coefficient between the remaining knock monitoring features and the first performance evaluation indicator is calculated; The remaining tapping monitoring features corresponding to the maximum Spearman correlation coefficient are selected as the second performance evaluation index, and iterative calculation and index selection are performed until a preset performance evaluation index quantity threshold is reached to obtain the final performance evaluation index information; Based on the final performance evaluation index, historical normal knocking monitoring features and historical abnormal knocking monitoring features corresponding to the performance evaluation index are extracted from the historical knocking monitoring data set to obtain first feature extraction information; According to the first feature extraction information, a feature interval is constructed with a normal knock monitoring feature as an upper limit and an abnormal knock monitoring feature as a lower limit, and an operating state category attribute of each historical knock monitoring feature is obtained through a historical knock monitoring data set; The performance level of gearbox gear knock is defined by combining the operating status category attributes of each historical knock monitoring feature and the constructed feature interval. The entropy weight method is used to set the evaluation weights of the performance evaluation indicators and construct a gear knock evaluation system.

[0008] In this solution, the gear knock evaluation system constructed is used to evaluate the performance of the historical knock monitoring instances in the historical knock monitoring data set and construct a training sample set, and a support vector machine is used to construct a gear knock evaluation model, which specifically includes: According to the constructed gear knock evaluation system, the performance of the historical knock monitoring instances in the historical knock monitoring data set is evaluated. According to the performance evaluation results, the historical knock monitoring instances corresponding to different performance levels are obtained and a training sample set is constructed. The gear knocking evaluation model was constructed by using support vector machine, and the improved sparrow search algorithm was introduced to optimize the penalty parameters and kernel parameters of the support vector machine, and the individual fitness values ​​in the initialized sparrow population were calculated. Introducing elite differential mutation in sparrow individual update, elite individuals are selected by calculating individual fitness, and differential vectors are calculated based on the distance and direction information between elite individuals; The calculated difference vector is weighted and added to the selected elite individual to obtain the added elite individual, the sparrow population is updated according to the added elite individual, and the optimal sparrow individual is obtained through iterative updating; Based on the optimal sparrow individual, the optimal penalty parameter and the optimal kernel parameter are obtained, and the optimal hyperplane is determined according to the optimal penalty parameter and the optimal kernel parameter to configure the support vector machine, and the constructed training sample set is imported into the configured support vector machine for training; After iterative training, the validation samples are used to test the classification evaluation performance of the support vector machine. When the performance meets the standard, the corresponding model configuration parameters are saved and the gear knock evaluation model that meets the expectations is output.

[0009] In this solution, the gear knock monitoring of the target gearbox is performed to obtain real-time gear knock monitoring information, the gear knock evaluation model is used to evaluate the gear knock of the target gearbox, and the real-time performance of the target gearbox is analyzed, which specifically includes: Performing gear knock monitoring on the target gearbox to obtain real-time gear knock monitoring information, using a wavelet threshold denoising algorithm to perform noise reduction processing on the real-time collected gear knock monitoring signal, and using a finite impulse response bandpass filter to extract the monitoring signal within a preset frequency band to obtain first processing result information; Performing feature extraction on the first processing result information to obtain the local maximum and local minimum of the real-time gear knock monitoring signal, and performing signal difference processing to construct the upper envelope and lower envelope of the real-time gear knock monitoring signal; Preset the envelope threshold interval, perform average calculation on the upper envelope and the lower envelope of the real-time tapping monitoring signal to obtain the envelope mean, and compare the calculated envelope mean with the preset envelope threshold interval; If the envelope mean is within a preset envelope threshold interval, the local signal trend corresponding to the envelope mean is removed from the real-time knock monitoring signal and a new real-time knock monitoring signal is formed; Repeat the envelope operation step until all real-time knock monitoring signals are processed, integrate all new real-time knock monitoring signals, and extract the real-time knock monitoring environment characteristics of the target gearbox through the real-time gear knock monitoring information, including operating environment characteristics and working condition environment characteristics; The integrated real-time knock monitoring signal is associated with the real-time knock monitoring environment characteristics to obtain first associated information; the first associated information is input into the gear knock evaluation model to perform gear knock evaluation on the target gearbox, and the performance of the target gearbox under the current knock monitoring environment is analyzed to obtain knock evaluation result information.

[0010] In this solution, the performance of the target gearbox is analyzed based on the knocking evaluation result information to determine whether it meets the current operating condition benchmark. If not, an abnormality cause analysis is performed, and a maintenance warning report is generated and pushed based on the abnormality cause analysis, specifically including: Preset a number of knock evaluation thresholds under different operating conditions, obtain knock evaluation result information, judge the knock evaluation result information with the corresponding knock evaluation threshold, and analyze whether the performance of the target gearbox meets the current operating condition benchmark; If the knock evaluation result information is less than the corresponding knock rating threshold, it means that the performance of the target gearbox under the current operating condition is unqualified, and the cause of the performance abnormality is analyzed; Based on data retrieval, historical gearbox abnormality instances under various operating conditions are obtained, abnormality cause characteristics of each historical gearbox abnormality instance are extracted, and each historical gearbox abnormality instance is divided into classification categories to obtain first division result information; Extract the abnormal knock monitoring features corresponding to each abnormal cause category according to the first division result information, and construct an abnormal cause knowledge graph with operating conditions, abnormal knock monitoring features, and abnormal causes as associated paths; The abnormal cause knowledge graph is represented by a graph through the MetaPath random walk algorithm, the graph representation of the abnormal cause knowledge graph is learned by using a graph neural network, and an abnormal cause analysis model is constructed; The first associated information is obtained and input into the abnormal cause analysis model to analyze the cause of abnormal performance of the target gearbox under the current operating condition, and a maintenance early warning report is generated and pushed according to the abnormal cause analysis.

[0011] A second aspect of the present invention provides a gearbox gear knock evaluation system based on data analysis, the system comprising: a memory, a processor, the memory containing a gearbox gear knock evaluation method program based on data analysis, the gearbox gear knock evaluation method program based on data analysis when executed by the processor implements the following steps: Obtain historical gear knock monitoring information of target gearboxes in different operating states and perform preprocessing, build a historical knock monitoring data set based on the preprocessing results, and use the historical knock monitoring data set to select evaluation indicators and build a gear knock evaluation system; According to the constructed gear knock evaluation system, the performance of historical knock monitoring instances in the historical knock monitoring data set is evaluated and a training sample set is constructed. The gear knock evaluation model is constructed using support vector machine. Performing gear knock monitoring on the target gearbox to obtain real-time gear knock monitoring information, using the gear knock evaluation model to perform gear knock evaluation on the target gearbox, analyzing the real-time performance of the target gearbox, and obtaining knock evaluation result information; Based on the knocking evaluation result information, it is analyzed whether the performance of the target gearbox meets the current operating condition benchmark. If not, an abnormality cause analysis is performed, and a maintenance warning report is generated and pushed based on the abnormality cause analysis.

[0012] The present invention discloses a gearbox gear knock evaluation method and system based on data analysis, including: obtaining historical gear knock monitoring information of target gearboxes in different operating states and constructing a historical knock monitoring data set; selecting evaluation indicators and constructing a gear knock evaluation system, and performing performance evaluation on historical knock monitoring instances in the historical knock monitoring data set; using a support vector machine to construct a gear knock evaluation model, inputting real-time gear knock monitoring information to perform gear knock evaluation on the target gearbox, and analyzing the real-time performance of the target gearbox; analyzing whether the performance of the target gearbox meets the current operating condition benchmark, and if not, performing abnormal cause analysis, and generating a maintenance warning report based on the abnormal cause analysis for push. Avoiding the gearbox performance being affected by gear problems, thereby reducing the use risk. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the embodiments or exemplary embodiments of the present invention, the drawings required for use in the embodiments or exemplary descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained according to the drawings without paying creative work.

[0014] Figure 1A flow chart of a transmission gear knock evaluation method based on data analysis provided by one embodiment of the present invention; Figure 2 A flow chart of a method for evaluating and analyzing real-time gear knocking in a gearbox provided by an embodiment of the present invention; Figure 3 A block diagram of a gearbox gear knock evaluation system based on data analysis provided by one embodiment of the present invention; The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0015] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0016] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0017] Figure 1 A flow chart of a transmission gear knock evaluation method based on data analysis provided by one embodiment of the present invention; like Figure 1 As shown, the present invention provides a flow chart of a gearbox gear knock evaluation method based on data analysis, comprising: S102, obtaining historical gear knock monitoring information of the target gearbox in different operating states and performing preprocessing, constructing a historical knock monitoring data set according to the preprocessing result, and selecting evaluation indicators and constructing a gear knock evaluation system using the historical knock monitoring data set; S104, performing performance evaluation on historical knock monitoring instances in the historical knock monitoring data set according to the constructed gear knock evaluation system and constructing a training sample set, and constructing a gear knock evaluation model using a support vector machine; S106, performing gear knock monitoring on the target gearbox to obtain real-time gear knock monitoring information, performing gear knock evaluation on the target gearbox using the gear knock evaluation model, analyzing the real-time performance of the target gearbox, and obtaining knock evaluation result information; S108, analyzing whether the performance of the target gearbox meets the current operating condition benchmark based on the knocking evaluation result information, and if not, performing abnormal cause analysis, and generating a maintenance warning report based on the abnormal cause analysis for push.

[0018] It should be noted that the present invention provides a gearbox gear knock evaluation method and system based on data analysis. First, historical gear knock monitoring information of the target gearbox under different operating conditions is obtained, and the information is preprocessed. Through the preprocessing results, a historical knock monitoring data set covering various operating conditions is constructed. Based on the constructed historical knock monitoring data set, the gear knock evaluation index selection and system construction are carried out. Mainly by analyzing the characteristic variables in the data set, such as the time-frequency characteristics, amplitude changes and impact power characteristics of the signal, the key indicators that can characterize the knock phenomenon are selected, and the gear knock evaluation system is built. Further construct a training sample set, and use the support vector machine algorithm to construct a gear knock evaluation model. The support vector machine can accurately divide the knock state under different working conditions by optimizing the classification boundary. Real-time gear knock monitoring is carried out, and the real-time knock monitoring information generated during the operation is collected. Then it is input into the constructed gear knock evaluation model for analysis and calculation, and the real-time performance of the target gearbox is evaluated, and then the knock evaluation result is obtained. Finally, based on the knock evaluation result, the performance of the target gearbox is further analyzed to see whether it meets the current operating condition benchmark. If it is found that there is a significant difference between the performance and the benchmark, the abnormal cause analysis is performed. By extracting the abnormal features of the knocking signal, the possible root cause of the fault is determined, and a maintenance warning report is generated accordingly, so that timely measures can be taken to ensure the normal operation of the gearbox. The accuracy of fault detection is improved, and the efficiency and pertinence of vehicle maintenance are significantly enhanced.

[0019] Further, in a preferred embodiment of the invention, the acquisition of historical gear knock monitoring information of the target gearbox in different operating states and preprocessing, and the construction of a historical knock monitoring data set based on the preprocessing results, specifically include: Acquiring historical gear knock monitoring information of a target gearbox in different operating states based on data retrieval, wherein the historical gear knock monitoring information includes vibration monitoring information of a gearbox housing, acoustic monitoring information when the gearbox knocks, and knock monitoring environment information; Preprocessing the acquired historical gear knock monitoring information, eliminating missing data and abnormal data in the historical gear knock monitoring data, and using the preprocessed historical gear knock monitoring data to form a first data set; Based on the first data set, the state parameters corresponding to each historical knocking monitoring data are obtained, wherein the state parameters include operation state parameters and working condition state parameters, and a data label is constructed according to the obtained state parameters; Using a clustering algorithm to cluster the historical percussion monitoring data in the first data set with data labels as categories, dividing the data sets into several different categories and merging them to output several category subsets; The operating status attributes of each category subset are extracted, and the normal knocking category subset and the abnormal knocking category subset are divided according to the extracted health status attributes, and a historical knocking monitoring data set is constructed.

[0020] It should be noted that the monitoring data of the target gearbox under different operating conditions is obtained by a data retrieval-based method. These historical gear knocking monitoring information include three types of key data: vibration monitoring information of the gearbox housing, acoustic monitoring information when the gear knocks, and knocking environment information during monitoring. Vibration information can reflect the mechanical vibration characteristics caused by knocking, acoustic information reveals the time-frequency characteristics of the knocking sound, and knocking environment information provides the necessary background working condition support for data analysis. After the preprocessing is completed, the first data set is constructed using these cleaned historical gear knocking monitoring data. Subsequently, the state parameters corresponding to each historical monitoring data are further extracted from the first data set. The state parameters include two parts: operating state parameters (such as speed, load, temperature and other operating conditions that directly affect the knocking phenomenon) and working state parameters (such as external influencing factors such as environmental noise level or installation accuracy). According to the extracted state parameters, corresponding data labels are generated for each data. These data labels are used to identify the operating state category corresponding to the monitoring information, thereby providing a basis for classification and clustering analysis. Based on the generated data labels, a clustering algorithm is used to analyze the historical knocking monitoring data in the first data set. The clustering algorithm divides the monitoring data with similar state parameters into several different categories of data sets to form multiple category subsets. Each subset represents a specific knocking mode or operating state, which effectively distinguishes the data characteristics and reveals the internal structure of the data. The operating state attributes of each category subset are extracted to classify and identify the monitoring information. By analyzing the relevant attributes of the health state, it is identified which categories belong to the knocking information under normal operation and divided into normal knocking category subsets. At the same time, the subsets of the remaining categories that reflect abnormal operation or gear knocking are classified as abnormal knocking category subsets. Finally, a complete historical knocking monitoring data set is constructed based on these subsets.

[0021] Furthermore, in a preferred embodiment of the invention, the selection of evaluation indicators and the construction of a gear knock evaluation system using a historical knock monitoring data set specifically include: Acquire a historical percussion monitoring data set, perform feature extraction on the historical percussion monitoring data set, obtain percussion monitoring features of each percussion category subset, perform matching analysis based on the percussion monitoring features, and obtain matching differences between each subset based on the matching analysis results; The matching difference is compared with a preset threshold, the knock monitoring feature whose matching difference is greater than the preset threshold is used as a first performance evaluation indicator, and the Spearman correlation coefficient between the remaining knock monitoring features and the first performance evaluation indicator is calculated; The remaining tapping monitoring features corresponding to the maximum Spearman correlation coefficient are selected as the second performance evaluation index, and iterative calculation and index selection are performed until a preset performance evaluation index quantity threshold is reached to obtain the final performance evaluation index information; Based on the final performance evaluation index, historical normal knocking monitoring features and historical abnormal knocking monitoring features corresponding to the performance evaluation index are extracted from the historical knocking monitoring data set to obtain first feature extraction information; According to the first feature extraction information, a feature interval is constructed with a normal knock monitoring feature as an upper limit and an abnormal knock monitoring feature as a lower limit, and an operating state category attribute of each historical knock monitoring feature is obtained through a historical knock monitoring data set; The performance level of gearbox gear knock is defined by combining the operating status category attributes of each historical knock monitoring feature and the constructed feature interval. The entropy weight method is used to set the evaluation weights of the performance evaluation indicators and construct a gear knock evaluation system.

[0022] It should be noted that in order to build an accurate gearbox gear knock evaluation system, by extracting the monitoring features of each knock category subset, matching analysis of the knock monitoring features is performed to evaluate the matching differences between the subsets. These differences quantify the similarity or difference of features of different categories, and are an important basis for screening performance evaluation indicators. Subsequently, the obtained matching differences are compared with the preset threshold. For features whose differences are greater than the preset threshold, they are selected as the first performance evaluation indicator, and by calculating the Spearman correlation coefficient between the remaining knock monitoring features and the first performance evaluation indicator, the remaining features with the most correlation with the first performance evaluation indicator are further identified. According to the sorting of the correlation coefficients, the remaining knock monitoring features corresponding to the maximum Spearman correlation coefficient are selected as the second performance evaluation indicator. In order to ensure the comprehensiveness and scientificity of the evaluation indicators, the feature set is continuously expanded through iterative calculation and indicator screening until the preset performance evaluation indicator quantity threshold is reached. Then, based on these indicators, the monitoring features of normal knock and abnormal knock categories are extracted from the historical knock monitoring data set to form the first feature extraction information. By analyzing the feature extraction information, a feature interval with a clear range is constructed, in which the normal knocking monitoring feature is set as the upper limit of the interval and the abnormal knocking monitoring feature is set as the lower limit of the interval. The feature interval provides a clear discrimination benchmark for performance evaluation. Subsequently, according to the operating state category attributes in the historical knocking monitoring data set and the operating state of each knocking monitoring feature, the knocking feature distribution is mapped to the constructed feature interval, thereby defining the performance level for the gear knocking of the gearbox. These levels describe the operating performance of gear knocking according to different state categories and feature distributions. After the performance level is determined, the entropy weight method is used to set a reasonable weight for each performance evaluation indicator. The entropy weight method is used to quantify the relative importance of each indicator in the evaluation to ensure the scientificity and fairness of the evaluation process. Finally, a complete gear knocking evaluation system is constructed based on the set weights and evaluation indicators.

[0023] Further, in a preferred embodiment of the invention, the gear knocking evaluation system constructed performs performance evaluation on the historical knocking monitoring instances in the historical knocking monitoring data set and constructs a training sample set, and uses a support vector machine to construct a gear knocking evaluation model, specifically including: According to the constructed gear knock evaluation system, the performance of the historical knock monitoring instances in the historical knock monitoring data set is evaluated. According to the performance evaluation results, the historical knock monitoring instances corresponding to different performance levels are obtained and a training sample set is constructed. The gear knocking evaluation model was constructed by using support vector machine, and the improved sparrow search algorithm was introduced to optimize the penalty parameters and kernel parameters of the support vector machine, and the individual fitness values ​​in the initialized sparrow population were calculated. Introducing elite differential mutation in sparrow individual update, elite individuals are selected by calculating individual fitness, and differential vectors are calculated based on the distance and direction information between elite individuals; The calculated difference vector is weighted and added to the selected elite individual to obtain the added elite individual, the sparrow population is updated according to the added elite individual, and the optimal sparrow individual is obtained through iterative updating; Based on the optimal sparrow individual, the optimal penalty parameter and the optimal kernel parameter are obtained, and the optimal hyperplane is determined according to the optimal penalty parameter and the optimal kernel parameter to configure the support vector machine, and the constructed training sample set is imported into the configured support vector machine for training; After iterative training, the validation samples are used to test the classification evaluation performance of the support vector machine. When the performance meets the standard, the corresponding model configuration parameters are saved and the gear knock evaluation model that meets the expectations is output.

[0024] It should be noted that, firstly, based on the previously constructed gear knocking evaluation system, a comprehensive performance evaluation is performed on the monitoring instances in the historical knocking monitoring data set. Through the performance evaluation results, historical knocking monitoring instances belonging to different performance levels can be obtained. These instances represent the distribution characteristics of gear knocking of gearboxes under various working conditions and performance conditions. Subsequently, a training sample set is constructed based on these instances to lay the foundation for subsequent model training. In order to improve the accuracy of gear knocking evaluation, support vector machine is selected as the core evaluation model, and an improved sparrow search algorithm is introduced to optimize the key parameters of the support vector machine. In the model optimization stage, the penalty parameters and kernel parameters of the support vector machine are regarded as optimization objects, and their optimal values ​​are determined by the sparrow search algorithm. Subsequently, in the update process of sparrow individuals, the elite differential mutation method is introduced. By calculating the distance and direction information between the selected elite individuals, the differential vector describing the difference in solutions within the population is obtained. The differential vector is weighted and added to the selected elite individuals to generate the updated elite individuals. In this way, the sparrow population can be gradually optimized and converge to a better solution space. After several iterations, the optimal sparrow individual is output, which represents the optimal solution of the current parameter search. Based on this, the optimal penalty parameter and kernel parameter are obtained, and the optimal hyperplane of the support vector machine is determined by these optimal parameters, thereby completing the parameter configuration of the support vector machine. Next, the previously constructed training sample set is input into the configured support vector machine for training. The support vector machine gradually improves the classification ability of gear knock characteristics by learning the pattern of training data. During the training process, the model will be continuously iteratively optimized, and the classification performance of the support vector machine will be evaluated using validation samples. The validation sample is used to detect the performance of the model on unknown data. When the model performance meets the preset classification criteria, it means that the optimization work has been completed. Finally, the model configuration parameters that meet the requirements are saved, and the optimized gear knock evaluation model is output at the same time. This model can accurately and efficiently classify and evaluate the gear knock performance of the gearbox, providing reliable technical support for the health management and maintenance decision of the gearbox.

[0025] Further, in a preferred embodiment of the invention, the gear knocking monitoring of the target gearbox is performed to obtain real-time gear knocking monitoring information, the gear knocking evaluation model is used to perform gear knocking evaluation on the target gearbox, and the real-time performance of the target gearbox is analyzed, specifically including: Performing gear knock monitoring on the target gearbox to obtain real-time gear knock monitoring information, using a wavelet threshold denoising algorithm to perform noise reduction processing on the real-time collected gear knock monitoring signal, and using a finite impulse response bandpass filter to extract the monitoring signal within a preset frequency band to obtain first processing result information; Performing feature extraction on the first processing result information to obtain the local maximum and local minimum of the real-time gear knock monitoring signal, and performing signal difference processing to construct the upper envelope and lower envelope of the real-time gear knock monitoring signal; Preset the envelope threshold interval, perform average calculation on the upper envelope and the lower envelope of the real-time tapping monitoring signal to obtain the envelope mean, and compare the calculated envelope mean with the preset envelope threshold interval; If the envelope mean is within a preset envelope threshold interval, the local signal trend corresponding to the envelope mean is removed from the real-time knock monitoring signal and a new real-time knock monitoring signal is formed; Repeat the envelope operation step until all real-time knock monitoring signals are processed, integrate all new real-time knock monitoring signals, and extract the real-time knock monitoring environment characteristics of the target gearbox through the real-time gear knock monitoring information, including operating environment characteristics and working condition environment characteristics; The integrated real-time knock monitoring signal is associated with the real-time knock monitoring environment characteristics to obtain first associated information; the first associated information is input into the gear knock evaluation model to perform gear knock evaluation on the target gearbox, and the performance of the target gearbox under the current knock monitoring environment is analyzed to obtain knock evaluation result information.

[0026] It should be noted that, first, the gear knocking information of the target gearbox is obtained in real time through the monitoring equipment, and the collected signal is denoised by the wavelet threshold denoising algorithm to reduce the interference of noise on the analysis result. Then, a finite impulse response bandpass filter is used to extract the monitoring signal within the preset frequency band to obtain the result of preliminary processing, that is, the first processing result information. Subsequently, the first processing result information is feature extracted, focusing on the local maximum and minimum values ​​of the signal. By calculating the difference between these values, the upper envelope and lower envelope of the real-time gear knocking monitoring signal are constructed to reflect the change trend and amplitude range of the signal. The envelope threshold interval is preset, and the envelope mean is calculated based on the upper envelope and the lower envelope. This envelope mean is compared with the preset envelope threshold interval to determine the current state of the signal. If the envelope mean falls within the preset interval, it means that this part of the signal may contain a certain analytical error. Therefore, the local trend corresponding to this part of the signal is removed from the original signal to form a new real-time knocking monitoring signal. The complex characteristic signal is decomposed into a set of intrinsic mode functions, which represent the local oscillation characteristics of the signal and contain more high-frequency components, effectively helping to analyze the local characteristics of the complex signal. Finally, the integrated real-time knock monitoring signal is associated with the extracted environmental characteristics to form the first associated information. This associated information contains comprehensive data of gearbox knock monitoring, including signal characteristics, environmental characteristics, etc., which are input into the gear knock evaluation model, and the gear knock evaluation is performed on the target gearbox. The performance under the current knock monitoring environment is analyzed, and the knock evaluation result information is obtained.

[0027] Further, in a preferred embodiment of the invention, the performance of the target gearbox is analyzed based on the knocking evaluation result information to determine whether it meets the current operating condition benchmark. If not, an abnormality cause analysis is performed, and a maintenance warning report is generated and pushed based on the abnormality cause analysis, specifically including: Preset a number of knock evaluation thresholds under different operating conditions, obtain knock evaluation result information, judge the knock evaluation result information with the corresponding knock evaluation threshold, and analyze whether the performance of the target gearbox meets the current operating condition benchmark; If the knock evaluation result information is less than the corresponding knock rating threshold, it means that the performance of the target gearbox under the current operating conditions is unqualified, and the cause of the performance abnormality is analyzed; Based on data retrieval, historical gearbox abnormality instances under various operating conditions are obtained, abnormality cause characteristics of each historical gearbox abnormality instance are extracted, and each historical gearbox abnormality instance is divided into classification categories to obtain first division result information; Extract the abnormal knock monitoring features corresponding to each abnormal cause category according to the first division result information, and construct an abnormal cause knowledge graph with operating conditions, abnormal knock monitoring features, and abnormal causes as associated paths; The abnormal cause knowledge graph is represented by a graph through the MetaPath random walk algorithm, the graph representation of the abnormal cause knowledge graph is learned by a graph neural network, and an abnormal cause analysis model is constructed; The first associated information is obtained and input into the abnormal cause analysis model to analyze the cause of abnormal performance of the target gearbox under the current operating condition, and a maintenance early warning report is generated and pushed according to the abnormal cause analysis.

[0028] It should be noted that after the gear knocking evaluation of the target gearbox is carried out, based on the evaluation results reflecting its real-time performance, it is judged against the preset knocking evaluation thresholds of different operating conditions to analyze whether its performance under the corresponding conditions is qualified. In order to carry out the analysis of the causes of performance abnormalities, historical gearbox abnormal instances under different operating conditions are first obtained. By extracting the abnormal cause features from the historical abnormal data, these features are regarded as the basis for classification, and all historical abnormal instances are divided to form a set of abnormal data corresponding to multiple categories. Next, according to the classification results, the knocking monitoring feature information corresponding to each abnormal cause category is extracted from the historical abnormal data. Including operating environment features and knocking features that are closely related to abnormal phenomena. According to the first division result information, the abnormal knocking monitoring features corresponding to each abnormal cause category are extracted, and the abnormal cause knowledge graph with operating conditions, abnormal knocking monitoring features, and abnormal causes as the associated paths is constructed. The MetaPath random walk algorithm is used to process the graph, and the feature vector representation of the graph structure is generated by the algorithm. Subsequently, the graph neural network (GNN) is used to learn these graph structures and construct an abnormal cause analysis model. Finally, the correlation information obtained by integrating the knocking monitoring information of the target gearbox with the knocking monitoring environment characteristics is input into the abnormal cause analysis model. The model is used to analyze the specific causes of the performance abnormality of the target gearbox under the current working conditions, and a detailed maintenance early warning report is generated. The report not only provides possible fault causes and diagnostic basis, but also pushes it to users or relevant personnel, providing comprehensive and efficient technical support for the operation and maintenance of the gearbox.

[0029] Figure 2 A flow chart of a method for evaluating and analyzing real-time gear knocking in a gearbox provided by an embodiment of the present invention; like Figure 2 As shown, the present invention provides a flow chart of a gearbox real-time gear knock evaluation and analysis method, including: S202, performing gear knock monitoring on the target gearbox to obtain real-time gear knock monitoring information, and performing gear knock evaluation on the target gearbox using the gear knock evaluation model to obtain knock evaluation result information; S204, judging the knocking evaluation result information and the corresponding knocking evaluation threshold, and analyzing whether the performance of the target gearbox meets the current operating condition benchmark; S206, if the knock evaluation result information is less than the corresponding knock rating threshold, it means that the performance of the target gearbox under the current operating condition is unqualified, and then the cause of the performance abnormality is analyzed; S208, analyzing the reasons for abnormal performance of the target gearbox under the current operating conditions using the abnormal reason analysis model, and generating a maintenance warning report based on the abnormal reason analysis for push.

[0030] Furthermore, in a transmission gear knock evaluation method based on data analysis provided by the present invention, the following steps are also included: The historical gear knock monitoring instances of gearboxes in different life states are obtained by using the big data network, and four life state categories, namely, extremely poor, good, normal and healthy, are defined. The historical gear knock monitoring instances obtained are initially classified into categories to obtain the first category classification information. Extracting the gear knock monitoring signals and corresponding operating condition characteristics of the historical knock monitoring instances under each life state category based on the first category classification information, and associating the operating condition characteristics with the gear knock monitoring signals as subsidiary attributes; The extracted gear knocking monitoring signal is subjected to feature extraction to obtain time-frequency features, and the corresponding signal power features are fitted and calculated based on the obtained time-frequency features. The time-frequency features and signal power features are input into the thermal model to output the thermal chromaticity features of the gear knocking signal of each life state category. Construct a Bayesian inference network, input the thermal and chromatic characteristics of the gear knocking signal of each life state category into the Bayesian inference network for training, output the transient prior distribution and probability likelihood function, and build a life state analysis model based on the transient prior distribution and probability likelihood function; Performing gear knock monitoring on the target gearbox, and inputting the monitoring results into the life state analysis model for analysis, determining the life state of the target gearbox, and obtaining life state analysis information; Based on the life status analysis information, it is determined whether the target gearbox needs to be repaired. If repair is required, the corresponding repair method is obtained in the preset search space according to the gear knocking monitoring result of the target gearbox for push and warning.

[0031] It should be noted that the gear knocking of the gearbox not only indicates its performance characteristics, but also represents the life status of the gearbox to a certain extent. Therefore, the gearbox gear knocking characteristics under different life states are analyzed through historical data, and the thermal model is used to further enhance the description of the characteristics, so as to construct a life status analysis model. The life status of the target gearbox is judged by the life status analysis model. Once it is determined that the target gearbox needs to be repaired, the matching repair method is immediately searched in the preset search space, and timely push and warning are carried out. The accuracy and efficiency of the evaluation are improved, and powerful technical suggestions are provided for the maintenance and care of the gearbox.

[0032] Figure 3 A gearbox gear knock evaluation system 3 based on data analysis is provided in one embodiment of the present invention. The system comprises: a memory 31 and a processor 32. The memory 31 contains a gearbox gear knock evaluation method program based on data analysis. When the gearbox gear knock evaluation method program based on data analysis is executed by the processor 32, the following steps are implemented: Obtain historical gear knock monitoring information of target gearboxes in different operating states and perform preprocessing, build a historical knock monitoring data set based on the preprocessing results, and use the historical knock monitoring data set to select evaluation indicators and build a gear knock evaluation system; According to the constructed gear knock evaluation system, the performance of historical knock monitoring instances in the historical knock monitoring data set is evaluated and a training sample set is constructed. The gear knock evaluation model is constructed using support vector machine. Performing gear knock monitoring on the target gearbox to obtain real-time gear knock monitoring information, using the gear knock evaluation model to perform gear knock evaluation on the target gearbox, analyzing the real-time performance of the target gearbox, and obtaining knock evaluation result information; Based on the knocking evaluation result information, it is analyzed whether the performance of the target gearbox meets the current operating condition benchmark. If not, an abnormality cause analysis is performed, and a maintenance warning report is generated and pushed based on the abnormality cause analysis.

[0033] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0034] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0035] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0036] A person of ordinary skill in the art can understand that: all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: a mobile storage device, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0037] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0038] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A gearbox gear knock evaluation method based on data analysis, characterized in that: include: Obtain historical gear knock monitoring information of target gearboxes in different operating states and perform preprocessing, build a historical knock monitoring data set based on the preprocessing results, and use the historical knock monitoring data set to select evaluation indicators and build a gear knock evaluation system; According to the constructed gear knock evaluation system, the performance of historical knock monitoring instances in the historical knock monitoring data set is evaluated and a training sample set is constructed. The gear knock evaluation model is constructed using support vector machine. Performing gear knock monitoring on the target gearbox to obtain real-time gear knock monitoring information, using the gear knock evaluation model to perform gear knock evaluation on the target gearbox, analyzing the real-time performance of the target gearbox, and obtaining knock evaluation result information; Based on the knocking evaluation result information, it is analyzed whether the performance of the target gearbox meets the current operating condition benchmark. If not, an abnormality cause analysis is performed, and a maintenance warning report is generated and pushed based on the abnormality cause analysis.

2. A gearbox gear knock evaluation method based on data analysis according to claim 1, characterized in that: The obtaining of historical gear knock monitoring information of the target gearbox in different operating states and preprocessing, and constructing a historical knock monitoring data set according to the preprocessing results, specifically includes: Acquiring historical gear knock monitoring information of a target gearbox in different operating states based on data retrieval, wherein the historical gear knock monitoring information includes vibration monitoring information of a gearbox housing, acoustic monitoring information when the gearbox knocks, and knock monitoring environment information; Preprocessing the acquired historical gear knock monitoring information, eliminating missing data and abnormal data in the historical gear knock monitoring data, and using the preprocessed historical gear knock monitoring data to form a first data set; Based on the first data set, the state parameters corresponding to each historical knock monitoring data are obtained, wherein the state parameters include operation state parameters and working condition state parameters, and a data label is constructed according to the obtained state parameters; Using a clustering algorithm to cluster the historical percussion monitoring data in the first data set with data labels as categories, dividing the data sets into several different categories and merging them to output several category subsets; The operating status attributes of each category subset are extracted, and the normal knocking category subset and the abnormal knocking category subset are divided according to the extracted health status attributes, and a historical knocking monitoring data set is constructed.

3. The gearbox gear knock evaluation method based on data analysis according to claim 1 is characterized in that: The use of historical knock monitoring data sets to select evaluation indicators and construct a gear knock evaluation system specifically includes: Acquire a historical knock monitoring data set, perform feature extraction on the historical knock monitoring data set, obtain knock monitoring features of each knock category subset, perform matching analysis based on the knock monitoring features, and obtain matching differences between each subset based on the matching analysis results; The matching difference is compared with a preset threshold, the knock monitoring feature whose matching difference is greater than the preset threshold is used as a first performance evaluation indicator, and the Spearman correlation coefficient between the remaining knock monitoring features and the first performance evaluation indicator is calculated; The remaining tapping monitoring features corresponding to the maximum Spearman correlation coefficient are selected as the second performance evaluation index, and iterative calculation and index selection are performed until a preset performance evaluation index quantity threshold is reached to obtain the final performance evaluation index information; Based on the final performance evaluation index, historical normal knocking monitoring features and historical abnormal knocking monitoring features corresponding to the performance evaluation index are extracted from the historical knocking monitoring data set to obtain first feature extraction information; According to the first feature extraction information, a feature interval is constructed with a normal knock monitoring feature as an upper limit and an abnormal knock monitoring feature as a lower limit, and an operating state category attribute of each historical knock monitoring feature is obtained through a historical knock monitoring data set; The performance level of gearbox gear knock is defined by combining the operating status category attributes of each historical knock monitoring feature and the constructed feature interval. The entropy weight method is used to set the evaluation weights of the performance evaluation indicators and construct a gear knock evaluation system.

4. The gearbox gear knock evaluation method based on data analysis according to claim 1 is characterized in that: The gear knock evaluation system constructed is used to evaluate the performance of the historical knock monitoring instances in the historical knock monitoring data set and construct a training sample set, and a support vector machine is used to construct a gear knock evaluation model, which specifically includes: According to the constructed gear knock evaluation system, the performance of the historical knock monitoring instances in the historical knock monitoring data set is evaluated. According to the performance evaluation results, the historical knock monitoring instances corresponding to different performance levels are obtained and a training sample set is constructed. The gear knocking evaluation model was constructed by using support vector machine, and the improved sparrow search algorithm was introduced to optimize the penalty parameters and kernel parameters of the support vector machine, and the individual fitness values ​​in the initialized sparrow population were calculated. Introducing elite differential mutation in the sparrow individual update, the elite individuals are selected by calculating the individual fitness, and the differential vector is calculated based on the distance and direction information between the elite individuals; The calculated difference vector is weighted and added to the selected elite individual to obtain the added elite individual, the sparrow population is updated according to the added elite individual, and the optimal sparrow individual is obtained through iterative updating; Based on the optimal sparrow individual, the optimal penalty parameter and the optimal kernel parameter are obtained, and the optimal hyperplane is determined according to the optimal penalty parameter and the optimal kernel parameter to configure the support vector machine, and the constructed training sample set is imported into the configured support vector machine for training; After iterative training, the validation samples are used to test the classification evaluation performance of the support vector machine. When the performance meets the standard, the corresponding model configuration parameters are saved and the gear knock evaluation model that meets the expectations is output.

5. The gearbox gear knock evaluation method based on data analysis according to claim 1, characterized in that: The gear knock monitoring of the target gearbox is performed to obtain real-time gear knock monitoring information, the gear knock evaluation model is used to perform gear knock evaluation on the target gearbox, and the real-time performance of the target gearbox is analyzed, specifically including: Performing gear knock monitoring on the target gearbox to obtain real-time gear knock monitoring information, using a wavelet threshold denoising algorithm to perform noise reduction processing on the real-time collected gear knock monitoring signal, and using a finite impulse response bandpass filter to extract the monitoring signal within a preset frequency band to obtain first processing result information; Performing feature extraction on the first processing result information to obtain the local maximum and local minimum of the real-time gear knock monitoring signal, and performing signal difference processing to construct the upper envelope and lower envelope of the real-time gear knock monitoring signal; Preset the envelope threshold interval, perform average calculation on the upper envelope and the lower envelope of the real-time tapping monitoring signal to obtain the envelope mean, and compare the calculated envelope mean with the preset envelope threshold interval; If the envelope mean is within a preset envelope threshold interval, the local signal trend corresponding to the envelope mean is removed from the real-time knock monitoring signal and a new real-time knock monitoring signal is formed; Repeat the envelope operation step until all real-time knock monitoring signals are processed, integrate all new real-time knock monitoring signals, and extract the real-time knock monitoring environment characteristics of the target gearbox through the real-time gear knock monitoring information, including operating environment characteristics and working condition environment characteristics; The integrated real-time knock monitoring signal is associated with the real-time knock monitoring environment characteristics to obtain first associated information; the first associated information is input into the gear knock evaluation model to perform gear knock evaluation on the target gearbox, and the performance of the target gearbox under the current knock monitoring environment is analyzed to obtain knock evaluation result information.

6. The gearbox gear knock evaluation method based on data analysis according to claim 1, characterized in that: The analysis based on the knocking evaluation result information whether the performance of the target gearbox meets the current operating condition benchmark, if not, the abnormality cause analysis is performed, and a maintenance warning report is generated and pushed according to the abnormality cause analysis, specifically including: Preset a number of knock evaluation thresholds under different operating conditions, obtain knock evaluation result information, judge the knock evaluation result information with the corresponding knock evaluation threshold, and analyze whether the performance of the target gearbox meets the current operating condition benchmark; If the knock evaluation result information is less than the corresponding knock rating threshold, it means that the performance of the target gearbox under the current operating condition is unqualified, and the cause of the performance abnormality is analyzed; Based on data retrieval, historical gearbox abnormality instances under various operating conditions are obtained, abnormality cause characteristics of each historical gearbox abnormality instance are extracted, and each historical gearbox abnormality instance is divided into classification categories to obtain first division result information; Extract the abnormal knock monitoring features corresponding to each abnormal cause category according to the first division result information, and construct an abnormal cause knowledge graph with operating conditions, abnormal knock monitoring features, and abnormal causes as associated paths; The abnormal cause knowledge graph is represented by a graph through the MetaPath random walk algorithm, the graph representation of the abnormal cause knowledge graph is learned by a graph neural network, and an abnormal cause analysis model is constructed; The first associated information is obtained and input into the abnormal cause analysis model to analyze the cause of abnormal performance of the target gearbox under the current operating condition, and a maintenance early warning report is generated and pushed according to the abnormal cause analysis.

7. A gearbox gear knock evaluation system based on data analysis, characterized in that: The system includes: a memory and a processor, wherein the memory contains a transmission gear knock evaluation method program based on data analysis, and when the transmission gear knock evaluation method program based on data analysis is executed by the processor, the following steps are implemented: Obtain historical gear knock monitoring information of target gearboxes in different operating states and perform preprocessing, build a historical knock monitoring data set based on the preprocessing results, and use the historical knock monitoring data set to select evaluation indicators and build a gear knock evaluation system; According to the constructed gear knock evaluation system, the performance of historical knock monitoring instances in the historical knock monitoring data set is evaluated and a training sample set is constructed. The gear knock evaluation model is constructed using support vector machine. Performing gear knock monitoring on the target gearbox to obtain real-time gear knock monitoring information, using the gear knock evaluation model to perform gear knock evaluation on the target gearbox, analyzing the real-time performance of the target gearbox, and obtaining knock evaluation result information; Based on the knocking evaluation result information, it is analyzed whether the performance of the target gearbox meets the current operating condition benchmark. If not, an abnormality cause analysis is performed, and a maintenance warning report is generated and pushed based on the abnormality cause analysis.

Images