Motorcycle power unit assembly line management system

Abstract

The invention provides a motorcycle power unit assembly line management system, relates to the field of intelligent detection, and adopts a signal processing and analysis technology based on deep learning to decompose operation sound signals of a motorcycle engine to be detected and extract and reinforce sound component features. Therefore, whether the running state of the motorcycle engine to be detected is abnormal or not is automatically judged according to the significant aggregation characteristics of the sound components in the full time domain, and a corresponding manual repair prompt signal is generated. Thus, the sound signals can be analyzed more deeply, fine fault features which may be omitted in a traditional method can be captured, and meanwhile, the objectivity of a detection result and the real-time performance of response are ensured through automatic sound signal processing.

Description

Technical Field

[0001] The present application relates to the field of intelligent detection, and more specifically, to a motorcycle power group assembly line management system. Background Art

[0002] The motorcycle power pack usually refers to the motorcycle engine and its related components. It is one of the most critical components of the motorcycle and is responsible for providing power to drive the motorcycle forward. Therefore, the operation status detection of the power pack is an indispensable part of the motorcycle power pack assembly line. Strict quality inspection can ensure that the power pack can operate stably under various conditions and reduce the risk of accidents.

[0003] However, traditional methods for detecting the operating status of motorcycle power groups usually rely on simple threshold judgments or auditory inspections, which may result in the inability to conduct in-depth analysis of the time-frequency characteristics, nonlinear characteristics, and complex patterns of sound signals. For example, certain faults may manifest themselves in the modulation characteristics or transient behavior of sound signals, and these subtle features are often difficult to capture through traditional auditory inspections or basic signal processing techniques. In addition, the traditional detection has a relatively low level of intelligence. Due to excessive reliance on the experience and auditory judgment of technicians, it is highly subjective, making it difficult to ensure the accuracy of the detection results, and it is also difficult to achieve rapid feedback and processing.

[0004] Therefore, there is a need for an optimized motorcycle power group assembly line management system. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a motorcycle power group assembly line management system, which uses deep learning-based signal processing and analysis technology to decompose the running sound signal of the motorcycle engine to be detected and extract and enhance the sound component features, so as to automatically judge whether there is an abnormality in the running state of the motorcycle engine to be detected according to the significant aggregation characteristics of each sound component in the full time domain, and generate a corresponding manual repair prompt signal. In this way, a more in-depth analysis of the sound signal can be performed, which is conducive to capturing subtle fault features that may be missed by traditional methods. At the same time, the automated sound signal processing ensures the objectivity of the detection results and the real-time response.

[0006] According to one aspect of the present application, a motorcycle power group assembly line management system is provided, which includes:

[0007] A motorcycle engine running sound signal acquisition module is used to acquire the running sound signal of the motorcycle engine to be detected;

[0008] A running sound signal decomposition module, used for performing CEEMDAN decomposition on the running sound signal to obtain a set of IMF components of the sound signal;

[0009] A sound signal IMF component feature extraction module, used for respectively passing each sound signal IMF component in the set of sound signal IMF components through a sound signal component feature extractor to obtain a set of sound signal IMF component feature vectors;

[0010] A sound signal IMF component feature reverse enhancement module, used for inputting the set of sound signal IMF component feature vectors into a self-supervised cluster center reverse feature enhancement selection network to obtain a set of reverse selected sound signal IMF component feature vectors;

[0011] A sound signal significant aggregation module, used for inputting the set of reversely selected sound signal IMF component feature vectors into a feature aggregation network based on feature energy local significance gating to obtain a sound signal significant aggregation representation vector as a sound signal significant aggregation representation feature;

[0012] The detection result generating module is used to obtain the detection result based on the significant aggregation representation features of the sound signal, and based on the detection result, determine whether to generate a manual rework prompt signal.

[0013] In the above-mentioned motorcycle power group assembly line management system, the sound signal IMF component feature extraction module is used to: respectively pass each sound signal IMF component in the set of sound signal IMF components through a sound signal component feature extractor based on a 1D-CNN model to obtain a set of sound signal IMF component feature vectors.

[0014] In the above-mentioned motorcycle power group assembly line management system, the sound signal IMF component feature reverse reinforcement module includes: a sound signal IMF component feature self-supervised clustering processing unit, which is used to perform self-supervised clustering processing based on energy measurement on the set of sound signal IMF component feature vectors to obtain the sound signal IMF component feature self-supervised cluster center representation vector; a sound signal IMF component feature anti-correlation coefficient calculation unit, which is used to calculate the anti-correlation coefficient between each sound signal IMF component feature vector in the set of sound signal IMF component feature vectors and the sound signal IMF component feature self-supervised cluster center representation vector to obtain a sequence of sound signal IMF component feature anti-correlation coefficients; a sound signal IMF component feature anti-correlation weight coefficient calculation unit, which is used to input the sequence of sound signal IMF component feature anti-correlation coefficients into a Sigmoid activation function to obtain the anti-correlation coefficient sequence of the sound signal IMF component feature. A sequence of normalized sound signal IMF component feature anti-correlation weight coefficients is obtained; a sound signal IMF component feature weighting unit is used to use each normalized sound signal IMF component feature anti-correlation weight coefficient in the sequence of normalized sound signal IMF component feature anti-correlation weight coefficients as a weight to weight each sound signal IMF component feature vector in the set of sound signal IMF component feature vectors to obtain a sequence of sound signal IMF component anti-enhancement feature vectors; a reverse selection sound signal IMF component feature generation unit is used to calculate the positional difference between each group of corresponding sound signal IMF component feature vectors and sound signal IMF component anti-enhancement feature vectors in the set of sound signal IMF component feature vectors and the sequence of sound signal IMF component anti-enhancement feature vectors to obtain the set of reverse selection sound signal IMF component feature vectors.

[0015] In the above-mentioned motorcycle power group assembly line management system, the sound signal IMF component feature self-supervised clustering processing unit includes: a sound signal IMF component feature energy measurement coefficient calculation subunit, which is used to determine the energy measurement coefficient of each sound signal IMF component feature vector based on the maximum value, mean and variance of each sound signal IMF component feature vector in the set of sound signal IMF component feature vectors to obtain a sequence of sound signal IMF component feature energy measurement coefficients; a sound signal IMF component feature activation energy measurement coefficient calculation subunit, which is used to input the sequence of sound signal IMF component feature energy measurement coefficients into a softmax function to obtain a sequence of sound signal IMF component feature activation energy measurement coefficients; a sound signal IMF component feature self-supervised cluster center representation generation subunit, which is used to use the sequence of sound signal IMF component feature activation energy measurement coefficients as a sequence of weights to calculate the position-weighted sum of the set of sound signal IMF component feature vectors to obtain the sound signal IMF component feature self-supervised cluster center representation vector.

[0016] In the above-mentioned motorcycle power group assembly line management system, the sound signal IMF component characteristic energy measurement coefficient calculation subunit is used to: extract the maximum value of the sound signal IMF component characteristic vector to obtain the sound signal IMF component characteristic maximum value; calculate the mean and variance of the sound signal IMF component characteristic vector respectively to obtain the sound signal IMF component characteristic mean and the sound signal IMF component characteristic variance; multiply the value obtained by adding the sound signal IMF component characteristic variance and the bias term by four to obtain the sound signal IMF component characteristic first energy measurement value; calculate the sound signal IMF The square of the difference between the maximum value of the component feature and the mean value of the IMF component feature of the sound signal is obtained to obtain the IMF component feature difference value of the sound signal; the modulated sound signal IMF component feature variance obtained by multiplying the sound signal IMF component feature variance by two, the value obtained by multiplying the bias term by two and the sound signal IMF component feature difference value are added to obtain the second energy measurement value of the IMF component feature of the sound signal; the first energy measurement value of the IMF component feature of the sound signal is divided by the second energy measurement value of the IMF component feature of the sound signal to obtain the energy measurement coefficient of the IMF component feature of the sound signal.

[0017] In the above-mentioned motorcycle power group assembly line management system, the sound signal IMF component feature anti-correlation coefficient calculation unit is used to: calculate the positional difference between the sound signal IMF component feature vector and the sound signal IMF component feature self-supervised cluster center representation vector to obtain the sound signal IMF component offset feature vector; calculate the covariance matrix between the sound signal IMF component feature vector and the sound signal IMF component feature self-supervised cluster center representation vector, and calculate the inverse of each eigenvalue in the covariance matrix to obtain the sound signal IMF component anti-correlation covariance matrix; calculate the transposed vector of the sound signal IMF component offset feature vector and the product of the sound signal IMF component anti-correlation covariance matrix and the sound signal IMF component offset feature vector to obtain the sound signal IMF component feature correlation value; calculate the square root of the sound signal IMF component feature correlation value to obtain the sound signal IMF component feature anti-correlation coefficient.

[0018] In the above-mentioned motorcycle power group assembly line management system, the sound signal significant aggregation module includes: a reverse selection sound signal IMF component characteristic energy level coefficient calculation unit, which is used to determine the characteristic energy level coefficient of each reverse selection sound signal IMF component characteristic vector in the set of reverse selection sound signal IMF component characteristic vectors based on the maximum value, minimum value, mean value and variance of each reverse selection sound signal IMF component characteristic vector in the set of reverse selection sound signal IMF component characteristic vectors to obtain a set of reverse selection sound signal IMF component characteristic energy level coefficients; a reverse selection sound signal IMF component characteristic energy level significance descriptor calculation unit, which is used to determine the scale of the local neighborhood and calculate the characteristic energy level of each reverse selection sound signal IMF component in the set of reverse selection sound signal IMF component characteristic energy level coefficients. The invention relates to a method for generating a sound signal significant aggregate representation vector comprising: a local neighborhood significance descriptor of a coefficient to obtain a set of reverse-selected sound signal IMF component feature energy level significance descriptors; a reverse-selected sound signal IMF component feature energy level significance descriptor mask unit, which is used to input the set of reverse-selected sound signal IMF component feature energy level significance descriptors into a mask module based on a gating function to obtain a set of mask-gated probabilistic reverse-selected sound signal IMF component feature energy level significance descriptors; and a sound signal significant aggregate representation generating subunit, which is used to use the set of mask-gated probabilistic reverse-selected sound signal IMF component feature energy level significance descriptors as a sequence of weights, and calculate the position-weighted sum of the set of reverse-selected sound signal IMF component feature vectors to obtain the sound signal significant aggregate representation vector.

[0019] In the above-mentioned motorcycle power group assembly line management system, the reverse selection sound signal IMF component characteristic energy level coefficient calculation unit is used to: respectively extract the maximum value and the minimum value in the reverse selection sound signal IMF component characteristic vector to obtain the reverse selection sound signal IMF component characteristic maximum value and the reverse selection sound signal IMF component characteristic minimum value; respectively calculate the mean and variance of the reverse selection sound signal IMF component characteristic vector to obtain the reverse selection sound signal IMF component characteristic mean and the reverse selection sound signal IMF component characteristic variance; add the reverse selection sound signal IMF component characteristic variance and the preset hyperparameter to obtain the reverse selection sound signal IMF component characteristic first energy level coefficient; calculate the reverse selection sound signal IMF component characteristic maximum value and the reverse selection sound signal IMF component characteristic variance; The square of the difference between the feature mean values ​​of the IMF components of the reverse-selected sound signal is calculated to obtain the maximum difference value of the IMF components of the reverse-selected sound signal; the square of the difference between the minimum feature value of the IMF components of the reverse-selected sound signal and the feature mean value of the IMF components of the reverse-selected sound signal is calculated to obtain the minimum difference value of the IMF components of the reverse-selected sound signal; the maximum difference value of the IMF components of the reverse-selected sound signal is added to the minimum difference value of the IMF components of the reverse-selected sound signal and the preset hyperparameter to obtain the second energy level coefficient of the IMF components of the reverse-selected sound signal; the first energy level coefficient of the IMF components of the reverse-selected sound signal is divided by the second energy level coefficient of the IMF components of the reverse-selected sound signal to obtain the energy level coefficient of the IMF components of the reverse-selected sound signal.

[0020] In the above-mentioned motorcycle power group assembly line management system, the reverse-selected sound signal IMF component feature energy level significance descriptor mask unit is used to: normalize each reverse-selected sound signal IMF component feature energy level significance descriptor in the set of reverse-selected sound signal IMF component feature energy level significance descriptors to obtain a set of normalized reverse-selected sound signal IMF component feature energy level significance descriptors; input each normalized reverse-selected sound signal IMF component feature energy level significance descriptor in the set of normalized reverse-selected sound signal IMF component feature energy level significance descriptors into a gating function for masking processing to obtain the set of mask-gated probabilistic reverse-selected sound signal IMF component feature energy level significance descriptors.

[0021] In the above-mentioned motorcycle power group assembly line management system, the detection result generation module is used to: input the sound signal significant aggregation representation vector into the classifier-based equipment performance detection module to obtain the detection result, and the detection result is used to indicate whether there is an abnormality in the operating state of the motorcycle engine to be detected; in response to the detection result that the operating state of the motorcycle engine to be detected is abnormal, generate a management result, and the management result is to generate the manual return repair prompt signal.

[0022] Compared with the prior art, the motorcycle power group assembly line management system provided by the present application uses deep learning-based signal processing and analysis technology to decompose the running sound signal of the motorcycle engine to be tested and extract and enhance the sound component features, so as to automatically judge whether there is an abnormality in the running state of the motorcycle engine to be tested based on the significant aggregation features of each sound component in the full time domain, and generate a corresponding manual repair prompt signal. In this way, the sound signal can be analyzed more deeply, which is conducive to capturing subtle fault features that may be missed by traditional methods. At the same time, the automated sound signal processing ensures the objectivity of the detection results and the real-time response. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0024] Figure 1 It is a system block diagram of a motorcycle power group assembly line management system according to an embodiment of the present application.

[0025] Figure 2 Schematic diagram of data flow of a motorcycle power group assembly line management system according to an embodiment of the present application.

[0026] Figure 3 It is a block diagram of a reverse enhancement module for IMF component characteristics of a sound signal in a motorcycle power group assembly line management system according to an embodiment of the present application.

[0027] Figure 4 It is a block diagram of a self-supervised clustering processing unit for IMF component features of sound signals in a motorcycle power group assembly line management system according to an embodiment of the present application.

[0028] Figure 5 It is a block diagram of a sound signal significant aggregation module in a motorcycle power group assembly line management system according to an embodiment of the present application. DETAILED DESCRIPTION

[0029] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0030] Motorcycle power pack is the general term for motorcycle engine and its related accessories. It plays a vital role in motorcycles and provides the necessary power for motorcycles. Therefore, ensuring that the operating status of the power pack is accurately detected is a key step to ensure the safe driving of motorcycles. Strict quality inspection processes help ensure that the power pack can work stably in different environments, thereby reducing potential accident risks.

[0031] However, there are some limitations in the current detection methods of the operating status of motorcycle power groups. Traditional methods often rely on basic threshold detection or auditory evaluation, which may not be able to deeply analyze the time-frequency characteristics, nonlinear characteristics, and complex signal patterns of sound signals. For example, some faults may appear in the modulation characteristics or transient changes of sound signals, and these subtle features are difficult to identify through traditional auditory inspection or basic signal processing technology. In addition, the intelligence level of traditional detection methods is limited, and they rely too much on the experience and auditory judgment of technicians, which not only increases the subjectivity of the detection results, but also affects the accuracy. At the same time, this method is difficult to achieve rapid feedback and processing, which limits the detection efficiency.

[0032] Based on this, the present application proposes an optimized motorcycle power group assembly line management system, which first arranges the components required for the motorcycle power group according to the established order and specifications, then performs an air blowing and dust removal operation on the components after the arrangement to remove the surface dust and impurities, and then pre-loads the components after the dust removal operation to ensure the accurate placement of the components, and then installs O-rings and screws to ensure the sealing and stability of the power group, and then inspects the installed semi-finished products. When the inspection is qualified, the semi-finished products enter the next step of shell installation, and the semi-finished products that fail the inspection will be manually repaired. The semi-finished products after manual repair need to be inspected again. If qualified, they will continue to enter the shell installation stage of the assembly line. The unqualified semi-finished products will be directly scrapped, and then the products after the shell installation will be functionally tested. The products that pass the functional test will enter the packaging and delivery stage.

[0033] Accordingly, when testing the motorcycle power group, it obtains the running sound signal of the motorcycle engine to be tested, and uses signal processing and analysis technology based on deep learning to decompose the running sound signal and extract and enhance the sound component features, so as to automatically judge whether there is any abnormality in the running state of the motorcycle engine to be tested based on the significant aggregation characteristics of each sound component in the full time domain, and generate a corresponding manual repair prompt signal. In this way, using deep learning technology, the system can perform a more in-depth time-frequency and nonlinear feature analysis of the sound signal, and capture subtle fault features that may be missed by traditional methods. At the same time, automated sound signal processing reduces the interference of human factors and ensures the objectivity of the test results and the real-time response.

[0034] Figure 1 It is a system block diagram of a motorcycle power group assembly line management system according to an embodiment of the present application.

[0035] Figure 2 Schematic diagram of data flow of a motorcycle power group assembly line management system according to an embodiment of the present application. Figure 1 and Figure 2 As shown, in the motorcycle power group assembly line management system 100, it includes: a motorcycle engine running sound signal acquisition module 110, used to acquire the running sound signal of the motorcycle engine to be detected; an running sound signal decomposition module 120, used to perform CEEMDAN decomposition on the running sound signal to obtain a set of sound signal IMF components; a sound signal IMF component feature extraction module 130, used to pass each sound signal IMF component in the set of sound signal IMF components through a sound signal component feature extractor to obtain a set of sound signal IMF component feature vectors; a sound signal IMF component feature reverse enhancement module 130. 40, used for inputting the set of sound signal IMF component feature vectors into the self-supervised cluster center inverse feature enhancement selection network to obtain a set of reversely selected sound signal IMF component feature vectors; the sound signal significant aggregation module 150, used for inputting the set of reversely selected sound signal IMF component feature vectors into the feature aggregation network based on feature energy local significance gating to obtain a sound signal significant aggregation representation vector as a sound signal significant aggregation representation feature; the detection result generation module 160, used for obtaining a detection result based on the sound signal significant aggregation representation feature, and based on the detection result, determining whether to generate an artificial rework prompt signal.

[0036] In the embodiment of the present application, the motorcycle engine running sound signal acquisition module 110 is used to acquire the running sound signal of the motorcycle engine to be detected. It should be understood that the running sound signal of the motorcycle engine to be detected contains sound frequency component information. Specifically, sound components of different frequencies can help identify specific faults. For example, too high a frequency may indicate wear or failure of certain components, while low-frequency abnormalities may be related to the overall unstable operation of the engine. In addition, the running sound signal also includes time domain characteristics such as the intensity and duration of the sound. The intensity and duration of the sound can reflect the stability of the engine under different working conditions. If the sound intensity fluctuates greatly, it may mean that the engine is under load changes or has a fault. In general, the running sound signal of the motorcycle engine to be detected contains rich frequency and time domain information, which provides an important basis for judging the running state of the engine. By conducting in-depth analysis of the running sound signal of the motorcycle engine to be detected, efficient fault detection of the motorcycle power group can be achieved.

[0037] In an embodiment of the present application, the operation sound signal decomposition module 120 is used to perform CEEMDAN decomposition on the operation sound signal to obtain a set of IMF components of the sound signal. Accordingly, it is considered that different local signals in the operation sound signal have different signal details and structural information. CEEMDAN is a signal processing technology based on empirical mode decomposition (EMD), which enhances the stability and effect of EMD by introducing auxiliary noise, and is suitable for processing nonlinear and non-stationary signals. The basic idea of ​​CEEMDAN is to add Gaussian white noise to the original signal, and then perform EMD decomposition on the new signal, and perform overall averaging on the multiple modal components generated to obtain a more stable and accurate decomposition result. This process will be repeated many times, and the amplitude of the white noise added each time will gradually decrease to ensure the accuracy of the final result. Therefore, in order to be able to more carefully analyze and process the temporal feature information and dynamic changes of each local signal, in the technical solution of the present application, the operation sound signal is decomposed by CEEMDAN to obtain a set of IMF components of the sound signal.

[0038] In an embodiment of the present application, the sound signal IMF component feature extraction module 130 is used to pass each sound signal IMF component in the set of the sound signal IMF components through a sound signal component feature extractor to obtain a set of sound signal IMF component feature vectors. Specifically, in an embodiment of the present application, the sound signal IMF component feature extraction module is used to: pass each sound signal IMF component in the set of the sound signal IMF components through a sound signal component feature extractor based on a 1D-CNN model to obtain a set of sound signal IMF component feature vectors. It should be understood that each sound signal IMF component in the set of the sound signal IMF components has a local information dependency within a local signal segment. Therefore, in the technical solution of the present application, each sound signal IMF component in the set of the sound signal IMF components is passed through a sound signal component feature extractor based on a 1D-CNN model to obtain a set of sound signal IMF component feature vectors. The essence of the sound signal component feature extractor based on the 1D-CNN model described here is a one-dimensional convolutional neural network model. By using a one-dimensional convolution kernel, it can effectively capture the local features in the IMF component of the sound signal, such as the change of specific frequency components, thereby effectively enhancing the ability to detect abnormal conditions of the engine.

[0039] In an embodiment of the present application, the sound signal IMF component feature reverse enhancement module 140 is used to input the set of sound signal IMF component feature vectors into the self-supervised cluster center reverse feature enhancement selection network to obtain a set of reverse selected sound signal IMF component feature vectors. Accordingly, considering that different sound signal IMF component features in the set of sound signal IMF component feature vectors have different degrees of correlation and influence with the essential central features of the sound IMF component. In order to be able to adaptively enhance each sound signal IMF component feature accordingly, in the technical solution of the present application, the set of sound signal IMF component feature vectors is input into the self-supervised cluster center reverse feature enhancement selection network to obtain a set of reverse selected sound signal IMF component feature vectors. Specifically, first, by measuring the energy distribution of each sound signal IMF component feature, the overall internal structure and pattern of the sound signal IMF component feature are captured and refined to calculate the sound signal IMF component feature self-supervised cluster center representation vector. Then, the anti-correlation coefficient between the sound signal IMF component feature vector and the self-supervised cluster center is calculated to quantify the feature difference between them, and the Sigmoid activation function is used to normalize and adaptively weight each feature vector to generate the sound signal IMF component anti-enhancement feature vector. Finally, the original sound signal IMF component feature vector and the sound signal IMF component anti-enhancement feature vector are subtracted by position to reversely filter out the sound signal IMF component features that are most relevant to the center feature, thereby obtaining the reverse selection sound signal IMF component feature vector containing feature expression capabilities.

[0040] Specifically, Figure 3 FIG. 1 is a block diagram of a reverse enhancement module for the IMF component characteristics of a sound signal in a motorcycle power group assembly line management system according to an embodiment of the present application. Figure 3As shown, the sound signal IMF component feature reverse enhancement module 140 includes: a sound signal IMF component feature self-supervised clustering processing unit 141, which is used to perform self-supervised clustering processing based on energy measurement on the set of the sound signal IMF component feature vectors to obtain the sound signal IMF component feature self-supervised cluster center representation vector; a sound signal IMF component feature anti-correlation coefficient calculation unit 142, which is used to calculate the anti-correlation coefficient between each sound signal IMF component feature vector in the set of the sound signal IMF component feature vectors and the sound signal IMF component feature self-supervised cluster center representation vector to obtain a sequence of sound signal IMF component feature anti-correlation coefficients; a sound signal IMF component feature anti-correlation weight coefficient calculation unit 143, which is used to input the sequence of sound signal IMF component feature anti-correlation coefficients into a Sigmoid activation function to obtain a normalized A sequence of anti-correlation weight coefficients of sound signal IMF component features; a sound signal IMF component feature weighting unit 144, used to use each normalized sound signal IMF component feature anti-correlation weight coefficient in the sequence of normalized sound signal IMF component feature anti-correlation weight coefficients as a weight, and weight each sound signal IMF component feature vector in the set of sound signal IMF component feature vectors to obtain a sequence of sound signal IMF component anti-enhancement feature vectors; a reverse selection sound signal IMF component feature generating unit 145, used to calculate the positional difference between each group of corresponding sound signal IMF component feature vectors and sound signal IMF component anti-enhancement feature vectors in the set of sound signal IMF component feature vectors and the sequence of sound signal IMF component anti-enhancement feature vectors to obtain the set of reverse selection sound signal IMF component feature vectors.

[0041] More specifically, Figure 4 FIG. 1 is a block diagram of a self-supervised clustering processing unit for IMF component features of a sound signal in a motorcycle power group assembly line management system according to an embodiment of the present application. Figure 4As shown, the sound signal IMF component feature self-supervised clustering processing unit 141 includes: a sound signal IMF component feature energy measurement coefficient calculation subunit 1411, which is used to determine the energy measurement coefficient of each sound signal IMF component feature vector in the set of the sound signal IMF component feature vectors based on the maximum value, mean and variance of each sound signal IMF component feature vector in the set of the sound signal IMF component feature vectors to obtain a sequence of sound signal IMF component feature energy measurement coefficients; a sound signal IMF component feature activation energy measurement coefficient calculation subunit 1412, which is used to input the sequence of sound signal IMF component feature energy measurement coefficients into a softmax function to obtain a sequence of sound signal IMF component feature activation energy measurement coefficients; a sound signal IMF component feature self-supervised cluster center representation generation subunit 1413, which is used to use the sequence of sound signal IMF component feature activation energy measurement coefficients as a sequence of weights to calculate the position-weighted sum of the set of the sound signal IMF component feature vectors to obtain the sound signal IMF component feature self-supervised cluster center representation vector.

[0042] More specifically, in the embodiment of the present application, the sound signal IMF component feature energy measurement coefficient calculation subunit is used to: extract the maximum value of the sound signal IMF component feature vector to obtain the maximum value of the sound signal IMF component feature; respectively calculate the mean and variance of the sound signal IMF component feature vector to obtain the sound signal IMF component feature mean and the sound signal IMF component feature variance; multiply the value obtained by adding the sound signal IMF component feature variance and the bias term by four to obtain the sound signal IMF component feature first energy measurement value; calculate the sound signal IMF component The square of the difference between the maximum value of the feature and the mean value of the IMF component feature of the sound signal is obtained to obtain the IMF component feature difference value of the sound signal; the modulated sound signal IMF component feature variance obtained by multiplying the IMF component feature variance of the sound signal by two is added to the value obtained by multiplying the bias term by two and the IMF component feature difference value of the sound signal to obtain the second energy measurement value of the IMF component feature of the sound signal; the first energy measurement value of the IMF component feature of the sound signal is divided by the second energy measurement value of the IMF component feature of the sound signal to obtain the energy measurement coefficient of the IMF component feature of the sound signal.

[0043] More specifically, in an embodiment of the present application, the sound signal IMF component feature anti-correlation coefficient calculation unit is used to: calculate the positional difference between the sound signal IMF component feature vector and the sound signal IMF component feature self-supervised cluster center representation vector to obtain the sound signal IMF component offset feature vector; calculate the covariance matrix between the sound signal IMF component feature vector and the sound signal IMF component feature self-supervised cluster center representation vector, and calculate the inverse of each eigenvalue in the covariance matrix to obtain the sound signal IMF component anti-correlation covariance matrix; calculate the transposed vector of the sound signal IMF component offset feature vector and the product of the sound signal IMF component anti-correlation covariance matrix and the sound signal IMF component offset feature vector to obtain the sound signal IMF component feature correlation value; calculate the square root of the sound signal IMF component feature correlation value to obtain the sound signal IMF component feature anti-correlation coefficient.

[0044] In the embodiment of the present application, specifically, the sound signal IMF component feature reverse enhancement module 140 is used to: input the set of sound signal IMF component feature vectors into the self-supervised cluster center reverse feature enhancement selection network, and process them with the following reverse feature enhancement formula to obtain the set of reverse selected sound signal IMF component feature vectors; wherein the reverse feature enhancement formula is:

[0045] X={x1,x2,...,x n}

[0046]

[0047]

[0048]

[0049] s ai =sigmoid[s i (x i ,v c )]

[0050] X f ={s a1 ·x1,s a2 ·x2,...,s an ·x n}

[0051]

[0052] Where, X is the set of IMF component feature vectors of the sound signal, x nis the nth sound signal IMF component feature vector in the set of sound signal IMF component feature vectors, x i is the set of the ith sound signal IMF component feature vector in the set of the sound signal IMF component feature vector, μ i and σ i 2 They represent the mean and variance of the eigenvector of the IMF component of the i-th sound signal, β is the bias term, max(x i ) represents the maximum value of the eigenvector of the IMF component of the i-th sound signal, e i is the energy measurement coefficient of the eigenvector of the IMF component of the ith sound signal, softmax(·) is the softmax function, n is the number of eigenvectors in the set of eigenvectors of the IMF component of the sound signal, v c is the center representation vector of the self-supervised cluster of the IMF component feature of the sound signal, (·) T is the transpose of the vector, S i is the covariance matrix between the feature vector of the IMF component of the ith sound signal and the center representation vector of the self-supervised cluster of the IMF component feature of the sound signal, s i (x i ,v c ) is the anti-correlation coefficient between the ith sound signal IMF component feature vector and the center representation vector of the sound signal IMF component feature self-supervised cluster, sigmoid(·) is the sigmoid function, s ai is the normalized sound signal IMF component feature anti-correlation weight coefficient corresponding to the ith sound signal IMF component feature vector, X f is a sequence of anti-enhancement feature vectors of the IMF components of the sound signal, is subtracted by position, and V is the set of IMF component feature vectors of the reversely selected sound signal.

[0053] In an embodiment of the present application, the sound signal significant aggregation module 150 is used to input the set of the reverse selected sound signal IMF component feature vectors into a feature aggregation network based on feature energy local significant gating to obtain a sound signal significant aggregation representation vector as a sound signal significant aggregation representation feature. It should be understood that the set of the reverse selected sound signal IMF component feature vectors is obtained by reverse reinforcement. Therefore, in order to be able to interact with each other based on significant features in the local signal domain to characterize the entire signal, in the technical solution of the present application, the set of the reverse selected sound signal IMF component feature vectors is input into a feature aggregation network based on feature energy local significant gating to obtain a sound signal significant aggregation representation vector. That is, the feature aggregation network based on feature energy local significant gating refines the feature representation by quantizing and aggregating the energy distribution in the local signal component domain to characterize the aggregated features of the entire signal component. In detail, the feature aggregation network based on feature energy local significant gating first calculates the energy level of each vector based on the maximum value, minimum value, mean and variance of each reverse selected sound signal IMF component feature vector to quantify the energy distribution of each feature. Next, the scale of the local neighborhood is determined, and the local neighborhood significance descriptor of the energy level coefficient of each reverse-selected sound signal IMF component feature vector is calculated to capture the energy distribution characteristics of each feature vector in the local neighborhood. Subsequently, the significance descriptor is input into the mask module based on the gating function to perform nonlinear probabilistic transformation through the gating function and adaptive feature screening through the mask function to achieve dynamic adjustment of feature significance, so as to screen out the most representative and discriminative sound signal IMF component features. Finally, the significance descriptor obtained after probabilization is used as a weight to calculate the weighted sum of the set of reverse-selected sound signal IMF component feature vectors to enhance the local significance and importance characteristics of the sound signal IMF component, and generate a comprehensive sound signal significance aggregation representation vector with rich expressiveness.

[0054] Specifically, Figure 5 FIG. 1 is a block diagram of a sound signal significant aggregation module in a motorcycle power group assembly line management system according to an embodiment of the present application. Figure 5As shown, the sound signal significant aggregation module 150 includes: a reverse selection sound signal IMF component characteristic energy level coefficient calculation unit 151, which is used to determine the characteristic energy level coefficient of each reverse selection sound signal IMF component characteristic vector in the set of reverse selection sound signal IMF component characteristic vectors based on the maximum value, minimum value, mean value and variance of each reverse selection sound signal IMF component characteristic vector in the set of reverse selection sound signal IMF component characteristic vectors to obtain a set of reverse selection sound signal IMF component characteristic energy level coefficients; a reverse selection sound signal IMF component characteristic energy level significance descriptor calculation unit 152, which is used to determine the scale of the local neighborhood, calculate the local neighborhood of each reverse selection sound signal IMF component characteristic energy level coefficient in the set of reverse selection sound signal IMF component characteristic energy level coefficients. A domain significance descriptor is used to obtain a set of reverse-selected sound signal IMF component feature energy level significance descriptors; a reverse-selected sound signal IMF component feature energy level significance descriptor mask unit 153 is used to input the set of reverse-selected sound signal IMF component feature energy level significance descriptors into a mask module based on a gating function to obtain a set of mask-gated probabilistic reverse-selected sound signal IMF component feature energy level significance descriptors; a sound signal significant aggregate representation generating subunit 154 is used to use the set of mask-gated probabilistic reverse-selected sound signal IMF component feature energy level significance descriptors as a sequence of weights, and calculate the position-weighted sum of the set of reverse-selected sound signal IMF component feature vectors to obtain the sound signal significant aggregate representation vector.

[0055] More specifically, in the embodiment of the present application, the reverse selection sound signal IMF component characteristic energy level coefficient calculation unit is used to: respectively extract the maximum value and the minimum value in the reverse selection sound signal IMF component characteristic vector to obtain the reverse selection sound signal IMF component characteristic maximum value and the reverse selection sound signal IMF component characteristic minimum value; respectively calculate the mean and the variance of the reverse selection sound signal IMF component characteristic vector to obtain the reverse selection sound signal IMF component characteristic mean value and the reverse selection sound signal IMF component characteristic variance; add the reverse selection sound signal IMF component characteristic variance and the preset hyperparameter to obtain the reverse selection sound signal IMF component characteristic first energy level coefficient; calculate the reverse selection sound signal IMF component characteristic maximum value and the reverse selection sound signal IMF component characteristic variance; The square of the difference between the characteristic mean values ​​of the IMF component of the selected sound signal is used to obtain the maximum difference value of the IMF component of the reverse selected sound signal; the square of the difference between the minimum characteristic value of the IMF component of the reverse selected sound signal and the characteristic mean value of the IMF component of the reverse selected sound signal is calculated to obtain the minimum difference value of the IMF component of the reverse selected sound signal; the maximum difference value of the IMF component of the reverse selected sound signal is added to the minimum difference value of the IMF component of the reverse selected sound signal and the preset hyperparameter to obtain the second energy level coefficient of the IMF component characteristic of the reverse selected sound signal; the first energy level coefficient of the IMF component characteristic of the reverse selected sound signal is divided by the second energy level coefficient of the IMF component characteristic of the reverse selected sound signal to obtain the energy level coefficient of the IMF component characteristic of the reverse selected sound signal.

[0056] More specifically, in an embodiment of the present application, the reverse-selected sound signal IMF component feature energy level significance descriptor mask unit is used to: normalize each reverse-selected sound signal IMF component feature energy level significance descriptor in the set of reverse-selected sound signal IMF component feature energy level significance descriptors to obtain a set of normalized reverse-selected sound signal IMF component feature energy level significance descriptors; input each normalized reverse-selected sound signal IMF component feature energy level significance descriptor in the set of normalized reverse-selected sound signal IMF component feature energy level significance descriptors into a gating function for masking processing to obtain the set of mask-gated probabilistic reverse-selected sound signal IMF component feature energy level significance descriptors.

[0057] In the embodiment of the present application, specifically, the sound signal significant aggregation module 150 is used to: input the set of the reverse selected sound signal IMF component feature vectors into a feature aggregation network based on feature energy local significance gating, and process it with the following feature aggregation formula to obtain the sound signal significant aggregation representation vector; wherein the feature aggregation formula is:

[0058] V={v1,v2,...,v k ,...,v n}

[0059]

[0060] w si =mask(w i )

[0061]

[0062] Y={y1,y2,...,y k ,...,y n},y i =w si ·v i

[0063]

[0064] Wherein, V is the set of the feature vectors of the IMF components of the reverse selected sound signal, n represents the number of feature vectors in the set of the IMF components of the reverse selected sound signal, μ vi and σ vi 2 are the mean and variance of the eigenvector of the IMF component of the i-th reverse selected sound signal, max(v i ) and min(v i ) are the maximum and minimum values ​​of the eigenvector of the IMF component of the i-th reverse selected sound signal, ε is the preset hyperparameter, e vi is the characteristic energy level coefficient of the eigenvector of the IMF component of the reverse selected sound signal, e vq is the characteristic energy level coefficient of the qth reverse selected sound signal IMF component eigenvector, Num r Indicates that the i The number of characteristic energy coefficients in the local neighborhood centered at ci is the reverse selected sound signal IMF component feature energy level significance descriptor corresponding to the i-th reverse selected sound signal IMF component feature vector, w i is the ith normalized reverse-selected sound signal IMF component feature energy level significance descriptor in the set of normalized reverse-selected sound signal IMF component feature energy level significance descriptors, mask(·) is masking processing, and w siis the ith masked gated probabilistic reverse selected sound signal IMF component feature energy level significance descriptor in the set of masked gated probabilistic reverse selected sound signal IMF component feature energy level significance descriptors, θ is a predetermined threshold, Y is the set of weighted feature vectors of the reverse selected sound signal IMF component, y i is the i-th reverse selected sound signal IMF component weighted feature vector in the set of reverse selected sound signal IMF component weighted feature vectors, v s A significant aggregation representation vector is generated for the sound signal.

[0065] In the embodiment of the present application, the detection result generation module 160 is used to obtain the detection result based on the significant aggregation representation feature of the sound signal, and to determine whether to generate an artificial rework prompt signal based on the detection result. Specifically, in the embodiment of the present application, the detection result generation module is used to: input the significant aggregation representation vector of the sound signal into the equipment performance detection module based on the classifier to obtain the detection result, and the detection result is used to indicate whether the running state of the motorcycle engine to be detected is abnormal; in response to the detection result that the running state of the motorcycle engine to be detected is abnormal, generate a management result, and the management result is to generate the artificial rework prompt signal. That is, the sound signal significant aggregation representation feature obtained by local saliency of the set of IMF component feature vectors of the reverse selection sound signal is used for classification processing, so as to automatically determine whether the running state of the motorcycle engine to be detected is abnormal, and generate the corresponding artificial rework prompt signal. In this way, using deep learning technology, the system can perform a more in-depth time-frequency characteristic and nonlinear characteristic analysis of the sound signal, and capture the subtle fault characteristics that may be missed by traditional methods. At the same time, the automated sound signal processing reduces the interference of human factors and ensures the objectivity of the detection results and the real-time response.

[0066] In particular, in a preferred example, when the set of sound signal IMF component feature vectors respectively expresses the one-dimensional local correlation features of each sound signal IMF component in the set of sound signal IMF components, the self-supervised cluster center inverse feature enhancement selection network will perform distribution enhancement based on local sample domain center self-supervision, so that when the feature aggregation network based on feature energy local significance gating performs aggregation based on local sample domain distribution energy significance under the global sample domain, the sound signal significant aggregation representation vector obtained has a distribution regression mode offset caused by the distribution energy aggregation difference, that is, it causes the sound signal significant aggregation representation vector to be input into the classifier-based device performance detection module to obtain an overflow of the detection result, affecting its accuracy.

[0067] Based on this, the sound signal significant aggregation representation vector is input into the classifier-based equipment performance detection module to obtain the detection result, which specifically includes: determining the abnormal probability value corresponding to the abnormality of the running state of the motorcycle engine to be detected obtained by inputting the sound signal significant aggregation representation vector into the classifier-based equipment performance detection module, and passing the sound signal significant aggregation representation vector through a probabilistic activation function to obtain a probabilistic sound signal significant aggregation representation vector; subtracting the probabilistic sound signal significant aggregation representation vector from a unit feature vector, and using the feature vector of the point subtraction result as the base to calculate a power function with the abnormal probability value as the exponent; and adding the power function vector to the probabilistic activation function. The sound signal significant aggregation representation vector is dot-multiplied to obtain the sound signal significant aggregation representation intermediate feature vector; the absolute value of each eigenvalue of the probabilistic sound signal significant aggregation representation vector is used as the base, a power function with the inverse of the abnormal probability value as the exponent is calculated, and all eigenvalues ​​are summed to obtain the sound signal significant aggregation representation bias eigenvalue; after multiplying the sound signal significant aggregation representation bias eigenvalue by a weight serving as a hyperparameter, the result is dot-multiplied with the sound signal significant aggregation representation intermediate feature vector to obtain an optimized sound signal significant aggregation representation vector; the optimized sound signal significant aggregation representation vector is input into the classifier-based device performance detection module to obtain a detection result.

[0068] Wherein, let the probabilistic sound signal significant aggregation representation vector be denoted as V, then the optimized sound signal significant aggregation representation vector V' is expressed as:

[0069]

[0070] Where V is the probabilistic sound signal significant aggregation representation vector, I is the unit feature vector, and v i ∈V, is the eigenvalue of the probabilistic sound signal significant aggregation representation vector V, ⊙ is the position point multiplication, is subtracted by location point, p is the abnormal probability value, is a power function that calculates the eigenvalues ​​of each position in the eigenvector with the abnormal probability value as the exponent, is added by location point, α is the weight as a hyperparameter, and V' is the optimized sound signal significant aggregation representation vector.

[0071] That is, for the sound signal significant aggregation representation vector, based on the convergence limit of the probability distribution of the Cauchy-Hadamard form power series, the feature set of the sound signal significant aggregation representation vector is represented by a power series based on the convergence probability, and a set convergence bounded bias based on each eigenvalue of the sound signal significant aggregation representation vector as a convergence radius representation is added, so as to avoid the invalid overflow of the classification result caused by the deviation of the distribution mode of the sound signal significant aggregation representation vector caused by the gradient propagation of the model parameters based on the convergence probability in the regression process of the equipment performance detection module based on the classifier under the unified probability convergence limit standard, thereby improving the accuracy of the detection result obtained by the equipment performance detection module based on the classifier. In this way, using deep learning technology, the system can perform a more in-depth time-frequency and nonlinear feature analysis of the sound signal, and capture subtle fault features that may be missed by traditional methods. At the same time, automated sound signal processing reduces the interference of human factors and ensures the objectivity of the detection results and the real-time response.

[0072] In summary, the motorcycle power group assembly line management system 100 based on the embodiment of the present application is explained, which uses deep learning-based signal processing and analysis technology to decompose the running sound signal of the motorcycle engine to be detected and extract and enhance the sound component features, so as to automatically judge whether there is an abnormality in the running state of the motorcycle engine to be detected according to the significant aggregation characteristics of each sound component in the full time domain, and generate a corresponding manual repair prompt signal. In this way, the sound signal can be analyzed more deeply, which is conducive to capturing subtle fault features that may be missed by traditional methods. At the same time, the automated sound signal processing ensures the objectivity of the detection results and the real-time response.

[0073] As described above, the motorcycle power group assembly line management system 100 according to the embodiment of the present application can be implemented in various terminal devices, such as a server for motorcycle power group assembly line management. In one example, the motorcycle power group assembly line management system 100 according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the motorcycle power group assembly line management system 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the motorcycle power group assembly line management system 100 can also be one of the many hardware modules of the terminal device.

[0074] Alternatively, in another example, the motorcycle power group assembly line management system 100 and the terminal device may also be separate devices, and the motorcycle power group assembly line management system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0075] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details of the above application are only for the purpose of illustration and ease of understanding, rather than limitation, and the above details do not limit the present application to the use of the above specific details to be implemented.

Claims

1. A motorcycle power group assembly line management system, characterized in that: include: A motorcycle engine running sound signal acquisition module is used to acquire the running sound signal of the motorcycle engine to be detected; A running sound signal decomposition module, used for performing CEEMDAN decomposition on the running sound signal to obtain a set of IMF components of the sound signal; A sound signal IMF component feature extraction module, used for respectively passing each sound signal IMF component in the set of sound signal IMF components through a sound signal component feature extractor to obtain a set of sound signal IMF component feature vectors; A sound signal IMF component feature reverse enhancement module, used for inputting the set of sound signal IMF component feature vectors into a self-supervised cluster center reverse feature enhancement selection network to obtain a set of reverse selected sound signal IMF component feature vectors; A sound signal significant aggregation module, used for inputting the set of reversely selected sound signal IMF component feature vectors into a feature aggregation network based on feature energy local significance gating to obtain a sound signal significant aggregation representation vector as a sound signal significant aggregation representation feature; The detection result generating module is used to obtain the detection result based on the significant aggregation representation features of the sound signal, and based on the detection result, determine whether to generate a manual rework prompt signal.

2. The motorcycle power group assembly line management system according to claim 1, characterized in that: The sound signal IMF component feature extraction module is used to: respectively pass each sound signal IMF component in the set of sound signal IMF components through a sound signal component feature extractor based on a 1D-CNN model to obtain a set of sound signal IMF component feature vectors.

3. The motorcycle power group assembly line management system according to claim 2, characterized in that: The sound signal IMF component feature reverse enhancement module includes: A sound signal IMF component feature self-supervised clustering processing unit, used for performing self-supervised clustering processing based on energy measurement on the set of sound signal IMF component feature vectors to obtain a sound signal IMF component feature self-supervised cluster center representation vector; A sound signal IMF component feature anti-correlation coefficient calculation unit, used to calculate the anti-correlation coefficient between each sound signal IMF component feature vector in the set of the sound signal IMF component feature vectors and the center representation vector of the sound signal IMF component feature self-supervised cluster to obtain a sequence of sound signal IMF component feature anti-correlation coefficients; A sound signal IMF component feature anti-correlation weight coefficient calculation unit, used for inputting the sequence of sound signal IMF component feature anti-correlation weight coefficients into a Sigmoid activation function to obtain a sequence of normalized sound signal IMF component feature anti-correlation weight coefficients; A sound signal IMF component feature weighting unit, configured to use each normalized sound signal IMF component feature anti-correlation weight coefficient in the sequence of normalized sound signal IMF component feature anti-correlation weight coefficients as a weight, and to weight each sound signal IMF component feature vector in the set of sound signal IMF component feature vectors to obtain a sequence of sound signal IMF component anti-enhancement feature vectors; The reverse selection sound signal IMF component feature generation unit is used to calculate the positional difference between each group of corresponding sound signal IMF component feature vectors and sound signal IMF component anti-enhancement feature vectors in the set of the sound signal IMF component feature vectors and the sequence of the sound signal IMF component anti-enhancement feature vectors to obtain the set of reverse selected sound signal IMF component feature vectors.

4. The motorcycle power group assembly line management system according to claim 3, characterized in that: The sound signal IMF component feature self-supervised clustering processing unit comprises: A sound signal IMF component feature energy measurement coefficient calculation subunit, used to determine the energy measurement coefficient of each sound signal IMF component feature vector in the set of sound signal IMF component feature vectors based on the maximum value, mean value and variance of each sound signal IMF component feature vector to obtain a sequence of sound signal IMF component feature energy measurement coefficients; A sound signal IMF component feature activation energy measurement coefficient calculation subunit, used for inputting the sequence of sound signal IMF component feature energy measurement coefficients into a softmax function to obtain a sequence of sound signal IMF component feature activation energy measurement coefficients; The sound signal IMF component feature self-supervised cluster center representation generating subunit is used to calculate the position-weighted sum of the set of sound signal IMF component feature vectors using the sequence of sound signal IMF component feature activation energy measurement coefficients as a sequence of weights to obtain the sound signal IMF component feature self-supervised cluster center representation vector.

5. The motorcycle power group assembly line management system according to claim 4, characterized in that: The sound signal IMF component characteristic energy measurement coefficient calculation subunit is used to: Extracting the maximum value of the IMF component feature vector of the sound signal to obtain the maximum value of the IMF component feature of the sound signal; Calculating the mean and variance of the sound signal IMF component feature vector respectively to obtain the sound signal IMF component feature mean and the sound signal IMF component feature variance; The value obtained by adding the characteristic variance of the IMF component of the sound signal and the bias term is multiplied by four to obtain a first energy measurement value of the characteristic of the IMF component of the sound signal; Calculate the square of the difference between the maximum value of the sound signal IMF component feature and the mean value of the sound signal IMF component feature to obtain the sound signal IMF component feature difference value; The second energy measurement value of the IMF component feature of the sound signal is obtained by adding the IMF component feature variance of the modulated sound signal obtained by multiplying the IMF component feature variance of the sound signal by two, the value obtained by multiplying the bias term by two, and the IMF component feature difference value of the sound signal; The first energy measurement value of the sound signal IMF component feature is divided by the second energy measurement value of the sound signal IMF component feature to obtain the sound signal IMF component feature energy measurement coefficient.

6. The motorcycle power group assembly line management system according to claim 5, characterized in that: The sound signal IMF component characteristic anti-correlation coefficient calculation unit is used to: Calculating the position difference between the sound signal IMF component feature vector and the sound signal IMF component feature self-supervised cluster center representation vector to obtain the sound signal IMF component offset feature vector; Calculate the covariance matrix between the sound signal IMF component eigenvector and the sound signal IMF component feature self-supervised cluster center representation vector, and calculate the inverse of each eigenvalue in the covariance matrix to obtain the sound signal IMF component anti-correlation covariance matrix; Calculate the transposed vector of the sound signal IMF component offset feature vector and the product of the sound signal IMF component anti-correlation covariance matrix and the sound signal IMF component offset feature vector to obtain the sound signal IMF component feature correlation value; The square root of the IMF component characteristic correlation value of the sound signal is calculated to obtain the IMF component characteristic anti-correlation coefficient of the sound signal.

7. The motorcycle power group assembly line management system according to claim 6, characterized in that: The sound signal significant aggregation module includes: A reverse-selected sound signal IMF component characteristic energy level coefficient calculation unit, used to determine the characteristic energy level coefficient of each reverse-selected sound signal IMF component characteristic vector in the set of reverse-selected sound signal IMF component characteristic vectors based on the maximum value, minimum value, mean value and variance of each reverse-selected sound signal IMF component characteristic vector in the set of reverse-selected sound signal IMF component characteristic vectors to obtain a set of reverse-selected sound signal IMF component characteristic energy level coefficients; A reverse-selected sound signal IMF component feature energy level significance descriptor calculation unit is used to determine the scale of the local neighborhood, calculate the local neighborhood significance descriptor of each reverse-selected sound signal IMF component feature energy level coefficient in the set of reverse-selected sound signal IMF component feature energy level coefficients to obtain a set of reverse-selected sound signal IMF component feature energy level significance descriptors; A reverse selection sound signal IMF component feature energy level significance descriptor mask unit, used for inputting the set of reverse selected sound signal IMF component feature energy level significance descriptors into a mask module based on a gating function to obtain a set of masked gated probabilistic reverse selected sound signal IMF component feature energy level significance descriptors; The sound signal significant aggregate representation generating subunit is used to calculate the position-weighted sum of the set of the reverse-selected sound signal IMF component feature energy level significance descriptors using the mask-gated probabilistic reverse selection as a sequence of weights to obtain the sound signal significant aggregate representation vector.

8. The motorcycle power group assembly line management system according to claim 7, characterized in that: The reverse selection sound signal IMF component characteristic energy level coefficient calculation unit is used to: Respectively extracting the maximum value and the minimum value in the IMF component feature vector of the reverse-selected sound signal to obtain the maximum value of the IMF component feature of the reverse-selected sound signal and the minimum value of the IMF component feature of the reverse-selected sound signal; Respectively calculating the mean and variance of the IMF component feature vector of the reverse selected sound signal to obtain the IMF component feature mean and the IMF component feature variance of the reverse selected sound signal; Adding the IMF component characteristic variance of the reverse selected sound signal and the preset hyperparameter to obtain the first energy level coefficient of the IMF component characteristic of the reverse selected sound signal; Calculate the square of the difference between the maximum value of the IMF component feature of the reverse-selected sound signal and the mean value of the IMF component feature of the reverse-selected sound signal to obtain the maximum difference value of the IMF component of the reverse-selected sound signal; Calculating the square of the difference between the minimum value of the IMF component characteristic of the reverse-selected sound signal and the mean value of the IMF component characteristic of the reverse-selected sound signal to obtain the minimum difference value of the IMF component of the reverse-selected sound signal; Adding the maximum difference value of the IMF component of the reverse-selected sound signal, the minimum difference value of the IMF component of the reverse-selected sound signal and the preset hyperparameter to obtain the second energy level coefficient of the IMF component feature of the reverse-selected sound signal; The first energy level coefficient of the reverse-selected sound signal IMF component characteristic is divided by the second energy level coefficient of the reverse-selected sound signal IMF component characteristic to obtain the energy level coefficient of the reverse-selected sound signal IMF component characteristic.

9. The motorcycle power group assembly line management system according to claim 8, characterized in that: The reverse selection sound signal IMF component feature energy level significance descriptor mask unit is used to: Normalizing each reverse-selected sound signal IMF component feature energy level significance descriptor in the set of reverse-selected sound signal IMF component feature energy level significance descriptors to obtain a set of normalized reverse-selected sound signal IMF component feature energy level significance descriptors; Each normalized inverse-selected sound signal IMF component feature energy level significance descriptor in the set of normalized inverse-selected sound signal IMF component feature energy level significance descriptors is input into a gating function for masking processing to obtain the set of masked gated probabilistic inverse-selected sound signal IMF component feature energy level significance descriptors.

10. The motorcycle power group assembly line management system according to claim 9, characterized in that: The detection result generating module is used for: Inputting the significant aggregation representation vector of the sound signal into a classifier-based equipment performance detection module to obtain the detection result, wherein the detection result is used to indicate whether the running state of the engine of the motorcycle to be detected is abnormal; In response to the detection result that the running state of the engine of the motorcycle to be detected is abnormal, a management result is generated, and the management result is the generation of the manual repair prompt signal.

Images