A planetary half-shaft gear detection method

Abstract

The invention discloses a planetary axle gear detection method, belonging to the field of planetary axle gear fault detection technology. The invention sets a vibration sensor on the differential to collect the straight and turning vibration signals of the vehicle at different speeds; performs EMD decomposition on the turning and straight vibration signals at the same speed, extracts the average envelope to obtain the turning envelope difference signal; adds them up in stages, calculates the difference energy mean and impact energy value of each level of synthetic difference signal; calculates the evolution value and constructs the feature vector according to the difference energy mean and impact energy value of the same level synthetic difference signal at multiple speeds; finally uses a multi-input BP neural network to process the feature vector and the evolution value to obtain the fault value of the planetary axle gear. The invention solves the problem of low accuracy of existing fault detection.

Description

Technical Field

[0001] The present invention relates to the technical field of planetary axle gear fault detection, and in particular to a planetary axle gear fault detection method. Background Art

[0002] In modern automobile transmission systems, planetary axle gears are key components, and their performance directly affects the vehicle's power transmission and driving stability. Existing fault detection technologies mainly rely on methods such as vibration analysis, temperature monitoring, and oil analysis. Although these traditional technologies can identify abnormal conditions of gears to a certain extent, they often have problems such as insufficient sensitivity, high false alarm rates, and inability to monitor in real time. Especially under complex working conditions, such as when the vehicle is driving straight and turning at different speeds, existing technologies are difficult to effectively distinguish between normal vibration and fault vibration, and there is a problem of low fault detection accuracy. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, a planetary half-shaft gear detection method provided by the present invention solves the problem of low fault detection accuracy in the prior art.

[0004] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a planetary half-shaft gear detection method, comprising the following steps:

[0005] S1. A vibration sensor is arranged on the differential to collect vibration signals of the vehicle when it is traveling straight and turning at different speeds, and obtain a plurality of straight vibration signals and a plurality of turning vibration signals;

[0006] S2. At the same speed, perform EMD decomposition on the turning and straight-moving vibration signals respectively, extract the average envelope under each decomposition, and obtain the turning envelope difference signal of multiple decompositions;

[0007] S3, adding the turning envelope difference signals decomposed multiple times at the same speed in different levels, and calculating the difference energy mean and impact energy value of the synthetic difference signals at each level;

[0008] S4, calculating the evolution value of each level of synthetic difference signal according to the difference energy mean and impact energy value of the same level synthetic difference signal at multiple speeds;

[0009] S5. constructing a feature vector of each level of synthetic difference signal according to the difference energy mean and impact energy value of the same level synthetic difference signal at multiple speeds;

[0010] S6. Use a multi-input BP neural network to process the characteristic vector and evolution value of the multi-level synthetic difference signal to obtain the fault value of the planetary half-shaft gear.

[0011] Further, S2 includes the following sub-steps:

[0012] S21, performing EMD decomposition on the turning vibration signal at each speed, extracting the average envelope of each decomposition process, and obtaining multiple turning average envelopes;

[0013] S22, performing EMD decomposition on the straight-moving vibration signal at each speed, extracting the average envelope of each decomposition process, and obtaining multiple straight-moving average envelopes;

[0014] S23. At the same speed, subtract the average envelope of the turning decomposed for the ith time from the average envelope of the straight line decomposed for the ith time to obtain a difference signal of the turning envelope decomposed for the ith time, where i is a positive integer.

[0015] Furthermore, S3 includes the following sub-steps:

[0016] S31, decompose the first ~ the second The first-order synthetic difference signal is obtained by adding the second-order decomposed turning envelope difference signals, where N is the number of turning envelope difference signals, This is a round down operation;

[0017] S32, the same speed down Decomposition~ The second-order decomposed turning envelope difference signals are added to obtain a second-order synthetic difference signal;

[0018] S33, adding the remaining decomposed turning envelope difference signals at the same speed to obtain a third-level synthetic difference signal;

[0019] S34. Calculate the difference energy mean and impact energy value for each level of synthetic difference signal.

[0020] Further, S34 includes the following sub-steps:

[0021] S341, dividing each level of synthesized difference signal into multiple segments, and calculating the difference energy value on each segment;

[0022] S342, taking the average of the difference energy values ​​of each segment to obtain the difference energy average;

[0023] S343. Filter out the difference energy values ​​that are greater than the difference energy mean value, and calculate the impact energy value.

[0024] Furthermore, the formula for calculating the difference energy value in S341 is: , where G j is the difference energy value of the jth segment, g j,m is the mth amplitude on the jth segment, j and m are positive integers, || is the absolute value, and M is the length of a signal segment;

[0025] The formula for calculating the impact energy value in S343 is: , where ζ impact is the impact energy value, G impact,n is the nth difference energy value greater than the mean difference energy value, G avg is the difference energy mean, n is a positive integer, and L is the number of difference energy values ​​greater than the difference energy mean.

[0026] Further, S4 includes the following sub-steps:

[0027] S41, calculating a first evolution coefficient according to a ratio of a second speed to a first speed composite difference signal of the same level in terms of difference energy mean and impact energy value, wherein the second speed is greater than the first speed;

[0028] S42, calculating a second evolution coefficient according to a ratio of the third speed to the second speed in terms of the difference energy mean and the impact energy value of the same-level synthetic difference signal, wherein the third speed is greater than the second speed;

[0029] S43: Add the first evolution coefficient and the second evolution coefficient to obtain an evolution value of the synthetic difference signal of this level.

[0030] Furthermore, the formula for calculating the first evolution coefficient in S41 is: , where θ 1 is the first evolution coefficient, G avg,2,k is the difference energy mean of the k-th level synthetic difference signal at the second speed, G avg,1,k is the difference energy mean of the k-th level synthetic difference signal at the first speed, ζ impact,2,k is the impact energy value of the k-th level synthetic difference signal at the second speed, ζ impact,1,k is the impact energy value of the k-th level synthetic difference signal at the first speed, k is a positive integer;

[0031] The formula for calculating the second evolution coefficient in S42 is: , where θ 2 is the second evolution coefficient, G avg,3,k is the difference energy mean of the k-th level synthetic difference signal at the third speed, ζ impact,3,k is the impact energy value of the kth level synthetic difference signal at the third speed.

[0032] Further, S5 includes the following sub-steps:

[0033] S51, adding the difference energy means of the same-level synthetic difference signals at multiple speeds to obtain a first eigenvalue;

[0034] S52, adding the impact energy values ​​of the same-level synthetic difference signals at multiple speeds to obtain a second characteristic value;

[0035] S53: Use the first eigenvalue and the second eigenvalue as elements to construct a eigenvector of the synthetic difference signal of this level.

[0036] Further, the multi-input BP neural network in S6 includes: a first fully connected layer, a second fully connected layer, a third fully connected layer, a multiplier M1, a multiplier M2, a multiplier M3 and a BP neural network;

[0037] The input end of the first fully connected layer is used to input the feature vector of the first-level synthetic difference signal;

[0038] The input end of the second fully connected layer is used to input the feature vector of the second-level synthetic difference signal;

[0039] The input end of the third fully connected layer is used to input the feature vector of the third-level synthetic difference signal;

[0040] A first input terminal of the multiplier M1 is connected to the output terminal of the first fully connected layer, and a second input terminal thereof is used to input the evolution value of the first-level synthetic difference signal;

[0041] A first input terminal of the multiplier M2 is connected to an output terminal of the second fully connected layer, and a second input terminal thereof is used to input an evolution value of the second-level synthetic difference signal;

[0042] A first input terminal of the multiplier M3 is connected to the output terminal of the third fully connected layer, and a second input terminal thereof is used to input the evolution value of the third-level synthetic difference signal;

[0043] The input end of the BP neural network is connected to the output end of the multiplier M1, the output end of the multiplier M2 and the output end of the multiplier M3 respectively, and the output end thereof serves as the output end of the multi-input BP neural network.

[0044] Furthermore, the expressions of the first fully connected layer, the second fully connected layer, and the third fully connected layer are all: , where S is the output of the fully connected layer, C r is the rth eigenvalue of the fully connected layer input, ω r C r The weight of b r C r The bias of , r is a positive integer ranging from 1 to 2.

[0045] The beneficial effects of the present invention are:

[0046] 1. The present invention collects vibration signals when going straight and turning at different speeds, performs EMD decomposition on the turning and straight vibration signals respectively, extracts the average envelope after each decomposition, subtracts the average envelope of turning from the average envelope of straight running, and obtains a turning envelope difference signal. When the vehicle is going straight and turning, the stress conditions and motion states of the planetary half-shaft gears are different, and the vibration signal characteristics generated are also different. When going straight, the planetary half-shaft gears are stationary relative to the half-shaft gears on the output shaft, and the planetary half-shaft gears do not rotate. Therefore, the complex vibration signal can be decomposed through EMD decomposition to obtain the average envelope. The turning envelope difference signal obtained by subtracting the average envelope after extracting the average envelope can amplify the signal characteristic changes caused by the fault, make the fault characteristics more prominent, and effectively distinguish the fault vibration caused by the planetary half-shaft gears.

[0047] 2. The present invention can analyze signal characteristics from different levels and further explore the fault information hidden in the signal through this hierarchical calculation of energy values. The difference energy mean reflects the overall energy distribution of the signal, while the impact energy value is more sensitive to the impact component in the signal and can highlight the impact characteristics caused by the fault. By calculating and analyzing different energy values, the signal characteristics in normal state and fault state can be more clearly distinguished, thereby improving the accuracy of fault detection and reducing the false alarm rate.

[0048] 3. The present invention calculates the evolution value based on the difference energy mean and impact energy value of the same-level synthetic difference signal at multiple speeds. The evolution value can reflect the state change trend of the planetary axle gear as the vehicle running state changes, and reflects the deterioration of its difference energy mean and impact energy value as the vehicle speed increases.

[0049] 4. The present invention uses a multi-input BP neural network to process the characteristic vectors and evolution values ​​of multi-level synthetic difference signals to obtain fault values, comprehensively considers multiple factors, accurately determines the fault state of the planetary half-shaft gear, and improves the accuracy of fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 is a flow chart of a planetary half-shaft gear detection method;

[0051] Figure 2 This is a schematic diagram of the structure of a multi-input BP neural network. DETAILED DESCRIPTION

[0052] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0053] like Figure 1 As shown, a planetary half-shaft gear detection method comprises the following steps:

[0054] S1. A vibration sensor is arranged on the differential to collect vibration signals of the vehicle when it is traveling straight and turning at different speeds, and obtain a plurality of straight vibration signals and a plurality of turning vibration signals;

[0055] S2. At the same speed, perform EMD decomposition on the turning and straight-moving vibration signals respectively, extract the average envelope under each decomposition, and obtain the turning envelope difference signal of multiple decompositions;

[0056] S3, adding the turning envelope difference signals decomposed multiple times at the same speed in different levels, and calculating the difference energy mean and impact energy value of the synthetic difference signals at each level;

[0057] S4, calculating the evolution value of each level of synthetic difference signal according to the difference energy mean and impact energy value of the same level synthetic difference signal at multiple speeds;

[0058] S5. constructing a feature vector of each level of synthetic difference signal according to the difference energy mean and impact energy value of the same level synthetic difference signal at multiple speeds;

[0059] S6. Use a multi-input BP neural network to process the characteristic vector and evolution value of the multi-level synthetic difference signal to obtain the fault value of the planetary half-shaft gear.

[0060] In this embodiment, speeds in the same speed range are regarded as the same speed. For example, the first speed range is 5km / h~10km / h, and any speed in the range of 5km / h~10km / h belongs to the same speed. The second speed range is 15km / h~20km / h, and any speed in the range of 15km / h~20km / h belongs to the same speed. The third speed range is 25km / h~30km / h, and any speed in the range of 25km / h~30km / h belongs to the same speed. Therefore, the "same speed" in step S2 refers to the speed in the same speed range. For example, if the straight speed is 5km / h and the turning speed is 8km / h, then the two belong to the same speed, and the average envelope can be subtracted.

[0061] In this embodiment, S2 includes the following sub-steps:

[0062] S21, performing EMD decomposition on the turning vibration signal at each speed, extracting the average envelope of each decomposition process, and obtaining multiple turning average envelopes;

[0063] S22, performing EMD decomposition on the straight-moving vibration signal at each speed, extracting the average envelope of each decomposition process, and obtaining multiple straight-moving average envelopes;

[0064] S23. At the same speed, the average envelope of the turn decomposed for the ith time is subtracted from the average envelope of the straight line decomposed for the ith time to obtain the difference signal of the turn envelope decomposed for the ith time, wherein i is a positive integer, and "subtraction" here refers to the subtraction operation of the amplitudes at corresponding time points (or sampling points) of the two average envelopes.

[0065] In the present invention, the average envelope is the average value of the upper envelope and the lower envelope during the i-th decomposition.

[0066] The differential includes: planetary side gears, input shaft, side gears (also known as sun gears) and output shaft, etc. The side gears are connected to the output shaft, the planetary side gears are connected to the input shaft, and the planetary side gears are placed between the two side gears. When driving straight, the planetary side gears and side gears rotate as a whole. When turning, the planetary side gears start to rotate on their own to adjust the different speeds of the left and right tires.

[0067] The present invention extracts the average envelope of each decomposition, which reflects the extreme value distribution of the turning vibration signal and the straight-line vibration signal. The average envelopes of the turning and straight-line are compared at the same speed to obtain a difference signal. Under normal conditions, there is an inherent difference between the two, but the planetary half-shaft gear fault will change this difference. Through this comparison, the signal change caused by the fault can be amplified, making the fault characteristics more prominent, improving the sensitivity to early faults and subtle faults, and facilitating the timely discovery of potential problems.

[0068] In this embodiment, when the straight-moving vibration signal and the turning vibration signal are decomposed for different times, for example, the straight-moving vibration signal is decomposed for 5 times and the turning vibration signal is decomposed for 6 times, only the average envelope of the first 5 times is used for subtraction.

[0069] In this embodiment, S3 includes the following sub-steps:

[0070] S31, decompose the first ~ the second The first-order synthetic difference signal is obtained by adding the second-order decomposed turning envelope difference signals, where N is the number of turning envelope difference signals, For the rounding down operation, "addition" here refers to the addition operation of the amplitudes at the corresponding time points (or sampling points) of the two turning envelope difference signals;

[0071] S32, the same speed down Decomposition~ The second-order decomposed turning envelope difference signals are added to obtain a second-order synthetic difference signal;

[0072] S33, adding the remaining decomposed turning envelope difference signals at the same speed to obtain a third-level synthetic difference signal;

[0073] S34. Calculate the difference energy mean and impact energy value for each level of synthetic difference signal.

[0074] The present invention divides the turning envelope difference signals decomposed multiple times into three levels. For example, when N is equal to 6, the turning envelope difference signals decomposed for the first time and the second time are added to obtain a first-level synthetic difference signal, the turning envelope difference signals decomposed for the third time and the fourth time are added to obtain a second-level synthetic difference signal, and the turning envelope difference signals decomposed for the fifth time and the sixth time are added to obtain a third-level synthetic difference signal.

[0075] The present invention adds the difference signals of the turning envelope in different levels, and can fuse the signal features at different scales. Synthesized difference signals at different levels have different sensitivities to fault features. By calculating the difference energy mean and impact energy value at each level, the performance of the fault at different scales can be highlighted.

[0076] In this embodiment, S34 includes the following sub-steps:

[0077] S341, dividing each level of synthesized difference signal into multiple segments, and calculating the difference energy value on each segment;

[0078] S342, taking the average of the difference energy values ​​of each segment to obtain the difference energy average;

[0079] S343. Filter out the difference energy values ​​that are greater than the difference energy mean value, and calculate the impact energy value.

[0080] In this embodiment, the formula for calculating the difference energy value in S341 is: , where G j is the difference energy value of the jth segment, g j,m is the mth amplitude on the jth segment, j and m are positive integers, || is the absolute value, and M is the length of a signal segment;

[0081] The formula for calculating the impact energy value in S343 is: , where ζ impact is the impact energy value, G impact,n is the nth difference energy value greater than the mean difference energy value, G avgis the difference energy mean, n is a positive integer, and L is the number of difference energy values ​​greater than the difference energy mean.

[0082] The present invention divides each level of synthetic difference signal into multiple segments and calculates the difference energy value separately, which can capture the energy distribution of the signal in different parts in detail. During the operation of the planetary half-shaft gear, the energy distribution of the vibration signal at different times or under different working conditions will change, and these changes can be reflected more accurately through segmented calculation. For example, in the early stage of wear, the gear may only show energy anomalies in certain specific periods or working conditions, and segmented calculation can detect these local anomalies in time.

[0083] The present invention calculates the difference energy mean to reflect the difference in vibration between turning and going straight. The present invention selects the difference energy value greater than the difference energy mean to calculate the impact energy value, which can effectively highlight the impact component in the signal. When a planetary half-shaft gear fails, such as cracks, broken teeth, etc., impact vibration will be generated, and its energy characteristics are manifested as an instantaneous energy peak, that is, greater than the energy mean under normal conditions. By screening and calculating these larger energy values, the impact information generated by the fault can be captured, and a more accurate judgment of the fault type and severity can be achieved.

[0084] In this embodiment, S4 includes the following sub-steps:

[0085] S41, calculating a first evolution coefficient according to a ratio of a second speed to a first speed composite difference signal of the same level in terms of difference energy mean and impact energy value, wherein the second speed is greater than the first speed;

[0086] S42, calculating a second evolution coefficient according to a ratio of the third speed to the second speed in terms of the difference energy mean and the impact energy value of the same-level synthetic difference signal, wherein the third speed is greater than the second speed;

[0087] S43: Add the first evolution coefficient and the second evolution coefficient to obtain an evolution value of the synthetic difference signal of this level.

[0088] In this embodiment, the formula for calculating the first evolution coefficient in S41 is: , where θ 1 is the first evolution coefficient, G avg,2,k is the difference energy mean of the k-th level synthetic difference signal at the second speed, G avg,1,k is the difference energy mean of the k-th level synthetic difference signal at the first speed, ζ impact,2,k is the impact energy value of the k-th level synthetic difference signal at the second speed, ζ impact,1,k is the impact energy value of the k-th level synthetic difference signal at the first speed, k is a positive integer;

[0089] The formula for calculating the second evolution coefficient in S42 is: , where θ 2 is the second evolution coefficient, G avg,3,k is the difference energy mean of the k-th level synthetic difference signal at the third speed, ζ impact,3,k is the impact energy value of the kth level synthetic difference signal at the third speed.

[0090] The present invention can reflect the change of the fault characteristics of the planetary half-shaft gear with speed by calculating the ratio of the difference energy mean and the impact energy value of the same-level synthetic difference signal at different speeds. When the vehicle is traveling at different speeds, the stress, wear and other conditions of the gears will be different, and the fault manifestations will also be different. This solution can capture these changes in fault characteristics caused by speed changes. For example, at low speeds, it may only show slight energy fluctuations, and at high speeds, the impact energy may increase significantly. The evolution value quantifies this change.

[0091] In this embodiment, S5 includes the following sub-steps:

[0092] S51, adding the difference energy means of the same-level synthetic difference signals at multiple speeds to obtain a first eigenvalue;

[0093] S52, adding the impact energy values ​​of the same-level synthetic difference signals at multiple speeds to obtain a second characteristic value;

[0094] S53: Use the first eigenvalue and the second eigenvalue as elements to construct a eigenvector of the synthetic difference signal of this level.

[0095] The present invention adds the difference energy mean values ​​and impact energy values ​​of the same-level synthetic difference signals at multiple speeds respectively to obtain the first eigenvalue and the second eigenvalue, which can comprehensively integrate the fault characteristic information of the planetary half-shaft gears under different speed conditions.

[0096] like Figure 2 As shown, the multi-input BP neural network in S6 includes: a first fully connected layer, a second fully connected layer, a third fully connected layer, a multiplier M1, a multiplier M2, a multiplier M3 and a BP neural network;

[0097] The input end of the first fully connected layer is used to input the feature vector of the first-level synthetic difference signal;

[0098] The input end of the second fully connected layer is used to input the feature vector of the second-level synthetic difference signal;

[0099] The input end of the third fully connected layer is used to input the feature vector of the third-level synthetic difference signal;

[0100] A first input terminal of the multiplier M1 is connected to the output terminal of the first fully connected layer, and a second input terminal thereof is used to input the evolution value of the first-level synthetic difference signal;

[0101] A first input terminal of the multiplier M2 is connected to an output terminal of the second fully connected layer, and a second input terminal thereof is used to input an evolution value of the second-level synthetic difference signal;

[0102] A first input terminal of the multiplier M3 is connected to the output terminal of the third fully connected layer, and a second input terminal thereof is used to input the evolution value of the third-level synthetic difference signal;

[0103] The input end of the BP neural network is connected to the output end of the multiplier M1, the output end of the multiplier M2 and the output end of the multiplier M3 respectively, and the output end thereof serves as the output end of the multi-input BP neural network.

[0104] In this embodiment, the expressions of the first fully connected layer, the second fully connected layer, and the third fully connected layer are all: , where S is the output of the fully connected layer, C r is the rth eigenvalue of the fully connected layer input, ω r C r The weight of b r C r The bias of , r is a positive integer ranging from 1 to 2.

[0105] In this embodiment, the weights and biases in the multi-input BP neural network can be obtained by training through the gradient descent method.

[0106] The present invention processes the feature vector of a first-level synthetic difference signal through each fully connected layer, and then multiplies the evolution value of the synthetic difference signal of the corresponding level with the output of the fully connected layer. The evolution value reflects the characteristic change of the synthetic difference signal of the same level at different speeds, and it contains the fault evolution information of the planetary half-shaft gear under different speed conditions. By multiplying with the output of the fully connected layer, the features at different speeds can be weighted and adjusted, so that when the network makes a fault judgment, it not only considers the fault features of the current level, but also considers the changing trend of these features with speed. For example, if the evolution value of a synthetic difference signal of a certain level is large at a certain speed, it means that the fault features of this level show obvious changes when the speed changes. The multiplication operation can enhance the network's attention to this feature, thereby more accurately judging the correlation between the fault and the speed.

[0107] The present invention collects vibration signals when going straight and turning at different speeds, performs EMD decomposition on the turning and straight vibration signals respectively, extracts the average envelope after each decomposition, subtracts the average envelope of turning from the average envelope of straight running, and obtains a turning envelope difference signal. When the vehicle is going straight and turning, the stress conditions and motion states of the planetary half-shaft gears are different, and the vibration signal characteristics generated are also different. When going straight, the planetary half-shaft gears are stationary relative to the half-shaft gears on the output shaft, and the planetary half-shaft gears do not rotate. Therefore, the complex vibration signal can be decomposed through EMD decomposition to obtain the average envelope. The turning envelope difference signal obtained by subtracting the average envelope after extracting the average envelope can amplify the signal characteristic change caused by the fault, make the fault characteristic more prominent, and effectively distinguish the fault vibration caused by the planetary half-shaft gears.

[0108] The present invention can analyze signal characteristics from different levels and further explore the fault information hidden in the signal through this hierarchical calculation of energy values. The difference energy mean reflects the overall energy distribution of the signal, while the impact energy value is more sensitive to the impact component in the signal and can highlight the impact characteristics caused by the fault. By calculating and analyzing different energy values, the signal characteristics in normal state and fault state can be more clearly distinguished, thereby improving the accuracy of fault detection and reducing the false alarm rate.

[0109] The present invention calculates the evolution value based on the difference energy mean and impact energy value of the same-level synthetic difference signal at multiple speeds. The evolution value can reflect the state change trend of the planetary half-shaft gear as the vehicle running state changes, and reflects the deterioration of its difference energy mean and impact energy value as the vehicle speed increases.

[0110] The present invention uses a multi-input BP neural network to process the characteristic vector and evolution value of a multi-level synthetic difference signal to obtain a fault value, comprehensively considers multiple factors, accurately determines the fault state of the planetary half-shaft gear, and improves the accuracy of fault detection.

[0111] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A planetary side shaft gear detection method, characterized in that: The following steps are involved: S1. A vibration sensor is arranged on the differential to collect vibration signals of the vehicle when it is traveling straight and turning at different speeds, and obtain a plurality of straight vibration signals and a plurality of turning vibration signals; S2. At the same speed, perform EMD decomposition on the turning and straight-moving vibration signals respectively, extract the average envelope under each decomposition, and obtain the turning envelope difference signal of multiple decompositions; S3, adding the turning envelope difference signals decomposed multiple times at the same speed in different levels, and calculating the difference energy mean and impact energy value of the synthetic difference signals at each level; S4, calculating the evolution value of each level of synthetic difference signal according to the difference energy mean and impact energy value of the same level synthetic difference signal at multiple speeds; S5. constructing a feature vector of each level of synthetic difference signal according to the difference energy mean and impact energy value of the same level synthetic difference signal at multiple speeds; S6. Use a multi-input BP neural network to process the characteristic vector and evolution value of the multi-level synthetic difference signal to obtain the fault value of the planetary half-shaft gear.

2. The planetary side gear detection method according to claim 1, characterized in that: The S2 comprises the following sub-steps: S21, performing EMD decomposition on the turning vibration signal at each speed, extracting the average envelope of each decomposition process, and obtaining multiple turning average envelopes; S22, performing EMD decomposition on the straight-moving vibration signal at each speed, extracting the average envelope of each decomposition process, and obtaining multiple straight-moving average envelopes; S23. At the same speed, subtract the average envelope of the turning decomposed for the ith time from the average envelope of the straight line decomposed for the ith time to obtain a difference signal of the turning envelope decomposed for the ith time, where i is a positive integer.

3. The planetary side gear detection method according to claim 1, characterized in that: The S3 comprises the following sub-steps: S31, decompose the first ~ the second The first-order synthetic difference signal is obtained by adding the second-order decomposed turning envelope difference signals, where N is the number of turning envelope difference signals, This is a round down operation; S32, the same speed down Decomposition~ The second-order decomposed turning envelope difference signals are added to obtain a second-order synthetic difference signal; S33, adding the remaining decomposed turning envelope difference signals at the same speed to obtain a third-level synthetic difference signal; S34. Calculate the difference energy mean and impact energy value for each level of synthetic difference signal.

4. The planetary side gear detection method according to claim 3, characterized in that: The S34 comprises the following sub-steps: S341, dividing each level of synthesized difference signal into multiple segments, and calculating the difference energy value on each segment; S342, taking the average of the difference energy values ​​of each segment to obtain the difference energy average; S343. Filter out the difference energy values ​​that are greater than the difference energy mean value, and calculate the impact energy value.

5. The planetary side gear detection method according to claim 4, characterized in that: The formula for calculating the difference energy value in S341 is: , where G j is the difference energy value of the jth segment, g j,m is the mth amplitude on the jth segment, j and m are positive integers, || is the absolute value, and M is the length of a signal segment; The formula for calculating the impact energy value in S343 is: , where ζ impact is the impact energy value, G impact,n is the nth difference energy value greater than the mean difference energy value, G avg is the difference energy mean, n is a positive integer, and L is the number of difference energy values ​​greater than the difference energy mean.

6. The planetary side gear detection method according to claim 1, characterized in that: The S4 comprises the following sub-steps: S41, calculating a first evolution coefficient according to a ratio of a second speed to a first speed composite difference signal of the same level in terms of difference energy mean and impact energy value, wherein the second speed is greater than the first speed; S42, calculating a second evolution coefficient according to a ratio of the third speed to the second speed in terms of the difference energy mean and the impact energy value of the same-level synthetic difference signal, wherein the third speed is greater than the second speed; S43: Add the first evolution coefficient and the second evolution coefficient to obtain an evolution value of the synthetic difference signal of this level.

7. The planetary side gear detection method according to claim 6, characterized in that: The formula for calculating the first evolution coefficient in S41 is: , where θ1 is the first evolution coefficient, G avg,2,k is the difference energy mean of the k-th level synthetic difference signal at the second speed, G avg,1,k is the difference energy mean of the k-th level synthetic difference signal at the first speed, ζ impact,2,k is the impact energy value of the k-th level synthetic difference signal at the second speed, ζ impact,1,k is the impact energy value of the k-th level synthetic difference signal at the first speed, k is a positive integer; The formula for calculating the second evolution coefficient in S42 is: , where θ2 is the second evolution coefficient, G avg,3,k is the difference energy mean of the k-th level synthetic difference signal at the third speed, ζ impact,3,k is the impact energy value of the kth level synthetic difference signal at the third speed.

8. The planetary side gear detection method according to claim 1, characterized in that: The S5 comprises the following sub-steps: S51, adding the difference energy means of the same-level synthetic difference signals at multiple speeds to obtain a first eigenvalue; S52, adding the impact energy values ​​of the same-level synthetic difference signals at multiple speeds to obtain a second characteristic value; S53: Use the first eigenvalue and the second eigenvalue as elements to construct a eigenvector of the synthetic difference signal of this level.

9. The planetary side gear detection method according to claim 1, characterized in that: The multi-input BP neural network in S6 includes: a first fully connected layer, a second fully connected layer, a third fully connected layer, a multiplier M1, a multiplier M2, a multiplier M3 and a BP neural network; The input end of the first fully connected layer is used to input the feature vector of the first-level synthetic difference signal; The input end of the second fully connected layer is used to input the feature vector of the second-level synthetic difference signal; The input end of the third fully connected layer is used to input the feature vector of the third-level synthetic difference signal; The first input terminal of the multiplier M1 is connected to the output terminal of the first fully connected layer, and the second input terminal thereof is used to input the evolution value of the first-level synthetic difference signal; The first input terminal of the multiplier M2 is connected to the output terminal of the second fully connected layer, and the second input terminal thereof is used to input the evolution value of the second-level synthetic difference signal; The first input terminal of the multiplier M3 is connected to the output terminal of the third fully connected layer, and the second input terminal thereof is used to input the evolution value of the third-level synthetic difference signal; The input end of the BP neural network is respectively connected to the output end of the multiplier M1, the output end of the multiplier M2 and the output end of the multiplier M3, and its output end serves as the output end of the multi-input BP neural network.

10. The planetary side gear detection method according to claim 9, characterized in that: The expressions of the first fully connected layer, the second fully connected layer and the third fully connected layer are all: , where S is the output of the fully connected layer, C r is the rth eigenvalue of the fully connected layer input, ω r C r The weight of b r C r The bias of r is used to identify the number of the eigenvalues ​​of the fully connected layer input, and r is a positive integer ranging from 1 to 2.

Images