Fault sound training method, fault type diagnosis method and device for main reducer

Abstract

This invention relates to the field of main reducer diagnostic technology, and discloses a fault sound training method, fault type diagnosis method, and apparatus for main reducers. The invention obtains fault sound training samples from the main reducer; extracts multiple fault sound training features corresponding to the training samples; inputs these features into a fault type diagnosis model for training, obtaining the fault type diagnosis result of the main reducer, and the loss between the fault type diagnosis result and the true label of the fault sound training samples. The loss is used to update the parameter values ​​of the fault type diagnosis model through a genetic optimization algorithm. Because this invention continuously trains the fault type diagnosis model, ensuring its sufficient accuracy, it ultimately achieves accurate diagnosis of main reducer fault types.

Description

Technical Field

[0001] This invention relates to the field of main reducer diagnostic technology, specifically to a fault sound training method, fault type diagnosis method and device for main reducers. Background Technology

[0002] The main gearbox is a critical component of a helicopter's transmission system. It typically consists of 3-4 stages of gear transmission, divided into main drive, rotor shaft, tail drive output, and accessory drive sections. Some light and small helicopters use a 2-stage main gearbox. Under high-power, high-speed, and high-load operating conditions, the input bevel gear system and planetary gear system of the main gearbox are prone to tooth surface wear, tooth contact fatigue, tooth bending fatigue, and even tooth breakage, significantly shortening the transmission system's lifespan. Therefore, early condition monitoring and fault diagnosis of the main gearbox are beneficial for improving the reliability of the helicopter transmission system.

[0003] In related technologies, abnormal state monitoring methods for main reducers include vibration monitoring, temperature monitoring, and acoustic emission monitoring. Vibration monitoring is limited by factors such as installation location, working environment, and vibration transmission, making it impossible to collect complete and effective fault vibration signals. Furthermore, vibration monitoring is only a point measurement and cannot accurately pinpoint the fault location. Temperature monitoring typically triggers an abnormal temperature signal when a fault occurs, but it cannot accurately locate the fault type in the main reducer before the fault actually occurs. Acoustic emission is highly sensitive to the properties of metallic materials; when the monitored object is complex and the environment varies greatly, the acoustic emission monitoring signal differs significantly. Moreover, when the monitored object is non-metallic or a composite material, the sound propagation path is complex, leading to insufficient accuracy in acoustic emission signal localization. Therefore, the above-mentioned methods for monitoring main reducer faults cannot guarantee the accuracy of fault type diagnosis. Summary of the Invention

[0004] This invention provides a fault sound training method, a fault type diagnosis method, and a device for a main reducer, in order to solve the problem of the inability to accurately monitor the fault type of the main detector.

[0005] In a first aspect, the present invention provides a fault sound training method for a main reducer, the method comprising: Obtain training samples of fault sounds from the main reducer; Extract multiple fault sound training features corresponding to the fault sound training samples; Multiple fault sound training features are input into the fault type diagnosis model for training to obtain the fault type diagnosis result of the main reducer, as well as the loss between the fault type diagnosis result and the true label of the fault sound training sample. The loss is used to update the parameter values ​​of the fault type diagnosis model through a genetic optimization algorithm.

[0006] In some optional implementations, multiple historical sound features include: the standard deviation, variance, skewness, peak value, mean, median, maximum value, minimum value, root mean square value, and frequency domain energy of the fault sound training samples. Multiple fault sound training features corresponding to the fault sound training samples are extracted, including: The fault sound training samples are decomposed into multiple sub-bands by wavelet packet decomposition according to the preset wavelet basis. Calculate the sound energy of each sub-band; Calculate the energy threshold for each sub-band; The fault sound training samples are denoised according to the energy threshold of each sub-frequency band. Extract the standard deviation, variance, skewness, peak value, mean, median, maximum value, minimum value, root mean square value, and frequency domain energy of the fault sound training samples after noise reduction.

[0007] In some optional implementations, the fault sound training method for the main reducer further includes: Obtain test samples of fault sounds from the main reducer; The accuracy of the fault type diagnostic model was tested using fault sound test samples. When the accuracy of the fault type diagnosis model does not meet the test requirements, the parameter values ​​of the fault type diagnosis model are updated using a genetic algorithm with loss function until the accuracy of the fault type diagnosis model meets the test requirements.

[0008] In some alternative implementations, the loss is used to update the parameter values ​​of the fault type diagnosis model via a genetic optimization algorithm, performed through the following steps: Based on the loss, a population of parameter values ​​and an initial population are created for the fault type diagnosis model. The population consists of multiple individuals. Calculate the fitness value for each individual based on the fitness function; Based on the fitness value of each individual, the fault type diagnosis model is trained iteratively until the fitness value of each individual reaches a preset threshold, and the optimal parameter values ​​of the fault type diagnosis model are output to obtain the trained fault type diagnosis model.

[0009] Secondly, the present invention provides a method for diagnosing fault types in a main reducer, the method comprising: Acquire the diagnostic sounds of the main reducer for faults; Extract multiple features of the fault sound corresponding to the fault sound to be diagnosed; Multiple sound features of the faults to be diagnosed are input into the fault type diagnosis model for diagnosis, and the fault type diagnosis results of the main reducer are obtained.

[0010] In some optional implementations, the fault type diagnosis model is a support vector machine model. Multiple acoustic features of the faults to be diagnosed are input into the fault type diagnosis model for diagnosis, resulting in the fault type diagnosis result of the main reducer, including: Multiple acoustic features of the faults to be diagnosed are input into a support vector machine model for classification and diagnosis, and the fault type diagnosis results of the main reducer are obtained.

[0011] Thirdly, the present invention provides a fault sound training device for a main reducer, the device comprising: The training sample acquisition module is used to acquire training samples of fault sounds from the main reducer. The training feature extraction module is used to extract multiple fault sound training features corresponding to the fault sound training samples; The training feature module is used to input multiple fault sound training features into the fault type diagnosis model for training, to obtain the fault type diagnosis result of the main reducer, and the loss between the fault type diagnosis result and the true label of the fault sound training sample. The loss is used to update the parameter values ​​of the fault type diagnosis model through a genetic optimization algorithm.

[0012] Fourthly, the present invention provides an electronic device, comprising: The memory and the processor are interconnected and communicate with each other. The memory stores computer instructions. The processor executes the computer instructions to perform the fault sound training method for the main reducer or the fault type diagnosis method for the main reducer in the first aspect or any embodiment of the first aspect.

[0013] Fifthly, the present invention provides a computer-readable storage medium storing computer instructions, the computer instructions being used to cause a computer to execute a fault sound training method for a main reducer or a fault type diagnosis method for a main reducer in the first aspect or any embodiment of the first aspect.

[0014] In a sixth aspect, the present invention provides a computer program product, including computer instructions, which are used to cause a computer to execute a fault sound training method for a main reducer or a fault type diagnosis method for a main reducer in the first aspect or any embodiment of the first aspect.

[0015] The technical solution of this invention has the following advantages: This invention discloses a fault sound training method, a fault type diagnosis method, and an apparatus for a main reducer. The method includes: acquiring fault sound training samples from the main reducer; extracting multiple fault sound training features corresponding to the training samples; inputting the multiple fault sound training features into a fault type diagnosis model for training, obtaining a fault type diagnosis result for the main reducer, and the loss between the fault type diagnosis result and the true label of the fault sound training samples. The loss is used to update the parameter values ​​of the fault type diagnosis model using a genetic optimization algorithm. Because this invention continuously trains the fault type diagnosis model, thereby ensuring its sufficient accuracy, it can ultimately achieve accurate diagnosis of main reducer fault types. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a flowchart of a fault sound training method for a main reducer according to an embodiment of the present invention; Figure 2 This is a flowchart of a fault sound diagnosis method for a main reducer according to an embodiment of the present invention; Figure 3 This is an overall block diagram of a fault sound training method for a main reducer according to an embodiment of the present invention. Figure 4 This is a schematic diagram of the training process of a fault sound diagnosis model according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the fitness curve for bevel gear fault identification according to an embodiment of the present invention; Figure 6 This is a schematic diagram illustrating the accuracy of bevel gear fault identification according to an embodiment of the present invention; Figure 7 This is a schematic diagram of a bevel gear fault identification confusion matrix according to an embodiment of the present invention; Figure 8 This is a structural block diagram of a fault sound training device for the main reducer according to an embodiment of the present invention; Figure 9 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0020] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] The main gearbox is a critical component of a helicopter's transmission system. It typically consists of 3-4 stages of gear transmission, divided into the main drive (including the main gearbox), rotor shaft, tail drive output, and accessory drive. Some light and small helicopters use a 2-stage main gearbox. The main functions of the main gearbox include deceleration, steering, and paralleling. It converts high-speed, low-torque engine power into low-speed, high-torque output. In the helicopter transmission system, the main gearbox also acts as a central load-bearing component, directly bearing all the forces and torques generated by the rotor and transmitting these forces to the fuselage. Under high-power, high-speed, and high-load operating conditions, the input stage bevel gear system and planetary gear system of the main gearbox are prone to tooth surface wear, tooth contact fatigue, tooth bending fatigue, and even tooth breakage, significantly shortening the service life of the transmission system. Therefore, early condition monitoring and fault diagnosis of the main gearbox are of significant economic and military importance for improving the reliability and performance of the helicopter transmission system, ensuring helicopter operational safety, and extending the service life of helicopter main gearbox components.

[0022] In related technologies, abnormal state monitoring methods for main reducers include vibration monitoring, temperature monitoring, and acoustic emission monitoring. Vibration monitoring, however, is limited by factors such as installation location, working environment, and vibration transmission, making it impossible to collect complete and effective fault vibration signals. Furthermore, vibration monitoring is only a point measurement and cannot accurately pinpoint the fault location. Temperature monitoring typically triggers an abnormal temperature signal when a fault occurs, but it cannot accurately locate the fault type in the main reducer before the fault actually occurs. Acoustic emission is highly sensitive to the properties of metallic materials; when the monitored object is complex and the environment varies greatly, the acoustic emission monitoring signal differs significantly. Moreover, when the monitored object is non-metallic or a composite material, the sound propagation path is complex, leading to insufficient accuracy in acoustic emission signal localization. Therefore, the above-mentioned methods for monitoring main reducer faults cannot guarantee the accuracy of main reducer fault diagnosis.

[0023] This embodiment provides a fault sound training method for a main reducer, which can be used in computer devices such as mobile phones, tablets, desktop computers, laptops, servers, etc. Figure 1 This is a flowchart of a fault sound training method for the main reducer according to an embodiment of the present invention, as shown below. Figure 1 As shown, the process includes the following steps: Step S101: Obtain training samples of fault sounds from the main reducer.

[0024] Specifically, training samples of fault sounds from key components of the main reducer are collected. These key components include, but are not limited to, bevel gears, first-stage planetary gears, first-stage sun gears, bearings, and shafts. In this embodiment, the fault sound training samples include, but are not limited to, training samples of bearing outer ring fault sounds, bearing inner ring fault sounds, bearing rolling element fault sounds, bearing cage fault sounds, bevel gear tooth breakage fault sounds, bevel gear crack fault sounds, first-stage planetary gear tooth breakage fault sounds, and first-stage planetary gear crack fault sounds.

[0025] Step S102: Extract multiple fault sound training features corresponding to the fault sound training samples.

[0026] In a specific example, multiple fault sound training features include: standard deviation, variance, skewness, peak value, mean, median, maximum value, minimum value, root mean square value, and frequency domain energy of the fault sound training samples.

[0027] Table 1 below shows the training features for multiple fault sounds extracted from the training samples. Table 1 lists the training features for multiple fault sounds.

[0028] Table 1 Training characteristics of multiple fault sounds

[0029] In some specific implementations, multiple fault sound training features corresponding to the fault sound training samples are extracted, including: Step a1: Perform wavelet packet decomposition on the fault sound training samples according to the preset wavelet basis to obtain multiple sub-frequency bands.

[0030] For example, with the preset wavelet basis as the db4 wavelet basis, wavelet packet decomposition is performed on the fault sound training samples to obtain multiple sub-frequency bands. Then, an L=8-level decomposition is performed to generate... Sub-band.

[0031] Step a2: Calculate the sound energy of each sub-band.

[0032] First, calculate the sound energy of each sub-band using the following formula.

[0033]

[0034] in, For the first j The energy of the k-th subband of the layer, For the first j The coefficients of the k-th sub-band of the layer. For the first j The coefficient length of the kth sub-band of the layer.

[0035] The average sound energy of each sub-band is then calculated using the following formula.

[0036]

[0037] in, The first frequency band of each sub-band j The average energy of the k-th sub-band of the layer, where L is the number of sub-bands. For the first j Layer i The sound energy of each frequency band.

[0038] Step a3: Calculate the energy threshold for each sub-band.

[0039] For example, ,in, Standard deviation For coefficients, The first frequency band of each sub-band j The average energy of the k-th sub-band of the layer.

[0040] Step a4: Perform noise reduction processing on the fault sound training samples according to the energy threshold of each sub-frequency band.

[0041] For example, if a faulty sound training sample is greater than or equal to the energy threshold of each sub-band, the faulty sound training sample is retained; if a faulty sound training sample is less than the energy threshold of each sub-band, the faulty sound training sample is deleted.

[0042] Step a5: Extract the standard deviation, variance, skewness, peak value, mean, median, maximum value, minimum value, root mean square value, and frequency domain energy of the fault sound training samples after noise reduction.

[0043] For the fault sound training samples after noise reduction, extract the standard deviation, variance, skewness, peak value, mean, median, maximum value, minimum value, root mean square value, and frequency domain energy corresponding to the fault sound training samples after noise reduction.

[0044] Step S103: Input multiple fault sound training features into the fault type diagnosis model for training to obtain the fault type diagnosis result of the main reducer, as well as the loss between the fault type diagnosis result and the true label of the fault sound training sample. The loss is used to update the parameter values ​​of the fault type diagnosis model through a genetic optimization algorithm.

[0045] In some specific implementations, the loss is used to update the parameter values ​​of the fault type diagnosis model through a genetic optimization algorithm, which is performed through the following steps: Step b1: Based on the loss value, create a population of parameter values ​​and an initial population for the fault type diagnosis model. The population consists of multiple individuals.

[0046] Step b2: Calculate the fitness value of each individual based on the fitness function.

[0047] Step b3: Based on the fitness value of each individual, the fault type diagnosis model is trained iteratively until the fitness value of each individual reaches the preset threshold, and the optimal parameter values ​​of the fault type diagnosis model are output to obtain the trained fault type diagnosis model.

[0048] Specifically, this embodiment optimizes the fault type diagnosis model through a genetic optimization algorithm, which helps to improve the diagnostic accuracy of the fault type diagnosis model.

[0049] The genetic optimization algorithm specifically requires designing a fitness function and calculating the fitness value of each individual. It iteratively trains the fault type diagnosis model until the fitness value of each individual reaches a preset threshold, outputting the optimal parameter values ​​for the fault type diagnosis model. If the fitness value of an individual does not reach the preset threshold, selection, crossover, and mutation operations are performed on the population to obtain offspring, and the fitness of the offspring population is calculated. This process continues to check whether the preset threshold is reached, iterating until a termination condition is met. Upon reaching the termination condition of the preset threshold, parameter decoding is performed, outputting the optimal parameter values, which include the optimal penalty factor and kernel function parameters from the genetic algorithm.

[0050] In some specific embodiments, the fault sound training method for the main reducer in this application further includes: Step c1: Obtain a fault sound test sample from the main reducer.

[0051] Step c2: Use fault sound test samples to test the accuracy of the fault type diagnosis model.

[0052] Step c3: When the accuracy of the fault type diagnosis model does not meet the test requirements, continue to use the loss to update the parameter values ​​of the fault type diagnosis model through a genetic algorithm until the accuracy of the fault type diagnosis model meets the test requirements.

[0053] Specifically, fault sound test samples are input into the fault type diagnosis model for testing, and test results are obtained.

[0054] This embodiment provides a method for diagnosing fault types in a main reducer, which can be used in computer equipment such as mobile phones, tablets, desktop computers, laptops, servers, etc. Figure 2 This is a flowchart of a fault type diagnosis method for the main reducer according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: Step S201: Obtain the sound of the fault to be diagnosed in the main reducer.

[0055] Step S202: Extract multiple fault sound features corresponding to the fault sound to be diagnosed.

[0056] For details on extracting multiple fault sound features corresponding to the fault sound to be diagnosed, please refer to steps a1-a5 above, which will not be repeated here.

[0057] Step S203: Input multiple sound features of the faults to be diagnosed into the fault type diagnosis model for diagnosis, and obtain the fault type diagnosis results of the main reducer.

[0058] Since the fault type diagnosis model is a pre-trained neural network model, inputting multiple sound features of the fault to be diagnosed into the fault type diagnosis model can quickly and accurately diagnose the fault type of the main reducer.

[0059] In some specific implementations, multiple acoustic features of the faults to be diagnosed are input into a support vector machine model for classification and diagnosis to obtain the fault type diagnosis results of the main reducer, including: Multiple acoustic features of the faults to be diagnosed are input into the support vector machine of the fault type diagnosis model for classification and diagnosis, and the fault type diagnosis results of the main reducer are obtained.

[0060] The fault type diagnosis model in this application embodiment can be a Support Vector Machine (SVM) model. Its core idea is to find the optimal hyperplane to maximize the margin between different classes of data. It can improve generalization ability by minimizing structural risk and effectively solves problems such as small sample size, nonlinearity, and local minima. It has been applied to pattern classification, regression prediction, probability estimation, etc. The classification accuracy of the SVM model is affected by the kernel function parameters. Constrained by the penalty factor C, the selection of these two parameters affects the SVM's generalization ability to the input sample set. During the SVM model building process, the values ​​of these two parameters are often subjectively chosen, making it difficult to select optimal values ​​in practical use. Kernel function parameters... Choosing the optimal values ​​for the penalty factor C and these two parameters has become a pressing issue. Therefore, this application, based on determining the loss of the fault type diagnosis model, further optimizes the parameter values ​​of the fault type diagnosis model using a genetic optimization algorithm, thereby improving the accuracy of the fault type diagnosis model.

[0061] like Figure 3 The diagram shown is an overall block diagram of the fault sound training method for the main reducer in this application.

[0062] For example, this application has been verified by simulated experimental data. Acoustic signals of key components of the main reducer simulation test bench were collected at a sampling frequency of 44100Hz and a sampling time of 30s. Nine types of fault tests were conducted, including bearing outer ring fault, bearing inner ring fault, bearing rolling element fault, bearing cage fault, bevel gear tooth breakage fault, bevel gear crack fault, first-stage planetary gear tooth breakage fault, first-stage planetary gear crack fault, and first-stage sun gear uneven wear fault. Feature extraction was performed on the fault sound training samples collected under each fault test to obtain multiple fault sound training features.

[0063] The collected fault sounds are automatically divided into training and testing sets in a 7:3 ratio. Ten time-domain features are used for feature construction: standard deviation, sample variance, range, skewness, kurtosis, mean, median, maximum, minimum, and root mean square. One frequency-domain feature is used: frequency amplitude. Therefore, a total of 11 features are used in training. The final fault type identification accuracy formula is shown below.

[0064]

[0065] To improve the accuracy of fault identification, To diagnose the number of correct labels, This is a sample of fault sound tests.

[0066] Taking the inner ring of a bearing as an example, the training set is automatically divided into 1930 samples and the test set into 820 samples, with a ratio of 7:3. The resulting fault sound training samples are: The fault sound test sample is The training process is as follows Figure 4 As shown.

[0067] Based on the actual debugging process, the final parameters of the genetic algorithm were set as follows: population size of 30, crossover probability of 0.9, mutation probability of 0.8, and termination generation of the genetic algorithm of 200. Then, according to the above parameters, fault identification was performed on bevel gear faults, planetary gear faults, sun gear faults, bearing outer ring faults, bearing inner ring faults, bearing rolling element faults, and bearing cage faults.

[0068] For the fault diagnosis of bevel gear tooth breakage and bevel gear cracking faults under nine operating conditions (300 r / min @ 0% load, 300 r / min @ 10% load, 300 r / min @ 20% load, 600 r / min @ 0% load, 600 r / min @ 10% load, 600 r / min @ 20% load, 900 r / min @ 0% load, 900 r / min @ 10% load, and 900 r / min @ 20% load), multiple fault sound features were first extracted from the above 18 sets of signals. These constructed fault sound features were then input into the fault sound diagnosis model for fault diagnosis. The dataset was automatically divided into training and test sets in a 7:3 ratio, with 3473 groups in the training set and 1476 groups in the test set. The parameters of the genetic algorithm were set as described above. After 200 iterations, the fitness curve and classification results of the fault sound diagnosis model were obtained as shown below. Figure 5 and Figure 6 As shown, the optimal parameters of the fault sound diagnosis model can also be obtained. Among them, the penalty factor c=523.35, g=722.95.

[0069] The fault conditions and label information corresponding to the confusion matrix are shown in Table 2.

[0070] Table 2. Correspondence between bevel gear fault conditions and labels

[0071] from Figure 5 , Figure 6 and Figure 7 As can be seen, for the fault sound signals under different working conditions of bevel gear tooth breakage and cracking, the fault sound diagnosis model has an accuracy rate of 1466 / 1476=99.32% for fault identification, which has a good identification effect.

[0072] from Figure 5 , Figure 6 and Figure 7 As can be seen, for the fault sound signals under different working conditions of bevel gear tooth breakage and cracking, the fault sound diagnosis model has an accuracy rate of 1466 / 1476=99.32% for fault identification, which has a good identification effect.

[0073] This application obtains fault sound training samples from the main reducer; extracts multiple fault sound training features corresponding to the fault sound training samples; inputs these multiple fault sound training features into a fault type diagnosis model for training, obtaining the fault type diagnosis result of the main reducer, and the loss between the fault type diagnosis result and the true label of the fault sound training samples. The loss is used to update the parameter values ​​of the fault type diagnosis model through a genetic optimization algorithm. Because this invention continuously trains the fault type diagnosis model, thereby ensuring that the fault type diagnosis model is sufficiently accurate, this invention can ultimately achieve accurate diagnosis of the fault type of the main reducer.

[0074] This embodiment also provides a fault sound training device for a main reducer, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0075] This embodiment provides a fault sound training device for a main reducer, such as... Figure 8 As shown, it includes: Training sample acquisition module 801 is used to acquire training samples of fault sounds from the main reducer. The training feature extraction module 802 is used to extract multiple fault sound training features corresponding to the fault sound training samples. The training feature training module 803 is used to input the multiple fault sound training features into the fault type diagnosis model for training, to obtain the fault type diagnosis result of the main reducer, and the loss between the fault type diagnosis result and the true label of the fault sound training sample. The loss is used to update the parameter values ​​of the fault type diagnosis model through a genetic optimization algorithm.

[0076] In some optional embodiments, the fault sound training device for the main reducer further includes: The test sample acquisition module is used to acquire fault sound test samples of the main reducer; The test sample testing module is used to test the accuracy of the fault type diagnostic model using fault sound test samples. The test parameter update module is used to continue updating the parameter values ​​of the fault type diagnosis model using a genetic algorithm with loss function when the accuracy of the fault type diagnosis model does not meet the test requirements, until the accuracy of the fault type diagnosis model meets the test requirements.

[0077] In some optional implementations, the loss in the training feature module updates the parameter values ​​of the fault type diagnosis model using a genetic optimization algorithm, which is performed through the following sub-modules: The population creation submodule is used to create a population of parameter values ​​for the fault type diagnosis model based on the loss and to initialize the population. The population consists of multiple individuals. The fitness calculation submodule is used to calculate the fitness value of each individual based on the fitness function; The optimal parameter optimization submodule is used to iteratively train the fault type diagnosis model based on the fitness value of each individual until the fitness value of each individual reaches a preset threshold, and outputs the optimal parameter values ​​of the fault type diagnosis model to obtain the trained fault type diagnosis model.

[0078] In some optional embodiments, the present invention also provides a fault type diagnosis device for a main reducer, the device comprising: The fault sound acquisition module is used to acquire the fault sounds of the main reducer that are to be diagnosed. The fault feature extraction module is used to extract multiple fault sound features corresponding to the fault sound to be diagnosed. The fault feature diagnosis module is used to input multiple sound features of the faults to be diagnosed into the fault type diagnosis model for diagnosis, and obtain the fault type diagnosis results of the main reducer.

[0079] In some optional implementations, the fault type diagnosis model is a support vector machine model. Multiple acoustic features of the faults to be diagnosed are input into the fault type diagnosis model for diagnosis, resulting in the fault type diagnosis result of the main reducer, including: Multiple acoustic features of the faults to be diagnosed are input into a support vector machine model for classification and diagnosis, and the fault type diagnosis results of the main reducer are obtained.

[0080] The fault type diagnosis device for the main reducer provided in this embodiment of the invention can execute the fault type diagnosis method for the main reducer provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0081] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0082] The following is a detailed reference. Figure 9 , Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0083] The following is a detailed reference. Figure 9 This diagram illustrates a suitable structural schematic for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 901, which can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) 902 or a program loaded from memory 908 into random access random access memory (RAM) 903. The RAM 903 also stores various programs and data required for the operation of the electronic device. The processor 901, ROM 902, and RAM 903 are interconnected via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0084] Typically, the following devices can be connected to I / O interface 905: input devices 906 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 907 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 908 including, for example, magnetic tapes, hard disks, etc.; and communication devices 909. Communication device 909 allows electronic devices to exchange data via wireless or wired communication with other devices. Although Figure 9 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0085] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 909, or installed from a memory 908, or installed from a ROM 902. When the computer program is executed by the processor 901, it performs the functions defined in the fault type diagnosis method for the main reducer according to embodiments of the present invention.

[0086] Figure 9 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0087] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that the computer, processor, microprocessor controller, or programmable hardware includes storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the fault type diagnosis method for the main reducer shown in the above embodiments is implemented.

[0088] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0089] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for training fault sounds in a main reducer, characterized in that, The method includes: Obtain training samples of fault sounds from the main reducer; Extract multiple fault sound training features corresponding to the fault sound training samples; The multiple fault sound training features are input into the fault type diagnosis model for training to obtain the fault type diagnosis result of the main reducer, and the loss between the fault type diagnosis result and the true label of the fault sound training sample. The loss is used to update the parameter values ​​of the fault type diagnosis model through a genetic optimization algorithm.

2. The method according to claim 1, characterized in that, The fault sound training features include: the standard deviation, variance, skewness, peak value, mean, median, maximum value, minimum value, root mean square value, and frequency domain energy of the fault sound training samples. Multiple fault sound training features corresponding to the fault sound training samples are extracted, including: The fault sound training samples are decomposed into multiple sub-frequency bands according to a preset wavelet basis. Calculate the sound energy of each sub-band; Calculate the energy threshold for each sub-band; The fault sound training samples are denoised according to the energy threshold of each sub-frequency band. Extract the standard deviation, variance, skewness, peak value, mean, median, maximum value, minimum value, root mean square, and frequency domain energy corresponding to the fault sound training samples after noise reduction processing.

3. The method according to claim 1, characterized in that, The method further includes: Obtain test samples of fault sounds from the main reducer; The accuracy of the fault type diagnostic model is tested using the fault sound test samples. When the accuracy of the fault type diagnosis model does not meet the test requirements, the parameter values ​​of the fault type diagnosis model are updated using the genetic algorithm based on the loss until the accuracy of the fault type diagnosis model meets the test requirements.

4. The method according to claim 1, characterized in that, The loss is used to update the parameter values ​​of the fault type diagnosis model through a genetic optimization algorithm, which is performed through the following steps: Based on the loss, a population of parameter values ​​and an initialization population are created for the fault type diagnosis model, wherein the population consists of multiple individuals; Calculate the fitness value for each individual based on the fitness function; Based on the fitness value of each individual, the fault type diagnosis model is trained iteratively until the fitness value of each individual reaches a preset threshold, and the optimal parameter values ​​of the fault type diagnosis model are output to obtain the trained fault type diagnosis model.

5. A method for diagnosing fault types in a main reducer, characterized in that, The method includes: Acquire the diagnostic sounds of the main reducer for faults; Extract multiple fault sound features corresponding to the fault sound to be diagnosed; The multiple sound features of the faults to be diagnosed are input into the fault type diagnosis model for diagnosis, and the fault type diagnosis results of the main reducer are obtained.

6. The method according to claim 5, characterized in that, The fault type diagnosis model is a support vector machine model. The multiple acoustic features of the faults to be diagnosed are input into the fault type diagnosis model for diagnosis, resulting in the fault type diagnosis results of the main reducer, including: The multiple sound features of the faults to be diagnosed are input into the support vector machine model for classification and diagnosis to obtain the fault type diagnosis result of the main reducer.

7. A fault sound training device for a main reducer, characterized in that, The device includes: The training sample acquisition module is used to acquire training samples of fault sounds from the main reducer. The training feature extraction module is used to extract multiple fault sound training features corresponding to the fault sound training samples; The training feature training module is used to input the multiple fault sound training features into the fault type diagnosis model for training, to obtain the fault type diagnosis result of the main reducer, and the loss between the fault type diagnosis result and the true label of the fault sound training sample. The loss is used to update the parameter values ​​of the fault type diagnosis model through a genetic optimization algorithm.

8. An electronic device, characterized in that, include: The system includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes the computer instructions to perform the fault sound training method for the main reducer or the fault type diagnosis method for the main reducer as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the fault sound training method for the main reducer according to any one of claims 1 to 6, or the fault type diagnosis method for the main reducer.

10. A computer program product, characterized in that, Includes computer instructions for causing a computer to execute the fault sound training method for the main reducer as described in any one of claims 1 to 6, or the fault type diagnosis method for the main reducer.

Images