CN113283386A - Equipment fault diagnosis method of coal mine underground coal mining machine based on knowledge distillation - Google Patents
Equipment fault diagnosis method of coal mine underground coal mining machine based on knowledge distillation Download PDFInfo
- Publication number
- CN113283386A CN113283386A CN202110686567.9A CN202110686567A CN113283386A CN 113283386 A CN113283386 A CN 113283386A CN 202110686567 A CN202110686567 A CN 202110686567A CN 113283386 A CN113283386 A CN 113283386A
- Authority
- CN
- China
- Prior art keywords
- mining machine
- loss function
- value
- coal mining
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The application provides a method for diagnosing equipment faults of a coal mine underground coal mining machine based on knowledge distillation, which comprises the following steps: identifying the coal mining machine fault signal data set by a teacher model, and learning attribute knowledge in the data set; selecting a data set, dividing an output value of a full-connection layer in the teacher model by a set parameter, and calculating based on a normalized exponential function to obtain a soft label value; inputting the data set number into the student model, repeating the same operation of the teacher model and obtaining the output value of the full connection layer; dividing the output value of the full-connection layer in the student model by the same set parameter as the teacher model, and calculating based on the normalized exponential function to obtain a relative entropy loss function; comparing the predicted value with the real label value to obtain a multi-classification cross entropy loss function; obtaining a mean value mixed loss function according to the relative entropy loss function and the multi-classification cross entropy loss function; the teacher model trains the student models through the mean mixing loss function.
Description
Technical Field
The application relates to the technical field of equipment fault diagnosis, in particular to an equipment fault diagnosis method of a coal mine underground coal mining machine based on knowledge distillation.
Background
The coal mining machine is one of core devices of a coal mining working face of a coal mine, a rocker arm of the coal mining machine is a key power component for cutting a coal layer of the coal mining machine, bears cutting load of the coal mining machine and nonlinear internal excitation of a rocker arm transmission system, and is a weak link of the coal mining machine. The environment of the coal mining working face is severe, gears of a rocker arm transmission system of the coal mining machine easily break down, coal mining efficiency is affected, economic losses of enterprises are caused, and casualty accidents are caused more seriously. The rocker arm transmission system has the characteristics of long transmission chain, multiple gear types, strong environmental noise and the like, so that great challenges are brought to the gear fault diagnosis. At present, gear fault diagnosis is mostly carried out based on a shallow deep learning model, and with the increase of the number of network layers, the weights between the layers cannot be updated, so that the precision is easy to be unchanged or reduced.
Therefore, there is a need to provide an improved solution to the above-mentioned deficiencies of the prior art.
Disclosure of Invention
The application aims to provide a method for diagnosing equipment faults of a coal mine underground coal mining machine based on knowledge distillation, so as to solve or relieve the problems in the prior art.
In order to achieve the above purpose, the present application provides the following technical solutions:
the application provides a method for diagnosing equipment faults of a coal mine underground coal mining machine based on knowledge distillation, which comprises the following steps: step S101, establishing a teacher network: training a teacher model, identifying the coal mining machine fault signal data set by the teacher model, and learning attribute knowledge in the coal mining machine fault signal data set; dividing the output value of the full-connection layer in the teacher model by a set parameter according to the fault signal data set of the coal mining machine, and calculating based on a normalized exponential function to obtain a soft label value; setting the value range of the parameter as [1, 20 ]; step S102, establishing a student network: inputting the coal mining machine fault signal data set into a student model, repeating the same operation of the teacher model and obtaining an output value of a full connection layer; dividing the output value of the full-connection layer in the student model by the set parameter same as that of the teacher model, calculating based on the normalized index function to obtain a predicted value, and comparing the predicted value with the soft label value to obtain a relative entropy loss function; comparing the predicted value with the real label value to obtain a multi-classification cross entropy loss function; obtaining a mean value mixed loss function according to the relative entropy loss function and the multi-classification cross entropy loss function; step S103, training a teacher-student heterogeneous network: the teacher model trains the student models through the mean mixing loss function.
Preferably, in step S102, the relative entropy loss function and the multi-class cross entropy loss function are added in equal proportion to obtain a mean value mixed loss function.
Preferably, the mean mixing loss function is:
KD Loss=KL(m2,n)·θ·T2+CE(Label,m1)·(1-θ)
KD Loss represents a mean value mixing Loss function, KL is relative entropy, CE is cross entropy, q is a result of distillation output by the teacher model, and m is1For the predicted values, m, for the soft label values in the student model2In the model of studentsIn the result of distillation of the hard predicted value, label is a real label of the data set, and theta is a proportionality coefficient and takes the value of 0.5; t is a set coefficient; according to the following formulae, respectively:
calculating relative entropy KL and cross entropy CE;
wherein p and q respectively represent the real distribution probability and the theoretical distribution probability of the data; q (x), p (x) respectively represent two probability distributions of x values; n is the total number of the data sample cloth; c represents the category of the data sample, and c is a positive integer; m is the number of data sample classes, M is a positive integer, yicA value of 0 or 1; p is a radical oficIs the probability of class c of the data sample.
Preferably, the shearer fault signal data set comprises: the straight gear of the rocker arm of the coal mining machine is normal, and vibration signals are generated in 5 states of abrasion, fracture, pitting corrosion and crack.
Preferably, vibration signals of the straight gear of the rocker arm of the coal mining machine in different states are collected by an acceleration sensor, wherein the acceleration sensor is connected with vibration signal collecting equipment.
Preferably, the acceleration sensor is mounted on the housing on the side of the spur gear of the rocker arm.
Preferably, the sampling frequency of the vibration signal acquisition equipment is set to be 12kHz, and the sensitivity is 500 mV/g.
Has the advantages that:
according to the technical scheme provided by the embodiment of the application, the teacher model identifies the coal mining machine fault signal data set, learns attribute knowledge in the coal mining machine fault signal data set and realizes training of the teacher model; selecting a coal mining machine fault signal data set, dividing an output value of a full connection layer in a teacher model by a set parameter, calculating based on a normalized exponential function to obtain a soft label value, inputting the coal mining machine fault signal data set into a student model, repeating the same operation of the teacher model, and obtaining the output value of the full connection layer; dividing the output value of the full-connection layer in the student model by the set parameter same as that of the teacher model, calculating based on the normalized index function to obtain a predicted value, and comparing the predicted value with the soft label value to obtain a relative entropy loss function; comparing the predicted value with the real label value to obtain a multi-classification cross entropy loss function; obtaining a mean value mixed loss function according to the relative entropy loss function and the multi-classification cross entropy loss function; the teacher model trains the student models through the mean value mixing loss function, so that the accuracy of the student models for the coal mining machine fault prediction is close to that of the teacher model through knowledge distillation, the parameter quantity and the running time consumption of the network model are reduced under the condition that the accuracy is guaranteed, and the efficiency of coal mining machine fault diagnosis is improved.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the application and, together with the description, serve to explain the application and are not intended to limit the application. Wherein:
fig. 1 is a schematic flow diagram of a method for diagnosing equipment failure of a coal mine underground mining machine based on knowledge distillation according to some embodiments of the present application;
FIG. 2 is a schematic illustration of a test rocker arm drive concept provided in accordance with some embodiments of the present application;
fig. 3 is a schematic structural diagram of a residual learning unit provided according to some embodiments of the present application.
Detailed Description
The present application will be described in detail below with reference to the embodiments with reference to the attached drawings. The various examples are provided by way of explanation of the application and are not limiting of the application. In fact, it will be apparent to those skilled in the art that modifications and variations can be made in the present application without departing from the scope or spirit of the application. For instance, features illustrated or described as part of one embodiment, can be used with another embodiment to yield a still further embodiment. It is therefore intended that the present application cover the modifications and variations of this invention provided they come within the scope of the appended claims and their equivalents.
Fig. 1 is a schematic flow diagram of a method for diagnosing equipment failure of a coal mine underground mining machine based on knowledge distillation according to some embodiments of the present application; as shown in fig. 1, the equipment fault diagnosis method for the coal mine underground coal mining machine based on knowledge distillation comprises the following steps:
step S101, establishing a teacher network:
training a teacher model, identifying the coal mining machine fault signal data set by the teacher model, and learning attribute knowledge in the coal mining machine fault signal data set;
dividing the output value of the full-connection layer in the teacher model by a set parameter according to the fault signal data set of the coal mining machine, and calculating based on a normalized exponential function to obtain a soft label value; setting the value range of the parameter as [1, 20 ];
in the embodiment of the application, the distillation process of the teacher model is to distill out the probability distribution in the complex network structure by setting a parameter T, and guide the compact model (student model) to train by using the probability distribution, wherein the value range of the parameter T is set to [1, 20 ].
In an embodiment of the present application, the attribute knowledge in the shearer fault signal dataset characterizes an internal correlation between the shearer's vibration signals and gear states.
In some optional embodiments, the shearer fault signal data set includes: the straight gear of the rocker arm of the coal mining machine is normal, and vibration signals are generated in 5 states of abrasion, fracture, pitting corrosion and crack. Further, vibration signals of the coal mining machine under different states of the straight gear of the rocker arm are acquired by utilizing an acceleration sensor, wherein the acceleration sensor is connected with vibration signal acquisition equipment and is arranged on a shell near the side of the straight gear of the rocker arm; the sampling frequency of the vibration signal acquisition equipment is set to be 12kHz, and the sensitivity is 500 mV/g.
In the embodiment of the application, a simulation experiment is carried out on the gear fault in the rocker arm transmission system by using a loading experiment table of the rocker arm of the coal mining machine. The coal mining machine rocker arm loading experiment table consists of an eddy current dynamometer, an accompanying rocker arm, a testing rocker arm, a connecting assembly, a coupler and a mechanical fastening device. The electric eddy current dynamometer is responsible for providing the load torque required by the rocker arm loading. The motor is arranged in the accompanying rocker arm, and the low-rotating-speed large torque can be converted into the high-rotating-speed small torque. The testing rocker arm contains a motor and is a power source for running the experiment table. A coupling connects the companion rocker arm and the test rocker arm together to transmit torque. The mechanical fastening device is used for supporting and fixing the test-accompanying rocker arm and the test rocker arm.
FIG. 2 is a schematic illustration of a test rocker arm drive concept provided in accordance with some embodiments of the present application; as shown in FIG. 2, Z1-Z14 are gears. The output shaft of the cutting part motor is connected with a gear Z1 through a slender flexible torque shaft, the output torque of the motor is transmitted to a 1 st-stage planetary reducer through gears Z1-Z8, a planet carrier of the 1 st-stage planetary reducer transmits power to a 2 nd-stage planetary reducer, the power output by the 2 nd-stage planetary reducer is transmitted to a square coupling sleeve, and finally, the power is transmitted to a cutting drum.
In the embodiment of the application, the loading experiment steps of the rocker arm of the coal mining machine are as follows:
(1) fix loaded two rocking arms on the support, will accompany the output of examination rocking arm and the input of electric eddy current dynamometer machine to be connected fixedly through coupling assembling, will accompany the input of examination rocking arm and the output of test rocking arm to be connected fixedly through the shaft coupling to suitably adjust, make each part centre of gyration of connecting fixedly keep unanimous as far as possible, prevent that the rocking arm from producing vibration.
(2) And the vibration signal acquisition of the coal mining machine rocker arm straight gear under different states is realized by utilizing the acceleration sensor. An acceleration sensor is mounted on the housing near the side of the rocker spur gear. Meanwhile, the acceleration sensor is connected with vibration signal acquisition equipment, the sampling frequency is set to be 12kHz, and the sensitivity is 500 mV/g.
(3) And after the loading platform and the test system are installed, a power supply is connected to the motor in the test rocker arm to carry out a loading experiment. The loading amount is adjusted to 50% by adjusting the loading amount button of the eddy current dynamometer, and the loading time of the loading amount is set to 1 h. And collecting the rocker arm vibration signal and storing the rocker arm vibration signal to a storage device within the loading time range.
(4) Repeating the steps, and respectively acquiring vibration signals of the rocker straight gear in 5 states of normal state, abrasion state, fracture state, pitting state and crack state.
In the embodiment of the application, the teacher model and the training set and the verification set data of the student model are acquired through the vibration signal acquisition equipment. Wherein, vibration data of each state of the gear under 50% loading is selected for experiment, and the length of the sample is set as 40000. Before model training, firstly, dividing one-dimensional original vibration signals in 5 states into 2455 total samples; then using cross validation of ten folds (10-fold), dividing 2455 total samples into 10 sample subsets with equal size; sequentially traversing 10 subsets, taking the former subset as a verification set and all the rest subsets as training sets each time, and training and evaluating the model; and finally, taking the average value of the 10 evaluation indexes as a final evaluation index.
In the embodiment of the application, after the vibration signal acquisition device acquires the training set and the verification set data, the vibration data needs to be preprocessed. Specifically, data normalization processing is performed on the one-dimensional original vibration signals in 5 different states. The normalization algorithm is selected whether the characteristic values after normalization can effectively distinguish the states of various gears of the current coal mining machine, namely whether the characteristic values after normalization of the gears in different states have obvious differences. The normalization processing model is shown in formula (1), and formula (1) is as follows:
y=(x-MinValue)/(MaxValue-MinValue)………………(1)
wherein, x and y are values before and after conversion respectively, and MaxValue and MinValue are maximum and minimum values of the sample respectively. In this case, the acceleration (a), which is a vibration characteristic index in 5 different gear states (normal, wear, pitting, cracking, tooth breakage), is preprocessed by applying a normalization algorithm formula.
Step S102, establishing a student network:
inputting the coal mining machine fault signal data set into a student model, repeating the same operation of the teacher model and obtaining an output value of a full connection layer;
dividing the output value of the full-connection layer in the student model by the set parameter same as that of the teacher model, then calculating based on the normalized index function to obtain a predicted value, and comparing the predicted value with the soft label value to obtain a relative entropy loss function; comparing the predicted value with the real label value to obtain a multi-classification cross entropy loss function; obtaining a mean value mixed loss function according to the relative entropy loss function and the multi-classification cross entropy loss function;
in the embodiment of the present application, the "softmax" output layer of the neural network converts the prediction result obtained by the previous convolutional layer into a probability value p. The layer will get a certain class of logit values Zi from the natural logarithm of the prediction result, and generate the probability pi of this class by comparing with the logit values Zj of all classes; knowledge distillation is carried out by modifying it and adding a set coefficient T to make the output layer generate a 'softened' probability vector qi: wherein q isiAs shown in equation (2), equation (2) is as follows:
wherein T is a setting parameter. As the T parameter increases, the corresponding distribution probability is gentler. And dividing the prediction output result of the teacher network by the set parameter T, and then performing softmax transformation to obtain the softened probability distribution qi for the subsequent calculation of KD Loss.
In the embodiment of the present application, in order for the student model to learn more about the knowledge of the teacher model, the relative Loss function KD Loss of mean mixture is used. The KD Loss is formed by proportionally combining a relative entropy Loss function and a multi-class cross entropy Loss function. Where the relative entropy loss function is used to measure the asymmetry of the difference between two probability distributions p and q, colloquially speaking, the different degrees of two events are defined. The mean mixing loss function is shown in equation (3), where equation (3) is as follows:
KD Loss=KL(m2,n)·θ·T2+CE(Label,m1)·(1-θ)……(3)
wherein KL is relative entropy, CE is cross entropy, q is the result of teacher model output after distillation, m1For soft predictions, m in student models2And (3) obtaining the result of distillation of hard syndromes in the student model, wherein label is a real label of the data set, and theta is a proportionality coefficient and takes the value of 0.5. Expressions of KL and CE are respectively expressed by formula (4) and formula (5), and formula (4) and formula (5) are as follows:
wherein p and q respectively represent the real distribution probability and the theoretical distribution probability of the data; q (x), p (x) respectively represent two probability distributions of x values; n is the total number of the data sample cloth; c represents the category of the data sample, and c is a positive integer; m is the number of data sample categories, and M is a positive integer; y isicAn indication variable (0 or 1), which is 1 if the class is the same as the class of sample i, and 0 otherwise; p is a radical oficRefers to the predicted probability that the observation sample i belongs to the class c.
In the present embodiment, when the coefficient θ is 0, KD Loss ═ CE (lab, m1), the mixing Loss function corresponds to a deep convolutional neural network using only the multi-class cross entropy Loss function without using knowledge distillation.
Step S103, training a teacher-student heterogeneous network:
the teacher model trains the student models through the mean mixing loss function.
In the embodiment of the application, the gradient is reduced by specifically calculating the mean value mixed Loss function KD Loss, the parameters of the student model are updated, and the accuracy of the fault diagnosis of the student model is improved.
In the embodiment of the application, model gradient disappearance and network degradation are eliminated by using the Resnet101 and the Resnet18 networks, and the identification accuracy of the networks is improved. Specifically, the accuracy of the Resnet101 and Resnet18 networks in fault diagnosis of equipment of the coal mining machine is improved through residual learning.
Fig. 3 is a schematic structural diagram of a residual learning unit provided according to some embodiments of the present application; as shown in fig. 3, Y represents the lower layer output, f (X) represents the upper layer input, and X directly establishes a correlation channel between the input and the output through identity mapping, so that the residual between the input and the output can be learned. Specifically, as shown in formula (6), formula (6) is as follows:
Y=F(x)+X…………………………(6)
in the embodiment of the application, the Resnet101 model is used as a teacher model to guide the training of the student model Resnet18 model, so that the Resnet18 model can effectively improve the efficiency of equipment fault diagnosis of the coal mining machine on the basis of ensuring the accuracy of equipment fault diagnosis of the coal mining machine. The parameter settings of Resnet101 and Resnet18 are shown in Table 1. Table 1 is as follows:
table 1 parameter setting table
The depth refers to the number of layers of the network model, and the larger the number of layers is, the deeper the network depth is, and the better the performance is; the size refers to the size of space required by model storage; the calculation force is the times of parameter operation of the model, and the larger the numerical value is, the larger the model parameter quantity is represented, and the more complex the model is.
In the present example, the Resnet101 network was used to guide the training of the Resnet18 model by knowledge distillation, dividing 2455 total samples into 10 equal-sized sample subsets; and sequentially traversing 10 subsets, wherein the former subset is taken as a verification set and all the rest subsets are taken as training sets to verify the model after knowledge distillation.
In the embodiment of the application, a teacher model is used for guiding the training of the student model, wherein the teacher model is a large-scale complex network model with large parameter quantity; the student model is a compact network model with small scale and small parameter quantity. And the effect of obtaining high accuracy by the small model is realized through induction training.
In the embodiment of the application, 5 types of faults of the coal mining machine are diagnosed, namely, the faults are normal, worn, pitting, cracked and broken of the gear, and the 5 types of data are classified by a machine learning method, so that the accuracy and the calculation efficiency are better.
In a multi-classification learning task, a multi-classification cross entropy loss function is adopted, so that overfitting of the model can be effectively prevented, meanwhile, the model is compressed by combining relative entropy in a knowledge distillation model, the accuracy rate and the recognition speed of the equipment fault detection model are improved, the quantity of network model parameters is reduced, and the consumption of running time is reduced.
In the embodiment of the application, on the basis of a multivariate cross entropy loss function, relative entropy is combined, parameters between the multivariate cross entropy loss function and the single loss function are fused to carry out iterative training on the model, the iterative training of the model is superior to that of the model with the single loss function, equal-proportion distribution is selected on the proportion distribution of the fusion of the multivariate cross entropy loss function and the single loss function, and on the basis of an original training model, the model parameters of knowledge distillation are introduced, so that the rapid convergence of the training model is better realized, and the accuracy is improved.
In the embodiment of the application, a knowledge distillation diagnosis model is constructed aiming at the problems that the conventional coal mining machine gear fault diagnosis method cannot automatically extract corresponding data, so that the fault diagnosis precision and efficiency of coal mining machine equipment are low. Firstly, after the vibration data are normalized, the data are input into a teacher model Resnet101, corresponding characteristics are automatically extracted, then the teacher model Resnet101 is used for training a student model Resnet18, and finally a result is output.
The above description is only a preferred embodiment of the present application and is not intended to limit the present application, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims (7)
1. A method for diagnosing equipment faults of a coal mine underground coal mining machine based on knowledge distillation is characterized by comprising the following steps:
step S101, establishing a teacher network:
training a teacher model, identifying the coal mining machine fault signal data set by the teacher model, and learning attribute knowledge in the coal mining machine fault signal data set;
dividing the output value of the full-connection layer in the teacher model by a set parameter according to the fault signal data set of the coal mining machine, and calculating based on a normalized exponential function to obtain a soft label value; setting the value range of the parameter as [1, 20 ];
step S102, establishing a student network:
inputting the coal mining machine fault signal data set into a student model, repeating the same operation of the teacher model and obtaining the output of an output layer;
dividing the output value of the full-connection layer in the student model by the set parameter same as that of the teacher model, calculating based on the normalized index function to obtain a predicted value, and comparing the predicted value with the soft label value to obtain a relative entropy loss function;
comparing the predicted value with the real label value to obtain a multi-classification cross entropy loss function;
obtaining a mean value mixed loss function according to the relative entropy loss function and the multi-classification cross entropy loss function;
step S103, training a teacher-student heterogeneous network:
the teacher model trains the student models through the mean mixing loss function.
2. The method for diagnosing the equipment failure of the coal mine underground coal mining machine based on the knowledge distillation as claimed in claim 1, wherein in the step S102, the relative entropy loss function and the multi-classification cross entropy loss function are added in equal proportion to obtain a mean value mixing loss function.
3. The method for diagnosing equipment failure of a coal mine underground coal mining machine based on knowledge distillation as claimed in claim 1, wherein the mean value mixing loss function is as follows:
KD Loss=KL(m2,n)·θ·T2+CE(Label,m1)·(1-θ)
KD Loss represents a mean value mixing Loss function, KL is relative entropy, CE is cross entropy, q is a result of distillation output by the teacher model, and m is1For the predicted values, m, for the soft label values in the student model2The hard prediction value in the student model is a distilled result, label is a real label of the data set, and theta is a proportionality coefficient and takes a value of 0.5; t is a set coefficient;
according to the following formulae, respectively:
calculating relative entropy KL and cross entropy CE;
wherein p and q respectively represent the real distribution probability and the theoretical distribution probability of the data; q (x), p (x) respectively represent two probability distributions of x values; n is the total number of the data sample cloth; c represents the category of the data sample, and c is a positive integer; m is the number of data sample classes, M is a positive integer, yicA value of 0 or 1; p is a radical oficIs the probability of class c of the data sample.
4. The method for diagnosing equipment failure of a coal mine underground mining machine based on knowledge distillation as claimed in any one of claims 1 to 3, wherein the coal mining machine failure signal data set comprises: vibration signals of a straight gear of a rocker arm of the coal mining machine under 5 states of normal state, abrasion state, breakage state, pitting corrosion state and crack state.
5. The method for diagnosing the equipment failure of the coal mine underground coal mining machine based on the knowledge distillation as claimed in claim 4, wherein vibration signals of the straight gear of the rocker arm of the coal mining machine under different states are collected by an acceleration sensor, and the acceleration sensor is connected with vibration signal collecting equipment.
6. The method for diagnosing equipment failure of a coal mine underground mining machine based on knowledge distillation as claimed in claim 5, wherein the acceleration sensor is installed on a housing of a rocker spur gear side.
7. The method for diagnosing the equipment failure of the coal mine underground coal mining machine based on the knowledge distillation as claimed in claim 5, wherein the sampling frequency of the vibration signal acquisition equipment is set to be 12kHz, and the sensitivity is 500 mV/g.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2021105724993 | 2021-05-25 | ||
CN202110572499 | 2021-05-25 |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113283386A true CN113283386A (en) | 2021-08-20 |
Family
ID=77285161
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110686567.9A Pending CN113283386A (en) | 2021-05-25 | 2021-06-21 | Equipment fault diagnosis method of coal mine underground coal mining machine based on knowledge distillation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113283386A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117056678A (en) * | 2023-10-12 | 2023-11-14 | 北京宝隆泓瑞科技有限公司 | Machine pump equipment operation fault diagnosis method and device based on small sample |
CN117473285A (en) * | 2023-12-27 | 2024-01-30 | 长春黄金设计院有限公司 | Intelligent operation and maintenance management system and method based on digital twinning |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106198000A (en) * | 2016-07-11 | 2016-12-07 | 太原理工大学 | A kind of rocker arm of coal mining machine gear failure diagnosing method |
CN110162018A (en) * | 2019-05-31 | 2019-08-23 | 天津开发区精诺瀚海数据科技有限公司 | The increment type equipment fault diagnosis method that knowledge based distillation is shared with hidden layer |
CN110659665A (en) * | 2019-08-02 | 2020-01-07 | 深圳力维智联技术有限公司 | Model construction method of different-dimensional features and image identification method and device |
CN110674880A (en) * | 2019-09-27 | 2020-01-10 | 北京迈格威科技有限公司 | Network training method, device, medium and electronic equipment for knowledge distillation |
CN111062951A (en) * | 2019-12-11 | 2020-04-24 | 华中科技大学 | Knowledge distillation method based on semantic segmentation intra-class feature difference |
CN112116030A (en) * | 2020-10-13 | 2020-12-22 | 浙江大学 | Image classification method based on vector standardization and knowledge distillation |
CN112286751A (en) * | 2020-11-24 | 2021-01-29 | 华中科技大学 | Intelligent diagnosis system and method for high-end equipment fault based on edge cloud cooperation |
-
2021
- 2021-06-21 CN CN202110686567.9A patent/CN113283386A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106198000A (en) * | 2016-07-11 | 2016-12-07 | 太原理工大学 | A kind of rocker arm of coal mining machine gear failure diagnosing method |
CN110162018A (en) * | 2019-05-31 | 2019-08-23 | 天津开发区精诺瀚海数据科技有限公司 | The increment type equipment fault diagnosis method that knowledge based distillation is shared with hidden layer |
CN110659665A (en) * | 2019-08-02 | 2020-01-07 | 深圳力维智联技术有限公司 | Model construction method of different-dimensional features and image identification method and device |
CN110674880A (en) * | 2019-09-27 | 2020-01-10 | 北京迈格威科技有限公司 | Network training method, device, medium and electronic equipment for knowledge distillation |
CN111062951A (en) * | 2019-12-11 | 2020-04-24 | 华中科技大学 | Knowledge distillation method based on semantic segmentation intra-class feature difference |
CN112116030A (en) * | 2020-10-13 | 2020-12-22 | 浙江大学 | Image classification method based on vector standardization and knowledge distillation |
CN112286751A (en) * | 2020-11-24 | 2021-01-29 | 华中科技大学 | Intelligent diagnosis system and method for high-end equipment fault based on edge cloud cooperation |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117056678A (en) * | 2023-10-12 | 2023-11-14 | 北京宝隆泓瑞科技有限公司 | Machine pump equipment operation fault diagnosis method and device based on small sample |
CN117056678B (en) * | 2023-10-12 | 2024-01-02 | 北京宝隆泓瑞科技有限公司 | Machine pump equipment operation fault diagnosis method and device based on small sample |
CN117473285A (en) * | 2023-12-27 | 2024-01-30 | 长春黄金设计院有限公司 | Intelligent operation and maintenance management system and method based on digital twinning |
CN117473285B (en) * | 2023-12-27 | 2024-03-19 | 长春黄金设计院有限公司 | Intelligent operation and maintenance management system and method based on digital twinning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Ravikumar et al. | Gearbox fault diagnosis based on Multi-Scale deep residual learning and stacked LSTM model | |
CN108320043B (en) | Power distribution network equipment state diagnosis and prediction method based on electric power big data | |
CN113283386A (en) | Equipment fault diagnosis method of coal mine underground coal mining machine based on knowledge distillation | |
CN109376620A (en) | A kind of migration diagnostic method of gearbox of wind turbine failure | |
CN103115789B (en) | Second generation small-wave support vector machine assessment method for damage and remaining life of metal structure | |
CN106092578A (en) | A kind of machine tool mainshaft bearing confined state online test method based on wavelet packet and support vector machine | |
CN110516305A (en) | Intelligent fault diagnosis method under small sample based on attention mechanism meta-learning model | |
CN106197999A (en) | A kind of planetary gear method for diagnosing faults | |
CN101806729A (en) | In-use lubricating oil quality rapid testing method | |
CN110533007B (en) | Intelligent identification and extraction method for bridge vehicle-mounted strain influence line features | |
CN104266677A (en) | Method for evaluating running health state of magnetic control paralleling reactor | |
CN112147432A (en) | BiLSTM module based on attention mechanism, transformer state diagnosis method and system | |
CN113076834B (en) | Rotating machine fault information processing method, processing system, processing terminal, and medium | |
CN113188794B (en) | Gearbox fault diagnosis method and device based on improved PSO-BP neural network | |
CN111428386B (en) | Elevator traction machine rotor fault diagnosis information fusion method based on complex network | |
CN112113755A (en) | Mechanical fault intelligent diagnosis method based on deep convolution-kurtosis neural network | |
CN115774851B (en) | Method and system for detecting internal defects of crankshaft based on hierarchical knowledge distillation | |
CN115982896B (en) | Bearing retainer service life detection method and device | |
CN109374293A (en) | A kind of gear failure diagnosing method | |
CN112329520B (en) | Truck bearing fault identification method based on generation countermeasure learning | |
Wang et al. | Numerical simulation of gears for fault detection using artificial intelligence models | |
CN110530639A (en) | A kind of bullet train axle box bearing failure diagnostic method | |
CN116720112A (en) | Machine pump fault diagnosis method and system based on machine learning | |
CN106339720A (en) | Automobile engine failure detection method | |
CN114757365A (en) | High-speed railway roadbed settlement prediction and early warning method based on deep learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |