CN112414446B - Data-driven transmission sensor fault diagnosis method - Google Patents
Data-driven transmission sensor fault diagnosis method Download PDFInfo
- Publication number
- CN112414446B CN112414446B CN202011201104.0A CN202011201104A CN112414446B CN 112414446 B CN112414446 B CN 112414446B CN 202011201104 A CN202011201104 A CN 202011201104A CN 112414446 B CN112414446 B CN 112414446B
- Authority
- CN
- China
- Prior art keywords
- sensor
- neural network
- fault
- value
- probabilistic neural
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D18/00—Testing or calibrating apparatus or arrangements provided for in groups G01D1/00 - G01D15/00
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Algebra (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a data-driven transmission sensor fault diagnosis method, which comprises the following steps: s1: establishing an optimal multiple linear regression model of the sensor; s2: after a fitting value is obtained through the sensor model, a residual error is made through a sensor signal value and the fitting value which are actually collected; s3: obtaining characteristic values of sensor signals by processing residual errors through wavelet transformation, wherein the characteristic values are used as input of a probabilistic neural network; s4: after the feature extraction is finished, identifying the features by using a probabilistic neural network; s5: and obtaining a fault diagnosis result of the sensor after the probabilistic neural network diagnosis. The invention is based on wavelet transformation combined with probability neural network method to extract the characteristics and diagnose the fault of the original vehicle sensor and the residual sequence of the sensor model, and comprehensively considers the relationship between the sensor and the whole clutch system, thereby being capable of diagnosing the fault of the sensor rapidly, accurately and efficiently.
Description
Technical Field
The invention belongs to the technical field of automobile fault diagnosis, and particularly relates to a data-driven transmission sensor fault diagnosis method.
Background
The sensor is used as a device for acquiring a transmission signal, and is an important bridge for communication between the transmission and the controller, and whether the sensor is normal or not directly influences the performance of the transmission, so that the sensor is important for fault diagnosis of the sensor.
There are three methods for diagnosing the failure of the sensor: the fault diagnosis method is based on a model, a rule and data driving. Under the condition that an accurate analysis model of the system can be obtained, the fault diagnosis method based on the model is most direct and effective, but for a complex system, the accurate analysis model is generally difficult to obtain, and in addition, the uncertainty of the model, the nonlinear characteristic of the system and the like can generate great influence on the diagnosis result. Some researchers use the odd-even equation method to diagnose the fault of the sensor in the electric power steering system, and for the uncertainty of the model, the adaptive threshold value is set to improve the robustness of the fault diagnosis, but the method is targeted at a linear time-invariant system, and the method is poor in applicability to a non-linear time-variant system such as a transmission. Researchers have studied a fault diagnosis method for a multi-input multi-output nonlinear model, which is used for processing the nonlinear problem by converting the fault diagnosis method into a plurality of one-dimensional linear equivalent models, so that the workload of modeling is simplified, but the influence of structures inside an analysis system on fault diagnosis is not analyzed, and the fault diagnosis needs to be considered when the model is simplified.
The Fault diagnosis method based on the rules requires that enough prior knowledge of Fault causes and Fault expressions is accumulated, then the knowledge is converted into reasoning rules, fault Tree Analysis (FTA), failure Mode and Effect Analysis (FMEA) and other methods are used for realizing Fault diagnosis, and the method has the advantages that the rules are easy to modify, and the defects that the knowledge is difficult to acquire. The sensor fault is diagnosed by a scholars through an FTA method, only 2 fault rules are considered, the number of the fault rules is up to 102, the workload is large, the transmission system is complex, the number of the sensors is large, and the fault rules can reach 300 under the method. The learners use the probability distribution method to process the uncertainty in the knowledge, but the method is only suitable for simple systems.
The data-driven fault diagnosis method does not need to establish an accurate analytical model and accumulate enough prior knowledge, and can finish the diagnosis of the sensor fault only by analyzing the change characteristics of the sensor signal and combining a related recognition algorithm.
Disclosure of Invention
The invention aims to overcome the defects of the existing transmission sensor fault diagnosis method based on a model and a rule, and provides a transmission sensor fault diagnosis method based on wavelet transformation and a probabilistic neural network.
The technical scheme of the invention is as follows: a data-driven transmission sensor fault diagnostic method characterized by: the method comprises the following steps:
s1: establishing an optimal multiple linear regression model of the sensor;
s2: after a fitting value is obtained through the sensor model, a residual error is made through a sensor signal value and the fitting value which are actually collected;
s3: obtaining characteristic values of sensor signals by processing residual errors through wavelet transformation, wherein the characteristic values are used as input of a probabilistic neural network;
s4: after the feature extraction is finished, identifying the features by using a probabilistic neural network;
s5: and obtaining a fault diagnosis result of the sensor after the probabilistic neural network diagnosis.
Further, in step S1, the multiple linear regression model is:
in the formula (I), the compound is shown in the specification,representative dependent variable v 1 、ν 2 ...ν n Respectively representing selected independent variables, formed by the combination of signals of rotating speed, gear, pre-shift and wheel speed, a 1 、a 2 ...a n Respectively, the coefficients representing the respective variables, and b represents the reference value of the output value.
Further, the wavelet transform processing procedure in step S3 is to pass the signal through a low pass filter and a high pass filter, respectively, so that the signal is decomposed into a low frequency part and a high frequency part, the low frequency part is called approximation coefficient, the high frequency part is called detail coefficient, the approximation coefficient and the detail coefficient are passed through the low pass filter and the high pass filter, and the approximation coefficient and the detail coefficient of the next layer are obtained, and the procedure is repeated until a predetermined number of decomposition layers is reached.
Further, the step S3 is that the probabilistic neural network is based on a radial basis neural network and combines a Bayesian decision rule.
Compared with the prior art, the invention has the following beneficial effects:
(1) The method is based on wavelet transformation combined with a probabilistic neural network method to perform feature extraction and fault diagnosis on the original vehicle sensor and the residual sequence of the sensor model, comprehensively considers the relationship between the sensor and the whole clutch system, and can perform fault diagnosis on the sensor quickly, accurately and efficiently.
(2) The complexity of a fault diagnosis method based on a model and rules is avoided, and the diagnosis is quickly and accurately carried out by processing the sensor data.
(3) And carrying out fault diagnosis from the perspective of a system, collecting sensor data of the real vehicle, and building a sensor model by using the transmission and the data of the whole vehicle, wherein a diagnosis result is suitable for the running condition of the real vehicle. .
Drawings
FIG. 1 is a flow chart of a data driven transmission sensor fault diagnostic method of the present invention;
FIG. 2 is a block diagram of a probabilistic neural network for a data driven transmission sensor fault diagnostic method of the present invention;
FIG. 3 is a graphical representation of the diagnostic accuracy as a function of the smoothing factor for a data driven transmission sensor fault diagnostic method of the present invention;
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments obtained by a person skilled in the art without making any inventive step are within the scope of protection of the present invention.
Referring to fig. 1 and fig. 2, the technical solution of the present invention is: as shown in step S1 in fig. 1, a sensor model is created using an odd-numbered oil pressure sensor in the twin clutch type automatic transmission as an example. The basic idea of the stepwise regression algorithm is to gradually introduce independent variables with obvious contribution to dependent variables and eliminate the independent variables with insignificant contribution to the dependent variables by calculating the contribution degree of the independent variables to the dependent variables, and repeat the process until all the independent variables with obvious contribution to the dependent variables are introduced and all the independent variables with insignificant contribution to the dependent variables are eliminated, so that the optimal multivariate linear regression model of the sensor can be established. The index for measuring the contribution degree is a correlation coefficient, and the correlation coefficients among independent variables and between the independent variables and dependent variables are calculated to obtain the following correlation coefficient matrix:
wherein r is a correlation coefficient and m represents a dependent variable. And selecting the independent variable with the largest contribution degree to the dependent variable from the independent variables which are not introduced as the independent variable to be introduced, and selecting the independent variable with the smallest contribution degree to the dependent variable from the introduced independent variables as the independent variable to be eliminated. The finally established sensor multivariate linear regression model is as follows:
in the formula (I), the compound is shown in the specification,representative dependent variable v 1 、ν 2 ...ν n Respectively representing the selected independent variable, and formed by the combination of signals of rotating speed, gear, pre-shift gear and wheel speed, a 1 、a 2 ...a n Respectively, the coefficients representing the respective variables, and b represents the reference value of the output value.
The input parameters of the multiple linear regression model of the rotating speed sensor are formed by combining signals of vehicle speed, gear, pre-shifting gear, clutch oil pressure and the like, the output is a fitted rotating speed value, and the modeling process is consistent with that of the oil pressure sensor.
S2: after a fitting value is obtained through a sensor model, a residual error is made through a sensor signal value and the fitting value which are actually collected and is used as input of fault diagnosis:
as shown in step S2 and step S3 in fig. 1, the residual error is processed by wavelet transform to obtain characteristic values of the sensor signal, and these characteristic values are used as input of the probabilistic neural network. The wavelet transform processes a signal through a low pass filter and a high pass filter, respectively, so that the signal is decomposed into a low frequency part and a high frequency part, the low frequency part is called an approximation coefficient, the high frequency part is called a detail coefficient, the approximation coefficient and the detail coefficient are continuously passed through the low pass filter and the high pass filter to obtain an approximation coefficient and a detail coefficient of a next layer, and the process is repeated until a predetermined decomposition layer number is reached. Assume that the input residual sequence is Δ y = [ Δ y ] 1 ,Δy 2 ,…,Δy n ]The number of decomposition layers is J, and the decomposition results in 2 according to the wavelet transformation principle J Sub-signal, sub-signal dimension d = n/2 J The approximation and detail coefficients are:
l J =Ga J-1
h J =Ha J-1
in the formula: l. the J 、h J Approximate coefficients and detail coefficients of the J-th layer, respectively; G. h is a low-pass filter and a high-pass filter, and is composed of wavelet functions; a is a J-1 Is a node value of the J-1 th layer. After the delta y is decomposed through wavelet transformation, the Shannon entropy of the sub-signals is extracted as the characteristics. The shannon entropy is an index for measuring the uncertainty of the signal, is one of the common characteristics of wavelet transform, and has the following calculation formula:
in the formula: s J (k) Is the Shannon entropy of the kth node of the J-th layer, k =1,2, \ 8230;, 2 J ,e J (i) Is the energy proportion of the ith data point in the node, and the calculation formula is as follows:
in the formula: a is J,k (i) Is the value of the ith data point in the kth node of the jth layer. Through the above formula, the characteristic vector of the Shannon entropy can be obtained
After the feature extraction is completed, the probabilistic neural network is used to identify the features, so as to diagnose the fault, as shown in step S4 in fig. 1. The probability neural network is based on the radial basis neural network, and the Bayesian decision rule is combined, so that the local optimal problem of the back propagation neural network and the problem that the radial basis neural network is sensitive to the radial basis function are solved, the training is easy, the convergence speed is high, and the probability neural network is very suitable for real-time processing. The general structure of a probabilistic neural network is shown in figure 2.
The probabilistic neural network comprises a four-layer structure, an input layer is used for receiving Shannon entropy from wavelet transformation, the number of neurons is the same as the length of an input vector, a hidden layer is a radial base layer, an activation function is a Gaussian kernel function, and the output can be expressed as:
in the formula: phi j (s) is the output of the jth hidden layer neuron, j =1,2, \8230;, n, n is the number of samples; sigma is a smoothing factor and plays a crucial role in network performance, and as shown in fig. 3, the optimal value of sigma cannot be too large or too small in a certain interval; w is a j Is the weight vector for the jth hidden layer neuron. The summation layer performs a weighted average of the outputs of the hidden layers, with each summation layer neuron representing a class. The summation layer outputs the result:
in the formula: p is i Is the weighted output of the ith class, i =1,2, \8230;, c, c is the number of sample classes(ii) a L is the number of hidden layer neurons pointing to the class i class. The output layer decides the output category according to Bayesian decision rule, the Bayesian decision aims to minimize the risk of misjudgment, and the risk function is defined as:
in the formula: r (c) i | s) is a risk of determining the input vector s as the ith class; lambda ij Judging the category j as the loss of the category i; p (c) j S) is a conditional probability of determining the input vector s as the class j, corresponding to the output of the jth neuron in the summation layer. If will be ij The loss defined as misclassification is 1 and the loss defined as correct classification is 0, the above formula becomes R (c) i |s)=1-P(c i S), to make R (c) i S) minimum, P (c) i | s) needs to be the largest, the output layer takes the largest category corresponding to the output of the summation layer, that is:
after the probabilistic neural network diagnosis is performed, a fault diagnosis result of the sensor is obtained, and the accuracy of the diagnosis result can be calculated by the following formula:
in the formula, z i Represents the actual sample class or classes of samples,representing the sample category obtained by diagnosis, and the number of the test samples is k.
It should be noted that, in this document, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrases "comprising a component of' 8230; \8230;" does not exclude the presence of additional identical elements in the process, method, article, or apparatus that comprises the element.
Although the embodiments of the present invention have been described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (1)
1. A data-driven transmission sensor fault diagnostic method characterized by: the method comprises the following steps:
s1: establishing an optimal multiple linear regression model of the sensor;
s2: after a fitting value is obtained through the sensor model, a residual error is made through a sensor signal value and the fitting value which are actually acquired;
s3: residual errors are processed through wavelet transformation to obtain characteristic values of sensor signals, the Shannon entropy of the sub-signals is extracted to serve as characteristics after wavelet transformation decomposition, and the characteristic values serve as input of a probabilistic neural network;
s4: after the feature extraction is finished, identifying the features by using a probabilistic neural network;
s5: obtaining a fault diagnosis result of the sensor after the probabilistic neural network diagnosis;
the step S1 is that a multiple linear regression model is as follows:
in the formula (I), the compound is shown in the specification,representing dependent variable v 1 、ν 2 ...ν n Respectively representing selected independent variables, formed by the combination of signals of rotating speed, gear, pre-shift and wheel speed, a 1 、a 2 ...a n Respectively representing the coefficients of the respective variables, b represents the reference value of the output value;
the input parameters of the multiple linear regression model of the revolution speed sensor are formed by combining signals of vehicle speed, gears, pre-shifting gears, clutch oil pressure and the like, the output is a fitted revolution speed value, and the modeling process is consistent with that of the oil pressure sensor;
the wavelet transform processing procedure of the step S3 is to make the signal pass through a low-pass filter and a high-pass filter respectively, so that the signal is decomposed into a low-frequency part and a high-frequency part, the low-frequency part is called an approximate coefficient, the high-frequency part is called a detail coefficient, the approximate coefficient and the detail coefficient are continuously passed through the low-pass filter and the high-pass filter to obtain the approximate coefficient and the detail coefficient of the next layer, and the procedure is continuously repeated until the preset decomposition layer number is reached;
and S3, the probabilistic neural network is based on the radial basis neural network and combines a Bayesian decision rule.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011201104.0A CN112414446B (en) | 2020-11-02 | 2020-11-02 | Data-driven transmission sensor fault diagnosis method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011201104.0A CN112414446B (en) | 2020-11-02 | 2020-11-02 | Data-driven transmission sensor fault diagnosis method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112414446A CN112414446A (en) | 2021-02-26 |
CN112414446B true CN112414446B (en) | 2023-01-17 |
Family
ID=74828504
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011201104.0A Active CN112414446B (en) | 2020-11-02 | 2020-11-02 | Data-driven transmission sensor fault diagnosis method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112414446B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115265609A (en) * | 2021-03-12 | 2022-11-01 | 青岛科技大学 | Method for diagnosing state of sensor in structural health monitoring system |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104677629A (en) * | 2014-10-28 | 2015-06-03 | 芜湖杰诺瑞汽车电器系统有限公司 | Fault detection method for vehicle transmission |
CN104318305B (en) * | 2014-10-30 | 2017-02-01 | 东北电力大学 | Inverter low-frequency noise fault diagnosis method based on wavelets and neural network |
KR101877127B1 (en) * | 2016-10-06 | 2018-07-10 | 국방과학연구소 | Apparatus and Method for detecting voice based on correlation between time and frequency using deep neural network |
CN107202952A (en) * | 2017-07-06 | 2017-09-26 | 北京信息科技大学 | Rotary kiln method for diagnosing faults, fault diagnosis GUI and system based on wavelet neural network |
CN109472003A (en) * | 2018-10-24 | 2019-03-15 | 江苏税软软件科技有限公司 | A kind of arithmetic of linearity regression applied to cost analysis |
CN110084106A (en) * | 2019-03-19 | 2019-08-02 | 中国地质大学(武汉) | Microgrid inverter method for diagnosing faults based on wavelet transformation and probabilistic neural network |
CN111191854A (en) * | 2020-01-10 | 2020-05-22 | 上海积成能源科技有限公司 | Photovoltaic power generation prediction model and method based on linear regression and neural network |
-
2020
- 2020-11-02 CN CN202011201104.0A patent/CN112414446B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN112414446A (en) | 2021-02-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111340238B (en) | Fault diagnosis method, device, equipment and storage medium of industrial system | |
CN109765053B (en) | Rolling bearing fault diagnosis method using convolutional neural network and kurtosis index | |
CN112161784B (en) | Mechanical fault diagnosis method based on multi-sensor information fusion migration network | |
CN108256556A (en) | Wind-driven generator group wheel box method for diagnosing faults based on depth belief network | |
CN110618353B (en) | Direct current arc fault detection method based on wavelet transformation + CNN | |
CN113188807B (en) | Automatic abs result judging algorithm | |
CN114234361A (en) | Central air-conditioning sensor fault detection method based on double noise reduction and fuzzy indexes | |
Li et al. | Gear pitting fault diagnosis using raw acoustic emission signal based on deep learning | |
Wang et al. | Fault diagnosis of centrifugal pump using symptom parameters in frequency domain | |
CN111881954A (en) | Transduction reasoning small sample classification method based on progressive cluster purification network | |
CN115147645A (en) | Membrane module membrane pollution detection method based on multi-feature information fusion | |
CN112504682A (en) | Chassis engine fault diagnosis method and system based on particle swarm optimization algorithm | |
CN112414446B (en) | Data-driven transmission sensor fault diagnosis method | |
CN102680646A (en) | Method of soft measurement for concentration of reactant in unsaturated polyester resin reacting kettle | |
CN111562109A (en) | Deep learning state identification and diagnosis method for mechanical equipment | |
CN115204272A (en) | Industrial system fault diagnosis method and equipment based on multi-sampling rate data | |
Haleem et al. | Improving the predictive response using ensemble empirical mode decomposition based soft sensors with auto encoder deep neural network | |
CN105823634A (en) | Bearing damage identification method based on time frequency relevance vector convolution Boltzmann machine | |
CN116044740B (en) | Pump fault diagnosis method based on acoustic signals | |
CN112149355A (en) | Soft measurement method based on semi-supervised dynamic feedback stack noise reduction self-encoder model | |
CN116822089A (en) | Data-driven motor internal disturbance analysis and modeling method | |
CN116644273A (en) | Fault diagnosis method and system based on interpretability multiplication convolution network | |
CN116933003A (en) | Remaining service life prediction method of unmanned aerial vehicle engine based on DaNet | |
CN113409213B (en) | Method and system for enhancing noise reduction of time-frequency diagram of fault signal of plunger pump | |
CN114252266A (en) | Rolling bearing performance degradation evaluation method based on DBN-SVDD model |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |