CN112985830A - Abs result automatic judging algorithm - Google Patents
Abs result automatic judging algorithm Download PDFInfo
- Publication number
- CN112985830A CN112985830A CN202110161343.6A CN202110161343A CN112985830A CN 112985830 A CN112985830 A CN 112985830A CN 202110161343 A CN202110161343 A CN 202110161343A CN 112985830 A CN112985830 A CN 112985830A
- Authority
- CN
- China
- Prior art keywords
- data
- abs
- layer
- vehicle
- network
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M17/00—Testing of vehicles
- G01M17/007—Wheeled or endless-tracked vehicles
-
- 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
- G06N3/088—Non-supervised learning, e.g. competitive learning
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Regulating Braking Force (AREA)
Abstract
The invention provides an abs result automatic judgment algorithm, which comprises the following steps of S1: acquiring measurement data and additional information related to the detection of the ABS braking performance of the automobile in the braking process; s2, preprocessing the acquired measurement data and additional information; s3: making sample data with a label; s4: the unsupervised deep learning network model is used for training, real vehicle measurement and corresponding indoor whole vehicle test data are accumulated by utilizing a cloud database, the brake execution conditions of vehicles with ABS on various road surfaces can be comprehensively detected, the vehicle brake safety is ensured, the traffic accidents are reduced, the practical value is high, and the unsupervised deep learning network is provided in the whole process from feature extraction to pattern recognition, so that the unsupervised deep learning network is suitable for automatically judging the ABS working state of the vehicle under the condition that class labels are lost.
Description
[ technical field ]
The invention relates to the technical field of ABS brake performance detection, in particular to an ABS result automatic judgment algorithm with a prominent application effect.
[ background art ]
With the increasing application of anti-lock brake systems (ABS) in automobiles, in the detection line of the entire automobile, devices for detecting the braking performance of the ABS of the automobile need to be added, and the devices can determine the working performance of the ABS by detecting the wheel speed, the automobile body speed, the pedal pressure, the pipeline pressure and the like of the automobile during braking. Because ordinary detection personnel do not have relevant professional knowledge, the ABS braking performance detection equipment of the automobile can automatically judge the working state of the ABS of the detected vehicle.
In massive monitoring data, because the occurrence probability of faults is lower than that of normal working conditions and manual marking of information is difficult, a large number of high-value and labeled sample sets are difficult to construct. How to perform effective state monitoring and fault identification under the condition of limited or lack of training sample label information is a key problem which is full of challenges and has important application value.
[ summary of the invention ]
In order to overcome the problems in the prior art, the invention provides an abs result automatic judgment algorithm with a prominent application effect.
The invention provides an abs result automatic decision algorithm, which comprises the following steps,
s1: acquiring measurement data and additional information related to the detection of the ABS braking performance of the automobile in the braking process;
s2, preprocessing the acquired measurement data and additional information;
s3: making sample data with a label;
s4: and (5) training an unsupervised deep learning network model.
Preferably, the measurement data in step S1 includes vehicle speed, wheel speed, pedal force, abs braking time, interval, brake line pressure data; the additional information includes vehicle type, vehicle number, vehicle age, road location and vehicle type technical parameters.
Preferably, the step S1 further includes dividing the acquired measurement data and additional information into structured data and unstructured data; the structured data comprises numerical data and a database; the unstructured data comprises character type data or a time domain oscillogram.
Preferably, the step S2 specifically includes the following steps, a 1: intercepting data of the length related to abs braking action in the measured data waveform time curve as original waveform data; a2: removing baseline drift noise by adopting a high-pass filter; a3: whether the noise is too high is determined based on a standard variance and a threshold value method, and when the noise is too high, a low-pass Butterworth filter is used for removing noise interference; a4: and (4) aiming at the characteristics of multi-source and isomerism of data sources, carrying out normalization processing on all data.
Preferably, in step S3, the measured data of the Abs braking system operating under different operating conditions have a certain similarity, and the time-domain distribution diagram of the fault data has a certain similarity, that is, the source domain and the target domain have a common part.
Preferably, in step S3, tags are added to the processed data under different working conditions and recorded as a Source Domain (SD); the evaluation conclusion of the overall state of the ABS comprises normal, fault and insufficient labels; and recording the preprocessed real-time acquired data as a target field (TargetDomain, TD for short).
Preferably, the unsupervised deep learning network model in the step S4 includes a primary network and a secondary network; the main network has a feedforward connection structure and a feedback connection structure and has local memory capacity, and the main network sequentially consists of an input layer, a bearing layer, an intermediate layer and an output layer; the secondary network is used for initialization of the primary network and comprises a visual layer and an implicit layer.
Preferably, the training process of step S4 is divided into two stages: the first step is that a bottom-up unsupervised training mode is adopted; secondly, adopting a top-down supervised learning mode; the bottom-up training is to perform layered training by using calibration-free data or calibration data; firstly, inputting a training sample, learning the weight of a first layer of a network until the weight reaches an n-1 th layer of a model, and taking the output of the model as the input of the n-1 th layer of the model to obtain neuron parameters of each layer, wherein the training process is unsupervised; the top-down learning is to further train the network through the data with the labels after the first training process is finished, so that the errors are transmitted layer by layer from top to bottom, the network parameters obtained by pre-training are finely adjusted, and finally, the network parameters of each layer of the model are determined, and the training process is supervised.
Preferably, the vehicle type and/or ABS braking system for accumulated abundant fault data extracts fault features comprehensively and effectively from multi-source and multi-structure measured data and additional information of the vehicle to be tested, such as different vehicle types, vehicle numbers, vehicle ages, road positions, numerical values, characters, pictures and the like, under the condition of limited or lacking training sample label information, so that the system is suitable for automatic judgment of the ABS working state of the vehicle under the condition of class label missing; the method comprises the following specific steps: A. taking unstructured data as the input of a convolutional neural network, and taking structured data as the input of a deep neural network; B. fully connecting neurons in a fully-connected layer of the CNN and a last hidden layer of the DNN with neurons in a first hidden layer of the feature fusion layer through a feature fusion layer comprising a plurality of hidden layers, wherein the fully-connected operation seamlessly integrates fault features extracted from unstructured data by the CNN and fault features extracted from structured data by the DNN; C. and inputting the output of the last hidden layer of the feature fusion layer into a softmax classifier, and classifying the faults.
Preferably, the implicit representative features are automatically extracted from different data sets under different working conditions, and a universal fault diagnosis model is established, so that the fault diagnosis of the abs system can be realized on the premise of changing the working conditions; the method comprises the following specific steps: A. acquiring ABS braking performance detection data, analyzing ABS detection results, and selecting m characteristic parameters capable of reflecting the working performance of the ABS of the automobile; B. making sample data with a label; C. carrying out data cleaning processing on n sample data (n-ns + nt) of the characteristic parameters; D. and constructing an unsupervised deep learning network model.
Compared with the prior art, the ABS result automatic judgment algorithm utilizes the cloud database to accumulate real vehicle measurement and corresponding indoor whole vehicle test data, and can comprehensively detect the brake execution conditions of the vehicle with the ABS on various road surfaces, thereby ensuring the safety of vehicle braking, reducing the occurrence of traffic accidents, and having higher practical value.
[ description of the drawings ]
FIG. 1 is a diagram of a database application system architecture of the present invention.
FIG. 2 is an unsupervised deep learning model of the present invention.
Fig. 3 is an ABS braking performance test result automatic determination algorithm 1 of the present invention.
Fig. 4 is the ABS braking performance test result automatic determination algorithm 2 of the present invention.
FIG. 5(a) is a vehicle speed/wheel speed comparison curve when the ABS of the present invention is operating normally.
Fig. 5(b) is a vehicle speed/wheel speed comparison curve at the time of ABS failure of the present invention.
FIG. 6(a) is a vehicle speed/wheel speed comparison curve for normal ABS operation on a road surface with the same adhesion coefficient according to the present invention.
FIG. 6(b) is a vehicle speed/wheel speed comparison curve for an ABS brake force insufficient on a road surface with the same adhesion coefficient according to the present invention.
FIG. 7 is a process flow of ABS brake data dimension reduction according to the present invention.
Fig. 8 is a schematic flow chart of an abs result automatic determination algorithm according to the present invention.
[ detailed description of the invention ]
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Referring to the following drawings, an abs result automatic determination algorithm according to the present invention is a database application system of an automobile abs braking fault diagnosis system, as described in fig. 1,
the method specifically comprises the following steps: private data interface 1, public data interface 2, Ethernet 3, brake platform electric control cabinet 4, abs brake detection platform 5, vehicle 6 under test, ODB adapter 7, station computer 8.
The station computer 8 is connected with the brake table electric control cabinet 4 and the ODB adapter 7 through a field bus (RS232/CAN bus and the like); and can be connected with the private data interface 1 and the public data interface 2 for communication through the Ethernet 3.
The other end of the OBD adapter 7 is connected with a vehicle-mounted OBD interface of the tested vehicle 6 through a vehicle-mounted diagnosis protocol (K line/CAN bus) and is communicated with an ABS Electronic Control Unit (ECU) on the OBD adapter.
The brake table electric control cabinet 4 is provided with an embedded control board card and an A/D signal conversion board card and is connected with an abs brake detection table 5 through a signal line.
By way of example, these components may have the following functions:
1. a station computer:
a. communicating with a brake table electric control cabinet, coordinating the tested vehicle and an abs brake detection table to detect, and receiving abs brake test data;
b. indicating the tested vehicle to brake, and receiving abs brake operation data of the tested vehicle through an OBD adapter;
c. embedding a deep learning algorithm model to realize database application of a deep neural network model and automatically judging abs braking performance of a detected vehicle;
d. a human-machine interface (HMI), a visual chart and the like are embedded so as to realize daily intelligent monitoring and interaction of industrial equipment;
e. and the data transmission device is connected with the private data interface and the public data interface for communication, and is used for issuing data and receiving data feedback.
2. abs braking detecting table and electric control cabinet for braking table
a. Performing abs braking detection;
b. acquiring abs braking test data;
c. and conditioning the acquired data into digital signals which can be recognized by a computer.
3. Tested vehicle and ODB adapter
a. Performing abs braking;
b. acquiring abs braking operation data;
c. and conditioning the acquired data into digital signals which can be recognized by a computer.
4. Private data interface and public data interface
a. Connecting the private cloud platform and the public cloud platform;
b. receiving test data and test results issued by a workstation computer;
c. receiving an abs braking data labeling sample set aiming at a specific vehicle type and vehicle number;
d. and feeding back the trained deep neural network model to the station computer and/or an abs braking labeling sample set required by the station computer to complete the training of the deep neural network model aiming at the specific vehicle type and vehicle number.
The invention relates to an automatic judgment algorithm for an ABS brake performance test result based on an unsupervised deep learning model, which specifically comprises the following steps as shown in the figure 2:
1. acquiring measurement data and additional information related to the detection of the ABS braking performance of the automobile in the braking process;
a. the detection data comprises measurement data of braking processes such as vehicle speed, wheel speed, pedal force, abs braking time, interval, brake pipeline pressure and the like;
b. the additional information includes, but is not limited to: vehicle type, vehicle number, vehicle age and road position; vehicle type technical parameters, for example: the system comprises a wheel base, a servicing quality, a tire specification, an ABS form, an ABS signal, power, torque and other power data, wherein the data are derived from a preset vehicle type and vehicle number technical parameter database;
c. in one embodiment of the invention, the acquired measurement data and additional information are distinguished into structured data and unstructured data. Structured data, e.g., numerical data, databases, etc.; unstructured data, such as text font data, time domain oscillograms, and the like;
2. the data preprocessing is carried out on the acquired measurement data and the additional information, and the method comprises the following steps:
a. intercepting data of the length related to abs braking action in the measured data waveform time curve as original waveform data;
b. removing baseline drift noise by adopting a high-pass filter;
c. whether the noise is too high is determined based on a standard variance and a threshold value method, and when the noise is too high, a low-pass Butterworth filter is used for removing noise interference;
d. aiming at the characteristics of multi-source and isomerism of data sources, all data are subjected to normalization processing;
3. sample data with a label is produced:
the measured data of the abs braking system running under different working conditions have certain similarity, and the time domain distribution diagram of the fault data has certain similarity, namely the source field and the target field have a common part, which is the premise of adopting an unsupervised deep learning network model in the invention.
b. In an embodiment of the invention, tags are added to the processed data under different working conditions and are recorded as a Source Domain (SD for short), for example, for the evaluation conclusion of the overall state of the ABS, the tags include normal, fault and deficiency; and recording the preprocessed real-time acquired data as a target field (TargetDomain, TD for short).
4. Training of unsupervised deep learning network models
a. The feature extraction of the input data is a very important step in a machine learning algorithm and is also a most time-consuming stage; before fault identification, characteristic indexes are extracted by means of human experience and a large amount of experimental research, the workload is large, and whether the selected characteristic indexes can completely express the operation mode of the abs braking process cannot be verified;
the unsupervised deep learning network model is adopted, the stage of characteristic design is skipped in the whole learning process, the input signals under different modes can be directly learned, and all characteristic information of the signals can be learned through layer-by-layer characteristic transformation in the learning process of the model, so that the global but not local optimal fault recognition result can be obtained;
the model may include: a primary network and a secondary network; the main network has a feedforward connection structure and a feedback connection structure and has local memory capacity, and consists of an input layer, a bearing layer, an intermediate layer and an output layer in sequence; the secondary network is used for initializing the main network and comprises a visual layer and an implicit layer;
the task of the model can be classification or regression, relating to the problem itself being modeled and the form of the label. The classification task in the invention can be automatic judgment of abs braking performance detection results; the regression task can be prediction of the next abs braking performance fault type and occurrence time of the vehicle to be detected according to the current detection result.
b. The training process of the unsupervised deep learning network model is divided into two stages: the first step is that a bottom-up unsupervised training mode is adopted; secondly, adopting a top-down supervised learning mode;
the bottom-up training is to perform layered training by using calibration-free data (or calibration data); firstly, inputting a training sample, learning the weight of a first layer of a network until the weight reaches an n-1 th layer of a model, and taking the output of the model as the input of the n-1 th layer of the model to obtain neuron parameters of each layer, wherein the training process is unsupervised;
the method is a characteristic learning process, and in the training process, due to the characteristics of capacity limitation, sparsity constraint and the like of the depth model, the model can learn all information contained in massive input data; the maximum difference between the initial value and the traditional neural network is that the initial value of the model is obtained in the internal structure of the learning training sample and is closer to the global optimum of the network, so that the initial value can obtain more accurate effect than the neural network;
the top-down learning is to further train the network through the data with the labels after the first training process is finished, so that the errors are transmitted layer by layer from top to bottom, the network parameters obtained by pre-training are finely adjusted, and finally, the network parameters of each layer of the model are determined, and the training process is supervised.
The training process of the unsupervised deep learning network model depends on big data, and the data acquisition mode in the invention comprises the acquisition through road test, bench test and vehicle-mounted tracking measurement, or the acquisition through the existing database or other ways.
An embodiment of an algorithm of the present invention, as shown in the following fig. 3, is aimed at extracting fault features from multi-source, multi-structure measured data of a vehicle to be tested and additional information from different vehicle types, vehicle numbers, vehicle ages, road positions, numerical values, characters, pictures and the like under the condition of limited or lack of training sample label information for vehicle types and/or ABS braking systems with accumulated abundant fault data in an all-round and effective manner, so that the method is suitable for automatic determination of ABS working state of a vehicle under the condition of class label missing; the method comprises the following specific steps:
1. firstly, taking unstructured data as the input of a convolutional neural network, and taking structured data as the input of a deep neural network;
a. the Convolutional Neural Network (CNN) consists of a convolutional layer, a sub-sampling layer and a full connection layer, and through operations such as convolution and pooling, the output of the full connection layer is the fault characteristics extracted from unstructured data;
b. the Deep Neural Network (DNN) has a plurality of hidden layers, wherein the first hidden layer extracts basic low-layer features from original data, and the subsequent hidden layer converts the basic low-layer features into more abstract high-layer features layer by layer, and the high-layer features can describe data distribution more accurately; the DNN can adaptively learn some rules of deep hiding from sample data without professional knowledge in a specific field; the output of the last hidden layer of the DNN is fault characteristics extracted from the structured data;
2. fully connecting neurons in a fully-connected layer of the CNN and a last hidden layer of the DNN with neurons in a first hidden layer of the feature fusion layer through a feature fusion layer comprising a plurality of hidden layers, wherein the fully-connected operation seamlessly integrates fault features extracted from unstructured data by the CNN and fault features extracted from structured data by the DNN;
a. if the output dimension of the CNN full-connection layer is NC, the feature vector Vc belongs to R1×Nc(Vc belongs to a space of dimension 1 XNc); let the output dimension of the last hidden layer of DNN be Nd, then its feature vector Vd belongs to R1×Nd(Vd belongs to a space of 1 × Nd dimensions); the feature vector Vin ∈ R constructed after the full join operation1×(Nc+Nd)(Vin belongs to a space of 1 × (Nc + Nd) dimensions);
b. using Vin as the input of the first hidden layer of the feature fusion layer, and performing fusion mapping on the feature vector Vin in a plurality of hidden layers of the feature fusion layer;
3. and finally, inputting the output of the last hidden layer of the feature fusion layer into a softmax classifier, and classifying the faults.
The invention also provides an algorithm embodiment, which aims to automatically extract implicit representative characteristics from different data sets under different working conditions and establish a universal fault diagnosis model, so that the fault diagnosis of an abs system can be realized on the premise of changing the working conditions, for example, the working condition simulation of an open road surface of an indoor test bench, the automatic judgment of vehicle-mounted real-time measurement data under a complex road surface and the like; the method comprises the following specific steps:
1. acquiring ABS braking performance detection data, analyzing ABS detection results, and selecting m characteristic parameters capable of reflecting the working performance of the ABS of the automobile; for example: selecting the slip rate, the adhesion coefficient utilization rate and the like as judgment indexes of ABS detection results;
a. the slip rate represents the difference degree between the speed and the wheel speed in the braking process of the automobile; the calculation formula is as follows:
wherein S represents a slip ratio, V represents a vehicle speed, r represents a wheel radius, and ω represents a wheel angular velocity; the control principle of the automobile abs on the braking process is to keep the automobile near the optimal slip ratio in the braking process so as to obtain larger ground adhesion;
b. adhesion coefficient utilization ratio: the road adhesion coefficient is the ratio of the adhesion to the normal pressure of the wheel, and can be regarded as the static friction coefficient between the wheel and the road, and the larger this coefficient is, the larger the available adhesion is, and the less the wheel is easy to slip. The adhesion coefficient utilization rate is the effective utilization degree of the maximum adhesion force of the whole vehicle to the ground in the braking process, and is the concrete embodiment of ABS control braking efficiency on a certain adhesion coefficient road surface;
2. making sample data with a label;
a. determining label parameters; for example, for the evaluation of the overall state of the ABS, a normal state is represented by (0, 1), a failure state is represented by (1, 0), and a state in which the braking force is insufficient is represented by (0, 0);
b. adding a label Ys to a sample data set Xs under different working conditions and recording as a Source Domain (SD) (see the following table);
table 4: source domain sample data
c. Marking the non-label sample data Xt collected in real time as a target field (TargetDomain, TD for short);
d. through the steps, obtaining a data input Xs of a source field SD, a label output Ys of the source field SD and a data input Xt of a target field TD; let Xs be Xt, Ys be Yt; and the edge distribution of the source data and the target data is different (i.e. P (Xs) ≠ P (Xt))), the condition distribution is different (i.e. Q (Ys |. Xs) ≠ Q (Yt |. Xt)); the problem to be solved is that a classifier trained by using source data predicts the label output Yt of the target data input Xt;
3. carrying out data cleaning processing on n sample data (n-ns + nt) of the characteristic parameters;
a. carrying out linear normalization processing on the non-binaryzation sample data to enable the non-binaryzation sample data to be distributed between 0 and 1; the calculation formula is as follows: x' ═ X-Xmin)/(Xmax-Xmin;
wherein X represents a certain sample value in a certain data attribute, Xmax represents a maximum value in a certain data attribute, Xmin represents a minimum value in a certain data attribute, and X' represents a numerical value after normalization processing;
b. removing noise contained in the normalized data through Fourier transform;
4. constructing an unsupervised deep learning network model, as shown in fig. 5;
a. the unsupervised deep learning network model is structurally the same as a traditional multilayer neural network, and comprises an input layer, a plurality of hidden layers and an output layer: the neurons of each layer are not connected, and the layers are all connected; for the abs braking system general fault diagnosis model, the node numbers of the input layer and the output layer respectively correspond to the input attribute and the category number of the data set;
the front part of the hidden layer is formed by stacking a plurality of layers of Automatic Encoders (AE), and a layer-by-layer training method is adopted during training, namely, the output of the previous layer is used as the input of the next layer for training in sequence; after extracting features, extracting public information between a source field and a target field hidden in a data set through edge distribution adaptation and condition distribution adaptation by each layer of automatic encoder to obtain feature representation for remarkably reducing the difference of edge distribution and condition distribution of the source field and the target field;
the last layer of the hidden layer is a classification layer representing expected output variables, preferably a Softmax classifier suitable for nonlinear multi-classification problems, the output of the Softmax classifier is probability values of corresponding samples belonging to different label states respectively, and the state with the maximum probability value is a final diagnosis result;
the training process is divided into two stages of pre-training and fine-tuning, wherein the pre-training adopts sample data of a source field SD or a target field TD as the input of a network, and the initialization of a plurality of layers of AE parameters at the front part is completed through a BP algorithm; when the AE parameters of each layer are pre-trained, the parameters of other layers are fixed and kept unchanged, and the fine adjustment adopts labeled source data
The whole network parameters including the classification layer are adjusted simultaneously through a BP algorithm, so that the discrimination performance of the network is optimal;
b. the auto-encoder AE comprises an output layer, a hidden layer and an input layer, wherein the output layer and the input layer have the same scale; the characteristic transformation process from the input layer to the hidden layer is called encoding, and the characteristic transformation process from the hidden layer to the output layer is called decoding;
the coding function is defined as f (x) Sf(Wx + p) and the decoding function is defined as g (h) Sg(WTh + q), wherein: sf、SgPreferably a sigmoid function, W representing a weight matrix between the input layer and the hidden layer, WTRepresenting a weight matrix between the hidden layer and the output layer; p represents the bias vector of the hidden layer; q represents a bias vector of the output layer; the AE parameter is recorded as theta;
suppose training sample set S ═ X1，…，XnThe process of training AE is essentially the process of training the parameter theta by using S; the specific method comprises the following steps:
JAE+spexpressing sparse self-coding, and L (x, y) is a reconstruction error function; beta is a weight coefficient for controlling the sparsity penalty term; rho is a sparsity parameter;
the abs braking system general fault diagnosis model operates in different working conditions, the edge distribution of the source data and the target data is inconsistent, and the distance of the edge distribution of the source domain and the target domain needs to be further shortened, and the specific method comprises the following steps:
JMindicating edge distribution adaptation, XS，XTRepresenting feature representations from a source domain and a target domain; learning the feature transformation matrix A by using edge distribution adaptation to obtain a new feature representation Z ═ ATX；
The training of the abs braking system general fault diagnosis model requires predicting the label of a target source by using a classifier trained by source data, considering the difference of condition distribution between the source data and the target data, minimizing the class condition probability to achieve the target of minimizing the condition probability, and still using the Maximum Mean Difference (MMD) to shorten the distance between the condition distribution of the source domain and the target domain, the specific method is as follows:
JCindicating adaptation of the condition distribution, XS，XTRepresenting a representation of features from a source domain and a target domain, C is a class conditional probability, Q (X)S|YS＝C)、Q(XT|YTC), C ∈ {1, 2 …, n }; learning a feature transformation matrix A by using conditional distribution adaptation to obtain a new feature representation Z ═ ATX；
e. Under the condition that the edge distribution and the condition distribution of the source data and the target data are different greatly, the classifier trained by the source data is used for predicting the output Yt of the target data, and the MMD distances of the edge distribution and the condition distributions of all classes need to be added for optimization, specifically:
f. The BP algorithm repeated iteration process comprises the following specific steps:
through pre-training, each layer of the model is regarded as a self-coding network, and the operations of coding and decoding are continuously carried out on input data until reaching an output layer of the depth model. Meanwhile, continuously calculating the error of the training sample by using a back propagation algorithm, and optimizing a loss function in each layer according to the gradient of the error to obtain an optimized weight and an optimized bias parameter;
wherein, W is the weight before adjustment, W' is the weight after adjustment, E is the error, eta is the learning rate;
calculating the error change sigma of two continuous iterations, and stopping the iteration process of the back propagation algorithm when the sigma is more than or equal to 0 and less than or equal to H;
finally, the error between the expected output and the actual output of the system is transmitted to each layer by utilizing back propagation again, so that the parameters of the integral model are optimized;
g. the Softmax classifier is trained through feature vectors output by a front multilayer automatic encoder; assuming that there are a total of k classification categories, the output of the Softmax classifier is a first-order probability matrix, and a probability value p is estimated for class labels from 1 to k, and the system equation is:
each row of the matrix is a parameter of the classifier corresponding to one classification label, and k rows are summed, and the loss function can be expressed as:
where l {. is an indicative function, i.e., when the value in the parenthesis is true, the function value is 1, otherwise it is 0.
The partial derivative function of the loss function over the parameter θ is as follows:
and obtaining the parameter value of the system by utilizing a gradient descent method according to the training sample, the loss function and the partial derivative function thereof.
5. Specifically, the construction steps of the universal fault diagnosis model for the abs braking system are as follows;
a. pre-training an algorithm model by using a label sample in a source field to obtain a weight and a bias parameter of the model;
b. respectively selecting the same number of sample data as input in the source field and the target field, adjusting the weight and the bias parameter of the model again, and simultaneously obtaining the corresponding characteristic representation of the data;
c. the feature representation of the source field is used for training a Softmax classifier to obtain a trained classification model,
d. taking the characteristic representation of the target field as the input of a Softmax classifier to obtain a classification label of each sample, thereby obtaining an automobile abs braking performance automatic judgment algorithm model;
next, an example of an ABS detection parameter and a detection method thereof that can be employed in the present invention will be described.
(example 1) a characteristic determination index using a deceleration ratio as an ABS detection result will be described.
Fig. 5 schematically shows a vehicle speed/wheel speed comparison curve at the time of ABS braking. The deceleration ratio is the ratio of the wheel deceleration to the vehicle body deceleration during braking, and under the normal operation of the ABS (FIG. 5(a)), the wheel deceleration is basically consistent with the vehicle body deceleration, and under the failure of the ABS (FIG. 5(b)), the wheel is rapidly locked, the wheel speed is reduced to zero in a short time, and the vehicle body speed is reduced relatively slowly, so that a large difference is formed between the two decelerations. Therefore, the larger the deceleration ratio, the worse the ABS adjustment capability;
(example 2) explanation will be given of the characteristic determination index using the braking deceleration as the ABS detection result.
Fig. 6 schematically shows a vehicle speed/wheel speed comparison curve during ABS braking under the same adhesion coefficient road surface, and the braking deceleration reflects the speed of vehicle speed reduction during braking, so the braking deceleration represents the braking effectiveness of the braking system to a certain extent. If a small deceleration value occurs on a road surface of the same adhesion coefficient, there may occur a case where the braking force is insufficient as shown in fig. 6(b) (in the case where the initial braking speed is smaller than that in fig. 6(a), the braking distance is 8 times that in fig. 6 (a));
(example 3) a method and an example of reconstructing raw data of ABS braking by a deep learning algorithm are described.
Fig. 7 schematically shows a flow of the dimension reduction processing of the raw data of the ABS braking using the automatic encoder AE.
The characteristic parameters reflecting the ABS working performance are more, and the ABS working performance is evaluated according to all the parameters, so that the training of the neural network is more complicated, and an accurate conclusion cannot be obtained possibly due to multiple correlations among the characteristic parameters. Therefore, for a desired abs brake system general fault diagnosis model, the data source can be first subjected to dimension reduction processing by using the automatic encoder AE, and the following beneficial effects are achieved:
the automatic encoder AE can automatically learn features from the data without labels, is a neural network taking reconstructed input information as a target, has stronger feature learning capability and can give better feature description than the original data;
compared with a traditional Principal Component Analysis (PCA) algorithm, the automatic encoder AE can represent both linear transformation and nonlinear transformation, and can be used as a layer for constructing a deep learning network. By setting proper dimensionality and sparse constraint, the self-encoder AE can learn more meaningful data projection than the technology of PCA and the like;
(example 4) an example of judging abs behavior by vehicle speed and wheel speed is explained.
1. Raw data preparation
1.1, first, the preparation part of the data.
In the practical operation of the embodiment, the body speed and the wheel speed are selected as the research objects. The method is mainly used for detecting the insufficient braking force by taking the initial braking speed of 40km/h as a reference, selecting the data sampling frequency as 1 point every 10 milliseconds, the sampling time as 30 seconds, and taking 1 point every 100 milliseconds in the last 20 seconds.
1.2 testing the working conditions
According to the test requirements of GB/T13594-.
1.3, tag data
Referring to fig. 5 and 6, evaluation labels for the overall state of the ABS, such as a normal state, a fault state, and an insufficient braking force, are added to sample data in the source domain.
2. Deep learning model training
2.1, establishing an abs braking performance automatic judgment algorithm model
And establishing a deep learning neural network consisting of a plurality of layers of automatic encoders, and obtaining a result of the abs braking performance automatic judgment hidden layer by adopting a Softmax regression algorithm as a top-layer classifier.
The number of the hidden layers is two, corresponding to the input layer of 1200 neurons, the first hidden layer has 50 neurons, and the second hidden layer has 25 neurons. And taking the output of the first hidden layer as the input of the second hidden layer, and obtaining a final reconstruction basis vector through the layer-by-layer learning of data.
The output layer is set to 2 neurons according to the number of classes. The normal state is represented by (0, 1), the failure state is represented by (1, 0), and the insufficient braking force state is represented by (0, 0).
2.2 occlusion handling of data
In order to enhance the robustness of the algorithm model, the input data needs to be subjected to shielding processing, wherein the shielding proportion is 0.25, namely, 25% of neuron node data of an input layer is uniformly distributed (qD distributed) to 0, and a new input sample x' is obtained; and then, taking the x' after the shielding treatment as the input of the model, expressing the sample x before shielding through deep learning of a multilayer neural network, and realizing the reconstruction of the original data, thereby realizing the characteristic self-expression with high robustness and improving the anti-noise capability of the algorithm.
2.3, Fine tuning
The unsupervised self-learning process lacks enough labels, so that the classification precision is not high, and the classification precision requirement of performance judgment cannot be met. In the fine adjustment process, source data with labels are adopted, the whole network parameters including the classification layer are adjusted based on a back propagation algorithm, the output residual error of the neuron is reduced, and then the weight and the bias parameters of the network model are optimized, so that the discrimination precision of the network reaches more than 99%.
As described above, several examples of the ABS detection parameters and the detection method thereof that can be used in the present invention have been described, but the ABS detection parameters and the detection method thereof that can be used in the present invention are not limited to the above examples, and a designer can freely design the ABS detection parameters and the detection method according to the type of vehicle, the ABS operating model, the indoor detection platform, the measurement and control system, the structural type of the relevant simulation mechanism, and the like.
Compared with the traditional ABS detection method, the method has the advantages that the cloud database is used for accumulating real vehicle measurement and corresponding indoor whole vehicle test data, and the braking execution conditions of the vehicle with the ABS on various road surfaces can be comprehensively detected, so that the braking safety of the vehicle is ensured, the occurrence of traffic accidents is reduced, and the method has higher practical value.
The above-described embodiments of the present invention do not limit the scope of the present invention. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.
Claims (10)
1. An abs result automatic decision algorithm, characterized by: comprises the following steps of (a) carrying out,
s1: acquiring measurement data and additional information related to the detection of the ABS braking performance of the automobile in the braking process;
s2, preprocessing the acquired measurement data and additional information;
s3: making sample data with a label;
s4: and (5) training an unsupervised deep learning network model.
2. The abs result automatic decision algorithm of claim 1, wherein: the measured data in the step S1 includes vehicle speed, wheel speed, pedal force, abs braking time, interval, and brake line pressure data; the additional information includes vehicle type, vehicle number, vehicle age, road location and vehicle type technical parameters.
3. The abs result automatic decision algorithm of claim 2, wherein: the step S1 further includes dividing the acquired measurement data and additional information into structured data and unstructured data; the structured data comprises numerical data and a database; the unstructured data comprises character type data or a time domain oscillogram.
4. The abs result automatic decision algorithm of claim 1, wherein: the step S2 specifically includes the following steps, a 1: intercepting data of the length related to abs braking action in the measured data waveform time curve as original waveform data; a2: removing baseline drift noise by adopting a high-pass filter; a3: whether the noise is too high is determined based on a standard variance and a threshold value method, and when the noise is too high, a low-pass Butterworth filter is used for removing noise interference; a4: and (4) aiming at the characteristics of multi-source and isomerism of data sources, carrying out normalization processing on all data.
5. The abs result automatic decision algorithm of claim 1, wherein: in step S3, the measured data of the Abs braking system operating under different conditions have a certain similarity, and the time-domain distribution diagram of the fault data has a certain similarity, that is, the source domain and the target domain have a common part.
6. The abs result automatic decision algorithm of claim 5, wherein: in step S3, tags are added to the processed data under different working conditions and recorded as a Source Domain (SD); the evaluation conclusion of the overall state of the ABS comprises normal, fault and insufficient labels; and recording the preprocessed real-time acquired data as a target field (TargetDomain, TD for short).
7. The abs result automatic decision algorithm of claim 1, wherein: the unsupervised deep learning network model in the step S4 includes a primary network and a secondary network; the main network has a feedforward connection structure and a feedback connection structure and has local memory capacity, and the main network sequentially consists of an input layer, a bearing layer, an intermediate layer and an output layer; the secondary network is used for initialization of the primary network and comprises a visual layer and an implicit layer.
8. The abs result automatic decision algorithm of claim 7, wherein: the training process of step S4 is divided into two stages: the first step is that a bottom-up unsupervised training mode is adopted; secondly, adopting a top-down supervised learning mode; the bottom-up training is to perform layered training by using calibration-free data or calibration data; firstly, inputting a training sample, learning the weight of a first layer of a network until the weight reaches an n-1 th layer of a model, and taking the output of the model as the input of the n-1 th layer of the model to obtain neuron parameters of each layer, wherein the training process is unsupervised; the top-down learning is to further train the network through the data with the labels after the first training process is finished, so that the errors are transmitted layer by layer from top to bottom, the network parameters obtained by pre-training are finely adjusted, and finally, the network parameters of each layer of the model are determined, and the training process is supervised.
9. The abs result automatic decision algorithm of claim 1, wherein: the system can comprehensively and effectively extract fault characteristics from multi-source and multi-structure measured data and additional information of the vehicle to be measured under the condition of limited or lacking training sample label information, such as different vehicle types, vehicle numbers, vehicle ages, road positions, numerical values, characters, pictures and the like, so that the system is suitable for automatic judgment of the ABS working state of the vehicle under the condition of lacking class labels; the method comprises the following specific steps: A. taking unstructured data as the input of a convolutional neural network, and taking structured data as the input of a deep neural network; B. fully connecting neurons in a fully-connected layer of the CNN and a last hidden layer of the DNN with neurons in a first hidden layer of the feature fusion layer through a feature fusion layer comprising a plurality of hidden layers, wherein the fully-connected operation seamlessly integrates fault features extracted from unstructured data by the CNN and fault features extracted from structured data by the DNN; C. and inputting the output of the last hidden layer of the feature fusion layer into a softmax classifier, and classifying the faults.
10. The abs result automatic decision algorithm of claim 1, wherein: the implicit representative features are automatically extracted from different data sets under different working conditions, and a universal fault diagnosis model is established, so that the fault diagnosis of an abs system can be realized on the premise of changing the working conditions; the method comprises the following specific steps: A. acquiring ABS braking performance detection data, analyzing ABS detection results, and selecting m characteristic parameters capable of reflecting the working performance of the ABS of the automobile; B. making sample data with a label; C. carrying out data cleaning processing on n sample data (n-ns + nt) of the characteristic parameters; D. and constructing an unsupervised deep learning network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110161343.6A CN112985830B (en) | 2021-02-05 | 2021-02-05 | Automatic abs result judging algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110161343.6A CN112985830B (en) | 2021-02-05 | 2021-02-05 | Automatic abs result judging algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112985830A true CN112985830A (en) | 2021-06-18 |
| CN112985830B CN112985830B (en) | 2023-06-23 |
Family
ID=76348107
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110161343.6A Active CN112985830B (en) | 2021-02-05 | 2021-02-05 | Automatic abs result judging algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112985830B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113156913A (en) * | 2021-02-05 | 2021-07-23 | 深圳大雷汽车检测股份有限公司 | ABS fault diagnosis system and method |
| CN113188807A (en) * | 2021-02-05 | 2021-07-30 | 深圳大雷汽车检测股份有限公司 | Abs result automatic judging algorithm |
| CN114155476A (en) * | 2022-02-07 | 2022-03-08 | 天津所托瑞安汽车科技有限公司 | AEB (automatic Emergency bank) accident scene identification method, device, equipment and medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001344590A (en) * | 2000-05-31 | 2001-12-14 | Fuji Electric Co Ltd | Neural network and method for learning the same and method for analyzing the same and method for judging abnormality |
| CN106482938A (en) * | 2016-10-14 | 2017-03-08 | 温州大学 | Brake fluid system multi-source fusion fault predicting method based on GA BP network |
| CN108956153A (en) * | 2018-04-27 | 2018-12-07 | 东华大学 | A kind of automobile anti-lock braking detection method based on RBF radial base neural net |
| CN109000935A (en) * | 2018-07-12 | 2018-12-14 | 清华大学深圳研究生院 | The determination method of new-energy automobile braking system performance |
| CN110751633A (en) * | 2019-10-11 | 2020-02-04 | 上海眼控科技股份有限公司 | Multi-axis cart braking detection method, device and system based on deep learning |
| CN111238825A (en) * | 2020-01-10 | 2020-06-05 | 东南大学 | Intelligent driving automatic emergency braking performance testing method for combined test pavement |
| US20200239026A1 (en) * | 2019-01-30 | 2020-07-30 | Perceptive Automata, Inc. | Navigating autonomous vehicles based on modulation of a world model representing traffic entities |
| CN111814966A (en) * | 2020-08-24 | 2020-10-23 | 国网浙江省电力有限公司 | Neural network architecture searching method, neural network application method, device and storage medium |
-
2021
- 2021-02-05 CN CN202110161343.6A patent/CN112985830B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001344590A (en) * | 2000-05-31 | 2001-12-14 | Fuji Electric Co Ltd | Neural network and method for learning the same and method for analyzing the same and method for judging abnormality |
| CN106482938A (en) * | 2016-10-14 | 2017-03-08 | 温州大学 | Brake fluid system multi-source fusion fault predicting method based on GA BP network |
| CN108956153A (en) * | 2018-04-27 | 2018-12-07 | 东华大学 | A kind of automobile anti-lock braking detection method based on RBF radial base neural net |
| CN109000935A (en) * | 2018-07-12 | 2018-12-14 | 清华大学深圳研究生院 | The determination method of new-energy automobile braking system performance |
| US20200239026A1 (en) * | 2019-01-30 | 2020-07-30 | Perceptive Automata, Inc. | Navigating autonomous vehicles based on modulation of a world model representing traffic entities |
| CN110751633A (en) * | 2019-10-11 | 2020-02-04 | 上海眼控科技股份有限公司 | Multi-axis cart braking detection method, device and system based on deep learning |
| CN111238825A (en) * | 2020-01-10 | 2020-06-05 | 东南大学 | Intelligent driving automatic emergency braking performance testing method for combined test pavement |
| CN111814966A (en) * | 2020-08-24 | 2020-10-23 | 国网浙江省电力有限公司 | Neural network architecture searching method, neural network application method, device and storage medium |
Non-Patent Citations (2)
| Title |
|---|
| 付昌星: "汽车制动性能检测***中神经网络的应用分析", 《汽车与驾驶维修(维修版)》, no. 07, pages 103 * |
| 杨宇等: "全参数动态学习深度信念网络在滚动轴承寿命预测中的应用", 《振动与冲击》, no. 10, pages 199 - 205 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113156913A (en) * | 2021-02-05 | 2021-07-23 | 深圳大雷汽车检测股份有限公司 | ABS fault diagnosis system and method |
| CN113188807A (en) * | 2021-02-05 | 2021-07-30 | 深圳大雷汽车检测股份有限公司 | Abs result automatic judging algorithm |
| CN113188807B (en) * | 2021-02-05 | 2024-05-03 | 深圳大雷汽车检测股份有限公司 | Automatic abs result judging algorithm |
| CN114155476A (en) * | 2022-02-07 | 2022-03-08 | 天津所托瑞安汽车科技有限公司 | AEB (automatic Emergency bank) accident scene identification method, device, equipment and medium |
| CN114155476B (en) * | 2022-02-07 | 2022-07-05 | 天津所托瑞安汽车科技有限公司 | AEB (automatic Emergency bank) accident scene identification method, device, equipment and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112985830B (en) | 2023-06-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113188807B (en) | Automatic abs result judging algorithm | |
| CN112985830A (en) | Abs result automatic judging algorithm | |
| CN110097755B (en) | Highway traffic flow state identification method based on deep neural network | |
| Ravikumar et al. | Gearbox fault diagnosis based on Multi-Scale deep residual learning and stacked LSTM model | |
| CN111666982B (en) | Electromechanical equipment fault diagnosis method based on deep neural network | |
| WO2019037557A1 (en) | Method for learning time sequence characteristics of locomotive operation | |
| CN112000084B (en) | Intelligent BIT design method of controller module based on 1D-CNN and GRU-SVM | |
| CN113392931B (en) | Hyperspectral open set classification method based on self-supervision learning and multitask learning | |
| US11954923B2 (en) | Method for rating a state of a three-dimensional test object, and corresponding rating system | |
| CN111414932A (en) | Classification identification and fault detection method for multi-scale signals of aircraft | |
| CN110991471B (en) | Fault diagnosis method for high-speed train traction system | |
| CN112766283B (en) | Two-phase flow pattern identification method based on multi-scale convolution network | |
| CN116610998A (en) | Switch cabinet fault diagnosis method and system based on multi-mode data fusion | |
| CN114218872A (en) | Method for predicting remaining service life based on DBN-LSTM semi-supervised joint model | |
| CN116311921A (en) | Traffic speed prediction method based on multi-spatial scale space-time converter | |
| CN112613542A (en) | Bidirectional LSTM-based enterprise decontamination equipment load identification method | |
| CN111967308A (en) | Online road surface unevenness identification method and system | |
| CN113283546B (en) | Furnace condition abnormity alarm method and system of heating furnace integrity management centralized control device | |
| CN114357372A (en) | Aircraft fault diagnosis model generation method based on multi-sensor data driving | |
| CN113156913A (en) | ABS fault diagnosis system and method | |
| CN114596726A (en) | Parking position prediction method based on interpretable space-time attention mechanism | |
| Zhang et al. | Cycle condition identification of loader based on optimized KNN algorithm | |
| CN117232817A (en) | Intelligent big data monitoring method of electric valve and Internet of things system | |
| CN115774942A (en) | Driving style identification model modeling and statistical method based on Internet of vehicles real vehicle data and SVM | |
| Zhang et al. | Kalman Filter-Based CNN-BiLSTM-ATT Model for Traffic Flow Prediction. |
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 |