CN114626474A - Vehicle power state detection method and system based on probabilistic neural network - Google Patents
Vehicle power state detection method and system based on probabilistic neural network Download PDFInfo
- Publication number
- CN114626474A CN114626474A CN202210276849.6A CN202210276849A CN114626474A CN 114626474 A CN114626474 A CN 114626474A CN 202210276849 A CN202210276849 A CN 202210276849A CN 114626474 A CN114626474 A CN 114626474A
- Authority
- CN
- China
- Prior art keywords
- layer
- power supply
- output
- neuron
- training
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 84
- 238000001514 detection method Methods 0.000 title claims abstract description 42
- 238000012549 training Methods 0.000 claims abstract description 108
- 210000002569 neuron Anatomy 0.000 claims abstract description 90
- 239000011159 matrix material Substances 0.000 claims abstract description 32
- 238000004364 calculation method Methods 0.000 claims abstract description 26
- 238000005070 sampling Methods 0.000 claims abstract description 25
- 238000000034 method Methods 0.000 claims abstract description 23
- 238000012545 processing Methods 0.000 claims description 14
- 230000014509 gene expression Effects 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 10
- 239000013598 vector Substances 0.000 claims description 5
- 238000012546 transfer Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 20
- 238000010586 diagram Methods 0.000 description 16
- 238000004088 simulation Methods 0.000 description 8
- 238000004146 energy storage Methods 0.000 description 6
- 238000003745 diagnosis Methods 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 239000003990 capacitor Substances 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000003111 delayed effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 240000007651 Rubus glaucus Species 0.000 description 1
- 235000011034 Rubus glaucus Nutrition 0.000 description 1
- 235000009122 Rubus idaeus Nutrition 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000007488 abnormal function Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Tests Of Electric Status Of Batteries (AREA)
Abstract
The application provides a vehicle power supply state detection method based on a probabilistic neural network, which comprises the following steps: s1, sampling and acquiring power supply fluctuation data of a vehicle power supply in three working states of flameout, normal running and engine starting; s2, extracting power supply fluctuation values in the power supply fluctuation data as training samples, respectively constructing a sample matrix A, B, C, and combining to form a training set D; s3, inputting the training set D into a probabilistic neural network for training to obtain a radial basis function center G, a threshold b and a weight IW of the neurons in the mode layer and a weight LW of the neurons in the summation layer; s4, respectively carrying out output calculation on the mode layer and the summation layer according to the parameters obtained by training, thereby obtaining the neuron output y of the output layer through calculation; and S5, predicting the vehicle power supply fluctuation state by using the trained probabilistic neural network. The method and the device have the effects of accurately predicting the fluctuation state type of the vehicle power supply in advance and being low in cost.
Description
Technical Field
The application relates to the technical field of power supply detection, in particular to a vehicle power supply state detection method and system based on a probabilistic neural network.
Background
The vehicle power supply has a plurality of different fluctuation conditions in the vehicle running working condition, the fluctuation is inevitable for the hardware of the vehicle storage battery with constant power, but due to the fluctuation of the power supply, the corresponding Electronic Control Unit (ECU) has abnormal functions, such as abnormal storage, system jamming and frequent false triggering of Electronic parts with standby batteries to switch to the standby battery for power supply.
In order to solve the above problems, a general design is that an energy storage device, such as a super capacitor, a small-capacity backup battery, and the like, is added to a secondary power supply loop of the ECU in a hardware redundancy manner, a system is powered down in a delayed manner, and meanwhile, a hard wire is combined to detect a power state, so as to protect stored information in the system and obtain more time to perform corresponding protection measures, and a basic schematic block diagram is shown in fig. 1. However, the above method cannot predict the fluctuation state type of the vehicle power supply in advance and accurately, so that the system does not have enough time for information protection such as storage, diagnosis and the like, and is also high in hardware cost.
Therefore, it is important to provide a method and a system for detecting the state of a vehicle power supply, which can predict the fluctuation type of the vehicle power supply in advance and accurately and have low cost.
Disclosure of Invention
In order to solve the technical problems that the fluctuation state type of a vehicle power supply cannot be predicted in advance and the hardware cost is high in vehicle power supply detection in the prior art, the application provides a vehicle power supply state detection method and system based on a probabilistic neural network, and the method and system are used for solving the technical problems.
According to a first aspect of the present application, a method for detecting a vehicle power state based on a probabilistic neural network is provided, which includes the following steps:
s1, sampling and acquiring power supply fluctuation data of a vehicle power supply in three working states of flameout, normal running and engine starting;
s2, extracting power supply fluctuation values in the power supply fluctuation data in three working states as training samples, respectively constructing sample matrixes A, B, C, and combining the sample matrixes A, B, C into a final training set D;
s3, inputting the training set D into a probabilistic neural network for training, wherein the probabilistic neural network comprises an input layer, a mode layer, a summation layer and an output layer, and the training is carried out to obtain a radial basis function center G, a threshold b and a weight IW of a neuron in the mode layer and a weight LW of the neuron in the summation layer;
s4, respectively carrying out output calculation on the mode layer and the summation layer according to the radial basis function center G, the threshold b and the weight IW of the neuron in the mode layer and the weight LW of the neuron in the summation layer, thereby obtaining the neuron output y of the output layer through calculation; and
and S5, predicting the vehicle power supply fluctuation state by using the trained probabilistic neural network.
According to the technical scheme, based on the fluctuation characteristics of the vehicle power supply, the probabilistic neural network is used as the classifier, the power supply fluctuation data of the vehicle power supply are continuously sampled, the vehicle power supply is detected, the accuracy and the real-time performance of power supply fluctuation event detection are improved, the power supply fluctuation category can be effectively detected under the condition of low cost, corresponding protection countermeasures are further made, and normal work of vehicle equipment such as a memory inside an ECU (electronic control Unit) and other power supply sensitive devices is guaranteed.
Preferably, the calculating process of the neuron output y of the output layer in step S4 specifically includes: carrying out output calculation on the mode layer to obtain neuron output a of the mode layeriOutputting the neurons of the pattern layer by aiMultiplying the weight LW of the neuron in the summation layer to obtain the neuron output n of the summation layeriFrom the neuron output n of the summation layeriAnd calculating to obtain the neuron output y of the output layer.
Preferably, in the step S2, the first m power fluctuation values in the power fluctuation data in three operating states are extracted as the training samples, and the training samples include (n-1) -dimensional input variables and 1-dimensional output variables, so as to construct m × n sample matrices A, B, C respectively, and the sample matrices A, B, C are merged into the training set D of i × j, where i is 3 × m and j is n.
Preferably, in step S3, a sample input matrix E is extracted from the training set D, the input layer receives the sample input matrix E and outputs an output matrix T, and expressions of the sample input matrix E and the output matrix T are specifically:
wherein d is a training sample, i is the number of the training samples, j-1 is the dimension of an input variable, the dimension of an output variable is 1, and ei (j-1) represents the (j-1) th dimension input variable of the ith training sample; ti1 represents the 1-dimensional output variable of the ith training sample.
Preferably, in step S3, the output matrix T is input into the pattern layer, each neuron in the pattern layer corresponds to one training sample, and an expression of a radial basis function center G of a neuron in the pattern layer is:
G＝E′
where E' is the transpose of the sample input matrix E.
Preferably, the pattern layer includes i neurons, and an expression of a threshold b corresponding to the i neurons in the pattern layer is as follows:
b1＝[b11 b12...b1i]
wherein, b1Set of thresholds b for i neurons in the pattern layer, b1iRepresenting the threshold of the ith neuron in the mode level, and spread is the speed of propagation of the radial basis function.
Preferably, the neuron output a of the pattern layeriThe specific calculation formula of (2) is:
ai＝exp(-||G-ei||2b1)
wherein e isi＝[ei1,ei2,…,ei(j-1)]，eiRepresenting the set of input vectors for all training samples.
Preferably, in step S3, the radial basis function center G of the neuron in the pattern layer is used as the weight IW of the neuron in the pattern layer, and the output matrix T is used as the weight LW of the neuron in the summation layer.
Preferably, the neuron output n of the summation layeriThe specific calculation formula of (2) is:
ni＝LWai
the specific calculation formula of the neuron output y of the output layer is as follows:
y＝compet(ni)
wherein, the comp is a competition transfer function.
According to a second aspect of the present application, a probabilistic neural network-based vehicle power state detection system is provided, including:
the sampling module is configured for sampling and acquiring power supply fluctuation data of a vehicle power supply in three working states of flameout, normal running and engine starting;
an operation processing unit configured to extract power supply fluctuation values in the power supply fluctuation data in three working states as training samples, respectively construct a sample matrix A, B, C, combine the sample matrix A, B, C into a final training set D, input the training set D into a probabilistic neural network for training, the probabilistic neural network comprises an input layer, a mode layer, a summation layer and an output layer, obtain a radial basis function center G, a threshold b and a weight IW of neurons in the mode layer and a weight LW of neurons in the summation layer through the training, respectively perform output calculation on the mode layer and the summation layer according to the radial basis function center G, the threshold b and the weight IW of the neurons in the mode layer and the weight LW of the neurons in the summation layer, and thereby calculate a neuron output y of the output layer, and simulating the training process of the probabilistic neural network to obtain a final training result.
The application provides a vehicle power supply state detection method and system based on a probabilistic neural network, wherein the probabilistic neural network is used as a classifier based on the fluctuation characteristics of a vehicle power supply, a sampling module is used for continuously sampling power supply fluctuation data of the vehicle power supply, and an operation processing unit is used for operation and analysis, so that the fluctuation category of the vehicle power supply can be predicted accurately in advance and judged in advance, and an ECU (electronic control unit) system can have enough time for information protection such as storage, diagnosis and the like; an energy storage device can be omitted on hardware, so that the BOM cost is saved; the whole system has simple structure and low cost, the operation processing unit can use the common MCU in the market, and the existing ECU hardware structure is used for realizing the detection function without additionally increasing the cost.
Drawings
The accompanying drawings are included to provide a further understanding of the embodiments and are incorporated in and constitute a part of this specification. The drawings illustrate embodiments and together with the description serve to explain the principles of the application. Other embodiments and many of the intended advantages of embodiments will be readily appreciated as they become better understood by reference to the following detailed description. The elements of the drawings are not necessarily to scale relative to each other. Like reference numerals designate corresponding similar parts.
FIG. 1 is a schematic block diagram of a prior art detection of a vehicle power state;
FIG. 2 is a block diagram of a probabilistic neural network based vehicle power state detection system according to an embodiment of the present application;
FIG. 3 is a topology diagram of a probabilistic neural network according to a specific embodiment of the present application;
FIG. 4 is a functional block diagram of a probabilistic neural network based vehicle power state detection system according to an embodiment of the present application;
FIG. 5 is a flowchart of a probabilistic neural network based vehicle power state detection method according to an embodiment of the present application;
FIG. 6 is a flow chart of training of a training set D in a probabilistic neural network according to an embodiment of the present application;
FIG. 7 is a diagram of simulation results of probabilistic neural network training data classification results and errors, according to an embodiment of the present application;
FIG. 8 is a diagram illustrating simulation results for predicting a vehicle power supply fluctuation state using a trained probabilistic neural network, in accordance with an embodiment of the present application;
FIG. 9 is a graph of a simulation result of a BP neural network predicting a prediction result and an error of a vehicle power supply fluctuation state after being trained with 30 training samples according to an embodiment of the present application;
fig. 10 is a diagram of simulation results of prediction results and errors of a BP neural network after training of 70 training samples according to an embodiment of the present application.
Description of reference numerals: 1. a sampling module; 2. an arithmetic processing unit; 3. an input layer; 4. a mode layer; 5. a summing layer; 6. and (5) outputting the layer.
Detailed Description
Features and exemplary embodiments of various aspects of the present application will be described in detail below, and in order to make objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail below with reference to the accompanying drawings and the embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application. It will be apparent to one skilled in the art that the present application may be practiced without some of these specific details. The following description of the embodiments is merely intended to provide a better understanding of the present application by illustrating examples thereof.
It is noted that, herein, relational terms such as first and second, and the like may be 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. The term "comprising", without further limitation, means that the element so defined is not excluded from the list of additional identical elements in a process, method, article, or apparatus that comprises the element.
Fig. 1 shows a schematic block diagram of detecting a vehicle power state in the prior art, as shown in fig. 1, in the prior art, an energy storage device, such as a super capacitor, a small-capacity backup battery, and the like, is added to a secondary power supply loop of an ECU in a hardware redundancy manner, a system is powered down in a delayed manner, and meanwhile, the power state is detected in combination with a hard wire, stored information in the system is protected, and more time is obtained to perform corresponding protection measures. On one hand, the method cannot predict the fluctuation state type of the vehicle power supply in advance and accurately, so that the system does not have enough time for information protection such as storage, diagnosis and the like; on the other hand, the energy storage device and the elements required by hard-line detection are required in terms of hardware requirements, and the cost is high.
According to a first aspect of the present application, a probabilistic neural network based vehicle power state detection system is presented.
Fig. 2 shows a block diagram of a vehicle power state detection system based on a probabilistic neural network according to an embodiment of the present application, fig. 3 shows a topological diagram of the probabilistic neural network according to a specific embodiment of the present application, and as shown in fig. 2 and fig. 3, the detection system includes:
the system comprises a sampling module 1, a data acquisition module and a data processing module, wherein the sampling module is configured for sampling and acquiring power supply fluctuation data of a vehicle power supply in three working states of flameout, normal running and engine starting;
the operation processing unit 2 is configured to extract power supply fluctuation values in power supply fluctuation data in three working states as training samples, respectively construct a sample matrix A, B, C, combine the sample matrix A, B, C into a final training set D, input the training set D into a probabilistic neural network for training, the probabilistic neural network comprises an input layer 3, a mode layer 4, a summation layer 5 and an output layer 6, obtain a radial basis function center G, a threshold b and a weight IW of a neuron in the mode layer 4 and a weight LW of the neuron in the summation layer 5 through training, respectively perform output calculation on the mode layer 4 and the summation layer 5 according to the radial basis function center G, the threshold b and the weight IW of the neuron in the mode layer 4 and the weight LW of the neuron in the summation layer 5, thereby calculate a neuron output y of the output layer 6, and simulate a training process of the probabilistic neural network, and obtaining a final training result.
Fig. 4 is a schematic block diagram illustrating an application of a vehicle power state detection system based on a probabilistic neural network according to an embodiment of the present application, and as shown in fig. 4, in the embodiment, the sampling module 1 is specifically an AD sampling circuit, and the AD sampling circuit samples and obtains a power supply fluctuation AD value (i.e., power supply fluctuation data) of a vehicle power supply in three operating states of flameout, normal driving, and engine starting. The arithmetic processing unit 2 is specifically an MCU, and in other embodiments, may also be a single chip or other processing units with certain arithmetic capability, such as a raspberry unit. The detection system is mounted on an ECU, and the ECU is mounted on a vehicle to detect the fluctuation state of the vehicle power supply.
According to the method and the device, based on the fluctuation characteristics of the vehicle power supply, the Probabilistic Neural Network (PNN) is used as the classifier, the fluctuation state of the power supply (namely the vehicle power supply) of the detection system is detected, the accuracy and the real-time performance of power supply fluctuation event detection are improved, the power supply fluctuation category can be effectively detected, corresponding protection countermeasures are further made, and the normal work of a storage inside the ECU and other power supply sensitive devices is guaranteed. An energy storage device can be omitted on hardware, so that the BOM cost is saved; the whole system has simple structure and low cost, the operation processing unit can use the common MCU in the market, and the existing ECU hardware structure is used for realizing the detection function without additionally increasing the cost.
According to a second aspect of the application, based on the detection system, a vehicle power state detection method based on a probabilistic neural network is provided.
Fig. 5 is a flowchart of a probabilistic neural network-based vehicle power state detection method according to an embodiment of the present application, and as shown in fig. 5, the detection method includes the following steps:
and S1, sampling and acquiring power supply fluctuation data of the vehicle power supply in three working states of flameout, normal running and engine starting.
In a specific embodiment, power supply fluctuation data of a vehicle power supply in three working states of flameout, normal running and engine starting are respectively collected through an AD sampling circuit, the power supply fluctuation data comprise a plurality of groups of power supply fluctuation AD values, and sampling parameters are as follows:
the sampling frequency f is 1 KHz;
each sampling time t is 3 s;
the sampling number k is 50.
S2, extracting power supply fluctuation values in the power supply fluctuation data under the three working states as training samples, respectively constructing sample matrixes A, B, C, and merging the sample matrixes A, B, C into a final training set D.
In a specific embodiment, the first m power supply fluctuation AD values in the power supply fluctuation data in three operating states are extracted as training samples, each training sample includes an (n-1) -dimensional input variable and a 1-dimensional output variable, that is, each training sample includes n-dimensional features, so as to respectively construct an m × n sample matrix A, B, C, and the sample matrices A, B, C are combined into an i × j training set D, D ═ a; b; c ], i.e., i ═ 3 × m, j ═ n.
And S3, inputting the training set D into a probabilistic neural network for training, wherein the probabilistic neural network comprises an input layer, a mode layer, a summation layer and an output layer, and the radial basis function center G, the threshold b and the weight IW of the neuron in the mode layer and the weight LW of the neuron in the summation layer are obtained through training.
And S4, respectively carrying out output calculation on the mode layer and the summation layer according to the radial basis function center G, the threshold b and the weight IW of the neuron in the mode layer and the weight LW of the neuron in the summation layer, thereby obtaining the neuron output y of the output layer through calculation.
In a specific embodiment, a Probabilistic Neural Network (PNN) is first created, which includes an input layer, a pattern layer, a summation layer, and an output layer, and then a training set D is input into the PNN for training. Fig. 6 shows a training flow chart of the training set D in the probabilistic neural network according to an embodiment of the present application, and as shown in fig. 6, the training process is specifically as follows:
a) and extracting a sample input matrix E from the training set D, inputting the sample input matrix E into an input layer, and outputting an output matrix T by the input layer.
In the specific embodiment, the training set D defines an i × j matrix, i rows represent a total of i training samples in the training set D, the data of the first (j-1) column in the j columns is the power fluctuation AD value of the extracted training sample, and the j-th column represents the state of the vehicle power supply, i.e., which state of the vehicle power supply is in the off state, the normal driving state and the engine start state. Specifically, in the present embodiment, in the data in the j-th column, "1" represents that the vehicle power supply is in the key-off state, "2" represents that the vehicle power supply is in the normal running state, and "3" represents that the vehicle power supply is in the engine start state. The expressions of the sample input matrix E and the output matrix T are specifically:
wherein d is a training sample, i is the number of the training samples, j-1 is the dimension of an input variable, the dimension of an output variable is 1, and ei (j-1) represents the (j-1) th dimension input variable of the ith training sample; t is ti1Representing the 1-dimensional output variable of the ith training sample.
b) And inputting an output matrix T output by the input layer into the mode layer, training to obtain a radial basis function center G, a threshold b and a weight IW of the neuron in the mode layer, inputting an output value of the mode layer into the summation layer, and training to obtain a weight LW of the neuron in the summation layer.
In a specific embodiment, each neuron in the pattern layer corresponds to one training sample, that is, an expression of a radial basis function center G corresponding to i neurons in the pattern layer is:
G＝E′
where E' is the transpose of the sample input matrix E.
The expression of the threshold b corresponding to i neurons in the mode layer is as follows:
b1＝[b11 b12...b1i]
wherein, b1Set of thresholds b for i neurons in the pattern layer, b1iRepresenting the threshold of the ith neuron in the mode layer, spread is the speed of propagation of the radial basis function.
The expression of the weight IW corresponding to the i neurons in the mode layer is as follows:
IW＝G
the expression of the weight LW corresponding to the neuron in the summation layer is as follows:
LW＝T
c) and respectively carrying out output calculation on the mode layer and the summation layer according to the radial basis function center G, the threshold b and the weight IW of the neuron in the mode layer obtained by training and the weight LW of the neuron in the summation layer, and calculating the neuron output of the output layer according to the output calculation result.
In a specific embodiment, the output calculation is performed on the mode layer to obtain the neuron output a of the mode layeriOutput a of each neuron in the mode layeriThe specific calculation formula of (A) is as follows:
ai＝exp(-||G-ei||2b1)
wherein e isi＝[ei1,ei2,…,ei(j-1)]，eiRepresenting a set of input vectors for i training samples.
General modeNeuronal output of layer aiMultiplying the weight LW of the neuron in the summation layer to obtain the neuron output n of the summation layeriThe specific calculation formula is as follows:
ni＝LWai
neuron output n according to summation layeriAnd calculating to obtain the neuron output y of the output layer by using a competitive transfer function (comp), wherein the specific calculation formula is as follows:
y＝compet(ni)
with continued reference to fig. 5, after step S4,
and S5, predicting the vehicle power supply fluctuation state by using the trained probabilistic neural network.
In a specific embodiment, the first 10 power supply fluctuation AD values in the power supply fluctuation data in the three operating states are extracted as training samples, where each training sample includes 10-dimensional input variables and 1-dimensional output variables, i.e., m is 10, n is 11, i is 3 m is 30, and j is n is 11. Fig. 7 is a diagram showing simulation results of classification results and errors of probabilistic neural network training data according to an embodiment of the present application, and as shown in fig. 7, 30 training samples for training the probabilistic neural network are used as input, the probabilistic neural network can accurately classify states "1" or "2" or "3" (flameout or normal driving or engine starting) of the vehicle power supply, and the error rate is 0, which reflects the effectiveness and accuracy of the probabilistic neural network in predicting the vehicle power supply fluctuation state. Fig. 8 is a diagram showing simulation results of predicting a vehicle power supply fluctuation state by using a trained probabilistic neural network according to an embodiment of the present application, and as shown in fig. 8, 10 new power supply fluctuation AD values (prediction samples) of a vehicle power supply are input to the trained probabilistic neural network, and the probabilistic neural network can predict a corresponding vehicle power supply fluctuation state "1" or "2" or "3" (flameout or normal driving or engine starting).
In particular embodiments, the predictive effect of other neural network algorithms on predicting vehicle power supply surge conditions is also evaluated. Specifically, a BP neural network is selected to evaluate the prediction effect of the BP neural network on predicting the fluctuation state of the vehicle power supply.
Firstly, a BP neural network is created, and the specific parameters are shown in the following table:
the input layer vector of the BP neural network is 10, the number of the neurons is 8, and the vector of the output layer is 3.
Training samples are then also input for training. Fig. 9 is a diagram illustrating a prediction result and an error simulation result of a BP neural network after being trained by 30 training samples according to an embodiment of the present application, and as shown in fig. 9, after being trained by 30 training samples, the BP neural network inputs 10 new power supply fluctuation AD values (prediction samples) of a vehicle power supply to the trained BP neural network, and finally the BP neural network can predict a corresponding vehicle power supply fluctuation state "1" or "2" or "3" (shutdown or normal driving or engine start), but the error is high. Fig. 10 is a diagram illustrating a prediction result and an error simulation result of the BP neural network after being trained by 70 training samples according to an embodiment of the present application, and as shown in fig. 10, after being trained by 70 training samples, the BP neural network inputs 10 new power supply fluctuation AD values (prediction samples) of the vehicle power supply to the trained BP neural network, and finally the BP neural network can predict a corresponding vehicle power supply fluctuation state "1" or "2" or "3" (flameout or normal driving or engine starting), and the error rate is 0.
Therefore, in practical applications, comparing the probabilistic neural network and the BP neural network can lead to the following conclusions:
1. the training sample amount required by the probabilistic neural network is small, and the number of the training samples required by the probabilistic neural network in the application is less than half of that of the training samples required by using a BP neural network, so that stable and accurate prediction can be realized;
2. the probability neural network has strong sample adding capability and can tolerate individual wrong samples, new training samples are added or removed from individual problematic samples in the fault diagnosis process, only corresponding mode layer units are needed to be added or reduced, for the BP neural network, training needs to be carried out again after the training samples are changed, weight connection is changed, and the method is equivalent to reestablishing the whole network.
In practical application, a vehicle power supply fluctuation state sample library is correspondingly required to be established, the content of the vehicle power supply fluctuation state sample library changes along with the increase and change of the vehicle power supply fluctuation state, and the advantage of strong adding capability of a probabilistic neural network sample can be fully embodied.
In a specific embodiment, since the present application uses the sampling MCU as an arithmetic processing unit to predict the vehicle power supply fluctuation state by using the probabilistic neural network algorithm, the frequency and memory usage required for operating the probabilistic neural network MCU are also evaluated:
1. and (3) evaluating the memory occupation:
in this application, a 10-bit precision AD conversion module is used, and the length of each power supply data is as follows:
Data Lenth＝2Bytes；
for 30 training samples, 10 data in each group, total number of samples data size:
Total Data Lenth＝600Bytes；
MCU master frequency evaluation:
a universal MCU with main frequency of 24MHz is used, and the time t for executing a single instruction is actually measured to be about 350 ns;
the probabilistic neural network initialization algorithm requires about 6300 instructions in total;
the time T required by initialization is approximately equal to 2.2ms, and the total initialization time (< 200ms) of the system is hardly influenced;
314 instructions needed to perform sample prediction;
the predicted required time T is approximately equal to 110us, and system resources cannot be occupied for a long time in practical application, so that other tasks are jammed.
In summary, the probabilistic neural network used in the present application requires an additional memory resource of 600Bytes for MCU configuration with a dominant frequency greater than 12MHz, and is suitable for most MCUs on the market.
The application provides a vehicle power supply state detection method and system based on a probabilistic neural network, wherein the probabilistic neural network is applied as a classifier based on the fluctuation characteristics of a vehicle power supply, an AD sampling circuit is used for continuously sampling power supply fluctuation data of the vehicle power supply, and an MCU (microprogrammed control unit) operation processing unit is used for performing operation analysis, so that the fluctuation category of the vehicle power supply can be predicted accurately in advance and judged in advance, and an ECU (electronic control unit) system of the vehicle can have enough time for information protection such as storage, diagnosis and the like; compared with the traditional detection mode, an energy storage device can be omitted in hardware, and BOM cost is saved; the whole detection system is simple in structure and low in cost, the operation processing unit can use a common MCU in the market, and then the existing ECU hardware structure is used for realizing the detection function, so that the cost is not additionally increased.
In the embodiments of the present application, it should be understood that the disclosed technical contents may be implemented in other ways. The above-described embodiments of the apparatus/system/method are merely illustrative, and for example, the division of the units may be a logical division, and in actual implementation, there may be another division, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or contributed to by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic disk, or an optical disk, and various media capable of storing program codes.
It is apparent that various modifications and variations can be made to the embodiments of the present application by those skilled in the art without departing from the spirit and scope of the application. In this way, the present application is also intended to cover such modifications and changes if they come within the scope of the claims of the present application and their equivalents. The word "comprising" does not exclude the presence of other elements or steps than those listed in a claim. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. Any reference signs in the claims shall not be construed as limiting the scope.
Claims (10)
1. A vehicle power state detection method based on a probabilistic neural network is characterized by comprising the following steps:
s1, sampling and acquiring power supply fluctuation data of a vehicle power supply in three working states of flameout, normal running and engine starting;
s2, extracting power supply fluctuation values in the power supply fluctuation data in three working states as training samples, respectively constructing sample matrixes A, B, C, and combining the sample matrixes A, B, C into a final training set D;
s3, inputting the training set D into a probabilistic neural network for training, wherein the probabilistic neural network comprises an input layer, a mode layer, a summation layer and an output layer, and the radial basis function center G, a threshold b and a weight IW of the neuron in the mode layer and the weight LW of the neuron in the summation layer are obtained through the training;
s4, respectively carrying out output calculation on the mode layer and the summation layer according to the radial basis function center G, the threshold b and the weight IW of the neuron in the mode layer and the weight LW of the neuron in the summation layer, thereby obtaining the neuron output y of the output layer through calculation; and
and S5, predicting the vehicle power supply fluctuation state by using the trained probabilistic neural network.
2. The detection method according to claim 1, wherein the calculation process of the neuron output y of the output layer in the step S4 specifically includes: carrying out output calculation on the mode layer to obtain neuron output a of the mode layeriOutputting the neurons of the mode layer as aiMultiplying the weight LW of the neuron in the summation layer to obtain the neuron output n of the summation layeriFrom the neuron output n of the summation layeriAnd calculating to obtain the neuron output y of the output layer.
3. The detection method according to claim 1, wherein in step S2, the first m power supply fluctuation values in the power supply fluctuation data in three operation states are extracted as the training samples, the training samples include (n-1) -dimensional input variables and 1-dimensional output variables, so as to construct m × n sample matrices A, B, C, respectively, and the sample matrices A, B, C are combined into the training set D of i × j, where i is 3 m, and j is n.
4. The detection method according to claim 3, wherein in the step S3, a sample input matrix E is extracted from the training set D, the input layer receives the sample input matrix E and outputs an output matrix T, and expressions of the sample input matrix E and the output matrix T are specifically:
wherein d is a training sample, i is the number of training samples, (j-1) is the dimension of an input variable, the dimension of an output variable is 1, ei(j-1)An (j-1) -dimensional input variable representing an ith training sample; t is ti1Representing the 1-dimensional output variable of the ith training sample.
5. The detection method according to claim 4, wherein in the step S3, the output matrix T is input into the pattern layer, each neuron in the pattern layer corresponds to one of the training samples, and the radial basis function center G of a neuron in the pattern layer is expressed as:
G＝E′
where E' is the transpose of the sample input matrix E.
6. The detection method according to claim 5, wherein the pattern layer comprises i neurons, and an expression of a threshold b corresponding to the i neurons in the pattern layer is as follows:
b1＝[b11 b12...b1i]
wherein, b1Set of thresholds b for i neurons in the pattern layer, b1iRepresenting the threshold of the ith neuron in the mode layer, spread is the speed of propagation of the radial basis function.
7. The detection method according to claim 6, wherein the neuron of the pattern layer outputs aiThe specific calculation formula of (A) is as follows:
ai＝exp(-||G-ei||2b1)
wherein e isi＝[ei1,ei2,…,ei(j-1)]，eiRepresenting the set of input vectors for all training samples.
8. The detection method according to claim 2, wherein in step S3, the radial basis function center G of the neuron in the pattern layer is used as the weight IW of the neuron in the pattern layer, and the output matrix T is used as the weight LW of the neuron in the summation layer.
9. The detection method according to claim 8, wherein the neuron of the summation layer outputs niThe specific calculation formula of (2) is:
ni＝LWai
the specific calculation formula of the neuron output y of the output layer is as follows:
y＝compet(ni)
wherein, the composition is a competition transfer function.
10. A probabilistic neural network-based vehicle power state detection system, comprising:
the sampling module is configured for sampling and acquiring power supply fluctuation data of a vehicle power supply in three working states of flameout, normal running and engine starting;
an operation processing unit configured to extract power supply fluctuation values in the power supply fluctuation data in three working states as training samples, respectively construct a sample matrix A, B, C, combine the sample matrix A, B, C into a final training set D, input the training set D into a probabilistic neural network for training, the probabilistic neural network comprises an input layer, a mode layer, a summation layer and an output layer, obtain a radial basis function center G, a threshold b and a weight IW of neurons in the mode layer and a weight LW of neurons in the summation layer through the training, respectively perform output calculation on the mode layer and the summation layer according to the radial basis function center G, the threshold b and the weight IW of the neurons in the mode layer and the weight LW of the neurons in the summation layer, and thereby calculate a neuron output y of the output layer, and simulating the training process of the probabilistic neural network to obtain a final training result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210276849.6A CN114626474A (en) | 2022-03-21 | 2022-03-21 | Vehicle power state detection method and system based on probabilistic neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210276849.6A CN114626474A (en) | 2022-03-21 | 2022-03-21 | Vehicle power state detection method and system based on probabilistic neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114626474A true CN114626474A (en) | 2022-06-14 |
Family
ID=81904745
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210276849.6A Pending CN114626474A (en) | 2022-03-21 | 2022-03-21 | Vehicle power state detection method and system based on probabilistic neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114626474A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101894185A (en) * | 2010-06-29 | 2010-11-24 | 北京航空航天大学 | Method for predicting life of small sample data object based on dynamic bipolar modified probabilistic neural network (MPNN) |
| CN109508653A (en) * | 2018-10-26 | 2019-03-22 | 南京邮电大学 | A kind of subjective and objective individual combat Emotion identification method merged based on EEG signals with psychology |
| CN110459292A (en) * | 2019-07-02 | 2019-11-15 | 南京邮电大学 | A kind of risk management stage division based on cluster and PNN |
| US20210095965A1 (en) * | 2019-09-26 | 2021-04-01 | Harbin Engineering University | Elman neural network assisting tight-integrated navigation method without GNSS signals |
| CN112637812A (en) * | 2020-12-16 | 2021-04-09 | 温州大学 | Vehicle-mounted cooperative communication relay selection method based on supervised machine learning |
| CN113391188A (en) * | 2021-04-22 | 2021-09-14 | 厦门大学 | Self-calibration system and method based on neural network |
-
2022
- 2022-03-21 CN CN202210276849.6A patent/CN114626474A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101894185A (en) * | 2010-06-29 | 2010-11-24 | 北京航空航天大学 | Method for predicting life of small sample data object based on dynamic bipolar modified probabilistic neural network (MPNN) |
| CN109508653A (en) * | 2018-10-26 | 2019-03-22 | 南京邮电大学 | A kind of subjective and objective individual combat Emotion identification method merged based on EEG signals with psychology |
| CN110459292A (en) * | 2019-07-02 | 2019-11-15 | 南京邮电大学 | A kind of risk management stage division based on cluster and PNN |
| US20210095965A1 (en) * | 2019-09-26 | 2021-04-01 | Harbin Engineering University | Elman neural network assisting tight-integrated navigation method without GNSS signals |
| CN112637812A (en) * | 2020-12-16 | 2021-04-09 | 温州大学 | Vehicle-mounted cooperative communication relay selection method based on supervised machine learning |
| CN113391188A (en) * | 2021-04-22 | 2021-09-14 | 厦门大学 | Self-calibration system and method based on neural network |
Non-Patent Citations (1)
| Title |
|---|
| 施雷红;陈强;: "基于PNN的煤炭输送机减速器齿轮故障诊断", 工业控制计算机, no. 06, 25 June 2012 (2012-06-25), pages 43 - 47 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Remaining useful life prediction using multi-scale deep convolutional neural network | |
| Iqbal et al. | Fault detection and isolation in industrial processes using deep learning approaches | |
| Roffo et al. | Infinite latent feature selection: A probabilistic latent graph-based ranking approach | |
| CN105224872B (en) | A kind of user's anomaly detection method based on neural network clustering | |
| Basak et al. | Mechanisms for integrated feature normalization and remaining useful life estimation using lstms applied to hard-disks | |
| Chen et al. | An efficient network behavior anomaly detection using a hybrid DBN-LSTM network | |
| Ruospo et al. | Evaluating convolutional neural networks reliability depending on their data representation | |
| Cummings et al. | Structured citation trend prediction using graph neural networks | |
| Li et al. | Remaining useful life prediction of turbofan engines using CNN-LSTM-SAM approach | |
| CN112485692A (en) | Battery health state estimation and residual life prediction method based on sparrow search and least square support vector machine | |
| CN114594378A (en) | Lithium battery SOH model training method, prediction method, device and equipment | |
| Huang et al. | Research on transformer fault diagnosis method based on GWO optimized hybrid kernel extreme learning machine | |
| CN114399321A (en) | Business system stability analysis method, device and equipment | |
| Xie et al. | Residual life prediction of lithium-ion batteries based on data preprocessing and a priori knowledge-assisted CNN-LSTM | |
| Gil et al. | Supervised SOM based architecture versus multilayer perceptron and RBF networks | |
| Zhong et al. | CAE‐WANN: A novel anomaly detection method for gas turbines via search space extension | |
| Sun et al. | A novel parameter-optimized recurrent attention network for pipeline leakage detection | |
| GOYAL | Early prediction of software fault-prone module using artificial neural network | |
| CN114626474A (en) | Vehicle power state detection method and system based on probabilistic neural network | |
| CN111461815A (en) | Order recognition model generation method, recognition method, system, device and medium | |
| Vidnerová et al. | Sensor data air pollution prediction by kernel models | |
| Liu et al. | State of health prediction of medical lithium batteries based on multi-scale decomposition and deep learning | |
| Dawood et al. | Power quality disturbance classification based on efficient adaptive Arrhenius artificial bee colony feature selection | |
| Sharma et al. | Hybrid Multi-Scale Convolutional Long Short-Term Memory Network for Remaining Useful Life Prediction and Offset Analysis | |
| Wang et al. | A conditional random field based feature learning framework for battery capacity 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 |