CN110162910A - A kind of hill start optimization method based on technology of Internet of things - Google Patents

Abstract

The invention discloses a slope optimization method based on Internet of Things technology, comprising: step 1, collecting environmental parameters of the position where the car is located, and calculating environmental impact factors; step 2, collecting the main cylinder pressure P and engine output torque of the car R _T and brake pedal opening α; step 3, input the collected parameters into the BP neural network model after normalization, as the input layer, after the BP neural network model training, the car ramp start process The throttle control valve is controlled and adjusted; Step 4, the control and adjustment signal is input into the fuzzy controller, and the output vector group representing the control and adjustment category is obtained, which is output as the adjustment answer. By monitoring the environmental parameters and driving parameters of the car during the ramp start process, the throttle control valve is controlled and adjusted during the ramp start process of the car to prevent the car from slipping and improve the success rate of the ramp start.

Description

A slope optimization method based on Internet of Things technology

technical field

The invention relates to a slope optimization method based on Internet of Things technology, which belongs to the field of automobiles.

Background technique

The term Internet of Things was first proposed by an American professor in 1999, and has developed from the initial radio frequency technology and sensor equipment to today's embedded systems, cloud computing and other more technologically intelligent technologies. Today's Internet of Things mainly relies on the interconnection of the Internet, communication networks, and wireless sensor networks to achieve intelligent control of various production fields. The Internet of Things is based on a certain Internet protocol. Objects, projects and other objects are set up with input and output hardware, and then use the signals of the software system and hardware to realize information interaction to achieve the purpose of intelligent control.

Car hill start means that the car starts on a slope with a certain angle. Due to the performance of the car itself, environmental factors and other factors, the car often rolls when it starts on a hill. Therefore, the hill start assist system came into being. , through interventional braking, hydraulic brake assist and transmission regulation to prevent the car from sliding down and avoid engine stalling. However, most of the existing auxiliary systems are proposed based on the ABS system, and cannot be used separately from the ABS system, and the existing auxiliary system optimization methods for hill starts cannot be adjusted according to environmental factors, and the success rate is low and the accuracy is poor. .

Contents of the invention

The present invention designs and develops a slope optimization method based on Internet of Things technology. By monitoring the environmental parameters and driving parameters of the vehicle during the slope start process, the accelerator control valve in the slope start process of the vehicle is controlled and adjusted to prevent The phenomenon of car slipping can be improved to improve the success rate of hill start.

Another object of the present invention is to adjust the throttle control valve of the car during the hill start process through BP neural network and fuzzy control, improve the output torque of the engine, and ensure that the car can start normally on the ramp without slipping.

Another object of the present invention is to improve the accuracy of throttle control and the success rate of hill starts by calculating the environmental impact factor as the input layer parameter and by calculating the outlet flow of the throttle control valve.

The technical scheme provided by the invention is:

A slope optimization method based on Internet of Things technology, including:

Step 1, collect the environmental parameters of the location where the car is located, and calculate the environmental impact factor;

Step 2, collecting the master cylinder pressure P, engine output torque R _T and brake pedal opening α of the vehicle;

Step 3, input the collected parameters into the BP neural network model after normalization, as the input layer, through the BP neural network model training, control and adjust the throttle control valve in the process of starting the car on a ramp;

Step 4: Input the control regulation signal into the fuzzy controller, obtain the output vector group representing the control regulation category, and output it as the regulation answer.

Preferably, the environmental parameters include ambient temperature T, ambient humidity RH, ramp gradient θ, wind force level W, and visibility S.

Preferably, said step three specifically includes:

Step 1. Obtain the ambient temperature T, ambient humidity RH, ramp gradient θ, wind force level W, and visibility S of the location of the car according to the sampling cycle, and calculate the environmental impact factor μ by cycle;

Step 2, collect the master cylinder pressure P of the automobile, the engine output torque R _T and the brake pedal opening α, and normalize together with the environmental influence factor μ, and determine that the input layer vector of the three-layer BP neural network is x＝{x ₁ ,x ₂ ,x ₃ ,x ₄ }; Among them, x ₁ is the master cylinder pressure coefficient, x ₂ is the engine output torque coefficient, x ₃ is the brake pedal opening coefficient, x ₄ is the environmental impact factor coefficient ;

Step 3, the input layer vector is mapped to the middle layer, the middle layer vector y={y ₁ , y ₂ ,...,y _m }; m is the number of middle layer nodes;

Step 4. Obtain the output layer vector o={o ₁ }; o ₁ is the opening coefficient of the throttle control valve.

Preferably, said step five specifically includes:

Comparing the throttle control valve opening coefficient with the preset throttle opening coefficient to obtain the throttle adjustment deviation signal, calculating the throttle deviation signal to obtain the throttle deviation change rate signal, amplifying the throttle deviation change rate signal and inputting it to the output of the fuzzy controller Adjustment level.

Preferably, the regulation level is I={I ₀ , I ₁ , I ₂ , I ₃ }, I ₀ means normal operation without adjustment, I ₁ means primary adjustment is required, I ₂ means secondary adjustment, I ₃ is an abnormal operation, and an alarm is given.

Preferably, the empirical formula of the environmental impact factor satisfies:

Among them, θ _max is the maximum value of slope slope in the measurement period, θ _min is the minimum value of slope slope, is the standard value of the ramp slope, is the standard wind rating, is the ambient temperature standard value, Ambient humidity standard value, is the standard value of visibility, λ ₁ is the first environmental correlation coefficient, and λ ₂ is the second environmental correlation coefficient.

Preferably, when I=I ₀ , λ ₁ =0.25-0.42, λ ₂ =0.38-0.65;

When I=I ₁ or I=I ₂ , λ ₁ =0.43-0.55, and λ ₂ =0.65-0.78.

Preferably, the empirical formula of the flow rate at the valve of the control valve is:

Among them, δ is the correction coefficient, ω ₀ is the basic flow rate of the valve, A ₁ is the cross-sectional area of the valve outlet, A ₂ is the cross-sectional area of the oil outlet pipe, k ₁ is the flow resistance coefficient of the pipeline, k ₂ is the shrinkage coefficient, C is the volume of the fuel tank, L is the volume of the oil outlet pipeline, P is the pressure of the main cylinder, Standard value of master cylinder pressure, α is the gradient of the slope, It is the standard value of slope slope.

Preferably, the normalization formula is:

Among them, x _j is the parameter in the input layer vector, and X _j are the measurement parameters: P, R _T , α, μ; X _jmax and X _jmin are the maximum and minimum values of the corresponding measurement parameters, respectively.

The beneficial effects of the present invention: by monitoring the environmental parameters and driving parameters of the car during the ramp start process, the accelerator control valve in the ramp start process of the car is controlled and adjusted to prevent the car from slipping and improve the success of the ramp start Rate. Through the BP neural network and fuzzy control, the throttle control valve is adjusted during the start of the car on the slope, and the output torque of the engine is increased to ensure that the car can start normally on the slope without slipping. By calculating the environmental impact factor as the input layer parameter, and by calculating the outlet flow of the throttle control valve, the accuracy of throttle control and the success rate of hill start are improved.

Detailed ways

The present invention will be described in further detail below, so that those skilled in the art can implement it with reference to the description.

The present invention provides a slope optimization method based on the Internet of Things, through BP neural network and fuzzy control, the throttle control valve of the car in the process of starting on the slope is adjusted, the output torque of the engine is increased, and the car can be operated normally on the slope. start without slipping the car; among them, the environmental information is monitored through the sensor, the brake pedal opening is monitored through the brake pedal opening sensor, and the control valve valve opening is controlled. Connection, by controlling the oil intake of the engine, and then controlling the output torque of the engine, the car can overcome the resistance and complete the ramp start without slipping.

The slope optimization method based on the Internet of Things specifically includes the following steps:

Step 1. Collect the environmental parameters of the location of the car and calculate the environmental impact factor

Among them, the environmental parameters include ambient temperature T, ambient humidity RH, ramp gradient θ, wind force level W, and visibility S;

Step 2, collecting the master cylinder pressure P, engine output torque R _T and brake pedal opening α of the vehicle;

Step 3, input the collected parameters into the BP neural network model after normalization, as the input layer, through the BP neural network model training, control and adjust the throttle control valve in the process of starting the car on a ramp;

Step 5: Input the control regulation signal into the fuzzy controller, obtain the output vector group representing the control regulation category, and output it as the regulation answer.

In another embodiment, the empirical formula of the environmental impact factor satisfies:

Wherein, the θ _max is the maximum value of the slope slope in the measurement period, and θ _min is the minimum value of the slope slope, is the standard value of the ramp slope, is the standard wind rating, is the ambient temperature standard value, Ambient humidity standard value, is the standard value of visibility, λ ₁ is the first environmental correlation coefficient, and λ ₂ is the second environmental correlation coefficient.

In another embodiment, the empirical formula for the flow rate at the control valve valve is:

Among them, δ is the correction coefficient, ω ₀ is the basic flow rate of the valve, A ₁ is the cross-sectional area of the valve outlet, A ₂ is the cross-sectional area of the oil outlet pipe, k ₁ is the flow resistance coefficient of the pipeline, k ₂ is the shrinkage coefficient, C is the volume of the fuel tank, L is the volume of the oil outlet pipeline, P is the pressure of the main cylinder, Standard value of master cylinder pressure, α is the gradient of the slope, It is the standard value of slope slope.

In step three, the network model is trained through the BP neural network, and the throttle control valve in the process of starting the car on a hill is controlled and adjusted, which specifically includes the following steps:

Step 1, establish a BP neural network model;

In the BP model, the neurons of each level are fully interconnected, and there is no connection between the neurons of each level, and the output of the neurons of the input layer is the same as the input, that is, o _i = _xi . The operating characteristics of neurons in the middle hidden layer and output layer are

o _pj =f _j (net _pj )

Among them, p represents the current input sample, ω _ji is the connection weight from neuron i to neuron j, o _pi is the current input of neuron j, o _pj is its output; f _j is nonlinear differentiable non-decreasing The function is taken as a S-type function, that is, f _j (x)=1/(1+e ^−x ).

The BP network architecture adopted by the present invention is composed of three layers. The first layer is the input layer, with n nodes in total, corresponding to n detection signals representing the working status of the equipment, and these signal parameters are given by the data preprocessing module. The second layer is the hidden layer, with a total of m nodes, which is determined in an adaptive manner by the training process of the network. The third layer is the output layer, with a total of p nodes, determined by the actual output response of the system.

The mathematical model of the network is:

Input layer vector: x=(x ₁ ,x ₂ ,…,x _n ) ^T

Middle layer vector: y=(y ₁ ,y ₂ ,…,y _m ) ^T

Output layer vector: z=(z ₁ ,z ₂ ,…,z _p ) ^T

In the present invention, the number of input layer nodes is n=4, and the number of output layer nodes is p=1. The number of hidden layer nodes m is estimated by the following formula:

The four parameters of the input signal are respectively expressed as: x ₁ is the pressure coefficient of the master cylinder, x ₂ is the engine output torque coefficient, x ₃ is the brake pedal opening coefficient, x ₄ is the environmental impact factor coefficient;

Since the data acquired by sensors belong to different physical quantities, their dimensions are different. Therefore, before the data is fed into the artificial neural network, the data needs to be normalized to a value between 0-1.

The normalization formula is:

Among them, x _j is the parameter in the input layer vector, and X _j are the measurement parameters: P, R _T , α, μ; X _jmax and X _jmin are the maximum and minimum values of the corresponding measurement parameters, respectively.

Specifically, for the master cylinder pressure P, after normalization, the master cylinder pressure coefficient x ₁ is obtained:

Among them, P _min and P _max are the minimum and maximum values of the master cylinder pressure respectively;

Similarly, for the engine output torque R _T , after normalization, the engine output torque coefficient x ₂ is obtained:

Among them, R _Tmin and R _Tmax are the minimum and maximum values of the engine output torque respectively;

Similarly, for the ramp gradient α, after normalization, the ramp gradient coefficient x ₃ is obtained:

Among them, α _min and α _max are the minimum and maximum values of the ramp slope;

Similarly, for the environmental impact factor, after normalization, the environmental impact factor coefficient x ₄ is obtained:

Among them, μ _min and μ _min are the minimum and maximum values of environmental impact factors, respectively;

Output layer vector o={o ₁ }; o ₁ is the opening coefficient of the throttle control valve.

Step 2, carry out BP neural network training

Obtain training samples according to historical experience data, and given the connection weight W _ij between the input node i and the hidden layer node j, the connection weight W _jk between the hidden layer node j and the output layer node k, the hidden layer The threshold θ _j of node j, and the threshold θ _k , W _ij , W _jk , θ _j , θ _k of the output layer node k are all random numbers between -1 and 1.

During the training process, the values of W _ij and W _jk are constantly revised until the system error is less than or equal to the expected error, and the training process of the neural network is completed.

(1) Training method

Each sub-network adopts a separate training method; when training, a set of training samples must first be provided, each of which is composed of an input sample and an ideal output pair, and when all the actual outputs of the network are consistent with their ideal outputs, it indicates that the training is over; Otherwise, by correcting the weights, the ideal output of the network is consistent with the actual output;

(2) Training algorithm

The BP network is trained using the Backward Propagation algorithm, and its steps can be summarized as follows:

Step 1: Select a network with a reasonable structure, and set the initial values of all node thresholds and connection weights.

Step 2: Calculate the following for each input sample:

(a) Forward calculation: for unit j of layer l

In the formula, It is the weighted sum of unit j information of layer l in the nth calculation, is the connection weight between unit j of layer l and unit i of the previous layer (that is, layer l-1), It is the working signal sent by the unit i of the previous layer (that is, the l-1 layer, the number of nodes is n _l-1 ); when i=0, let is the threshold of unit j in layer l.

If the activation function of unit j is a sigmoid function, then

and

If neuron j belongs to the first hidden layer (l=1), then

If neuron j belongs to the output layer (l=L), then we have

And e _j (n) = x _j (n) - o _j (n);

(b) Reverse calculation error:

for the output unit

hidden unit

(c) Correction weight:

η is the learning rate.

Step 3: Input new samples or samples of a new cycle until the network converges, and the input order of samples in each cycle should be randomly sorted during training.

The BP algorithm uses the gradient descent method to find the extreme value of the nonlinear function, and there are problems such as falling into local minimum and slow convergence speed. A more effective algorithm is the Levenberg-Marquardt optimization algorithm, which makes the learning time of the network shorter and can effectively suppress the network from being trapped in a local minimum. Its weight adjustment rate is selected as

Δω＝(J ^T J+μI) ^-1 J ^T e;

Among them, J is the Jacobian matrix of the differential of the error to the weight value, I is the input vector, e is the error vector, and the variable μ is an adaptively adjusted scalar, which is used to determine whether the learning is completed according to the Newton method or the gradient method .

When the system is designed, the system model is a network that has only been initialized, and the weights need to be learned and adjusted according to the data samples obtained during use. For this purpose, the self-learning function of the system is designed. When the learning samples and quantity are specified, the system can conduct self-learning to continuously improve the network performance;

As shown in Table 1, a set of training samples and the values of each node during the training process are given

Step 3. Input the obtained output layer vector into the fuzzy controller to obtain a vector group representing the optimal adjustment category, as follows:

Compare the throttle control valve opening coefficient o ₁ with the preset throttle opening coefficient The throttle control valve opening adjustment deviation signal e ₁ is obtained by comparison, the throttle control valve opening deviation signal is calculated to obtain the throttle control valve opening deviation change rate signal ec ₁ , the throttle control valve opening deviation change rate signal is amplified and input To the fuzzy controller output adjustment level I = {I ₀ , I ₁ , I ₂ , I ₃ }, I ₀ means normal operation without adjustment, I ₁ means first-level adjustment is required, I ₂ means second-level adjustment, I ₃ For abnormal operation, an alarm is issued.

Among them, the actual variation range of e ₁ is [-1,1], the discrete universe is {-5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5}, I The discrete domain of discourse is {0,1,2,3}, then the quantization factor k ₁ =5/1;

Define the fuzzy self and the membership function, and divide the throttle control valve opening change rate signal into 7 fuzzy states: PB (positive big), PM (positive middle), PS (positive small), ZR (zero), NS (negative small) , NM (negative medium), NB (negative large), combined with experience, the membership function table of the air conditioning regulation change rate signal e ₁ is obtained, as shown in Table 2:

Table ₂ Membership function table of throttle control valve opening change rate signal e1

The fuzzy reasoning process must perform complex matrix operations, and the amount of calculation is very large. It is difficult to implement the reasoning online to meet the real-time requirements of the control system. Experience can summarize the preliminary control rules of the fuzzy controller. The fuzzy controller defuzzifies the output signal according to the fuzzy value obtained, and obtains the valve opening adjustment level I, and obtains the fuzzy control lookup table. Since the domain of discussion is discrete, The fuzzy control rules can be expressed as a fuzzy matrix. Using single-point fuzzification, the I control rules are shown in Table 3.

table 3

Through the BP neural network and fuzzy control, the throttle control valve is adjusted during the start of the car on the slope, and the output torque of the engine is increased to ensure that the car can start normally on the slope without slipping.

Although the embodiment of the present invention has been disclosed as above, it is not limited to the use listed in the specification and implementation, it can be applied to various fields suitable for the present invention, and it can be easily understood by those skilled in the art Therefore, the invention is not limited to the specific details and embodiments shown and described herein without departing from the general concept defined by the claims and their equivalents.

Claims (9)

1. a kind of hill start optimization method based on technology of Internet of things characterized by comprising

Step 1: the environmental parameter of the position where acquisition automobile, and calculate Environmental Factors；

Step 2: master cylinder pressure P, the engine output torque R of acquisition automobile_TAnd brake pedal aperture α；

Step 3: the parameter of acquisition is input in BP neural network model after normalization, as input layer, by BP mind Through network model training, the modulator pressure regulator during car ramp starting is controlled to adjust；

Step 4: being input to signal is controlled to adjust in fuzzy controller, the output vector group for indicating to control to adjust classification is obtained, It is exported as answer is adjusted.

2. the hill start optimization method according to claim 1 based on technology of Internet of things, which is characterized in that the environmental parameter Including environment temperature T, ambient humidity RH, hill gradient θ, wind scale W and visibility S.

3. the hill start optimization method according to claim 2 based on technology of Internet of things, which is characterized in that the step 3 tool Body includes:

Step 1, according to the sampling period, obtain environment temperature T, ambient humidity RH, the hill gradient θ, wind-force of automobile position Grade W and visibility S, and Environmental Factors μ is calculated by the period；

Step 2, master cylinder pressure P, the engine output torque R for acquiring automobile_TAnd brake pedal aperture α, and with the environment shadow It rings and determines that the input layer vector of three layers of BP neural network is x={ x because μ is normalized together₁,x₂,x₃,x₄}；Wherein, x₁ For master cylinder pressure coefficient, x₂Engine output torque coefficient, x₃For brake pedal aperture coefficient, x₄For Environmental Factors coefficient；

Step 3, the input layer DUAL PROBLEMS OF VECTOR MAPPING to middle layer, the middle layer vector y={ y₁,y₂,…,y_m}；M is middle layer Node number；

Step 4 obtains output layer vector o={ o₁}；o₁For Throttle Opening Control valve opening coefficient.

4. the hill start optimization method according to claim 3 based on technology of Internet of things, which is characterized in that the step 4 tool Body includes:

Throttle Opening Control valve opening coefficient and preset accelerator open degree coefficients comparison are obtained into throttle adjustment deviation signal, throttle is inclined Difference signal is input to after amplification fuzzy by throttle deviation variation rate signal is calculated, by throttle deviation variation rate signal Controller output adjusts grade.

5. the hill start optimization method according to claim 4 based on technology of Internet of things, which is characterized in that the adjusting grade For I={ I₀,I₁,I₂,I₃, I₀To operate normally, without adjusting, I₁To need to carry out level-one adjusting, I₂For second level adjusting, I₃For It is operating abnormally, alarms.

6. the hill start optimization method according to claim 5 based on technology of Internet of things, which is characterized in that the environment influences The empirical equation of the factor meets:

Wherein, θ_maxTo measure period internal ramp gradient maximum value, θ_minFor hill gradient minimum value,For the standard of hill gradient Value,For standard air grade,For ambient temperature level value,Ambient humidity standard value,For visibility standards value, λ₁ For first environment related coefficient, λ₂For second environment related coefficient.

7. the hill start optimization method according to claim 6 based on technology of Internet of things, which is characterized in that

Work as I=I₀When, λ₁=0.25~0.42, λ₂=0.38~0.65；

Work as I=I₁Or I=I₂When, λ₁=0.43~0.55, λ₂=0.65~0.78.

8. the hill start optimization method according to claim 7 based on technology of Internet of things, which is characterized in that the control valve valve The empirical equation of flow velocity at door are as follows:

Wherein, δ is correction coefficient, ω₀For valve basis flow velocity, A₁For area of section at valve export, A₂For fuel-displaced tube section face Product, k₁For pipeline flow resistance coefficient, k₂For constriction coefficient, C is volume of fuel tank, and L is oil outlet pipe volume, and P is master cylinder pressure,For Master cylinder pressure standard value, α are hill gradient,For hill gradient standard value.

9. the hill start optimization method according to claim 8 based on technology of Internet of things, which is characterized in that the normalization is public Formula are as follows:

Wherein, x_jFor the parameter in input layer vector, X_jRespectively measurement parameter: P, R_T,α,μ；X_jmaxAnd X_jminIt is respectively corresponding Maximum value and minimum value in measurement parameter.