CN-121697600-B - Pressure estimation method of pneumatic braking system based on data and physical fusion model

CN121697600BCN 121697600 BCN121697600 BCN 121697600BCN-121697600-B

Abstract

The invention discloses a pressure estimation method of a pneumatic braking system based on a data and physical fusion model, which comprises the steps of collecting air source pressure, braking air chamber pressure and an ABS electromagnetic valve control instruction through a bench test, constructing a time sequence sliding window sample, carrying out normalization processing, constructing a physical information neural network model comprising a long-short-period memory network and a full-connection layer, outputting a braking air chamber pressure estimation value, calculating to obtain the temperature of the braking air chamber, the mass flow of air flowing through a valve port and the volume of the air chamber, calculating a physical law residual according to a differential equation of a pressure change rate, forming a total loss function together with data fitting loss and physical residual loss in training, adopting an Adam optimizer to iteratively update network parameters until the model converges, and deploying the trained model on a braking system controller, so that the braking air chamber pressure can be estimated in an online high-precision, strong-robustness and physical consistency mode only according to the air source pressure and the valve instruction.

Inventors

PENG JINXIN
XIAO FENG
XU CHANGHE
WAN LIEN
JIN LIQIANG
PENG SILUN
LI JIANHUA
ZHANG XU

Assignees

吉林大学

Dates

Publication Date: 20260508
Application Date: 20260224

Claims (10)

1. The pressure estimation method of the pneumatic braking system based on the data and physical fusion model is characterized by comprising the following steps of: The method comprises the steps of S1, collecting air source pressure, brake chamber pressure and ABS electromagnetic valve control instructions under different brake working conditions based on bench test, and taking the brake chamber pressure as a label, defining time step and window length, and constructing a time sequence sliding window sample; s2, constructing a physical information neural network model comprising a long-term memory network layer and a full-connection layer, wherein an input end of the physical information neural network model receives a time sequence sliding window sample, and an output end outputs an estimated value of the pressure of a brake chamber; S3, substituting the estimated value of the brake chamber pressure into a dynamic estimation model of the brake chamber temperature to calculate the brake chamber temperature, substituting the estimated value of the brake chamber pressure into a gas mass flow calculation model to calculate the gas mass flow; S4, calculating the pressure change rate of the air chamber according to a differential equation of the pressure change rate of the air chamber based on the temperature of the air chamber, the mass flow of the air and the volume of the air chamber, calculating the numerical derivative of the estimated value of the pressure of the air chamber with respect to time, and calculating the difference between the pressure change rate of the air chamber and the numerical derivative to obtain a residual error of a physical rule; S5, constructing a total loss function comprising a physical rule residual error, and carrying out iterative updating on physical information neural network model parameters to minimize the total loss function; And S6, deploying the trained physical information neural network model in a pneumatic braking system controller, and estimating the pressure of a braking air chamber on line by the pneumatic braking system controller according to the air source pressure acquired in real time and an ABS electromagnetic valve control instruction.
2. The method for estimating pressure of a pneumatic brake system based on a data and physical fusion model of claim 1, wherein in step S3, the estimated value of the pressure of the brake chamber is substituted into a dynamic estimation model of the temperature of the brake chamber, and the temperature of the brake chamber is calculated, which specifically comprises: The dynamic estimation model of the brake chamber temperature is based on a temperature and pressure coupling relation established by a thermodynamic variable process equation, and the expression is as follows: ; Wherein, the For the brake chamber temperature, As an estimate of the brake chamber pressure, The pressure of the air is set to be the atmospheric pressure, In order to be at the temperature of the environment, Is a polytropic index.
3. The method for estimating pressure of a pneumatic brake system based on a data and physical fusion model according to claim 1, wherein in step S3, an estimated value of brake chamber pressure is substituted into a gas mass flow calculation model, and a gas mass flow is calculated, specifically: The gas mass flow calculation model is established according to a gas isentropic flow theory and a Bernoulli equation, and the expression is as follows: ; Wherein, the In order to achieve a gas mass flow rate, Is the effective flow area of the ABS electromagnetic valve port, And Respectively the absolute pressures of the upstream and the downstream, As a flow coefficient of the water, the water is mixed with water, As a function of gas flow, by pressure ratio And (3) determining: When (when) When (1): ; When (when) When (1): ; Wherein, the Is the air insulation index.
4. The method for estimating pressure of a pneumatic brake system based on a data and physical fusion model according to claim 1, wherein in step S3, the estimated value of the brake chamber pressure is substituted into a dynamic segment model of the brake chamber volume, specifically: The dynamic segmentation model of the brake chamber volume has the expression: ; Wherein, the For the volume of the brake chamber, For the initial dead zone volume, For the maximum volume of the push rod after full push-out, In order to overcome the preload force of the return spring, Is the effective area of the air chamber membrane, In order to return the spring rate, The pressure is preloaded for the brake chamber diaphragm, The corresponding pressure value is obtained when the push rod reaches the maximum stroke.
5. The method for estimating the pressure of the pneumatic brake system based on the data and physical fusion model of claim 1, wherein step S4 is characterized in that the air chamber pressure change rate is calculated according to an air chamber pressure change rate differential equation based on the brake air chamber temperature, the air mass flow and the brake air chamber volume, specifically: Mass flow of gas Obtaining the air quality of the brake chamber according to the mass conservation and the ideal gas state equation : ; Combining the volume derivative, substituting the volume derivative into a differential equation of the pressure change rate of the air chamber, and calculating the pressure change rate of the air chamber The expression is: ; when the ABS solenoid control command is a boost command, When the control command of the ABS electromagnetic valve is a pressure reducing command, ; Wherein, the As an estimate of the brake chamber pressure, For the actual brake chamber pressure, For the volume of the brake chamber, For the air insulation index (air insulation index), Is a gas constant; in order to be at the temperature of the environment, Is the brake chamber temperature.
6. The method for estimating pressure of a pneumatic brake system based on a data and physical fusion model according to claim 1, wherein in step S5, the total loss function includes a data loss term and a physical loss term, specifically: Data loss term The expression is as follows for the error between the estimated value of the brake chamber pressure output by the physical information neural network and the actual brake chamber pressure label: ; Physical loss term And (4) calculating a physical rule residual error for the step S4, wherein the expression is as follows: ; The total loss function is: ; Wherein, the As the weight coefficient of the light-emitting diode, As an estimate of the brake chamber pressure, For the actual brake chamber pressure, Is the rate of change of the pressure in the air chamber.
7. The method for estimating pressure of a pneumatic brake system based on a data and physical fusion model according to claim 1, wherein in step S5, the iterative updating of the physical information neural network model parameters is performed to minimize a total loss function, specifically: And (3) carrying out iterative updating on the physical information neural network model parameters by adopting an Adam optimizer, inputting the parameters as a gradient of a total loss function on the physical information neural network model parameters, outputting the parameters as updating quantity of the physical information neural network model parameters, updating the weight and the bias of the physical information neural network model, and repeating the iterative updating until the physical information neural network model converges.
8. The method for estimating a pressure of a pneumatic brake system based on a model of data and physics fusion of claim 7, wherein a first step is provided The physical information neural network model parameters during the iteration are as follows The total Loss function is Loss, then the gradient is: ; Wherein, the As the derivative of the total loss function with respect to the current parameter, For model parameters related to neural networks Is a gradient of (2); The Adam optimizer calculates first and second moment estimates of the gradient using exponential moving averages, respectively: ; Wherein, the For the first moment of the estimation, For the second moment estimation, And In order for the attenuation coefficient to be a factor, Representing the element-by-element square of the gradient, Representing the time; Introducing bias correction, and calculating first moment estimation after bias correction And second moment estimation : ; Updating the neural network model parameters according to the corrected moment estimation: ; Wherein, the In order for the rate of learning to be high, Is a numerical stability constant.
9. An electronic device, comprising a memory, a processor and a computer program stored in the memory and operable on the processor, wherein the processor implements a method for estimating a pressure of a pneumatic brake system based on a data and physical fusion model as defined in any one of claims 1 to 8 when the computer program is executed by the processor.
10. A computer-readable storage medium having stored thereon a computer program for causing a computer to execute a method for estimating a pressure of a pneumatic brake system based on a model of data and physical fusion according to any one of claims 1 to 8.

Description

Pressure estimation method of pneumatic braking system based on data and physical fusion model Technical Field The invention relates to the technical fields of vehicle engineering and artificial intelligence application, in particular to a pressure estimation method of a pneumatic braking system based on a data and physical fusion model. Background The pneumatic braking system of the commercial vehicle is generally composed of components such as an air source, a pipeline, a joint, a braking valve/relay valve, an ABS electromagnetic valve, a braking air chamber and the like. In the braking process, the pressure of a braking air chamber is a core state quantity which forms braking force and realizes pressurization, pressure maintaining and depressurization regulation, and the dynamic change of the pressure is directly influenced on the braking response speed, the braking stability and the safety. To achieve fine control and health monitoring, the controller typically needs to acquire or estimate key node pressures and their law of variation. The existing pressure acquisition mode mainly comprises direct measurement, mechanism model estimation and data driving estimation, wherein the direct measurement needs to arrange pressure sensors at a plurality of nodes, but is limited by cost, installation space, reliability and maintenance complexity, and a real vehicle can only arrange the sensors at an air source or a few positions, so that the pressure of an intermediate chamber is not measurable. The mechanism model estimation usually establishes a state equation based on compressible gas flow, throttle orifice flow and valve dynamics, but the model needs more accurate valve port effective area, flow coefficient, response time constant, leakage coefficient and other parameters, and the parameters drift due to manufacturing differences, wear aging, temperature change, pipeline leakage and the like, so that estimation errors are increased. Although the pure data driving method can fit complex nonlinearity, the method has strong dependence on the coverage of training data, and is easy to generate physical inconsistent results in the scenes of working condition extrapolation, sectional flow switching, valve dead zone hysteresis, noise disturbance and the like, and has insufficient robustness and interpretability. In real vehicle application, the common condition is that only the air source pressure and the brake chamber pressure can be measured, and control signals such as ABS air inlet valve and exhaust valve control instructions can be obtained, and key pressure states such as front valve, rear valve and middle pipe section of the valve are difficult to directly measure. Meanwhile, differences exist between the simulation model and a real vehicle system in the aspects of valve characteristics, leakage, volume, pipeline damping and the like, so that the model relying on single simulation or single data is difficult to stably adapt. Therefore, a method for estimating the pressure of the pneumatic braking system with high precision, strong robustness and physical consistency is urgently needed by embedding the physical mechanisms such as an ideal gas state equation, control volume mass conservation, orifice flow, valve opening and closing actions and the like into a learning process in a residual constraint mode under the condition of a small quantity of measurable signals. Disclosure of Invention The invention aims to provide a pressure estimation method of a pneumatic braking system based on a data and physical fusion model, which aims to solve the problems of low estimation precision, poor robustness, physical inconsistency, excessive dependence on a sensor and the like when a data driving or mechanism model is simply relied on in the prior art. In order to achieve the above purpose, the technical scheme provided by the invention is a pneumatic braking system pressure estimation method based on a data and physical fusion model, comprising the following steps: The method comprises the steps of S1, collecting air source pressure, brake chamber pressure and ABS electromagnetic valve control instructions under different brake working conditions based on bench test, and taking the brake chamber pressure as a label, defining time step and window length, and constructing a time sequence sliding window sample; s2, constructing a physical information neural network model comprising a long-term memory network layer and a full-connection layer, wherein an input end of the physical information neural network model receives a time sequence sliding window sample, and an output end outputs an estimated value of the pressure of a brake chamber; S3, substituting the estimated value of the brake chamber pressure into a dynamic estimation model of the brake chamber temperature to calculate the brake chamber temperature, substituting the estimated value of the brake chamber pressure into a gas mass flow calculation model to calculate th