Reverse Calculation And Classification Of Typical Road Surface Spectrum Of Test Site Based On Measured Load Spectrum

The road surface is the input when the car is driving,which directly affects the service life of vehicle parts and the ride comfort.At present,the main methods to study the impact of the road on the car are real-vehicle test method and simulation method.The automobile test field provides various typical road incentives for actual vehicle tests,but it has disadvantages such as a long test period,low efficiency,large investment and manpower.Most of the simulation methods use standard road spectrum or random road load spectrum for simulation analysis.Although the road surface incentives is provided,there are still disadvantages such as the inability to reflect the vibration characteristics of the vehicle under actual operating conditions and the large error in the simulation results.Therefore,based on the actual measured load spectrum,this thesis carries out the reverse and classification of the typical road surface of the test field.On the one hand,it can provide the simulation software with the real road surface spectrum of the test field,so that the simulation value is closer to the actual vehicle test value.On the other hand,it provides in-depth analysis of the vehicle test field’s pavement structure which is also of great significance.First of all,combined with acceleration sensor,cable displacement sensor and GPS,the actual vehicle load spectrum collection plan was formulated,the test system was set up,and the actual vehicle load spectrum collection was completed.The vehicle speed and logic switch were used to extract the load spectrum of each typical road surface.And the original collected data is processed by spectrum analysis,filtering and removing trend items.In order to obtain the velocity and displacement signals,the collected acceleration signals were integrated in the time-frequency and frequency domains respectively,and the integration results were compared with the measured displacement data.The results showed that the accuracy of the frequency domain integration results was higher.Secondly,based on the measured load spectrum,the method of identifying the suspension parameters of the test car was studied,the 1/4 suspension model and the differential equation of motion were established,the MATLAB/Simulink suspension model was built,and the sprung and unsprung masses acceleration signal were obtained by simulation.Used the recursive least square method to identify the suspension parameters of the simulation model.The results showed that the identification error was within 0.1%,which verified the rationality of the parameter identification method.Inputting the measured abnormal noise road load spectrum to identify the stiffness of suspension and damping parameters of the frame.Then,a four-wheeled road surface time-domain model was established using the method of filtering white noise,and verify the accuracy of the road surface model,Using the road surface model as a road surface incentive,a "road-vehicle" model was established considering the interaction between the road surface and the vehicle.,Simulating the model with reference to actual measured conditions to obtain vehicle response signals under different levels of roads.According to the principle of BP neural network,the design of the neural network is completed,combined with the BP neural network,establishing the nonlinear mapping relationship between the vehicle response and the road unevenness,and the training of the BP neural network was completed.The simulation data verifies the accuracy of the road surface method based on the reversed way of neural network on "roadvehicle model.Finally,the measured three road load spectra are used as the input of the neural network,and the typical roads of the test field are output,and the grading characteristic parameters of each road are extracted to classify the roads,and the reasons for the differences of the roads are discussed.The reversed comprehensive road surface was used as the model road surface incentive to simulate,and the actual sprung and unsprung accelerations were selected to verify the simulation.The results showed that the simulated acceleration and the measured acceleration had a small error,verifying the rationality of parameter identification results and the reversed results.