A complete vehicle exhaust emission test system
By using neural network prediction model and correction factor calculation module in the vehicle operating exhaust emission test system, the emission factors are predicted and corrected, and the accuracy of the system decreases when the number of tests and usage time increases, and the exhaust emission test is improved accuracy.
Patent Information
- Application Number
- CN202411016758.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-29
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-07-29
AI Technical Summary
The existing vehicle-condition exhaust emission testing system cannot be effectively corrected when the number of tests and usage time increases, resulting in a decrease in the accuracy of the test results and the accurate measurement of exhaust emissions cannot be guaranteed.
Adopt neural network prediction model and correction factor calculation module to collect and calculate emission factor data, predict emission factors at future moments, and calculate correction factors, and feedback to emission factor calculation module to correct the test system to prevent data deviation and improve the test system. Accuracy.
The neural network prediction model improves the accuracy of data prediction, and the correction factor calculation module promptly corrects the test system to prevent data deviations and improves the accuracy of the exhaust emission test system.
Smart Images

Figure CN118569890B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of vehicle exhaust emission, and more specifically to a whole vehicle operating condition exhaust emission testing system. Background Art
[0002] With the sustained development of the national economy and the continuous improvement of living standards, the number of cars has shown a rapid growth trend. This phenomenon has aggravated the increase of automobile exhaust emissions, increased the load on the urban road network, and caused more traffic congestion problems, which further increased automobile exhaust emissions. With the increasing attention paid to environmental protection, exhaust emission and fuel consumption standards have been continuously tightened, and the demand for exhaust emission measurement in the product development process has increased rapidly.
[0003] Different road conditions will lead to differences in many parameters such as exhaust temperature, driving behavior, altitude, etc., and the differences in these parameters will directly affect emissions. Therefore, road conditions are the most complex factor affecting the test results of the portable emission test system (PEMS) for the whole vehicle. Most of the existing whole vehicle operating condition exhaust emission test systems adopt the operating condition method, using a vehicle chassis test bench to perform emission detection under simulated car driving conditions. There are currently three methods: steady-state conditions, transient conditions and simple transient conditions. The transient condition method simulates the actual operating conditions on the vehicle chassis test bench, and detects emissions under engine start-up, stop, rapid acceleration and other conditions through pre-selected set conditions. This method has the advantages of low false positives and misjudgment rates.
[0004] However, during the continuous testing process, as the number of tests and the system usage time increase, the existing exhaust emission testing system will have deviations during the testing process. The existing exhaust emission testing system is unable to correct the testing system and cannot guarantee the accuracy of the exhaust emission testing system. Summary of the invention
[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a vehicle exhaust emission test system to solve the problems existing in the above-mentioned background technology.
[0006] The present invention provides the following technical solutions: a vehicle exhaust emission test system under working conditions, comprising a collection module, a reading module, an emission factor calculation module, an emission prediction module, a correction factor calculation module and a result output module;
[0007] The acquisition module is used to collect target data through an acquisition device and transmit it to the emission factor calculation module;
[0008] The reading module is used to read the operating condition data in the computer control module and transmit it to the emission factor calculation module;
[0009] The emission factor calculation module is used to calculate the emission factor through the emission factor calculation formula and transmit it to the emission prediction module and the result output module;
[0010] The emission prediction module is used to input the emission factor data into the constructed neural network prediction model, predict the emission factor at the next t moments and transmit it to the correction factor calculation module;
[0011] The correction factor calculation module calculates the correction factor based on the emission factor tested in real time and the predicted emission factor and transmits it to the result output module and the emission factor calculation module;
[0012] The result output module is used to output data and then perform human-computer interaction.
[0013] Preferably, the emission factor calculation formula is: , where pf i is the fuel emission factor of the i-th pollutant gas, C i is the mass concentration of the i-th pollutant gas, A i is the real-time mass flow rate of the i-th pollutant gas, T i is the test cycle time of the i-th pollutant gas, is the correction factor of the i-th pollutant gas, and B is the real-time fuel consumption.
[0014] Preferably, the specific method of constructing the neural network model is:
[0015] Initialize the basic architecture of the neural network model, which includes an input layer, two LSTM units, a Dropout layer, a fully connected layer, a learnable activation function, an overall loss function, an optimizer, a regularizer, and an output layer. The output layer includes n nodes.
[0016] Each LSTM unit includes a memory unit and three control gates, which are input gate, forget gate and output gate. A fully connected layer is added between the LSTM units.
[0017] The Dropout layer randomly discards a certain proportion of nodes from the neural network model temporarily, which can effectively prevent overfitting.
[0018] Preferably, the learnable activation function formula is: , where j is the input data, g 1 is a learnable slope parameter that controls the slope of the linear transformation of j, g 2 is a learnable offset parameter that controls the offset of the Siogmoid function;
[0019] Uncertainty modeling is introduced into the loss function to obtain the overall loss function LOSS: , where u is the quantile, indicating the quantile level in the loss calculation; u∈(0,1), ε true is the real target value, indicating the actual comprehensive performance index, is the lower quantile of the forecast distribution, is the upper quantile of the prediction distribution, || + is a positive function.
[0020] Preferably, the optimizer adopts an adaptive optimizer that adaptively adjusts the learning rate, and the optimization formula is: ,in, The LSTM neural network model is trained with the number of iterations being λ+1. The parameters of the layer, The LSTM neural network model is trained with the number of iterations being λ. The parameters of the layer, It is the LSTM neural network model The learning rate of the layer, v λ is the exponential moving average of the squared gradients for training iterations λ, θ is a small constant to prevent the denominator from being zero, is the parameter gradient for the training iteration number λ.
[0021] Preferably, the ,but ,in, is the value of the overall loss function when the number of training iterations is λ, is the value of the overall loss function when the number of training iterations is λ-1, and X is the reduction parameter.
[0022] Preferably, the regularizer uses L2 norm regularization to control the complexity of the neural network model and prevent overfitting;
[0023] The n output layers are used to output the emission factors of n pollutants at t moments in the future.
[0024] Preferably, the calculation formula of the correction factor is: , where Yp i The predicted emission factor for the i-th pollutant gas; pf i With Yp i All data are from the same moment.
[0025] Technical effects and advantages of the present invention:
[0026] The present invention is advantageous for predicting emission factors by providing a neural network prediction model and a correction factor calculation module. By constructing a neural network prediction model, the accuracy of data prediction is improved. At the same time, the correction factor is calculated by the correction factor calculation module and fed back to the emission factor calculation module to correct the test system, thereby preventing the exhaust emission test system from having data deviations during the test process as the number of tests and the system usage time increase. The data can be corrected in time to improve the accuracy of the exhaust emission test system. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] Figure 1 This is a flow chart of the vehicle exhaust emission test system of the present invention. DETAILED DESCRIPTION
[0028] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the present invention. In addition, the forms of the various structures recorded in the following embodiments are only examples. The vehicle exhaust emission test system involved in the present invention is not limited to the various structures recorded in the following embodiments. All other implementations obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0029] The present invention provides a vehicle exhaust emission test system, comprising a collection module, a reading module, an emission factor calculation module, an emission prediction module, a correction factor calculation module and a result output module;
[0030] The acquisition module is used to collect target data through an acquisition device and transmit it to the emission factor calculation module; the acquisition device includes but is not limited to a gas mass flow meter and a gas concentration detector; the target data includes but is not limited to the real-time mass flow and real-time mass concentration of each pollutant gas in the exhaust gas emitted by the whole vehicle;
[0031] The reading module is used to read the operating condition data in the computer control module and transmit it to the emission factor calculation module; the operating condition data includes but is not limited to the fuel consumption data of the whole vehicle;
[0032] The emission factor calculation module is used to calculate the emission factor through the emission factor calculation formula and transmit it to the emission prediction module and the result output module;
[0033] The emission prediction module is used to input the emission factor data into the constructed neural network prediction model, predict the emission factor at the next t moments and transmit it to the correction factor calculation module;
[0034] The correction factor calculation module calculates the correction factor based on the emission factor tested in real time and the predicted emission factor and transmits it to the result output module and the emission factor calculation module;
[0035] The result output module is used to output data and then perform human-computer interaction.
[0036] In this embodiment, it should be specifically stated that the emission factor calculation formula is: , where pf i is the fuel emission factor of the i-th pollutant gas, C i is the mass concentration of the i-th pollutant gas, A i is the real-time mass flow rate of the i-th pollutant gas, T i is the test cycle time of the i-th pollutant gas, is the correction factor of the i-th pollutant gas, and B is the real-time fuel consumption; the initial setting value of the correction factor is 1, until the correction factor module feeds back the correction factor value, the correction factor value calculated by the correction factor module is used for substitution in the calculation;
[0037] For example, if the start time of the test vehicle for testing the emission of the i-th pollutant gas is T o The test vehicle stops at T z , then the test cycle time of the i-th pollutant gas is In this embodiment, it should be specifically explained that the specific method of constructing the neural network model is:
[0038] Initialize the basic architecture of the neural network model, which includes an input layer, two LSTM units, a Dropout layer, a fully connected layer, a learnable activation function, an overall loss function, an optimizer, a regularizer, and an output layer. If the total number of pollutant gas types to be tested is n, the output layer includes n nodes;
[0039] The preferred neural network model in this embodiment is the LSTM neural network model;
[0040] Each LSTM unit includes a memory unit and three control gates, which are input gate, forget gate and output gate. A fully connected layer is added between the LSTM units.
[0041] The Dropout layer randomly discards a certain proportion of nodes from the neural network model temporarily, which can effectively prevent overfitting;
[0042] The learnable activation function formula is: , where j is the input data, g 1 is a learnable slope parameter that controls the slope of the linear transformation of j, g 1 The value of will affect the gradient of the Sigmoid function in the input space, thereby affecting the response speed of the Sigmoid function to the input. 1The learning of g can adapt the model to different ranges of activation values in the data. 2 is a learnable offset parameter that controls the offset of the Siogmoid function, allowing the model to adapt to different baselines or offsets in the input data. 2 Learning can enable the model to better model the overall pattern of the data;
[0043] g 1 and g 2 The specific value of is obtained through model training; in deep learning, training is performed through the back-propagation algorithm and optimizer, which adjusts the parameters in the network so that the model achieves better performance on the training data;
[0044] Uncertainty modeling is introduced into the loss function to obtain the overall loss function LOSS: , where u is the quantile, indicating the quantile level in the loss calculation; u∈(0,1), determines whether the loss is calculated for the lower or upper quantile, ε true is the real target value, indicating the actual comprehensive performance index, is the lower quantile of the forecast distribution, is the upper quantile of the prediction distribution, || + is the positive function, which ensures that only when the difference between the predicted value and the true value is positive will it contribute to the loss;
[0045] The optimizer adopts an adaptive optimizer that adaptively adjusts the learning rate, and the optimization formula is: ,in, The LSTM neural network model is trained with the number of iterations being λ+1. The parameters of the layer, The LSTM neural network model is trained with the number of iterations being λ. The parameters of the layer, It is the LSTM neural network model The learning rate of the layer, choose different learning rates according to different layers, v λ is the exponential moving average of the squared gradients for training iterations λ, θ is a small constant to prevent the denominator from being zero, is the parameter gradient for the training iteration number λ;
[0046] like ,but ,in, is the value of the overall loss function when the number of training iterations is λ, is the value of the overall loss function when the number of training iterations is λ-1, and X is the reduction parameter;
[0047] The regularizer uses L2 norm regularization to control the complexity of the neural network model and prevent overfitting;
[0048] The n output layers are used to output the emission factors of n pollutants at t moments in the future.
[0049] In this embodiment, it should be specifically explained that the calculation formula of the correction factor is: , where Yp i The predicted emission factor for the i-th pollutant gas; pf i With Yp i All data are from the same time. For example, if pf i is the emission factor for the fifth hour, then Yp i It corresponds to the predicted emission factor for the 5th hour; the accuracy of the test system results is ensured by correction through the correction factor.
[0050] In this embodiment, it should be specifically explained that the acquisition module collects data and transmits it to the emission factor calculation module, the reading module reads the data and transmits it to the emission factor calculation module, the emission factor calculation module receives the data from the acquisition module and the reading module, calculates the emission factor and transmits it to the emission prediction module and the result output module, the emission prediction module receives the data from the emission factor calculation module, makes a prediction, and transmits the prediction result to the correction factor calculation module, the correction factor calculation module is used to receive the prediction result data of the emission prediction module, calculate the correction factor and then feed it back to the emission factor calculation module, and transmit it to the result output module at the same time, the result output module is used to receive the data from the correction factor calculation module and the emission factor calculation module, and display the data interactively.
[0051] In the present embodiment, it should be specifically explained that the difference between the present embodiment and the prior art lies mainly in that the present embodiment is provided with a neural network prediction model and a correction factor calculation module, which is beneficial to predicting emission factors. By constructing a neural network prediction model, the accuracy of data prediction is improved. At the same time, the correction factor is calculated by the correction factor calculation module and fed back to the emission factor calculation module to correct the test system, thereby preventing the exhaust emission test system from having data deviations during the test process as the number of tests and the length of time the system is used increase. The data can be corrected in time to improve the accuracy of the exhaust emission test system.
[0052] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
[0053] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
Claims
1. A vehicle exhaust emission test system, characterized in that: It includes a collection module, a reading module, an emission factor calculation module, an emission prediction module, a correction factor calculation module and a result output module; The acquisition module is used to collect target data through an acquisition device and transmit it to the emission factor calculation module; The reading module is used to read the operating condition data in the computer control module and transmit it to the emission factor calculation module; The emission factor calculation module is used to calculate the emission factor through the emission factor calculation formula and transmit it to the emission prediction module and the result output module; The emission prediction module is used to input the emission factor data into the constructed neural network prediction model, predict the emission factor at the next t moments and transmit it to the correction factor calculation module; The correction factor calculation module calculates the correction factor based on the emission factor tested in real time and the predicted emission factor and transmits it to the result output module and the emission factor calculation module; The result output module is used to output data and then perform human-computer interaction; The emission factor calculation formula is: , where pf i is the fuel emission factor of the i-th pollutant gas, C i is the mass concentration of the i-th pollutant gas, A i is the real-time mass flow rate of the i-th pollutant gas, T i is the test cycle time of the i-th pollutant gas, is the correction factor of the i-th pollutant gas, and B is the real-time fuel consumption; The specific method of constructing the neural network model is as follows: Initialize the basic architecture of the neural network model, which includes an input layer, two LSTM units, a Dropout layer, a fully connected layer, a learnable activation function, an overall loss function, an optimizer, a regularizer, and an output layer. The output layer includes n nodes. Each LSTM unit includes a memory unit and three control gates, which are input gate, forget gate and output gate. A fully connected layer is added between the LSTM units. The Dropout layer randomly discards a certain proportion of nodes from the neural network model temporarily, which can effectively prevent overfitting; The learnable activation function formula is: , where j is the input data, g1 is a learnable slope parameter that controls the slope of the linear transformation of j, and g2 is a learnable offset parameter that controls the offset of the Siogmoid function; Uncertainty modeling is introduced into the loss function to obtain the overall loss function LOSS: , where u is the quantile, indicating the quantile level in the loss calculation; u∈(0,1), ε true is the real target value, indicating the actual comprehensive performance index, is the lower quantile of the forecast distribution, is the upper quantile of the prediction distribution, || + is a positive function; The optimizer adopts an adaptive optimizer that adaptively adjusts the learning rate, and the optimization formula is: ,in, The LSTM neural network model is trained with the number of iterations being λ+1. The parameters of the layer, The LSTM neural network model is trained with the number of iterations being λ. The parameters of the layer, It is the LSTM neural network model The learning rate of the layer, v λ is the exponential moving average of the squared gradients for training iterations λ, θ is a small constant to prevent the denominator from being zero, is the parameter gradient for the training iteration number λ; Said ,but ,in, is the value of the overall loss function when the number of training iterations is λ, is the value of the overall loss function when the number of training iterations is λ-1, and X is the reduction parameter; The regularizer uses L2 norm regularization to control the complexity of the neural network model and prevent overfitting; The n output layers are used to output the emission factors of n pollutants at t moments in the future; The calculation formula of the correction factor is: , where Yp i The predicted emission factor for the i-th pollutant gas; pf i With Yp i All data are from the same moment.
