Vehicle axial flow cooling fan flow and efficiency prediction method based on GA-BP neural network
Through the method based on GA-BP neural network, the weight and bias of the BP neural network are optimized, and the flow and efficiency prediction model of automotive axial flow cooling fans is constructed, which solves the problems of complex measurement methods and high resource consumption in the existing technology. It achieves efficient and accurate flow and efficiency prediction, and is suitable for industrial applications.
Patent Information
- Application Number
- CN202411133219.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-19
- Publication Date
- 2024-11-26
AI Technical Summary
Among the existing automotive axial flow cooling fans flow and efficiency measurement methods, experimental measurement costs are high, long cycles are long, and operation is complex. Numerical simulation requires a large amount of computing resources, which is difficult to apply to different models and is difficult to apply on a large scale.
Using a method based on GA-BP neural network, through data acquisition, preprocessing, BP neural network construction and genetic algorithm optimization, a prediction model for vehicle axial flow cooling fans flow and efficiency are constructed, and the weight and bias of BP neural network are optimized by genetic algorithm. to improve prediction accuracy and generalization capabilities.
It greatly reduces the prediction time of the flow and efficiency of the automotive axle flow cooling fan, simplifies operations, saves manpower and material resources, can apply different models, is suitable for industrial applications, and provides more accurate prediction results.
Smart Images

Figure CN119021892A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of vehicle axial flow cooling fans, and in particular to a method for predicting flow rate and efficiency of a vehicle axial flow cooling fan based on a GA-BP neural network. Background Art
[0002] The cooling fan is a machine that inputs air. The automotive axial flow cooling fan is an important branch of the cooling fan. This fan has the characteristics of high efficiency, small size and easy layout. The passenger car powertrain cooling system is mainly composed of cooling fans, radiators, expansion tanks, pipelines and other accessories.
[0003] As the only component in the cooling system that can actively supply air to ensure the heat dissipation performance of the radiator, the performance of the automotive axial flow cooling fan is very important. The performance evaluation indicators of automotive axial flow cooling fans mainly include flow rate, efficiency, etc.
[0004] In existing research, the measurement methods for the flow rate and efficiency of automotive axial cooling fans mainly include experimental measurement methods and numerical simulation methods. Traditional experimental methods have high measurement costs, long cycles, and complex operations, consuming a lot of manpower and material resources, and the same experimental device is difficult to apply to different models, making it difficult to be applied on a large scale in industry. The emergence of numerical simulation methods reduces the computing cost, but requires very complex calculations and consumes huge computing resources.
[0005] Therefore, a new solution to the above problems needs to be proposed. Summary of the invention
[0006] The purpose of the present invention is to provide a method for predicting the flow rate and efficiency of an automotive axial cooling fan based on a GA-BP neural network. The flow rate and efficiency of the axial cooling fan are taken as the research objects. After selecting appropriate neural network feature input and output, the obtained feature data is substituted into the neural network for training. The trained neural network shows good prediction accuracy and generalization on the test set. This method greatly reduces the time for calculating the flow rate and efficiency of the automotive axial cooling fan, so as to solve the technical problems raised in the background technology.
[0007] To achieve the above object, the present invention provides the following technical solution: A method for predicting flow rate and efficiency of an axial flow cooling fan for a vehicle based on a GA-BP neural network, comprising at least the following steps:
[0008] S1: Automotive axial cooling test fan and simulation model. The automotive axial cooling test fan needs to pass the test and verify the rationality of the simulation model. After the rationality of the simulation model is verified, the corresponding fan flow and efficiency under various fan parameters can be obtained by simulating the model;
[0009] S2: data collection and extraction, which includes two schemes for obtaining parameter data, flow rate and efficiency data of a vehicle axial cooling test fan or a simulation model;
[0010] S3: Data preprocessing, normalizing the sample data to avoid the problem of slow network training convergence caused by large sample data differences;
[0011] S4: Set BP neural network parameters and build a BP neural network model;
[0012] S5: Use genetic algorithm to optimize BP neural network parameters, select appropriate population size, crossover operation probability, mutation operation probability and maximum number of evolution iterations, and use genetic algorithm to find the optimal initial threshold and weight of BP neural network, so that BP neural network can overcome the defects of random selection of initial threshold and weight, start training and iteration with the optimal solution, and obtain the final prediction result;
[0013] S6: Optimizing the BP neural network model based on the GA algorithm to obtain an initial GA-BP model;
[0014] S7: Verify the GA-BP model and explore the model usability;
[0015] S8: The optimal GA-BP model is used to predict the flow rate and efficiency data of the automotive axial cooling fan, that is, to obtain the predicted value.
[0016] Further, the data collection and extraction in S2 includes data obtained by performing experiments on the test fan and data obtained by performing experimental simulations on the simulation model;
[0017] The parameter data include rotation speed, blade installation angle, number of blades, radial clearance, blade chord length, curvature radius and hub ratio.
[0018] Furthermore, the S3 at least includes the following steps:
[0019] By normalizing the data, each variable is limited to a certain range, thus eliminating the influence;
[0020] To facilitate weighted comparative analysis, the normalization formula is as follows:
[0021]
[0022] Where: x * is the normalized value, x i is the training data, x min 、x max are the minimum and maximum values in the data respectively.
[0023] Furthermore, the S4 at least includes the following steps:
[0024] The number of neurons m in the input layer of the BP neural network is set to 7, and the number of neurons n in the output layer of the BP neural network is set to 2;
[0025] The number of neurons in the hidden layer of the BP neural network is calculated according to the empirical formula The range of a is a constant from 1 to 10, and the range is determined by repeated comparison through trial and error to determine the optimal number of neurons in the hidden layer of the model;
[0026] Based on the collected sample data of flow rate, efficiency and parameters of automotive axial cooling fans;
[0027] The model is constructed by taking parameters such as rotation speed, blade installation angle, number of blades, radial clearance, blade chord length, curvature radius, hub ratio, etc. as input variables and flow rate and efficiency as output variables;
[0028] Perform network training, which includes two stages:
[0029] The first stage is to propagate forward along the neural network to obtain the output value;
[0030] In the second stage, the error is propagated back along the neural network to correct the weights and threshold biases;
[0031] After two stages, the neural network training is completed. The training is repeated many times until the error is less than the target value. The weights and threshold bias are fixed, and the BP neural network model is constructed.
[0032] Further, the S6 at least includes the following steps:
[0033] S61: Initializing the BP neural network, determining the number of input layer nodes, the number of hidden layer nodes, and the number of output layer nodes of the BP neural network;
[0034] S62: Initial weights and thresholds, preprocessing the input data to obtain initial values, encoding the initial values through the GA algorithm, and establishing a fitness function to optimize the weights and thresholds of the BP neural network;
[0035] S63: Roulette wheel selection method is used to select the individual with the highest fitness in the population with coded population data;
[0036] S64: Perform a crossover operation on the individuals with the highest fitness in the selected group to obtain several new individuals;
[0037] S65: performing mutation operation on a number of the new individuals to obtain a new population;
[0038] S66: updating the new population, calculating the fitness, and replacing the individuals with low fitness with individuals with high fitness to form a new population;
[0039] S67: Repeat S61 to S66 until the fitness is greater than a preset threshold or the number of iterations reaches a preset number, then stop the iteration and obtain the optimal weight and threshold.
[0040] Furthermore, in order to evaluate the model prediction error and accuracy of the GA-BP model in S6 using two evaluation methods, mean absolute error MAE and mean absolute percentage error MAPE, the model usability was explored.
[0041] Further, the S8 at least includes the following steps:
[0042] Based on S7, whether the error meets the preset accuracy or whether the iteration reaches the maximum number;
[0043] If the calculation error is satisfied, the algorithm converges;
[0044] The weight and threshold of the last iteration are the optimal solution. If the calculation error is not satisfied, retraining and prediction are performed.
[0045] Until the calculation error meets the design requirements, the BP neural network at this time is the optimal BP neural network optimized by the genetic algorithm, that is, the optimal GA-BP model;
[0046] Based on the optimal GA-BP model, the flow rate and efficiency data of the automotive axial cooling fan are predicted, that is, the predicted value is obtained.
[0047] Compared with the prior art, the present invention has the following beneficial effects:
[0048] 1. In the GA-BP neural network of the present invention, the genetic algorithm optimizes the weights and biases of the BP neural network, and finds the optimal combination of weights and biases by simulating the biological evolution process, thereby improving the prediction performance and generalization ability of the BP neural network.
[0049] 2. The model of the present invention has a strong convergence speed and simple parameter adjustment. Compared with the existing methods for measuring fan flow and efficiency, this method is simpler to operate.
[0050] 3. The present invention integrates the flow rate, efficiency and parameter sample data of automotive axial-flow cooling fans for the first time. After data screening and integration, a flow rate and efficiency prediction model for automotive axial-flow cooling fans is constructed based on a BP neural network optimized by a genetic algorithm. This model can predict the flow rate and efficiency of different automotive axial-flow cooling fans.
[0051] 4. Compared with traditional experimental measurement methods and numerical simulation methods, the model of the present invention can save a lot of manpower and material resources, as well as huge computing resources, and can be applied to different models and can be widely used in industry.
[0052] 5. The present invention will further optimize the model in terms of sample quantity, variable factors and structural algorithms in the future, and can provide a more accurate flow and efficiency prediction method for automotive axial cooling fans. BRIEF DESCRIPTION OF THE DRAWINGS
[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.
[0054] Figure 1 It is the structural diagram of BP neural network of the present invention;
[0055] Figure 2 This is the flow chart of the genetic algorithm optimizing the BP neural network of the present invention. DETAILED DESCRIPTION
[0056] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.
[0057] See also Figure 1 - Figure 2 , Flow rate and efficiency prediction method of automotive axial cooling fan based on GA-BP neural network.
[0058] 1. First, select the automotive axial flow cooling test fan, establish the automotive axial flow cooling test fan simulation model, and verify the accuracy of the simulation model through experiments.
[0059] 2. Obtain parameter data, flow rate, and efficiency data of the automotive axial cooling test fan or simulation model. The parameter data include speed, blade installation angle, number of blades, radial clearance, blade chord length, curvature radius, and hub ratio.
[0060] 3. Normalize the data; Since the dimensions and units of the characteristic variables in the data sample are different, it will affect the results of data analysis. Through data normalization, each variable is limited to a certain range, such as between 0 and 1, thereby eliminating the impact of dimensions and units, which is convenient for weighted comparative analysis. The normalization formula is as follows:
[0061]
[0062] Where: x * is the normalized value, x i is the training data, x min 、x max are the minimum and maximum values in the data respectively.
[0063] 4. Input the normalized data into the genetic algorithm.
[0064] 5. Set BP neural network parameters: Based on the collected data of flow rate, efficiency and various parameters of automotive axial cooling fans, the speed, blade installation angle, number of blades, radial clearance, blade chord length, curvature radius, hub ratio and other parameters are used as input variables, and the flow rate and efficiency are used as output variables to build the model; therefore, the number of neurons in the network input layer m is 7, and the number of neurons in the output layer n is 2. The number of neurons in the hidden layer is calculated according to the empirical formula (a is a constant from 1 to 10) to determine the range, and repeatedly compare by trial and error to determine the optimal number of neurons in the hidden layer of the model. BP neural network is currently widely used and is a multi-layer feedforward neural network trained by error back propagation. BP neural network consists of input layer, hidden layer and output layer (see Figure 1 ), the number of neurons in the input layer is m, the number of neurons in the output layer is n, and the number of neurons in the hidden layer is q. The initial connection weights between the input layer neurons and the hidden layer neurons, and between the hidden layer neurons and the output layer neurons are W i j 、V j k , the threshold biases of hidden layer neurons and output layer neurons are ω k , where i = 1, 2, 3 ... m, j = 1, 2, 3 ... q, k = 1, 2, 3 ... n, and the activation function is f (x) = 1 / (1 + e -x ). Network training is divided into two stages. The first stage is to propagate along the neural network forward to obtain the output value; the second stage is to propagate the error backward along the neural network to correct the weights and threshold bias. After two stages, the neural network training is completed. The training is repeated many times until the error is less than the target value. The weights and threshold bias will be fixed and the model construction is completed.
[0065] 6. Use genetic algorithms to optimize BP neural network parameters: BP neural networks have initial weights and thresholds that are randomly selected. Different initial values will cause large deviations in the results, leading to problems such as difficulty in convergence, long training time, or falling into local extreme values, which also leaves a lot of room for optimization for such neural networks. Selecting the appropriate population size, crossover operation probability, mutation operation probability, and maximum number of evolutionary iterations, and optimizing them through genetic algorithms is an ideal solution. Genetic algorithms are a method that simulates the biological evolution mechanism in nature to search for the optimal solution to a problem. It uses mathematical means to map the problem-solving process to the selection, crossover, and mutation process of chromosomes in biological evolution, thereby screening out chromosomes with higher fitness to form a new generation of populations. This process is iterated continuously until the preset target conditions are met, and finally the chromosome with the highest fitness is obtained, which is considered the optimal solution to the problem. Figure 2 The process of combining genetic algorithm with BP neural network is demonstrated. In particular, genetic algorithm shows significant advantages in optimizing BP neural network. It determines the optimal initial threshold and weight for BP neural network through global search, effectively overcoming the performance fluctuation of traditional BP neural network caused by random initialization parameters. Training with the optimal parameters optimized by genetic algorithm as the starting point not only accelerates the convergence speed of the network, but also significantly improves the prediction accuracy. Therefore, BP neural network optimized by genetic algorithm becomes the core algorithm of this model construction.
[0066] 7. This application uses two evaluation methods to evaluate the model prediction value error and accuracy and explore the model usability.
[0067] Mean absolute error (MAE)
[0068] MAE is the average value of the absolute error between the model prediction value and the actual value. The absolute value is used to avoid the offset of positive and negative errors. It reflects the average deviation distance between the predicted value and the actual value. When there is no error, the value is 0. The calculation formula is as follows:
[0069]
[0070] Where: MAE is the mean absolute error; y′ i is the i-th predicted value; y i is the ith actual value.
[0071] Mean absolute percentage error MAPE
[0072] MAPE is the average error rate between the predicted value and the actual value, reflecting the accuracy of the predicted value. The smaller the value, the higher the accuracy. The calculation formula is as follows:
[0073]
[0074] Where: MAPE is the mean absolute percentage error; y′ i is the i-th predicted value; y i is the ith actual value.
[0075] 8. Determine whether the error meets the preset accuracy or whether the iteration reaches the maximum number. If the calculation error is met, the algorithm converges, and the weight and threshold of the last iteration are the optimal solution. If the calculation error is not met, retrain and predict until the calculation error meets the design requirements. The BP neural network at this time is the optimal BP neural network optimized by the genetic algorithm. The flow rate and efficiency data of the automotive axial cooling fan predicted at this time are the predicted values.
[0076] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.
Claims
1. A method for predicting flow rate and efficiency of automotive axial cooling fans based on GA-BP neural network, characterized by: At least the following steps are included: S1: Automotive axial cooling test fan and simulation model. The automotive axial cooling test fan needs to pass the test and verify the rationality of the simulation model. After the rationality of the simulation model is verified, the corresponding fan flow and efficiency under various fan parameters can be obtained by simulating the model; S2: data collection and extraction, which includes two schemes for obtaining parameter data, flow rate and efficiency data of a vehicle axial cooling test fan or a simulation model; S3: Data preprocessing, normalizing the sample data to avoid the problem of slow network training convergence caused by large sample data differences; S4: Set BP neural network parameters and build a BP neural network model; S5: Use genetic algorithm to optimize BP neural network parameters, select appropriate population size, crossover operation probability, mutation operation probability and maximum number of evolution iterations, and use genetic algorithm to find the optimal initial threshold and weight of BP neural network, so that BP neural network can overcome the defects of random selection of initial threshold and weight, start training and iteration with the optimal solution, and obtain the final prediction result; S6: Optimizing the BP neural network model based on the GA algorithm to obtain an initial GA-BP model; S7: Verify the GA-BP model and explore the model usability; S8: The optimal GA-BP model is used to predict the flow rate and efficiency data of the automotive axial cooling fan, that is, to obtain the predicted value.
2. The method for predicting flow rate and efficiency of an automotive axial cooling fan based on a GA-BP neural network according to claim 1, characterized in that: The data collection and extraction in S2 includes data obtained by performing experiments on the test fan and data obtained by performing experimental simulations on the simulation model; The parameter data include rotation speed, blade installation angle, number of blades, radial clearance, blade chord length, curvature radius and hub ratio.
3. The method for predicting flow rate and efficiency of an automotive axial cooling fan based on a GA-BP neural network according to claim 1, characterized in that: The S3 at least includes the following steps: By normalizing the data, each variable is limited to a certain range, thus eliminating the influence; To facilitate weighted comparative analysis, the normalization formula is as follows: Where: x * is the normalized value, x i is the training data, x min 、x max are the minimum and maximum values in the data respectively.
4. The method for predicting flow rate and efficiency of an automotive axial cooling fan based on a GA-BP neural network according to claim 1, characterized in that: The S4 at least comprises the following steps: The number of neurons m in the input layer of the BP neural network is set to 7, and the number of neurons n in the output layer of the BP neural network is set to 2; The number of neurons in the hidden layer of the BP neural network is calculated according to the empirical formula The range of a is a constant from 1 to 10, and the range is determined by repeated comparison through trial and error to determine the optimal number of neurons in the hidden layer of the model; Based on the collected sample data of flow rate, efficiency and parameters of automotive axial cooling fans; The model is constructed by taking parameters such as rotation speed, blade installation angle, number of blades, radial clearance, blade chord length, curvature radius, hub ratio, etc. as input variables and flow rate and efficiency as output variables; Perform network training, which includes two stages: The first stage is to propagate forward along the neural network to obtain the output value; In the second stage, the error is propagated back along the neural network to correct the weights and threshold biases; After two stages, the neural network training is completed. The training is repeated many times until the error is less than the target value. The weights and threshold bias are fixed, and the BP neural network model is constructed.
5. The method for predicting flow rate and efficiency of an automotive axial cooling fan based on a GA-BP neural network according to claim 1, characterized in that: The S6 at least comprises the following steps: S61: Initializing the BP neural network, determining the number of input layer nodes, the number of hidden layer nodes, and the number of output layer nodes of the BP neural network; S62: Initial weights and thresholds, preprocessing the input data to obtain initial values, encoding the initial values through the GA algorithm, and establishing a fitness function to optimize the weights and thresholds of the BP neural network; S63: Roulette wheel selection method is used to select the individual with the highest fitness in the population with coded population data; S64: Perform a crossover operation on the individuals with the highest fitness in the selected group to obtain several new individuals; S65: performing mutation operation on a number of the new individuals to obtain a new population; S66: updating the new population, calculating the fitness, and replacing the individuals with low fitness with individuals with high fitness to form a new population; S67: Repeat S61 to S66 until the fitness is greater than a preset threshold or the number of iterations reaches a preset number, then stop the iteration and obtain the optimal weight and threshold.
6. The method for predicting flow rate and efficiency of an automotive axial cooling fan based on a GA-BP neural network according to claim 1, characterized in that: In S7, the GA-BP model is evaluated for the prediction error and accuracy of the model using two evaluation methods: mean absolute error (MAE) and mean absolute percentage error (MAPE).
7. The method for predicting flow rate and efficiency of an automotive axial cooling fan based on a GA-BP neural network according to claim 1, characterized in that: The S8 at least comprises the following steps: Based on S7, whether the error meets the preset accuracy or whether the iteration reaches the maximum number; If the calculation error is satisfied, the algorithm converges; The weight and threshold of the last iteration are the optimal solution. If the calculation error is not satisfied, retraining and prediction are performed. Until the calculation error meets the design requirements, the BP neural network at this time is the optimal BP neural network optimized by the genetic algorithm, that is, the optimal GA-BP model; Based on the optimal GA-BP model, the flow rate and efficiency data of the automotive axial cooling fan are predicted, that is, the predicted value is obtained.
