A method for optimizing the transition fillet of balanced elbow based on fatigue analysis

Abstract

The present invention discloses a method for optimizing the transition fillet of a balancing elbow based on fatigue analysis, comprising the following steps: establishing a finite element model for fatigue analysis of a balancing elbow in finite element analysis software; uniformly selecting 20 to 40 values ​​from the adjustable ranges of the diameters of two transition fillets that need to be optimized to form 900 to 1000 combinations, and performing simulation calculations on each combination to obtain the fatigue life of the balancing elbow; selecting two transition fillet diameters as inputs and fatigue life as outputs to establish a neural network model; using an MIV method to find the fillet that has a greater impact on the fatigue life of the two fillets; uniformly selecting 90 to 100 values ​​from the adjustable range of the diameter of the fillet that has a greater impact on the life and calculating the corresponding fatigue life of the balancing elbow, and establishing a new neural network model with the transition fillet diameter as input and the fatigue life as output; establishing an optimization model for the transition fillet of the balancing elbow based on the newly constructed neural network model; and solving the optimization model using a genetic algorithm to obtain the optimal diameter of the transition fillet.

Description

Technical Field

[0001] The present invention relates to a method for optimizing a balance elbow transition fillet, and in particular to a method for optimizing a balance elbow transition fillet by using a neural network model and a genetic algorithm based on fatigue analysis. Background Art

[0002] The balancing elbow is a critical component of tracked armored vehicles. It connects the vehicle body and road wheels, transferring the substantial impact energy generated by the up-and-down motion of the road wheels to the torsion bar, thereby reducing the impact forces on the vehicle body. Fatigue damage or fracture in the balancing elbow can have devastating consequences, potentially even resulting in a loss of combat maneuverability for the armored vehicle. This places higher demands on the design of the balancing elbow. The transition radius of the balancing elbow, which forms the transition structure between the balancing elbow body and various parts of the balancing elbow, such as the roadwheel axle and spline shaft, is a highly susceptible location to fatigue damage. Optimizing the transition radius of the balancing elbow based on fatigue analysis can effectively improve the fatigue life of the balancing elbow.

[0003] The main optimization methods for the transition radius of the balanced elbow are: experimental comparison method, simulation comparison method, and topology optimization method. The experimental comparison method and simulation comparison method require a large number of experiments and simulation calculations, which cost a lot of time and money. They can only provide the optimal range or relatively optimal value of the transition radius of the balanced elbow, but cannot provide the theoretical optimal value. Although the topology optimization method can provide the theoretical optimal value, it needs to build an optimization model based on the simulation model. For complex structures, the modeling workload is large, the calculation is not easy to converge, and the optimization results are difficult to obtain. Therefore, it is necessary to design an optimization method for the transition radius of the balanced elbow that can easily complete the solution and provide the theoretical optimal solution. Summary of the Invention

[0004] In order to achieve the purpose of the present invention, the present invention provides a balanced elbow transition fillet optimization method that uses a neural network model to establish an optimization model based on fatigue analysis and uses a genetic algorithm to solve the optimization model.

[0005] The method for optimizing the transition fillet of a balanced elbow using a neural network model and a genetic algorithm based on fatigue analysis comprises the following steps:

[0006] (1) Establish a finite element model for fatigue analysis of the balanced elbow in finite element analysis software;

[0007] (2) Select two transition fillets A and B that need to be optimized, and evenly select 20 to 40 values ​​from the adjustable range of the fillet diameters of fillets A and B to form M combinations, 1000 ≥ M ≥ 900; simulate each combination as a case, simulate a shooting process to obtain the balance elbow response, and calculate the balance elbow fatigue life N based on the stress-strain history of the balance elbow. There are a total of M groups of data;

[0008] (3) Select the diameters of the transition fillets A and B as inputs, select the fatigue life of the balanced elbow N as output, and establish a 2-input 1-output BP neural network model based on M sets of data, denoted as N = f N (Φ A ,Φ B );

[0009] (4) Use the MIV method to find the fillet that has a greater impact on fatigue life among the two fillets, and record it as fillet C;

[0010] (5) Evenly select H values ​​from the adjustable range of the fillet C diameter, 100 ≥ H ≥ 90, and take each value as a case for simulation calculation. Simulate a shooting process to obtain the balanced elbow response and fatigue life N. There are a total of H groups of data;

[0011] (6) Select the fillet diameter C as input and the balanced elbow fatigue life N as output. Based on the H group of data, a 1-input 1-output BP neural network model is established, denoted as N = f N (Φ C );

[0012] (7) Establish an optimization model for the transition radius of the balanced elbow: Max:N=f N (Φ C ), Φ C ∈[Φ C1 ,Φ C2 ];

[0013] (8) Using genetic algorithm to solve the optimal diameter Φ of the transition fillet CM .

[0014] Furthermore, in step (1), finite element analysis software Ansys Workbench is used.

[0015] Furthermore, in step (2), the values ​​uniformly selected in the adjustable range of the diameter of the fillet A include the two end points of the range, and the values ​​uniformly selected in the adjustable range of the diameter of the fillet B include the two end points of the range.

[0016] Furthermore, in step (3), the neural network model established has 200 hidden layers, and the mean square error between the actual output value and the predicted value does not exceed 0.00001.

[0017] Furthermore, in step (4), it is considered that the fillet corresponding to a larger value of |MIV| has a greater impact on the fatigue life of the balancing elbow, which is recorded as fillet C.

[0018] Furthermore, in step (5), the values ​​uniformly selected from the adjustable range of the fillet C diameter include both end points of the range.

[0019] Furthermore, in step (8), a genetic algorithm is used to solve the optimization model of the balanced elbow transition fillet. The population size of the genetic algorithm is set to 50, the crossover probability is 0.8, the mutation probability is 0.1, and the termination evolution generation of the genetic operation is 1000. Finally, the optimal fillet diameter Φ is obtained. CM .

[0020] The optimal value of the transition fillet diameter of the balancing elbow can be obtained by this method, which will effectively improve the fatigue life of the balancing elbow.

[0021] The present invention will be described in detail with reference to the embodiments and accompanying drawings to provide a more in-depth understanding of the objectives, features and advantages of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It represents a method for optimizing the transition radius of a balancing elbow using a neural network model and a genetic algorithm based on fatigue analysis. DETAILED DESCRIPTION

[0023] Reference Figure 1 In the middle process, a finite element model for fatigue analysis of the balanced elbow was established in the finite element analysis software Ansys Workbench.

[0024] Select two transition fillets that need to be optimized, record them as fillet A and fillet B respectively, and determine the adjustable range of fillet A and B diameters [Φ A1 ,Φ A2 ] and [Φ B1 ,Φ B2 ]. In [Φ A1 ,Φ A2 ] uniformly select 30 diameter values, including Φ A1 and Φ A2 , in [Φ B1 ,Φ B2 ] uniformly select 30 diameter values, including Φ B1 and Φ B2 900 possible combinations were created by selecting one from each of the 30 values ​​for the fillet A diameter and the 30 for the fillet B diameter. A simulation was performed on one of these combinations, simulating the response of the balancing elbow during a single firing process. The fatigue life N of the balancing elbow was calculated based on the stress-strain history of the balancing elbow, yielding a total of 900 data sets.

[0025] Selecting the diameter of transition fillet A and the diameter of transition fillet B as input, and the fatigue life of the balanced elbow as output, a 2-input 1-output BP neural network model was established on the MATLAB platform based on 900 sets of data, denoted as N = f N (Φ A ,Φ B), the established neural network model has 200 hidden layers, and the mean square error between the actual output value and the predicted value does not exceed 0.00001.

[0026] The MIV method is used to determine the fillet that has a greater impact on the fatigue life of the two fillets. It is believed that the fillet corresponding to the larger |MIV| value has a greater impact on the fatigue life of the balanced elbow and is recorded as fillet C.

[0027] Adjustable range of fillet C diameter [Φ C1 ,Φ C2 ] uniformly select 90 values, including Φ C1 and Φ C2 , taking each value as a case for simulation calculation, 90 balanced elbow fatigue lives can be obtained through simulation calculation.

[0028] Select the transition fillet C diameter as input, select the balanced elbow fatigue life as output, and establish a 1-input 1-output BP neural network model, denoted as N = f N (Φ C ).

[0029] Based on the constructed BP neural network model, the optimization model of the balanced elbow transition radius is established as follows:

[0030] Max:N=f N (Φ C )

[0031] Φ C ∈[Φ C1 ,Φ C2 ]

[0032] The genetic algorithm is used to solve the optimization model of the balanced elbow transition fillet. The population size of the genetic algorithm is set to 50, the crossover probability is 0.8, the mutation probability is 0.1, and the termination evolution generation of the genetic operation is 1000. Finally, the optimal fillet diameter Φ is obtained. CM .

Claims

1. A method for optimizing the transition fillet of a balanced elbow based on fatigue analysis, characterized in that: The transition fillet optimization method comprises the following steps: (1) Establish a finite element model for fatigue analysis of the balanced elbow in finite element analysis software; (2) Select two transition fillets A and B that need to be optimized, and evenly select 20 to 40 values ​​from the adjustable range of the fillet diameters of fillets A and B to form M combinations, 1000 ≥ M ≥ 900; simulate each combination as a case, simulate a shooting process to obtain the balance elbow response, and calculate the balance elbow fatigue life N based on the stress-strain history of the balance elbow. There are a total of M groups of data; (3) Select the diameters of the transition fillets A and B as inputs, select the fatigue life of the balanced elbow N as output, and establish a 2-input 1-output BP neural network model based on M sets of data, denoted as N = f N (Φ A ,Φ B ); (4) Use the MIV method to find the fillet that has a greater impact on fatigue life among the two fillets, and record it as fillet C; (5) Evenly select H values ​​from the adjustable range of the fillet C diameter, 100 ≥ H ≥ 90, and take each value as a case for simulation calculation. Simulate a shooting process to obtain the balanced elbow response and fatigue life N. There are a total of H groups of data; (6) Select the fillet diameter C as input and the balanced elbow fatigue life N as output. Based on the H group of data, a 1-input 1-output BP neural network model is established, denoted as N = f N (Φ C ); (7) Establish an optimization model for the transition radius of the balanced elbow: Max:N=f N (Φ C ), Φ C ∈[Φ C1 ,Φ C2 ]; (8) Using genetic algorithm to solve the optimal diameter Φ of the transition fillet CM .

2. The method for optimizing the balance elbow transition fillet based on fatigue analysis according to claim 1, characterized in that: In the step (1), the finite element analysis software Ansys Workbench is used.

3. The method for optimizing the balance elbow transition fillet based on fatigue analysis according to claim 1, characterized in that: In the step (2), the values ​​uniformly selected in the adjustable range of the diameter of the fillet A include the two end points of the range, and the values ​​uniformly selected in the adjustable range of the diameter of the fillet B include the two end points of the range.

4. The method for optimizing the balance elbow transition fillet based on fatigue analysis according to claim 1, characterized in that: In the step (3), the neural network model established has 200 hidden layers, and the mean square error between the actual output value and the predicted value does not exceed 0.00001.

5. The method for optimizing the balance elbow transition fillet based on fatigue analysis according to claim 1, characterized in that: In step (4), it is considered that the fillet corresponding to the larger MIV value has a greater impact on the fatigue life of the balancing elbow, and is recorded as fillet C.

6. The method for optimizing the balance elbow transition fillet based on fatigue analysis according to claim 1, characterized in that: In the step (5), the values ​​uniformly selected from the adjustable range of the fillet C diameter include both end points of the range.

7. The method for optimizing the balance elbow transition fillet based on fatigue analysis according to claim 1, characterized in that: In step (8), a genetic algorithm is used to solve the optimization model of the balanced elbow transition fillet. The population size of the genetic algorithm is set to 50, the crossover probability is 0.8, the mutation probability is 0.1, and the termination evolution generation of the genetic operation is 1000. Finally, the optimal fillet diameter Φ is obtained. CM .

Images