Purpose: How to get a lighter and stronger anti-rolling torsion bar has become a barrier for the development of high-speed railway vehicles. The purpose of this paper is to realize the multi-objective optimization of an anti-rolling torsion bar with a Modified Non-dominated Sorting Genetic Algorithm III (MNSGA-III), which aims to obtain a better design scheme of an anti-rolling torsion bar device. Design/methodology/approach: First, the Non-dominated Sorting Genetic Algorithm III (NSGA-III) uses a simulated binary crossover (SBX) operator and a polynomial mutation operator, while the MNSGA-III algorithm proposed in this paper introduces an arithmetic crossover and an adaptive mutation operator to change the crossover and mutate operator in NSGA-III. Second, two algorithms are tested by ZDT3, ZDT4 functions. Both algorithms set the same population size and evolutionary generation, and then compare the results of NSGA-III and MNSGA-III. Finally, MNSGA-III is applied to the multi-objective model of an anti-rolling torsion bar which is established by taking the mass and stiffness of the torsion bar as the optimization object. After that, it obtains the Pareto solution set by solving the multi-objective model with MNSGA-III. The only optimal solution selected from the Pareto solution set is compared with the traditional design scheme of an anti-rolling torsion bar. Findings: The MNSGA-III converges faster than NSGA-III. Besides, MNSGA-III has better diversity of Pareto solutions than NSGA-III and is closer to the ideal Pareto frontier. Comparing with the results before the optimization, it shows that the volume of the anti-rolling torsion bar reduces by 1.6 percent and the stiffness increases by 3.3 percent. The optimized data verifies the effectiveness of this method proposed in this paper. Originality/value: The simulated binary crossover operator and polynomial mutation operator of NSGA-III are changed into an arithmetic crossover operator and an adaptive mutation operator, respectively, which improves the optimization performance of the algorithm.

Purpose: Most of the previous work on reliability analysis was based on the traditional reliability theory. The calculated results can only reflect the reliability of components at a specific time, which neglects the uncertainty of load and resistance over time. The purpose of this paper is to develop a time-dependent reliability analysis approach based on stochastic process to deal with the problem and apply it to the structural design of railway vehicle components. Design/methodology/approach: First, the parametric model of motor hanger for electric multiple unit (EMU) is established by ANSYS parametric design language, and its structural stress is analyzed according to relevant standards. The Latin hypercube method is used to analyze the sensitivity of the structure, and the uncertainty parameters (sizes and loads) which have great influence on the structural strength are determined. The D-optimal experimental design is carried out to establish the polynomial response surface function, which characterizes the relationship between uncertainty parameters and structural stress. Second, the Poisson stochastic process is adopted to describe the number of loads acting, and the Monte Carlo method is used to obtain the load acting history according to its probability distribution characteristics. The load history is introduced into the response surface function and the uncertainty of other parameters is considered at the same time, and the stress history of the motor hanger is obtained. Finally, the degradation process of structural resistance is described by a Gamma stochastic process, and the time-dependent reliability of the motor hanger is calculated based on the reliability theory. Findings: Time and the uncertainties of parameters have great impact on reliability. The results of reliability decrease with time fluctuation are more reasonable, stable and credible than traditional methods. Practical implications: In this paper, the proposed method is applied to the structural design of the motor hanger for EMU, which has a good guiding significance for accurately evaluating whether if the design meets the reliability requirements. Originality/value: The value of this paper is that the method takes both the randomness of load over time and the uncertainty of structural parameters in the design and manufactures process into consideration, and describes the monotonous degradation characteristics of structural resistance. At the same time, the time-dependent reliability of mechanical components is calculated by a response surface method. It not only improves the accuracy of reliability analysis, but also improves the analysis efficiency and solves the problem that the traditional reliability analysis method can only reflect the static reliability of components.

Purpose: Mechanical products usually require deterministic finite element analysis in the design phase to determine whether their structures meet the requirements. However, deterministic design ignores the influence of uncertainties in the design and manufacturing process of mechanical products, leading to the problem of a lack of design safety or excessive redundancy in the design. In order to improve the accuracy and rationality of the design results, a robust design method for structural reliability based on an active-learning marine predator algorithm (MPA)–backpropagation (BP) neural network is proposed.Design/methodology/approach: The MPA was used to obtain the optimal weights and thresholds of a BP neural network, and an active-learning function applicable to neural networks was proposed to efficiently improve the prediction performance of the BP neural network. On this basis, a robust optimization design method for mechanical product reliability based on the active-learning MPA-BP model was proposed. Random moving quadrilateral sampling was used to obtain the sample points required for training and testing of the neural network, and the reliability sensitivity corresponding to each sample point was calculated by subset simulated significant sampling (SSIS). The total mass of the mechanical product and the structural reliability sensitivity of the trained active-learning MPA-BP model output were taken as the optimization objectives, and a multi-objective reliability-robust optimization design model was constructed, which was solved by the second-generation non-dominated ranking genetic algorithm (NSGA-II). Then, the dominance function was used in the obtained Pareto solution set to make a dominance-seeking decision to obtain the final reliability-robust optimization design solution. The feasibility of the proposed method was verified by a reliability-robust optimization design example of the bogie frame.Findings: The prediction error of the active-learning MPA-BP neural network was smaller than those of the particle swarm optimization (PSO)-BP, marine predator algorithm (MPA)-BP and genetic algorithm (GA)-BP neural networks under the same basic parameter settings of the algorithm, which indicated that the improvement strategy proposed in this paper improved the prediction accuracy of the BP neural network. To ensure the reliability of the bogie frame, the reliability sensitivity and total mass of the bogie frame were reduced, which not only realized the lightweight design of the bogie frame, but also improved the reliability and robustness of the bogie.Originality/value: The MPA algorithm with a higher optimization efficiency was introduced to find the weights and thresholds of the BP neural network. A new active-learning function was proposed to improve the prediction accuracy of the MPA-BP neural network.

Purpose: In a structural optimization design-based single-level response surface, the number of optimal variables is too much, which not only increases the number of experiment times, but also reduces the fitting accuracy of the response surface. In addition, the uncertainty of the optimal variables and their boundary conditions makes the optimal solution difficult to obtain. The purpose of this paper is to propose a method of fuzzy optimization design-based multi-level response surface to deal with the problem. Design/methodology/approach: The main optimal variables are determined by Monte Carlo simulation, and are classified into four levels according to their sensitivity. The linear membership function and the optimal level cut set method are applied to deal with the uncertainties of optimal variables and their boundary conditions, as well as the non-fuzzy processing is carried out. Based on this, the response surface function of the first-level design variables is established based on the design of experiments. A combinatorial optimization algorithm is developed to compute the optimal solution of the response surface function and bring the optimal solution into the calculation of the next level response surface, and so on. The objective value of the fourth-level response surface is an optimal solution under the optimal design variables combination. Findings: The results show that the proposed method is superior to the traditional method in computational efficiency and accuracy, and improves 50.7 and 5.3 percent, respectively. Originality/value: Most of the previous work on optimization was based on single-level response surface and single optimization algorithm, without considering the uncertainty of design variables. There are very few studies which discuss the optimization efficiency and accuracy of multiple design variables. This research illustrates the importance of uncertainty factors and hierarchical surrogate models for multi-variable optimization design.

This paper develops a material-structure integrated design and optimization method based on a multiscale approach for the lightweight design of CFRP car doors. Initially, parametric modeling of RVE is implemented, and their elastic performance parameters are predicted using the homogenization theory based on thermal stress, exploring the impact of RVE parameters on composite material performance. Subsequently, a finite element model of the CFRP car door is constructed based on the principle of equal stiffness, and a parameter transfer across microscale, mesoscale, and macroscale levels is achieved through Python programming. Finally, the particle generation and updating strategies in the Multi-Objective Particle Swarm Optimization (MOPSO) algorithm are improved, enabling the algorithm to directly solve multi-constraint and multi-objective optimization problems that include various composite material layup process constraints. Case study results demonstrate that under layup process constraints and car door stiffness requirements, plain weave, twill weave, and satin weave composite car doors achieve weight reductions of 15.85%, 14.54%, and 15.35%, respectively, compared to traditional metal doors, fulfilling the requirements for a lightweight design. This also provides guidance for the lightweight design of other vehicle body components.