人工智能算法在铁道车辆动力学仿真中的应用进展

Application advances of artificial intelligence algorithms in dynamics simulation of railway vehicle

梳理了人工智能算法在铁道车辆系统动力学仿真中的应用实例和国内外相关文献,概述了铁道车辆动力学仿真中常用的机器学习和深度学习算法,归纳和评述了2种学习算法在铁道车辆系统动力学建模与仿真中的应用分类;从铁道车辆系统动力学建模、动力学性能预测与动力学性能优化等方面入手,详细讨论了人工智能算法应用在力元建模和仿真、轨道不平顺预测、运行平稳性预测、噪声预测、侧风安全性预测、运行安全性预测、悬挂优化、轮轨匹配优化、结构优化以及主动与半主动控制等领域的优势和局限性,指出了现阶段人工智能算法在动力学仿真应用中主要面临的训练样本缺乏、泛化能力不够、可解释性欠缺等问题;展望了今后人工智能算法和车辆系统动力学交叉研究的发展方向和重点研究内容。研究结果表明:融合经典力学和人工智能算法结合的混合建模理论可作为之后的重点研究方向;人工智能算法对解决随机动力学中的随机不确定性,提高随机动力学的性能具有较大的应用潜力;通过人工智能算法与优化算法相结合来实现动力学性能优化,可充分发挥人工智能算法的优势。

The application examples and domestic and foreign literatures using artificial intelligence algorithm for railway vehicle system dynamics simulation were reviewed. The machine learning and deep learning algorithms commonly used in railway vehicle dynamics simulation were summarized, and the application classifications of the 2 algorithms in railway vehicle system dynamics modelling and simulation were concluded and interpreted. According to railway vehicle system dynamics modelling, dynamics performance prediction and dynamics performance optimization, the advantages and limitations of applying artificial intelligence algorithms in force-elements modelling and simulation, track irregularity prediction, running stability prediction, noise prediction, crosswind safety prediction, running safety prediction, suspension optimization, wheel-rail matching optimization, structure optimization, and active and semi-active control were discussed in detail. The problems of applications of artificial intelligence algorithms in railway dynamics simulation were lack of training samples, generalization ability and interpretability. The development directions and key research contents of the interdisciplinary research between artificial intelligence and vehicle system dynamics were given. Research result shows that the hybrid modelling theory combining classical mechanics and artificial intelligence algorithms can be as a key research direction in the future. There is great potential to use the artificial intelligence algorithms to solve the random uncertainty in stochastic dynamics and improve the performance of stochastic dynamics. The artificial intelligence algorithms combinated with optimization algorithms can exploit their advantages in the dynamics performance optimization. 5 tabs, 12 figs, 77 refs.