基于深度学习和TRIZ的牵引变流器增效设计

Efficiency Design of Traction Inverters Based on Deep Learning and TRIZ

在牵引变流器系统设计中,各子系统间存在复杂的多元技术矛盾,而依赖手动查表的传统创新方法难以有效解决这些矛盾。对此,文章将深度学习算法引入发明问题解决理论(TRIZ)中,通过研究基于深度学习和TRIZ原理相结合的协同创新方法,构建TRIZ-CRNN模型,挖掘创新案例中隐含的发明原理,从而实现牵引变流器的协同创新。实验结果显示,TRIZ-CRNN模型在TRIZ创新案例数据集上的识别精度达到99.5%。这验证了该协同创新方法的可行性,其能实现对TRIZ创新机制的智能建模,为牵引变流器技术创新提供理论支撑及解决方案;同时,由于该设计方法基于TRIZ-CRNN的人机交互软件,其不仅提升了计算机辅助创新系统的操作智能化水平,而且还优化了TRIZ-CRNN算法的应用。

In the design of traction inverter system,there are complex and multiple technical contradictions among various subsystems.Traditional innovative method relies on manual table look-up,which is difficult to effectively deal with multiple technical issues.Due to the above technical difficult,this paper introduces deep learning algorithm into the theory of invention problem solving(Teoriya Resheniya Izobreatatelskikh Zadatch,TRIZ),concentrates on the collaborative innovation method based on deep learning and TRIZ principles.Through the construction of TRIZ-CRNN model,innovation mechanism contained in innovation cases is explored to realize the collaborative innovation of traction inverter.The TRIZ-CRNN model achieves 99.5%recognition accuracy on the TRIZ innovation case data set,which verifies the feasibility of the collaborative innovation method combining deep learning and TRIZ to be used in innovation mechanism modeling,and provides effective solution and theoretical support for traction inverter technology innovation.In addition,this paper designs human-computer interaction software based on TRIZ-CRNN to improve the operation intelligence of computer aided innovation system and optimize the application feasibility of TRIZ-CRNN algorithm.