Data-driven product design evaluation method based on multi-stage artificial neural network

Abstract

The application of the new generation of IT technology in the manufacturing industry has brought a rich data foundation for the research of product life cycle management. However, in the product design stage, the existing design evaluation methods have not yet effectively used the data-driven advantages, which leads to over-reliance on personal experience and inefficiency. This paper presents a new data-driven product design evaluation (PDE) method to address this problem. First, a PDE model considering multi-stage evaluation indicators of the product life cycle is established. Second, to accurately and quickly realize the nonlinear mapping between the PDE model indicators, an improved multi-stage artificial neural network (ANN) based on a hybrid optimization algorithm combining particle swarm optimization and Adam (PSO-Adam) is proposed. The PSO is employed for a rapid convergence during the initial stage of a global search, while Adam is used to optimizing the training parameters around the global optimum adaptively. Finally, comparison experiments of the smartphone design are carried out to demonstrate the effectiveness of the multi-stage ANN and the PSO-Adam. The results demonstrate the proposed approach can help designers comprehensively consider design parameters and make fast and accurate design evaluation.