A holistic method of complex product development based on a neural network-aided technological evolution system

Kang Wang,Runhua Tan,Qingjin Peng,Fanfan Wang,Peng Shao,Zhuoli Gao

DOI: https://doi.org/10.1016/j.aei.2021.101294

IF: 8.8

2021-04-01

Advanced Engineering Informatics

Abstract:<p>Product complexity increases along with the increase of product functions. Effective methods are required in the complex product development. Axiomatic design can effectively reduce product complexity, but there are deficiencies in the usage process and the innovation stimulation. This paper proposes a holistic method for the development of complex products by using a neural network-aided technological evolution process to control negative effects of complexity. For the first time, engineering parameters are used as the intermediary between the complexity and technological evolution. Problematic engineering parameters are extracted from a current reality tree model through a natural language process. An artificial neural network is trained to predict the technological evolution law based on existing successful samples. Appropriate analogical objects are formed in an invention principle case library through the similarity analysis of engineering parameters. The optimal scheme is developed objectively by using the coefficient variation method and Dempster combination rule. The proposed method is applied to develop a pipe cutting machine with a granted patent for its feasibility and effectiveness.</p>

engineering, multidisciplinary,computer science, artificial intelligence

What problem does this paper attempt to address?

The paper attempts to address several key issues in complex product design:

1. **Increased Complexity**: As product functions increase, the complexity of the product also increases. This complexity is not only reflected in the increase in the number of components and multi-level connections but may also be hidden within the product, making it difficult to detect.

2. **Faster Iteration Speed**: The iteration speed between different generations of products is accelerating, requiring breakthrough improvements to attract new users, but determining the direction of improvement becomes difficult.

3. **Difficulty in Data Collection and Analysis**: A large number of intelligent attributes (such as patents) can serve as references for product design, but collecting and analyzing this data is very challenging.

Existing complex product design methods are either too macro to guide specific designs or too micro to form general methods. In particular, Axiomatic Design (AD), although it can effectively reduce product complexity, has shortcomings in its use and is difficult to inspire innovation.

To address these issues, the paper proposes an integrated method based on neural network-assisted technology evolution, aiming to improve the design efficiency and innovation of complex products by controlling the negative impact of complexity. Specifically, this method introduces engineering parameters as intermediaries for product complexity and technological evolution for the first time, extracts problem engineering parameters through natural language processing, and uses artificial neural networks to predict technological evolution patterns. Additionally, the method combines similarity analysis of the invention principle case library and solution evaluation methods, and is ultimately applied to the design of a metal tube cutting machine, proving its feasibility and effectiveness.