Predicting Sequential Design Decisions Using the Function-Behavior-Structure Design Process Model and Recurrent Neural Networks
Predicting Sequential Design Decisions Using the Function-Behavior-Structure Design Process Model and Recurrent Neural Networks
复制标题
DOI：
10.1115/1.4049971
复制
发表时间：
2021-08
影响因子：
3.3
通讯作者：
M. H. Rahman;Charles Xie;Zhenghui Sha
中科院分区：
文献类型：
--
作者：
M. H. Rahman;Charles Xie;Zhenghui Sha
在工程系统设计中，设计人员反复在不同的设计阶段之间来回探索设计空间，并搜索满足所有设计约束的最佳设计解决方案。对于复杂的设计问题，人类已经表现出惊人的能力，有效地减少设计空间的维数，并迅速收敛到一个合理的范围内的算法介入，并继续搜索过程。因此，建模人类设计师如何在这样的顺序设计过程中做出决策，可以帮助发现有益的设计模式，策略和算法，这对于开发嵌入人类智能的新算法以增强计算设计至关重要。在本文中，我们开发了一种基于深度学习的方法来建模和预测设计人员在系统设计环境中的顺序决策。这种方法的核心是集成用于设计过程表征的功能-行为-结构（FBS）模型和用于深度学习的长短期记忆单元（LSTM）模型。这种方法是证明在两个案例研究的太阳能系统的设计，其预测精度进行评估基准上的几个常用的模型顺序的设计决策，如马尔可夫链模型，隐马尔可夫链模型，和随机序列生成模型。结果表明，该方法优于其他传统的模型。这意味着在系统设计任务期间，设计人员很可能依赖于过去设计决策的短期和长期记忆来指导他们在设计过程中的未来决策。我们的方法可以支持人机交互的设计，一般适用于其他设计环境，只要设计动作的序列数据是可用的。
In engineering systems design, designers iteratively go back and forth between different design stages to explore the design space and search for the best design solution that satisfies all design constraints. For complex design problems, human has shown surprising capability in effectively reducing the dimensionality of design space and quickly converging it to a reasonable range for algorithms to step in and continue the search process. Therefore, modeling how human designers make decisions in such a sequential design process can help discover beneficial design patterns, strategies, and heuristics, which are essential to the development of new algorithms embedded with human intelligence to augment the computational design. In this paper, we develop a deep learning-based approach to model and predict designers’ sequential decisions in the systems design context. The core of this approach is an integration of the function-behavior-structure (FBS) model for design process characterization and the long short-term memory unit (LSTM) model for deep leaning. This approach is demonstrated in two case studies on solar energy system design, and its prediction accuracy is evaluated benchmarking on several commonly used models for sequential design decisions, such as the Markov Chain model, the Hidden Markov Chain model, and the random sequence generation model. The results indicate that the proposed approach outperforms the other traditional models. This implies that during a system design task, designers are very likely to rely on both short-term and long-term memory of past design decisions in guiding their future decision-making in the design process. Our approach can support human–computer interactions in design and is general to be applied in other design contexts as long as the sequential data of design actions are available.