Making design decisions under uncertainties: probabilistic reasoning and robust product design

Abstract

Making design decisions is characterized by a high degree of uncertainty, especially in the early phase of the product development process, when little information is known, while the decisions made have an impact on the entire product life cycle. Therefore, the goal of complexity management is to reduce uncertainty in order to minimize or avoid the need for design changes in a late phase of product development or in the use phase. With our approach we model the uncertainties with probabilistic reasoning in a Bayesian decision network explicitly, as the uncertainties are directly attached to parts of the design artifact′s model. By modeling the incomplete information expressed by unobserved variables in the Bayesian network in terms of probabilities, as well as the variation of product properties or parameters, a conclusion about the robustness of the product can be made. The application example of a rotary valve from engineering design shows that the decision network can support the engineer in decision-making under uncertainty. Furthermore, a contribution to knowledge formalization in the development project is made.

A Knowledge-Based Engineering System for the Planning of Networked Rainwater Harvesting and Distribution Systems

“…Thus, the purpose of KBE is twofold: One is to automate the design or adaptation of products and their components based on artificial intelligence, e.g., a reasoning system, which is able to consider the design context, e.g., given requirements or restrictions [27,29]. The other is the automatic discovery of design knowledge about an artifact, e.g., through systematic investigation and evaluation of the sensitivities of design parameter changes and their dependencies with respect to the resulting product's properties [30,31]. Both qualify KBE as toolbox for realizing digital twins and thus imply its utility for operating smart service systems [32,33].…”

Modeling uncertain requirements

“…Consequently, the number of relevant literature reduces drastically, when focusing on publications which aim to describe or record irreducible uncertainty in attribute values of technical requirements. Like Yu et al (2013), many of the publications present approaches how to deal with uncertain requirements in the development process (see e.g., Kang et al (2018), Foith-Förster et al (2016, Gembarski et al (2021)). Thereby, the description of the uncertain requirements is driven by the chosen approach and not the information which must be encoded.…”

“…They aim to deal with requirement uncertainties, which stem from external sources (e.g., gas prices) and model them probabilistically. Gembarski et al (2021) use Bayesian decision networks to minimize design changes in late phases of product development. Whittle et al (2010) propose a requirements modeling language which addresses uncertainty for self-adaptive, mechatronic systems via fuzzy sets.…”

“…Thereby, all three characteristics coexist in engineering design. As such, all of the previously mentioned approaches fail to describe uncertain attributes in technical requirements properly, because they assume them to be either probabilistic (see e.g., Letier et al (2014), Gembarski et al (2021 or of a fuzzy nature (see e.g., Whittle et al (2010)).…”

The improvement of the intelligent decision support system for forecasting non-linear non-stationary processes

“…ISSN 2664-9969 decision options (alternatives), modeling of problem situations and scenarios, assessments of risks and uncertainties, selection and construction of the best solution [31][32][33][34][35]. However, these questions are related to the complex intellectual tasks, burdened by both uncertainty and the need to involve significant amounts of structured and unstructured data, and knowledge from various relevant subject areas [36][37][38][39][40].…”