Knowledge-based methods for evaluation of engineering changes

Abstract

Engineering Changes (ECs) are an integral part of a product's lifecycle. A proposed EC can affect several lifecycle-wide components. Detailed evaluation of each proposed EC or its effect is time-consuming and inefficient. Therefore, enterprises plan detailed evaluation of only those EC effects that might have a high cost impact. Currently, domain experts decide which effects should undergo a detailed evaluation process. Such an approach relies heavily on personal experience and is less reliable. To address this problem, this research develops a systematic knowledge-based method for determining whether a proposed EC effect has high impact and would require a detailed evaluation. Only some of the large number of EC attributes are important for retrieving past ECs, which can be used to evaluate the impact of a proposed EC. This research formulates the problem of determining important EC attributes as a multi-objective optimization problem. Information-theoretic concepts are used to define measures for quantifying importance of an attribute subset. The domain knowledge and the information in EC database are combined to estimate probability distributions, which are required in computation of measures. An Ant Colony Optimization (ACO)-based search approach is developed for efficiently locating the important attribute set. Utilizing past EC knowledge to predict the impact of proposed EC effect requires an approach to compute similarity between ECs. The second part of this research presents an approach to compute similarity between ECs that are defined using disparate attributes. Since the available information is probabilistic, the measures of information are utilized for defining measures to compute similarity between two attribute values or ECs. The semantics associated with attribute values are utilized to compute similarity between attribute values. In the last part of this research, an approach is developed to predict impact of proposed EC effect based on the similar past ECs. The approach incorporates a technique to quantify differences between important attribute values in proposed EC and a similar past EC. The Bayes's rule is used to determine differences in impact value from the differences in attribute values. The probability values required in the Bayes's rule are determined based on the minimum cross entropy principle.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 8 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.