Toward Automatically Assessing the Novelty of Engineering Design Solutions

We propose to automate the assessment of novelty during the early design stages, where the number of solution alternatives is high, and these alternatives are flexible. The novelty of a solution is measured as its distance from a reference product database. The distance is calculated by comparing the solution and a reference at different levels of abstraction, as defined by a casuality model called SAPPhIRE. The fundamental instrument of this comparison is a text similarity metric, which is integrated with the novelty score, a support for which is developed as an Android application, named IdeA-Novelty. The application of the algorithm upon 20 solutions reveals that the measurement is computationally expensive. The expert's ratings for the same solution set seem to vary drastically. We take this observation as the basis for an extensive discussion on novelty, an operational definition for which is proposed toward the end.