Automated Process Optimization in Manufacturing Systems Based on Static and Dynamic Prediction Models

摘要: A key aspect in predictive maintenance is the early recognition of product downtrends and a proper reaction on it, to reduce production waste avoid machine failures, components destruction, risks for operators. We propose an approach automated optimization process parameters manufacturing systems order automatically compensate possible quality at stage. This should significantly or even manual (reaction) efforts operators which are often time-intensive laborious. Such recognized by prediction models quality, extracted from data come two different variants: (1) static mappings established based (machining) parameter settings through combination new hybrid variant design experiment (DoE), cross-correlation analysis, data-driven mapping construction; (2) dynamic forecast respect time-series trends values measured during on-line production, being able properly recognize undesired changes dynamics happening (unexpectedly) process. These will have property be self-adapt evolve over time recordings; they employ generalized (flexible) evolving fuzzy (GEFS) combined with incremental update latent variable space obtained partial least squares (PLS). Both types can then used as surrogate within multi-objective, evolutionary important target criteria, relies fast co-evolution strategy. Several results micro-fluidic chip included demonstrate applicability performance proposed methods discuss open challenges.