Hybrid modelling through latent differential-regression analysis (LDRA) for predicting long-term equipment degradation in the chemical process industry

Abstract

Long-term equipment degradation have a major impact on plant safety, operation and economy. We present an hybrid modeling approach to predict and analyse this type of phenomena in chemical process industries (CPI). A new predictive methodology is introduced, called Latent Differential Regression Analysis (LDRA), that automatically extracts the dynamical elements that have an impact on the evolution of the degradation mode. A knowledge-based feature is also developed to act as the surrogate degradation index. The hybrid approach is tested using real data from an industrial site, where fouling takes place in several heat exchangers located in the plant. The methodology is general and can be applied to other long-term degradation modes common in the CPI, such as catalyst deactivation, corrosion, mechanical degradation of packing beds and catalysts, coking, etc. The proposed modeling approach based on LDRA and a case-dependent Equipment Health Indicator can find wide application in CPI.