(769c) Anomaly Detection and Diagnosis In Continuous Pharmaceutical Manufacturing: An Ontological Approach
The development and manufacturing of pharmaceutical products are governed by strict safety regulations since defective products would have a devastating consequence in public healthcare. As a result, pharmaceutical products undergo rigorous product testing to meet strict product quality specifications. With the push towards continuous manufacturing to reduce production costs, the identification and immediate correction of deviations from desired operating conditions become more important. Timely anomaly detection and diagnosis can potentially avoid the progression of abnormal events and thus reduce productivity loss without compromising the quality and safety of the products. Additionally, an automated fault detection and diagnosis system reduces the reliance of plant operation on human operators, which according to industrial statistics results in 70% of industrial accidents .
While several fault detection and diagnosis frameworks have been proposed in the literature, an approach that combines intelligent trend monitoring using Qualitative Trend Analysis (QTA) with causal reasoning based on Signed Digraphs (SDG) has been shown to be quite successful in the past. in diagnosing incipient faults . However, these frameworks generally require the user to be familiar with programming languages and are difficult to customize. To facilitate the application of SDG and QTA to detect and diagnose faults a framework is being developed within the Purdue Ontology for Pharmaceutical Engineering (POPE) environment which allows users to perform fault detection and diagnosis more easily.. The vision is to call upon the framework developed in , through POPE, to perform fault diagnosis without direct user involvement. Thus, the user need only provide mechanistic models (represented in ontologies) that describe the unit operation along with qualitative relationships within model variables, and a stream of data to classify faults, and subsequently detect and diagnose the faults.
The incorporation of the fault detection and diagnosis framework into POPE would allow direct access to a built-in library of control strategies; this would then enable rapid fault detection and diagnosis and immediate identification of the appropriate control methodology to correct the fault(s). For example, if a higher than desired ribbon density is detected; the possible causes are deviation in pressure, feed speed or motor malfunction. If through diagnosis, the most likely cause is determined to be a malfunction of the pressure transducer. The fault-tolerant control strategy would then provide new controller parameters to bring density back to desired value so as not to interrupt operations. We will present results of this approach for industrially inspired case studies.
1. Venkatasubramanian, V., R. Rengaswamy, K. Yin, and S. N. Kavuri (2003). A review of process fault detection and diagnosis Part I: Quantitative model-based methods. Computers and Chemical Engineering, 27,293-311
2. Maurya, M.R., R. Rengaswamy, V. Venkatasubramanian (2007). A signed directed graph and qualitative trend analysis ? based Framework for incipient fault diagnosis. Chemical Engineering Research and Design, 85(A10) 1407-1422