(164e) Exceptional Events Management for Continuous Pharmaceutical Manufacturing: Application & Integration Into POPE and DeltaV

The process analytical technology (PAT) initiative advanced by the FDA gives the pharmaceutical industry an opportunity to apply various systems engineering tools, such as fault detection and diagnosis that are commonplace in other industries. Fault detection and diagnosis, which deals with the timely remedy of process abnormalities, can potentially avoid the progression of exceptional events and thus reduce material and productivity loss while ensuring the safety and quality of the products. Additionally, fault detection and diagnosis, reduces the reliance of plant operation on human operators, which according to statistics results in seventy percent of industrial accidents [1].

We have previously presented our Exceptional Events Management (EEM) framework as applied to the operation of roller compaction [2]. In that study, we demonstrated the efficacy of our EEM framework in detecting and diagnosing material property related faults that if left alone would have progressed to an undesirable situation. Furthermore, it was shown that the control system, at times, responded in an adverse manner thus further aggravating the system ? validating the fact that control systems alone are unable to deal with faults that we refer to as exceptional events.

In continuous manufacturing, a fault that arises in one process/equipment is bound to propagate throughout the downstream processes. Thus, the effects of the exceptional event would be distributed temporally as well as spatially. This complicates fault diagnosis as it is harder to determine the fault's point of origin or whether the effect is in fact due to multiple faults occurring simultaneously. Thus, we explore the application of EEM for a continuous pharmaceutical line consisting of a feeder, blender and roller compactor in series. A key portion of EEM, fault detection and diagnosis, involves the use of a low resolution qualitative method known as Signed Directed Graphs (SDG) to roughly detect and diagnose a possible candidate set of faults, followed by a higher resolution, more discriminatory method known as Qualitative Trend Analysis (QTA) that determines the fault that is most likely occurring. The subsequent portion of EEM is mitigation of detected exceptional events. The Purdue Ontology for Pharmaceutical Engineering (POPE), which has been extensively developed at Purdue [3], is used to facilitate the detection, diagnosis and mitigation. Finally, in order to make this product commercially viable, the EEM framework is being integrated into DeltaV.

In addition to detecting and diagnosing exceptional events, our EEM framework includes mitigation capabilities in which corrective measures are suggested and/or automated, whenever possible. Though the application has been relatively unexplored in the pharmaceutical domain, the benefits of having a system that diagnoses a fault and allows for corrective measures provide a strong incentive for its development and implementation [4].

References:

1. Venkatasubramanian, V., A review of process fault detection and diagnosis Part I: Quantitative model-based methods, Computers & Chemical Engineering. 2003. p. 293.

2. Hamdan, I.M., Anomaly Detection and Diagnosis in Continuous Pharmaceutical Manufacturing: An Ontological Approach, AIChE Annual Meeting. 2008.

3. Hailemariam, L. M., Application of Ontology in Pharmaceutical Product Development: The POPE Experience, AIChE Annual Meeting. 2008.

4. Gudi, R., Monitoring Bioprocesses, Pharmaceutical Manufacturing. 2008.