ABSTRACT
Plan: Define the problem statement. The company was growing amino acids in a multistep reactor process for ultimate delivery to its pharmaceutical customers. Contaminates were frequently found in the batch at the end of the manufacturing process. Do: Identify necessary and sufficient conditions for the mode of failure. Using Substance Field Modeling, several possible contamination entry ports were identified (such as valve pumps for adding amino acids, nutrients, or stabilizers). Because the reactor system is closed (or controlled), these entry ports were determined to be the only possible pathway for contaminants to enter the system. Study: Study the system to identify conditions present during failure, and eliminate potential causes until mode is verified. Each condition for contamination was evaluated over the processing of many amino acid batches, and only normal
n
n
n
104
trace quantities of bacteria were identified, so no introduction of assignable contamination was verified. But batches were still failing because of measured contamination levels at the end of the process. With all the entry points in the closed system ruled out as contamination pathways, the company had no choice but to posit that, maybe, employees were responsible for sabotaging the batches. Given their FDA-approved quality system and other failure detection methods did not reveal the cause of contamination, the company considered installing cameras to monitor employee behavior. Prior to this, however, the cumulative effects of the system were analyzed using Substance Field Modeling (SFM). This is a natural progression in TRIZ: first consider the additive (independent) dynamics of a system, and then consider the cumulative (interdependent) dynamics. Act: Assign cause, and create cause and corrective action. In biological terms, the previous work pointed to the idea of a “bacterial quorum,” whereby trace contaminants from different parts of a system can combine and interact to create an overall contamination level that is greater than the sum of its parts. After further testing and evaluation using SFM, it was discovered that trace contaminants were, in fact, interacting to produce a quorum adequate enough to destabilize the batch. Bacterial counts were taken before introduction for each component entry point, and a continuous bacterial monitoring system was established in real time for the entire process. After two or three low bacteria count components were added, the bacterial level escalated rapidly to the point of failure. Although each step of the process functioned without flaw, interaction inside the reactor was causing the failure. Corrective actions were tested and installed, and the problem was solved.
