(430d) Customer Feedback Control: Design and Implementation on a Nanocomposite Manufacturing Process

Actual feedback from customers (in the form of qualitative/quantitative data) is the most directly relevant indicator of manufactured product performance in end-use—whether or not the product actually meets customer requirements as indicated by the manufacturing process specifications. Consequently, it is necessary to use such customer feedback data directly and explicitly within a control scheme in order to take rational corrective action when the product fails to perform in end-use as specified. However, to date, such a control scheme does not exist, as a result of a variety of challenges. In this presentation, we address these challenges and discuss the design and implementation of such a customer feedback regulator, which utilizes customer feedback information (z) (fundamentally binary, since the product either performs acceptably, or not) to determine appropriate target set-points for product properties (w). Specifically, we introduce techniques for modeling the relationship between measured product properties (continuous independent variables, w) and the customer feedback data (binary dependent variable, z); we also introduced concepts of “target/design probability of acceptance” and “achievable probability of acceptance”; and then develop theoretical results for determining optimal adjustments to the product property set-points needed to achieve the target probability of acceptance, given property measurements and customer feedback data. The performance of the control scheme is illustrated in simulation via an example involving the manufacture of an industrial polymer nanocomposite for a packaging film application. The simulation results show how, in response to an unmeasured disturbance that compromised the product performance in end use, the customer feedback regulator sequentially adapted the operating conditions and successfully increased the probability of acceptance to the desired level.