(302b) Customer Feedback Controller Design and Implementation

While customers are primarily concerned about product
attributes and especially product performance in end-use, manufacturers on the
other hand can only determine product attributes indirectly, if at all, during
the production process. Since it is impossible to determine actual
product performance until the product is actually deployed in its end-use, in
the absence of any direct measurements of surrogate properties, the only option
has been to treat measurable process variables (y) as acceptable indicators of
product attributes. In traditional manufacturing processes, therefore, the most
common control configuration consists of a base regulatory scheme for
regulating these process variables (y) by adjusting process manipulated
variables (u).  But this is not always effective. In some cases, it has become
customary to augment this basic scheme with a higher-level scheme whereby the
set-points for the process variables are manipulated to control product quality
variables (q).  More recently, Garge et al. [1, 2] proposed an inference-based
control scheme that extends these higher-level schemes further, achieving
end-use property (w) control by manipulating the set-points for the product
quality variables. The overall control scheme, which involves a three-layer,
multivariable cascade control structure, has been demonstrated on an industrial
pilot scale process. However, even a perfect implementation of such advanced
multi-layer cascade control schemes cannot guarantee actual product
performance. Especially with tailor-made products designed for specific
performance, consistently meeting customer demands requires a comprehensive
control scheme that will take actual product performance in end-use explicitly
into consideration. As with standard feedback control where actual measurements
of desired process outputs are used to adjust process inputs, it is necessary,
under such a scheme, to obtain customer feedback on product performance and
employ such information directly to take rational corrective action if the
product does not perform in end-use precisely as designed.

In ref [3] we proposed such a ?customer
feedback regulator? which utilizes customer feedback information (z) to
determine appropriate set-point targets for product attributes (w), to be
implemented in conjunction with the inference-based end-use property controller
of Garge et al. [1, 2]. The key issues in designing this customer
feedback regulator were discussed and illustrated with an example. However, the
proposed method is restricted to cases where the relation between π(w),
the probability that z=1, and the product attributes (w), is sigmoidal.  In
this work, we extend the results of [3] to the more general case (where the
relationship between π(w) and w is not restricted to be sigmoidal) and
propose a control strategy that is generally applicable. The control scheme's performance
is illustrated with a simulation case-study of a prototypical industrial
plastic manufacturing process.

Reference

1.      S.C.
Garge, M. D. Wetzel, and B. A. Ogunnaike, On the Control of Reactive Extrusion
Processes, AIChE Conference, 2006;

2.      S.C.
Garge, M. D. Wetzel, and B. A. Ogunnaike, An Inference-based Scheme for
Controlling Product End-use Properties of Reactive Extrusion Processes, In
preparation (2009).

3.      S.C.
Garge, and B. A. Ogunnaike, Customer Feedback Control for Assuring Acceptable
Product End-Use Performance, AIChE Conference, 2007.