Performance enhancement of manufacturing processes through mathematical models and observational inference of process physics
Abstract
Methods for solving performance improvement problems in manufacturing processes, such as, for example, a need to improve product quality, or to reduce cycle time are discussed. Two approaches are adopted. First, when the process output and the performance measure can be related to the control inputs and process parameters through a mathematical model, such problems can be posed as those of functional minimization. In this case, the control input that minimizes a specified performance index, subject to the constraint of the process model, is sought and determined mostly using convex minimization techniques. As an illustration of such methods, we present a path planning problem in automated spray coating, where the objective is to determine the path of a spray gun that causes the most uniform coat, when coating a given surface. Nevertheless, as is the case in many manufacturing processes, such mathematical relationships between the control inputs and process output are difficult to formulate. This motivates the second approach, where recourse is sought to extensive experimentation, whereby the process dynamics are thoroughly studied and characterized. Though the formal optimization of process parameters is precluded by the absence of a mathematical model, performance improvement is obtained by feedback control laws that are derived based on a good understanding of process physics. As a typical example, we present studies on improving the quality and machining time of an electrical discharge machining (EDM) process. The complexity of the process prevents the specification of useful mathematical models; nevertheless, by performing appropriately designed experiments, we characterize different phases of the process dynamics. By recognizing the phase that contributes to reduction in quality and increase in cycle time, we construct a nonlinear feedback control law driving a fast actuator that appropriately regulates the process in this phase, achieving the desired improvement. The experimental characterization, and the control of the EDM process are discussed in the context of an actual industrial implementation, where the above feedback control law results in a significant reduction in the machining time, while maintaining a high product quality.
Degree
Ph.D.
Advisors
Kashyap, Purdue University.
Subject Area
Electrical engineering|Mechanical engineering|Industrial engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.