Distributed control synthesis for manufacturing systems using customers' decision behaviour for mass customisation.
The mass customisation manufacturing (MCM) paradigm has created a problem in manufacturing control implementation, as each individual customer has the potential to disrupt the operations of production. The aim of this study was to characterise the manufacturing effects of customers’ decisions in product configuration, in order to research steady state control requirements and work-in-process distributions for effective MCM operations. A research method involving both analytic and empirical reasoning was used in characterising the distributed control environment of manufacturing systems involved in MCM. Sequences of job arrivals into each manufacturing system, due to customers’ decisions in product configuration, were analysed as Bernoulli processes. A customer model based on this analysis captured the correlation in product configuration decisions over time. Closed form analytic models were developed from first principles, which described the steady state behaviour of flow controlled manufacturing systems under generalised clearing policy and uncorrelated job arrival sequences. Empirical analysis of data sets achieved through discrete event simulation was used in adjusting the models to account for more complex cases involving multiple job types and varying correlation. Characteristic response surfaces were shown to exist over the domains of manufacturing system load and job arrival sequence correlation. A novel manufacturing flow control method, termed biased minimum feedback (BMF) was developed. BMF was shown to posses the capability to distribute work-in-process within the entire manufacturing facility through work-in-process regulation at each manufacturing system, so as to increase the performance of downstream assembly stations fed from parallel upstream processing stations. A case study in the production of a configurable product was used in presenting an application for the models and methods developed during this research. The models were shown to be useful in predicting steady state control requirements to increase manufacturing performance.