On restart of automated manufacturing systems

Doctoral thesis, 2015

Highly automated manufacturing systems have gained industrial popularity for their ability to combine high product volumes with high product quality. The high cost of investment in combination with many linked manufacturing systems in a factory, requires that the production runs smoothly with high utilization of the resources and that stoppages are avoided. One major reason for stoppages is the occurrence of errors. A wide variety of possible faults, such as badly fixated parts, broken actuators, and teething problems in the system, may cause errors that lead to an unsynchronization between the control system and the physical system that consequently lead to production stoppages. The succeeding error recovery is often a complex and thereby time consuming process that typically requires operator involvement. To plan for restart after errors already during the development of the system would therefore greatly support the online restart process and reduce the time the production is undesirably stopped. The common industrial practice to deal with such non-intended progress is to extend the control system with tailor-made solutions to account for foreseen errors. This extension is both time consuming and there is no guarantee that all relevant errors are handled. To support both offline and online work with restart of the production, this thesis proposes a method for automatic calculation of restart states. These restart states are states in the control system from where it is correct to resynchronize the control and the physical systems so that the production can be resumed, irrespectively if the error is foreseen or unforeseen. The method aids during the development of the system by letting the developer focus on modeling the nominal production and on specifying (un-)desired behavior during the restarted production, and then automatically retrieve the correct restart states for all control states. The production can thus be designed for restart. Based on these precalculated restart states, the online restart process is reduced to a semi-automatic process where an operator can be supported with instructions for how to correctly resynchronize the control and the physical systems in a selected restart state. In addition to a thorough theoretical presentation of the supervisory control theory based method, the thesis also describes a proof of concept implementation in a lab manufacturing system and a validation on an industrial windscreen mounting station. The implementation shows that restart after unforeseen errors is enabled and exemplifies the operator support during the online restart process. From the validation it can be concluded that the underlying model is sufficient and that employment of the proposed method would have improved the restart processes after historical production stoppages.