A Simulation-Based Optimisation Method for PLC Systems

Doktorsavhandling, 2012

Tuning of process parameters such as time constants, cam values, velocities and robot paths, in order to increase capacity utilisation, is today a challenging manual on-line task in many automated manufacturing systems. On-line methods interfere with production and will cause unwanted downtime, which indeed reduces capacity utilisation. The literature offers virtual manufacturing and simulation-based optimisation as an approach, but without handling time-synchronised control functions, e.g. motion and feedback control. This thesis presents a simulation-based optimisation method for PLC systems, able to handle time-synchronised control functions. A programmable logic controller (PLC) is a commonly used type of industrial control system that is capable of handling all control functions in a manufacturing system. The approach presented in this thesis is, however, independent of the type of control systems used. Hence, PLC is used as a general name for all control systems, managing all discrete event, continuous feedback, motion, supervisory, as well as safety control functions. A generic simulation-based method for tuning of process parameters has been formulated. Various goals are attainable by a multi-criteria optimisation approach. The idea is to use a time-synchronised hardware-in-the-loop simulation including real PLCs. By this approach, the method provides a distinct advantage as it involves all complex control functions by using the real PLC code. An additional benefit is that all tuned process parameters can be directly transferred to the PLCs in the manufacturing plant. To achieve a feasible simulation-based optimisation method for PLC systems, a new Combined Lipschitzian and Simplex (CoLiS) optimisation algorithm has been established. Complex control functions in industrial manufacturing systems cause conditions such as highly non-linear functions with multiple local optima, a considerable number of parameters and long evaluation times. All these conditions are managed by the non-gradient global CoLiS algorithm. The CoLiS algorithm starts with a global search and then switches over to local convergence. Additionally, all local optima determined during the global search are selected and then constitute starting points in separate local optimisation instances. To verify the formulated method’s suitability in industrial applications, and the effectiveness of the new CoLiS algorithm, an optimisation case study has been performed. Improved production performance, both in terms of increased production rate and smoother robot motions, was reached in an automotive sheet-metal press line.