Cognitive parameter adaption in regular control structures using process knowledge for parameter adaption

The colour control system of an offset printing machine is one example, where modern information processing technologies allow an improved process control and higher resource efficiency. It is not possible to measure the printing quality during production start. So no regular closed loop control can be used. For better system behaviour a simulation model is integrated to calculate the printing quality at any time. To get an optimal process performance, a high simulation quality must be ensured, which includes a compensation of process simulation inaccuracies as well as variable influences. Therefore a cognitive system is installed, which measures the most important influences like the used paper and many other process parameters. After each production the right model parameters will be calculated by identification algorithms. So a data set with influences and parameters is available. For the next production run the best-fitting parameters for the simulation model can be calculated by a Neural Network. Additionally wear and deposits, which change the machine's performance, can be compensated. The simulation accuracy and the process control quality rises, which enables a faster run-up. Savings of paper, ink, energy and time allow an economic application of this control concept.