The high complexity of manufacturing processes and the continuously growing

amount of data lead to excessive demands on the users with respect to process monitoring, data analysis and fault detection. For these reasons, problems and faults are often detected too late, maintenance intervals are chosen too short and optimization potential for higher output and

更多

increased energy eciency is not suciently used. A possibility to cope with these challenges

is the development of self-learning assistance systems, which identify relevant relationships

by observation of complex manufacturing processes so that failures, anomalies and need for

optimization are automatically detected. The assistance system developed in the present work

accomplishes data acquisition, process monitoring and anomaly detection in industrial and

agricultural processes. The assistance system is evaluated in three application cases: Large

distillation columns, agricultural harvesting processes and large-scale sorting plants. In this

paper, the developed infrastructures for data acquisition in these application cases are described as well as the developed algorithms and initial evaluation results.

收起