Plant performance evaluation in complex industrial applications

In large-scale industrial plants, investment and maintenance decisions should be based on quantitative knowledge about the factors that reduce overall performance. This is impossible without sound and objective methods for producing the necessary performance information. The need for performance assessment spans from low-level control loops to the business process level. Since the number of control loops in an industrial plant can be in the order of hundreds or thousands, it is infeasible to monitor their performance manually. This thesis proposes a set of performance indices for two-dimensional web processes, found, for example, in the paper, metal, and plastic industries. A scaling function approach where different performance indices are scaled to the interval 0…100 is introduced in order to enable the creation of a hierarchical performance assessment framework. A simulation example and two large-scale industrial applications illustrate the creation and usage of this performance evaluation framework. When the values of all performance indices are in the same range and are interpreted in the same way, evaluating performance becomes less demanding and time-consuming. This thesis demonstrates that a number of low-level performance indices can be aggregated into a small amount of high-level indices which still contain the relevant information. To assess process performance, it suffices to monitor the high-level indices, and only if an anomaly is detected, there is need to investigate low-level performance information. This thesis shows that by processing and combining low-level performance data coming from different sources, it is possible to obtain performance information that is easier to interpret for humans. This information can then be utilized in decision making. The performance assessment systems created for the industrial applications are in production use at the moment of publishing the thesis