Integrating Exception Handling in Machine Development

Sammanfattning: In modern batch plants, alarm floods overwhelming the operator is a common problem of ever increasing severity. As plant hardware increase in complexity and performance, measures such as downtime and meantime between failures, increase in importance and the need for a well-functioning exception handling routine therefore grows critical. In order to minimize downtime and hence boost production efficiency, it is important that faulty or unexpected behaviour is noted early and diagnosed accurately. If it is, the machine or its operator can deal with it fast and accurately, perhaps even while production continues. If it is not, however, it basically means two bad things, the first being unnecessary production stops and the second a flooded alarm list. The alarm list is the screen on which all exceptions are listed and which is meant to tell the process operator what is wrong and what he should do about it. If no care is taken about it, the list will be flooded since one key exception will cause several others, leaving the operator with the quite unpleasant task of identifying what went wrong and how he should fix it. This master's thesis, carried out at Tetra Pak in Lund, Sweden, presents a well-structured exception handling method and a way of linking it to a workflow. It uses advantages of reusability, modularisation and linking to almost any structural model such as a Function Means Tree (used in WCE) or a UML-model. The linking has several advantages minimizing the work and improving evolutionary possibilities. Examples are being made for illustrational purposes but no implementational efforts or issues are addressed. A well-defined structure for information flow is also suggested to provide modular thinking that, for example, prepares the possibility to collect data for statistical analysis and such.