Expected Utilities for Decision Making and Steering of Measurement Processes

摘要

One main goal of structural safety monitoring is minimizing the risk of unexpected collapses of artificial objects and geologic hazards. But nowadays, the classical methodology in many engineering disciplines and mathematically founded decisions are usually based on probabilities and significance levels, which have more or less no reference to the practical applications. In this study, a new concept which is based on the utility theory is aimed to be introduced to the current methodology. The new developed methodology should make decisions from the views of reality, economy and safety. Therefore, it allows the consideration of risk/consequences in decision making processes in order to meet the real requirements. In this concept, the utility is a measure of preferences over some set of goods. The goods can be all kinds of things like products or services that satisfy human wants. The new methodology extends the idea and judges each possible decision with a utility value, and the final decision would be the sum of all possible outcomes. In this case, the final decision is not a predominant one from all possible outcomes, but an overall consideration of all possibilities. Compared to the classical approach, risk/consequences can be judged directly based on utility values, so that the new methodology is more straightforward and concrete. Furthermore, the new developed methodology can be extended for u alternatives, which fit better to the reality with usually more than two possible outcomes. The influencing factors are also analysed and discussed within the illustration and discussions of results at each developing step. Based on the new developed methodology, a particular steering process for the measurements is designed by adopting the advantages of decision making with the consideration of expected utilities. Finally, it leads for the whole monitoring project to the minimum risks/consequences. It is of crucial importance when the monitoring project has especially time limitation, tight budget, etc. A set of formulas for the determination of significance levels with the given utility values are established. Thus, the significance level is not an intuitive choice any more, but a mathematically derived value. Then, the defined significance levels for control and optimization of the measurement process are used. In this way, the overall risk of a measurement project can be minimized based on possible occurring consequences of a wrong decision. From all the derived results, it can be seen that it is possible to make more rational decision with the consideration of the expected utilities, and the methodology steers the measurement processes by reducing the risk at each monitoring iteration and finally leads for the whole monitoring project to the maximum benefits. The significance levels can be derived with the aid of utility values, which avoids the intuitive choices. The developed methodology is transferred to three different applications, and to steering of a slide slope monitoring process. Modelling of measurement uncertainty and improving the reliability of the utility values should be further studied and improved in future research work.