Sensor deployment based on fuzzy graph considering heterogeneity and multiple-objectives to diagnose manufacturing system

In this paper, a methodology is presented to generate an optimized sensor deployment deciding sensor types, numbers, and locations to accurately monitor fault signatures in manufacturing systems. Sensor deployment to robustly monitor operation parameters is the corner stone for diagnosing manufacturing systems. However, current literature lacks investigation in methodologies that handle heterogeneity among sensor properties and consider multiple-objective optimization involved in the sensor deployment. We propose a quantitative fuzzy graph based approach to model the cause–effect relationship between system faults and sensor measurements; analytic hierarchy process (AHP) was used to aggregate the heterogeneous properties of the sensor–fault relationship into single edge values in fuzzy graph, thus quantitatively determining the sensor's detectability to fault. Finally sensor–fault matching algorithms were proposed to minimize fault unobservability and cost for the whole system, under the constraints of detectability and limited resources, thus achieving optimum sensor placement. The performance of the proposed strategy was tested and validated on different manufacturing systems (continuous or discrete); various issues discussed in the methodology were demonstrated in the case studies. In the continuous manufacturing case study, the results illustrated that compared with signed directed graph (SDG), the proposed fuzzy graph based methodology can greatly enhance the detectability to faults (from SDG's 0.699 to fuzzy graph's 0.772). In the discrete manufacturing case study, results from different optimization approaches were compared and discussed; the detectability of sensors to faults also increased from SDG's 0.61 to fuzzy graph's 0.65. The two case study results show that the proposed approach overcame the qualitative approach such as signed directed graph's deficiency on handling sensor heterogeneity and multiple objectives; the proposed approach is systematic and robust; it can be integrated into diagnosis architecture to detect faults in other complex systems.