(204q) On the Design of Robust Sensor Networks

The sensor network design problem (SNDP) consists in selecting the set of process variables to be measured, which is optimal with respect to some specified criteria and simultaneously satisfies certain information requirements of the system under analysis.

 Most advances in the area of SND have been focused on the monitoring of processes operating under normal conditions. However the optimal location of instruments to effectively diagnose plant faults has central importance for safety, environmental protection and process economy.

 Regarding the SND formulations for fault diagnosis, Bhushan and Rengaswamy (2000) developed a strategy based on the use of signed directed graphs to find a set of instruments that ensures the observability and the highest resolution of all failures, under the assumption of the occurrence of single/multiple faults at a time. Then the same authors presented a reliability formulation which uses quantitative information of fault occurrence and sensor failure probabilities to determine sensor locations given a fixed budget of the instrumentation project (Bhushan and Rengaswamy, 2000). Later on, they analyzed the maximum-reliability and minimum-cost DPs and addressed a lexicographic optimization procedure to combine those two objectives (Bhushan and Rengaswamy^, 2002). On the other hand, Bagajewicz et al.(2004) presented a MILP formulation to obtain minimum cost SNs for the same kind of constraints, which are stated using matrix algebra concepts.

 The robustness of the network to uncertaintities/errors in the underlying cause-effect model and probability data were next considered by Bhushan et al.(2008). More distributed networks were preferred to incorporate robustness to modeling errors. Regarding the treatment of inaccurate probability data, the main idea was to ensure that the constraints involving uncertain data were far being active at the optimal solution.

 The occurrence of some particular abnormal events in chemical plants may originate significant economic losses, dangerous working conditions and huge environmental damage. Those events are considered as key faults in this work. They should be necessarily identified, even if one or more of the instruments that reveal the occurrence of the key fault are temporarily unavailable. With this purpose, the concept of Resolution Degree of a key failure is introduced as a constraint of the SNDP. Furthermore robust design formulations are provided and their application to case studies extracted from the literature allows the comparison of the new structural methodology with the existing ones.

 References

 Bhushan M, Rengaswamy R. Design of sensor network based on the signed directed graph of the process for efficient fault diagnosis. Ind. Eng. Chem. Res. 2000; 39:  999-1019.

Bhushan M, Rengaswamy R. Design of sensor location based on various fault diagnostic observability and reliability criteria. Comp. Chem. Eng. 2000; 24: 735-741.

Bhushan M, Rengaswamy R.  Comprehensive design of a sensor network for chemical plants based on various diagnosability and reliability criteria. 1. Framework. Ind. Eng. Chem. Res. 2002; 41: 1826-1839.

Bagajewicz M, Fuxman a, Uribe a. Instrumentation network design and upgrade for process monitoring and fault detection. AIChE J 2004; 50: 1870-1880.

Bhushan M, Narasimhan s, Rengaswamy R Robust sensor network design for fault diagnosis. Comp. Chem. Eng. 2008; 32: 1067-1084.