1. The proposed architecture is shown in Fig. 2.11At the lowest level, we have a distributed network of HAS's which have knowledge of the process model in which they partake. Each HAS detects normal and abnormal behaviors of itself and its measurand/environment using information in its own domain. It also provides necessary information for sensor fusion to both the lower HAS-level and the middle process-level. The lower HAS-Level sensor fusion aims at verifying whether the sensor works well independent of other sensors and provides reasonable behavior interpretations for the sensor and measurand. This is primarily a function of the HAS model. The HAS-Bus-Level (Process-Level) sensor fusion aims at combining higher-level qualitative information from all HASs in the Sensor Bus associated with the same process. This operation further improves the interpretations of behaviors of each sensor and measurand. A final Process-Bus-Level fusion combines qualitative information linking associated processes. Again, this information further improves interpretation of behaviors of the sensors and measurand. It is noted that improving interpretations implies learning at the sensor and process levels and updating knowledge bases. The overall network should evolve, as