Method of Structural Monitoring for Systems with an Imperfect Sensor Network

Abstract

There are important objects behind which constant supervision (monitoring) is necessary. According to regulations, on such objects the structured systems of monitoring and management of engineering systems of buildings and constructions are subject to obligatory installation. They provide supervision over the basic systems of object, using a network of geodetic and geotechnical sensors. This network, being the basic source of data about an object state, itself has defects. Nowadays not monitoring of limiting values, but revealing of tendencies becomes the purpose of the stress-strain analysis. With highest resolution and highest recording rate of current instruments the small deformations, caused by the daily temperature changes, can observed. These disturbances are present in the readings of all sensors and, therefore, may appear in the cross-correlation relations. If correlation dependences of a sensor or group of sensors with the majority of others is weakened it testifies to abnormal behavior either sensor, or object. Propose the method of supervision, which map changes in object-network state (behavior) into subspace of correlation characteristics of groups (associations) of sensors with strong correlations. The method provides supervision of changes in object-network state, it is insensitive to the spread in biases and gains of sensors, up most of defects in the network, at the same time, it is identifies and locates drifting and freezing sensors. Presents the results using the method for long-term monitoring constructions covered sports arena.