Due to the poor instantaneity and adaptability

the current methods for sensors fault detection are still less applied in practical project. Based on the decentralized intelligent control system of building

a decentralized algorithm for the faults of slow drift in sensors is proposed. The basic idea of the method is to detect the consistency of variables under the equality and inequality constraints established through physic principle. In the decentralized control system

each sensor is fitted with a decentralized controller such that it become a sensor-agent

which can communicate with other sensors and thus operate collaboratively to check out the faulty nodes assited by Hopfield network. Unlike traditional centralized and distributed algorithms

this algorithm does not need to build a central node

regardless of increasing hardware or time redundancy

with good parallelism and simplicity. In the case of secondary chilled water pump system

this method is verified effective to detect the fault of sensors through simulation result.