Reza Dianat

detection, Identification and control of sensors for pressure, temperature, light, etc. are critical. In general, a faulty sensor can be inaccurate in identifying the said variables and affect how humans use the system. It is therefore important to identify and diagnose fault. This thesis examines faults that can affect the data. This faults can be divided into sensing error by the sensors themselves or network communication channel error which can be of wired, wireless, optical fiber, etc. fault to be investigated include Constant and random fault with constant mutation caused by the sensor itself and noise fault due to data passing through LoRa channel with Gaussian noise. The mean calculation is used to detect the constant , the entropy feature extraction method is used to detect the random fault with the constant mutation, and the ARIMA and ARMAX estimation models are used to detect the noise fault. MATLAB software evaluates these methods and calculates True Positive and false Positive for sensors for different variables. Finally, the number of broken bits per data pass through LoRa protocol for different expansion coefficients is examined. In this way, fault are detected, which are provided to the main control center to handle these errors.