Present day high end vehicles contain as many as 100 microprocessors and have more than 100 million lines of code . Electronic Control Units (ECUs) use numerous sensors spread across various sub-systems to perform a variety of functions such as stability control, adaptive cruise control, power train control, autonomous driving, etc. The accuracy of sensors degrades over the vehicle lifetime. The control logic of vehicles uses sensor measurements to detect the vehicle state and give appropriate commands to control the vehicle. Faults in sensors cause wrong feedback to be sent to the control logic and subsequently commands given by the control logic become faulty. Thus the effect of a sensor fault can be observed in subsystems controlled by the control logic. So advanced diagnosis techniques are required to detect the sensor degradation. In this work, a comparison is made between two different automotive sensor health monitoring methods used to diagnose sensor faults. The methods that are compared are: lookup table based method and machine learning (using multiple linear regression) based method. The comparison will enable one to choose the appropriate method by considering the attributes required for a particular application like accuracy, amount of computation required, simplicity, etc. Sensor faults are detected by comparing the current sensor output with a nominal no-fault sensor block output at the particular operating point of the system. The lookup table method was implemented in Matlab while the machine learning method was implemented using R programming language. The proposed methods are illustrated on the case example of motor fault in a Hybrid Electric Vehicle to analyze their efficacy. They have been found to be valid for all drive cycles and can be used for diagnosing sensor faults in real time.