On-board model based condition monitoring of an automotive spark ig- nition engine is still a challenging task for automotive industry. The diagnostic system aims to enhance fuel efficiency and to reduce harmful exhaust emissions. Among various subsystems of gasoline engine, air intake system holds prime impor- tance as it is responsible to ensure proper air and fuel proportions in combustion mixture. This subsystem exhibits highly nonlinear behavior due to its components like throttle body, intake manifold etc. Health monitoring of such nonlinear system cannot be performed by conventional diagnosis methods. That is why On-Board Diagnostic (OBD-II) standard kits do not have the provision to diagnose vari- ous air intake system faults. These faults include air leakages in intake manifold, clogged air filter, reduced throttle body efficiency and certain sensor faults. This manuscript presents a novel nonlinear health monitoring scheme based on sliding mode theory for on-board diagnostics of air intake system. Sliding mode theory is extensively used in fault diagnosis methodologies. Sliding mode observers based on nonlinear dynamics deliver robust platform for the estimation of un-measurable system variables. The estimation of such parameters can be exploited for fault diagnosis of dynamical systems. In this dissertation, second order sliding mode observers are designed for air intake system. The designed observers are used to estimate un-measurable and critical parameters/states. Five of the estimated crit- ical parameters are: frictional torque, combustion efficiency, volumetric efficiency, air filter discharge coefficient and throttle discharge coefficient. These parameters are estimated from a two state nonlinear model of gasoline engine based on inlet manifold pressure and rotational speed dynamics. These parameters are extremely helpful in engine modeling, controller design and fault diagnosis/prognosis. An- other contribution of this thesis is the development of virtual sensors for air intake system. Pressure dynamics are estimated from crankshaft sensor measurements and vice-versa. The outlined parameters and virtual sensors are used to moni- tor various functions of air intake system. These functions cannot be routinely sensed/monitored by any sensor. The estimation of afore-mentioned parameters has been conducted under healthy and faulty operating conditions to generate residuals. These residuals are evaluated to identify/classify any malfunction in air intake system. A detailed procedure for three fault diagnostic schemes have been discussed. These scheme require no extra sensor/hardware for their evaluation, only conventional on-board diagnostics (OBD) equipments are mandatory. The validation of novel estimation and diagnostic scheme is performed on production vehicle engine equipped with engine control unit compliant to OBD-II standards. It has been shown experimentally that the above discussed faults have been timely identified. The proposed fault diagnosis scheme has the potential for online im- plementation as it operates sample-by-sample on OBD-II measurements.