Evaluation of algorithms for temperature estimation in a crankcase

Ongoing research in emission reduction requires accurate load detection for combustion engines with a limited number of sensors. Therefore, fast estimation of load temperature is essential. Temperature measurements are influenced by the thermal properties of the sensor itself as well as its position and mounting method. In this paper, the transient thermal behavior of the engine's crankcase and the temperature sensor for load detection is modeled as a low-pass transfer function. The unknown parameters of the transfer function are identified from experimental measurements. Maximum likelihood estimation and Kalman filtering are used to estimate the original temperature from disturbed measurements. Estimator performance is evaluated via simulations of randomized test scenarios in a Monte Carlo fashion. Influences of the model errors, the measurement noise and the estimation window-time are investigated.