We propose an inference-based method to extract combustion indicators for closed loop combustion control and diagnosis. Two main combustion indicators (CI) are computed: the energy released during combustion and the middle of combustion. The proposed method rely on the use of the engine block transfer function combined with Bayesian inference. The CI extraction method is divided in two steps: (1) off-line analysis, which consists in characterising the knock sensor and approximating the engine block transfer function and (2) on-line analysis where a cost function given by the Bayesian inference is used in addition with vibration signals and thermodynamical models for CI estimation.
展开▼