Abstract
A diesel particulate filter system is very effective for reducing the amount of particulate matter in the exhaust emission of diesel engines. This system catches particulate matter in its filter and therefore, with time, the filter will be fully loaded such that it cannot be used. The filter must then be regenerated by burning the accumulated particulate matter. To effectively control the filter regeneration, an accurate and reliable estimation of particulate accumulation is required. However, due to unknown factors in filter dynamics and operating conditions, it is not easy to estimate particulate accumulation analytically. A map of particulate accumulation, can actually be constructed by conventional methods, nevertheless it requires such a large amount of data that it is very difficult to apply. To overcome this problem, it is proposed, in this work, to employ a feedforward neural network which is widely known for its capability to model nonlinear systems. The experimental results show that particulate accumulation can be estimated with the desired precision for a wide range of operating conditions. This subsequently makes regeneration control flexible enough that the system can easily be used.
