Neural Network System

The neural network misfire monitor uses a dedicated microprocessor in the PCM along with crankshaft position, (36-tooth wheel or 40-tooth wheel on a V-10), camshaft position, and engine load to determine engine misfire. A neural network is a different way of computing that uses a large number of simple processing elements with a high degree of interconnection to process complex information. The processing elements have adaptive characteristics (coefficients) that must be learned through a process called training. During training, the network is fed a sample set of data that consists of the inputs along with the desired output (misfire/no misfire). The network coefficients are recursively optimized so that the correct output is generated from the set of inputs and error is minimized. Once the coefficients have been learned, the network can process real data. The neural network size is 23 nodes and 469 coefficients. The engine off natural vacuum evaporative system monitor also uses the same microprocessor. Profile correction software is also used on neural network system to learn and correct for mechanical inaccuracies in crankshaft tooth spacing under de-fueled engine conditions (requires up to three 60 to 40 mph no-braking decelerations after Keep Alive Memory has been reset). These learned corrections improve the high RPM capability of the monitor. The misfire monitor is not active until a profile has been learned.