Real-time identification of the draft system using neural network

Making a good model is one of the most important aspects in the field of a control system. If one makes a good model, one is now ready to make a good controller for the system. The focus of this thesis lies on system modeling, the draft system in specific. In modeling for a draft system, one of the most common methods is the “least-square method”; however, this method can only be applied to linear systems. For this reason, the draft system, which is non-linear and a time-varying system, needs a new method. This thesis proposes a new method (the MLS method) and demonstrates a possible way of modeling even though a system has input noise and system noise. This thesis proved the adaptability and convergence of the MLS method.