In the first chapter, we have proved that the gradient of the objective function at the working point is the static gain of the system. Obviously, the key to the problem is how to calculate this static gain. Generally, if the structure of the system’s dynamic model is known, the parameter identification method can be used to obtain parameters in it, and therefore the static gain could be calculate by means of the model. However, in real process, the dynamic model will change with time and variation of working point, and how to obtain the model structure and track the change of structure is a difficult task. On the other hand, the calculation of static gain indirectly by dynamic model may lead to large errors, and no theory can calculate such errors accurately at present.