Although classical control and modern control look completely different, they are essentially the same. Both of them establish the mathematical model of the controlled object, and then design a controller in the form of the mathematical model to achieve the control purpose. The only difference is that the former establishes a differential equation or pid control, while the latter establishes a state equation, similar to the linear differential equations in mathematics. Optimal control, robust control and other issues, whether based on classical control theory or modern control theory, are inseparable from the essence-the mathematical model of the controlled object must be established, including the foc control proposed in 2002.

Intelligent control adopts a new way of thinking. It adopts the way of human thinking, establishes a logical model, and uses a control method similar to the human brain to control.

I happen to be doing this homework, so take my homework as an example to illustrate.

In order to control the water temperature of water heater, in the traditional control method, whether it is to establish differential equation or state equation, it is necessary to establish mathematical models between heating power and water temperature, and between heating power change rate and water temperature change rate. Whether seeking the optimal control method or extracting the sensitivity of variables for robust control, a mathematical model is always needed.

Intelligent control takes a completely different approach. For example, with a fuzzy controller, we first blur all the variables, and change the water temperature, water temperature change rate, heating power and heating power change rate into linguistic variables through a certain algorithm, such as describing them as large, relatively large, slightly larger, just right, slightly smaller, relatively small and very small. Then make a rule base like people, if the water temperature ... and the rate of change of the water temperature ... and then heat the power ... and then make a table of the rule base and the fuzzy relationship through some algorithms, and input it into the flash of the computing chip to work.

For another example, using artificial neural network, we simulated a computer neural network by imitating the human brain. Each artificial neuron contains many weighted axons, a cell body with excitation judgment and several dendrites with output function, and then many artificial neurons are connected according to a certain logical division. Then, a large amount of information collected by the water heater is added to the input and output terminals respectively, and the connection weights and excitation factors between artificial neurons are changed through a certain algorithm, so that all known training quantities can finally meet this artificial neural network. Under the condition of sufficient training,

Generally speaking, ordinary control is completely based on mathematics, while intelligent control is based on logical rules, using human thinking model or empirical model and even imitating human neural network for training and learning.

The advantages and disadvantages are obvious. Whether it is classical control or modern control, there is an obvious advantage-the stable point is very accurate, even if the overshoot and the fastest response time are flawed, it is always extremely accurate in the equilibrium position. However, no matter whether Zadeh or mcdaniel's theory is used in intelligent control, there are inevitably inaccurate components, but it is no doubt that many controlled objects are not so easy to establish equations, and even if many controlled objects can establish equations, they cannot be discussed in the current mathematical theory. For example, a boiler master often decides how much coal to add according to the temperature and whether the furnace wall is hot. It is easy for him to sum up experience, but it is obviously impossible to ask the boiler master to establish an equation.

When used in practical engineering, it is often a combination of intelligent control and ordinary control. For example, in a control model, first make a decision algorithm to judge whether it reaches the equilibrium position, and if not, use an intelligent controller to make it respond quickly. After the overshoot and other parameters reach a certain range, only the classic controller or the modern controller can be switched to run.

The above is only a summary of engineering and research experience, which can be written as material and refused to be reproduced without permission.