JP3005732B2 - Device control method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of controlling a device which enables a process control with optimized data by constructing a neural network.
For example, the present invention can be applied to a temperature controller for controlling the temperature of a furnace.

[0002]

2. Description of the Related Art In the prior art, in order to use the data of the result of auto-tuning as it is for process control, optimum values of P (proportional band), I (integration time) and D (differential time) parameters of different control systems are used. Conversion was difficult. That is,
A so-called on-line learning is carried out in each case based on the signal from the control object, and auto-tuning is performed in accordance with the signal. Thereafter, based on the value obtained as a result of the auto-tuning, the PID is executed following the auto-tuning. Is calculated, the transfer coefficient of the control system (the parameter that specifies the transfer function of the control system)
Value) is fixed and the optimal PID for all control systems
It was impossible to set a constant. Therefore, since the most average value is set in a compromised manner, even if a good control result is obtained in one system, overshoot occurs in the other system or the start-up is delayed, so that the PI
It was necessary to adjust the D parameter.

[0003]

As described above, the problem to be solved by the present invention is that the execution of online learning every time a signal from a control target is received complicates the control. If the transfer coefficient of the control system at the time of calculation is fixed, the optimum P
This makes it impossible to set an ID constant.

[0004] The present invention has been made in view of the above-mentioned points, and an object thereof is to construct a neural network which has learned an excellent control result which is almost completely fine-tuned in advance, and this neural network is constructed. An object of the present invention is to provide a control method of an apparatus which enables a different system to be optimally controlled with a network-optimized data.

[0005]

According to a control method of an apparatus of the present invention, a PID parameter obtained by automatic tuning is used as input information and a PID parameter finely adjusted by a manual operation so as to be optimal for a control system is output. As information, a neural network is created by repeatedly performing auto-tuning and fine-tuning of PID parameters with a plurality of control systems having different transfer coefficients. Based on this neural network, PID parameters obtained by auto-tuning are used as input information. Optimal P
The ID parameter is determined.

[0006]

According to the control method of the device of the present invention, a neural network is constructed from data obtained by the automatic tuning and the fine adjustment. Incorporating this constructed neural network into a control device such as a temperature controller,
Based on this, PID parameters are determined, and optimal control can be performed.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a method for controlling an apparatus according to the present invention will be described with reference to FIGS. First, auto-tuning by many different control systems and sampling of control results are performed. That is, auto-tuning is performed in a system having a certain transfer coefficient . Then, PID parameters are calculated and controlled based on the result of the auto tuning. The parameters are finely adjusted manually while observing the control results, and the optimum parameters Pm, Im, and Dm obtained by the manual fine adjustment are obtained. The amount of change, which is the difference between the optimal PID parameters obtained by manual fine adjustment, that is, Pm, Im, and Dm, and the PID parameters obtained by auto tuning is calculated, and the PID parameters obtained by auto tuning are calculated. The PID parameter and the amount of change obtained by the fine adjustment are used as data for the neural network. In this case, the PID parameter obtained by the auto tuning may be taught as input information to the neural network, and the PID parameter obtained by fine adjustment may be taught as output information, or the change amount may be taught as output information. May be. This operation is performed by a different control system, and a large amount of data is sampled. This state is shown in FIG.

Next, after the above-mentioned PID parameters and the like for different control systems are obtained, these sampling data are taught to a neural network, and the neural network
Upon completion of the learning of the network, a neural network is constructed in advance. Next, the completed neural network is incorporated into another control device that does not originally have a neural network learning function, for example, a temperature controller.
Then, using this neural network, the optimal P
The ID is determined and the device is controlled.

Next, an example of a case suitable for a temperature controller for controlling the temperature of a furnace will be described. In FIG. 2 , 1 is a temperature controller having a built-in neural network, 2 is a furnace to be controlled, 3 is a heater of the furnace 2, 4 is an amplifier for driving the heater 3 based on the temperature controller 1, and 5 is a furnace of the furnace 2. This is a temperature sensor for detecting a temperature, and data of the temperature sensor 5 is inputted to the temperature controller 1.

In order to actually perform the control, first, as shown in the flowchart of FIG. 3, auto tuning is performed at the discretion of the user (steps S1, S2). When the auto tuning operation is completed, the PID inside the temperature controller 1
The parameters are calculated (step S3), and the finely adjusted parameters Pm, Im, and Dm are calculated, and the calculated values are used as PID parameters (step S4). Then, the temperature of the furnace 2 is adjusted by PID calculation based on the parameters (step S5).

FIG. 4B is a diagram showing a control result by the conventional auto tuning, and FIG. 4A is a diagram showing a control result by the present invention.

[0012]

According to the control method of the apparatus according to the present invention, the following effects can be obtained because the apparatus is configured as described above.

1. According to the present invention, auto-tuning
Input and output PID parameters obtained by
Since a neural network is built as information,
The information amount is small, and an optimum PID can be obtained. Accordingly, good control can be achieved with a simple configuration, and effective parameters can be obtained with minimum cost and time, so that there is an effect that it is particularly suitable for a small-sized control device. 2. Conventionally, P which was manually operated by humans
Fine adjustment of the ID constant is automatically performed. Therefore, the adjustment time is shortened, and a good control result can be easily obtained even by a person without specialized knowledge.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram showing a procedure for constructing a neural network according to the present invention.

FIG. 2 is a block diagram showing an example in which the present invention is applied to a furnace temperature controller.

FIG. 3 is a flowchart showing a state of calculation and control of furnace parameters.

FIG. 4 is a characteristic diagram showing an example of the control result.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Temperature controller 2 Furnace 3 Heater 4 Temperature sensor

Claims (1)

(57) [Claims] 1. A PID obtained by auto-tuning
Parameters as input information, as output information the PID parameters finely adjusted manually to optimize the control system, and a fine adjustment of the auto-tuning and the PID parameters in a plurality of control systems having different transmission coefficients Repeat to create a neural network,
A method of controlling an apparatus, comprising: determining an optimal PID parameter using a PID parameter obtained by auto-tuning as input information based on the neural network.