JP2000257839A - Monitoring/controlling method for melting furnace slag discharge outlet - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a monitoring/controlling method capable of easily controlling a slag discharge outlet where modeling with numerals is impossible and wasteful time is long without relying upon judgement of an operator. SOLUTION: A future behavior after the passage of wasteful time is predicted using a neural network taking as an input signal measurable process data of a melting furnace. Predicted values of the resulting process data are combinmed with each other as parameters together with past and present process data, and an expert system judges stability of the slag discharge outlet. When the situation is judged stable, a furnace is operated at a rated value while when unstable, the rated value is altered to operate the furnace such that it is stabilized.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a method for monitoring and controlling a slag discharge port of a melting furnace.

[0002]

2. Description of the Related Art As one of the recent sludge treatment technologies, a system for reducing the volume and detoxifying slag by melting sludge incineration ash is becoming widespread. However, the melting characteristics change due to the change in the properties of the ash to be charged, and it is not easy to stabilize the condition of the slag discharge port (gate). That is, even if the furnace temperature and the furnace pressure are kept constant, the properties of the ash to be charged change, so that the gate is blocked or the slag is solidified vertically at the lower part of the gate so that stable discharge cannot be performed. is there. For this reason, a monitoring camera is installed at the lower part of the melting furnace to monitor the sprue. However, since it is impossible to predict when stability will be lost, the burden on the operator is not small.

[0003] The inventors of the present invention have studied the use of a numerical model of a melting furnace to predict its behavior from process data such as furnace temperature and furnace pressure, and to perform automatic control of the melting furnace. However, some process data, such as the properties of ash, heat of dissolution, and heat of taking out slag, required for numerical modeling cannot be measured in real time, so it has proved difficult to numerically model the melting furnace. . Moreover, since the melting furnace has a considerably long dead time, for example, 15 to 30 minutes, from operating the combustion conditions to changing the condition of the gate, it is necessary to perform appropriate control based on the measured process data. In this respect, it was difficult. Therefore, conventionally, the furnace operation is performed based on the judgment of the operator, and it is inevitable that there is an individual difference in the judgment.

[0004]

SUMMARY OF THE INVENTION The present invention solves the above-mentioned conventional problems, and includes a process data that cannot be measured. Therefore, a numerical model cannot be formed. Slag discharge (gate) of difficult melting furnace
Has been made in order to provide a method for monitoring and controlling the melting furnace slag discharge port which can be controlled stably without relying on the judgment of the operator.

[0005]

SUMMARY OF THE INVENTION The present invention has been made in order to solve the above-mentioned problems. The present invention relates to a method for neurally predicting the future behavior of a melting furnace after a dead time which cannot be numerically modeled because it includes unmeasurable process data. The stability of the slag outlet is determined by an expert system that combines the predicted values of the obtained process data with the past and present process data as parameters together with the prediction by the network. It is characterized by operating a furnace. It is preferable to adopt a self-learning type neural network and update the process data in order to ensure generalization. Further, as the parameters of the expert system, the change of the gate opening degree, the falling state of the solidified slag, and the change of the gate temperature can be selected.

According to the present invention, the behavior of a melting furnace that cannot be numerically modeled by using a neural network is predicted from measurable process data, and a predicted value of future process data after a lapse of dead time from the present is obtained. .
And multiple of the predicted values of the obtained process data,
Since the stability is determined by the expert system as a parameter together with the past and present process data, it is possible to control the melting furnace by predicting the future behavior after a lapse of dead time. Therefore, the gate can be controlled stably without relying on the operator's personal judgment.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below. FIG. 1 is a sectional view of a melting furnace.
Is a swirl melting section, 3 is a slag discharge port (gate), 4 is an exhaust gas outlet, and 5 is a burner. The input ash is blown tangentially from the vicinity of the tip of the swirling melting portion 2, and is discharged from the slag discharge port 3 as molten slag. A gate exhaust gas outlet 6 is provided at the slag discharge port 3 to heat the slag discharge port 3.

A thermal image camera 7 is provided below the furnace main body 1 to take a thermal image of the slag discharge port 3. The thermal image is displayed on the monitor 9 via the thermal image display 8. The thermal image data is input to a computer, for example, a workstation 12 via the image processing device 10 and the communication device 11, and data exchange is performed with the instrumentation control system 13. The instrumentation control system 13 receives measurable process data and outputs a control signal according to the present invention. Note that this control signal does not necessarily need to be used for automatic furnace control, and it is possible for an operator to operate the furnace in accordance with the control signal.

The measurable process data includes the flow rate of the sprue exhaust gas, the flow rate of the main exhaust gas, the furnace pressure, the fuel flow rate, the combustion air flow rate, the ash supply rate, etc., as well as the spout opening degree and the solidification monitored by the thermal image camera 7. There is slag growth degree. In the present invention, the behavior of the furnace is predicted by a neural network using these process data as input signals. FIG. 2 shows a flow for quantifying the gate opening degree and the solidified slag growth degree.

The neural network is a model of the function of the nervous system of an organism. Here, a model represented by the following equation is applied.

(Equation 1) Here, k (= 0, 1, 2,...) Is a time step, and y is an output. In addition, y (k), ..., y (k-n + 1) is input to the neural network, and a set of data with y (k + 1) as the output is learned by the network, and an approximate function of the function f Was constructed with a neural network. The neural network employed here is a three-layer (input layer, intermediate layer, output layer) neural network based on the back propagation method.

Although the learning data of the neural network is collected from the actual plant, generalization becomes a problem because the process changes with time. Therefore, it is preferable to use a self-learning neural network and update the process data. In practice, a method of collecting learning data in the background, calculating a new neural network, and then switching can be adopted. In the present invention, such a neural network is used to predict a future behavior after a lapse of dead time from the present.

The basic flow of the furnace control employed in the present invention is as shown in FIG. That is, the stability is determined while operating at the rated value, and if it is determined that the operation is unstable, the operation is performed by changing the rated value by a certain width, and if it is still unstable, the fluctuation is temporary or permanent. Then, if the fluctuation is permanent, the rated value itself is replaced with a new rated value. This flow corresponds to the daily operation method of the operator.

Then, it is determined whether the future behavior after the elapse of the dead time predicted by the neural network is stable or unstable. If the behavior is predicted to be unstable, the prediction is made. By performing the operation for avoiding instability at present, the gate can be controlled stably while avoiding the delay of the process due to dead time. However, if the judgment of the stability depends on the operator, individual differences occur. Therefore, the expert system is used in the present invention.

In this expert system, the obtained predicted values of the process data are combined with the past and present process data, and the stability is determined using the result of the linear approximation of the change tendency as a parameter. Skilled operators predict future changes from past and present gate temperatures, gate openings, and solidified slag growth, so here we create a matrix as shown in the following table that combines these three parameters. did.

[0015]

[Table 1]

If the tendency of the change of the process data is applied to the matrix shown in this table, the stability of the future behavior of the furnace after the elapse of the dead time can be determined at the same level as that of the skilled operator. And if it is determined that it becomes unstable in this state, the rated value will be changed,
The control amounts of the furnace interfere with each other, and changing one changes the other control amounts. Therefore, in this embodiment, each control amount is changed in the following procedure. Change furnace pressure. Each exhaust gas flow rate is not changed if it is within the specified range. If stabilization is not achieved even when the furnace pressure reaches the upper limit, the gate burner flow rate is increased. If stabilized, restore as much as possible.

FIG. 4 shows the results of actual monitoring and control of the melting furnace by the method of the present invention described above. The dashed line is the predicted value according to the invention and the solid line is the measured value. Here, after the stable operation was continued for 0.5 hour, the operating condition was changed to an unstable state, and the monitoring and control system of the present invention was started from that state to stabilize the gate. Since it was known that the dead time of this furnace was 15 minutes, control was performed in a 5-minute cycle.

When the supervisory control system of the present invention is started,
The gate exhaust gas temperature was falling, the gate opening degree was small, and the solidified slag was in a state of no change.
For this reason, the expert system determined that it was unstable, and increased the furnace pressure. Further, thereafter, when the opening of the gate was increased and the growth of the solidified slag was reduced, it was determined that the fluctuation was stable and temporary, and the operation of restoring the furnace pressure was performed. As described above, by performing the monitoring control according to the present invention, the stable control of the gate can be performed. In addition, the predicted value according to the present invention indicated by the broken line and the measured value indicated by the solid line agreed well, and it was confirmed that the prediction accuracy according to the present invention was high.

[0019]

As described above, according to the present invention, numerical modeling cannot be performed, and the slag discharge port (grate) of the melting furnace having a long dead time can be stably provided without relying on the operator's judgment. There are advantages that can be controlled.

[Brief description of the drawings]

FIG. 1 is a sectional view of a melting furnace showing an embodiment of the present invention.

FIG. 2 is a flow sheet showing a flow of quantifying a gate state.

FIG. 3 is a flow sheet showing a flow of basic control.

FIG. 4 is a graph showing the result of actual monitoring control of a melting furnace.

[Explanation of symbols]

1 Furnace main body, 2 swirl melting section, 3 slag discharge port (gate), 4 exhaust gas outlet, 5 burner, 6 gate exhaust gas outlet, 7 thermal image camera, 8 thermal image display device, 9 monitoring monitor, 10 image processing device, 11 Communication equipment, 12 workstations, 13 Instrumentation control system

 ────────────────────────────────────────────────── ─── Continuing from the front page (72) Inventor Toshiaki Naito 2-56, Suda-cho, Mizuho-ku, Nagoya-shi, Aichi Japan Insulators Co., Ltd. F-term (reference) 3K061 NB27 NB30 4K045 AA04 BA10 DA01 RC11 4K056 AA05 AA19 FA01 FA11 FA12 FA13

Claims (3)

[Claims] 1. A neural network predicts the future behavior of a melting furnace that cannot be numerically modeled because it includes unmeasurable process data, and predicts the future process behavior of the obtained process data by past and present values. A method for monitoring and controlling the slag discharge of a melting furnace, wherein the stability of the slag discharge is determined by an expert system combined as a parameter together with the process data, and in the case of instability, the furnace is operated so as to stabilize the slag discharge. . 2. The method according to claim 1, wherein a self-learning type neural network is employed and the process data is updated to ensure generalization. 3. The method of monitoring and controlling a slag discharge port of a melting furnace according to claim 1, wherein a change in the opening of the gate, a falling state of the solidified slag, and a change in the gate temperature are selected as parameters of the expert system.