JP2552907B2 - Cooperative distributed control method - Google Patents

Description

Detailed Description of the Invention [Industrial applications]

The present invention relates to a cooperative distributed control system in which a plurality of control devices cooperate with each other to control a plant.

[Prior art]

FIG. 6 is a system configuration diagram of a system to which a conventional cooperative distributed control method is applied. In the figure, 1 is a plant to be controlled, 2 is process data of the plant 1, and a control command signal is output. Input / output signal processor, 3a
Is a control device that performs control processing of the process data and outputs a control command signal to the plant 1 via the input / output signal processing device 2, and 3b is a standby device that switches to control the plant 1 when the control device 3a fails. The control devices 4 are communication paths for exchanging necessary information between the control devices 3a or the standby control device 3b. Next, the operation will be described. Normally, each control device 3a
Receives necessary process data from the subsystem in charge of control among the subsystems of the plant 1 via the input / output signal processing device 2, performs control processing, and then inputs / outputs a control command signal if necessary. The signal is output to the subsystem via the signal processing device 2. Further, when it is desired to know the process data of another subsystem that is not in charge of control, the process data of another subsystem is indirectly transmitted by communicating with another control device 3a using the communication path 4. Can be obtained. Here, when the control device 3a detects a failure using the self-diagnosis function, the control right is switched to the standby control device 3b,
The control of the process 1 is not hindered.

[Problems to be Solved by the Invention]

Since the conventional cooperative distributed control method is configured as described above, if the standby control device 3b also fails, the subsystem that is the control target becomes uncontrollable, and normally the failure rate of the control device 3a Since the standby control device 3b is not used so much, and when the plant 1 becomes large-scale and the number of subsystems increases, a large number of standby control devices 3b are required and the cost of the subsystem becomes expensive. was there. The present invention has been made to solve the above problems, even without a redundant system such as the standby control device 3b,
An object of the present invention is to obtain a cooperative distributed control system that does not hinder the control of the plant 1 when the control device 3a fails. Similar prior arts are disclosed in JP-A-63-115201 and JP-A-63-120302.

[Means for Solving the Problems]

In the coordinated distributed control system according to the present invention, when a certain control device fails, the faulty control device informs another normal control device via the communication path that the other normal control device has failed. According to a predetermined sharing rule, the sharing of control for the subsystem of the plant controlled by the failed controller is reconfigured.

[Action]

The coordinated distributed control system according to the present invention allocates the subsystem of the plant controlled by this control device to the control of another appropriate control device when the control device fails.

【Example】

An embodiment of the present invention will be described below with reference to the drawings. In FIG. 1, reference numeral 1 denotes a plant to be controlled, which is divided into a large number of subsystems for each suitable control unit. twenty one
˜23 is an input / output signal processing device that inputs the process data of the plant 1 and outputs a control command signal, 31-33 performs the control processing of the process data, and outputs the control command signal to the input / output signal processing devices 21-23. Control device 4 for outputting to the plant 1 via the communication path 4 for communicating necessary information between the control devices 31-33, and 5 is an input / output signal processing device 21-23 for any subsystem of the plant 1. Is a plant data communication path for obtaining the process data output by. Next, the operation will be described. First, each control device 31-33
In the above, a control process for one subsystem is called a task. Each of the control devices 31 to 33 is in charge of task processing for a plurality of tasks, and after the process data necessary for these task processes are obtained by the input / output signal processing devices 21 to 23 via the plant data communication path 5, It is transmitted to the control devices 31 to 33. Since these input / output processes are performed in a fixed time cycle, each control device 31-33 must complete all task processing within that time cycle. In FIG. 1, if a control device 32 fails, a failure is detected by a self-diagnosis device (not shown) provided in the control device 32, and the communication path 4 is transmitted to other control devices 31, 33, .... The failure is notified via. At the same time, according to a predetermined sharing protocol, the sharing of the task that the failed control device 32 is processing is reconfigured among the other control devices 31, 33, .... As a result of the reconfiguration of the task sharing, the task of the controller 32 that has failed is the other controller 31, 33, ...
.. are distributed and assigned to each control device 31, 33 ,. As a result, the function of the failed control device 32 is backed up and the function of the entire system is maintained. FIG. 2 is a diagram for explaining the concept of backup of the failed control device 32. In the figure, the shaded portions indicate tasks in each of the control devices 31 to 33. In FIG. 2 (b), the faulty control device 32 was in charge of 2
It is shown that one task has been transferred to the control of the other control devices 31 and 33, respectively. Input / output signal processor
21-23 are omitted. Here, the process of assigning a task according to the sharing rule that defines the sharing of tasks is called task scheduling. Further, the task processing time for the charge is called a load, and when this load exceeds the one cycle time described above, it is called an excessive load. When a multiple failure occurs in which the normal control device 31 further fails, notification of the failure and task scheduling are performed in the same manner, tasks are distributed among the remaining normal control devices 33, ..., And the function of the entire system is maintained. . The task scheduling and task sharing after reconfiguration must be completed within one cycle time of control. Backup by cooperative distribution is possible as long as one of the control devices does not become overloaded. The task scheduling method includes, for example, the following three types of implementation methods. A method that determines the allocation according to the failure mode (static scheduling method). The static scheduling method is effective when there are few types of failure modes, that is, when the number of control devices is small. A combination of the failure control devices 31, 32, 33, ... Is set as a failure mode, and task scheduling for the normal control devices 31, 32, 33, .. deep. The number of failure mode types is ( _n C ₁ + _n C ₂ + ... + _n C _p ), _{where n} is the number of control devices and p is the maximum number of failures to be considered. The task allocation obtained by scheduling is assigned to each control device 31, 32, 33, ...
Is an internal storage device (not shown) in advance as data
Record it in. If a failure occurs, normal control device 31,3
3 ... obtains a new failure mode from the current failure mode and the name of the controller that has just broken, and selects task assignments corresponding to that failure mode. After that, each control device 31, 33,
...... controls the subsystem according to the selected task assignment, that is, performs task processing. Method for obtaining optimal processing distribution at the time of failure (centralized scheduling method) When the number of control devices increases, the static scheduling method has a large storage capacity for storing task allocation, which is disadvantageous in terms of cost. The centralized scheduling method is a method in which, when a failure is detected, the control devices 31, 32, 33, ... Communicate with each other to obtain task allocation. Since it is not necessary to store the task assignment for each failure, the controller 31, 32,
The storage capacity of 33, ... Can be reduced compared to the static scheduling method. The centralized scheduling method is performed by the following procedure. When a failure is detected in a control device 32, the occurrence of the failure is communicated to all the control devices 31, 33, .... b. From the total load of the entire system and the number of control devices,
31,33, ... seeks the size of the load that it takes. c. Select the combination of tasks that corresponds to the load to be taken. d. Count unselected tasks to the next controller. By repeating the processing of ec and d, the task assignment of all control devices is determined. f. Each controller 31, 33, ... Performs task processing according to the obtained task assignment. In the case of multiple failures, the load per control device increases according to the number of failed devices, and backup is possible within a range that does not result in excessive load. In reality, even if combined like tasks, it does not match the burden of responsibility, so
It suffices that the load actually accepted falls within the allowable error from the theoretical responsibility load, and this allowable error is set within the range of 1/2 to 1 of the load per task. A method that constantly reconfigures the processing distribution (distributed scheduling method). In the method, each control device 31, 32, 33, ... Has to have a time for scheduling immediately after a failure as a margin in a normal state. As the number of tasks increases, the margin must be increased, so the number of control devices must be increased. In the distributed scheduling method, regardless of whether or not there is a failure, information about mutual loads is locally communicated between about several control devices, and task allocation is performed so that the mutual loads become even. Local task scheduling has the advantage that the time required for one scheduling is short. The details of the scheduling will be described below. When a failure is detected in a certain control device 32, the failure occurrence is notified only to the control devices 31, 33, ... Surrounding the failed control device 32. The task which the failed control device 32 was in charge of is distributed among the control devices 31, 33, ... Surrounding it. Then, the load of the control devices 31, 33, ... With more tasks becomes larger than that of the control devices 31, 33, ..
It is distributed to 31,33, .... After several times of scheduling, the load remaining among the control devices finally becomes stable in a state of task distribution in which the loads are even. Local task scheduling between two controllers is performed based on the following principle. Controller 31, 32, 33,
When the loads at time t of ... are X _i , X _{i + 1} , and α are appropriate coefficients, the load α (X _i
The load on each control device at time t + 1 after the task corresponding to −X _{i + 1} ) has moved is derived by the following equation. The convergence of X _t in t → ∝ occurs when all the eigenvalues of A corresponding to X _i −X _{i + 1} exist in the unit circle or in the left half plane. The eigenvalue corresponding to X _i −X _{i + 1} is 1-2
α, and α existing in the unit circle is 0 ≦ α ≦ 1. That is, at this time, the local scheduling becomes stable. Similarly, the range of coefficient values between a plurality of control elements can be determined. Local scheduling is performed by the following procedure using this coefficient α. Communicate with the other control device 31, 32, 33, ... About the current load status. b. Find the magnitude α (X _{i + 1} −X _i ) of the moving load. c. Find the combination of tasks that corresponds to the amount of moving load. d. The control device 31, 33, ... having a large load notifies the task to be moved. e. Each control device 31, 33, ... Processes the newly assigned task. FIG. 3 is a system configuration diagram of another embodiment of the present invention. In this embodiment, the process data is also passed through the communication path 4, and the scheduling methods 1 to 3 are applied as in the above embodiments. FIG. 4 is a system configuration diagram of a system in which the input / output signal processing device 2 is provided for each of the control devices 31 to 33. Also in this case, the scheduling methods 1 to 3 are applied as in the above-described embodiment. Further, FIG. 5 is a system configuration diagram of a system in which adjacent control devices 31 to 33 are directly connected to each other without forming the communication path 4 into a bus configuration. In this case, it is suitable for applying the distributed scheduling method described above. The controlled object of the coordinated distributed control method according to the present invention is not limited to the plant, but can be diverted to mechanical devices such as robots, electronic devices, and other devices that have conventionally used a redundant system.

【The invention's effect】

As described above, according to the present invention, the coordinated distributed control system uses the subsystem that the failed control device was in charge of controlling, and the load that the own device can bear on each control device that has been notified that the failure has occurred. By assigning a part or all of the control of the faulty device within the range of the above, it is assigned under the control of another normal control device, so a redundant system such as a standby control device is prepared for each control device. The number of control devices required for the entire system can be reduced, and a system with improved durability against multiple failures can be obtained. Furthermore, since the load is leveled among the devices that have been notified that the device has failed, one device does not bear a load close to the limit, and the device that has failed next There is an effect that all the devices can secure a margin for processing when accepting the load of.

[Brief description of drawings]

1 and 3 to 5 are system configuration diagrams of a system to which an embodiment of the cooperative distributed control system according to the present invention is applied, and FIG. 2 is an explanatory diagram of a concept of backup of a control device, and FIG. FIG. 1 is a system configuration diagram of a system to which a conventional cooperative distributed control method is applied. 1 is a plant, 2, 21-23 are input / output signal processing devices, 31-33
Is a control device, 4 is a communication path, and 5 is a plant data communication path. In the drawings, the same reference numerals indicate the same or corresponding parts.

Claims (1)

(57) [Claims] 1. A plant to be controlled is divided into a large number of subsystems, the large number of subsystems are divided into a plurality of groups, and each group is controlled by a controller via an input / output signal processor. In the cooperative distributed control method in which each of the control devices controls the plant by exchanging information using a communication path to which each of the control devices is connected, an arbitrary control device among the control devices is When a failure of the own device is detected and the fact that the device has failed is transmitted to each of the other control devices via the communication path, the range of the load that the own device can bear to each control device that has received the notification of the failure. By assigning a part or all of the control of the faulty device within the control system, the subsystem of the set that was the control target of the faulty control device was assigned to another control device as the control target, and the fault occurred. Characterized in that the load amounts of the respective devices are communicated with each other by using the communication path between the control devices having received the transmission of the control load from the control device having a large control load amount to the control device having a small control load amount. Cooperative distributed control method.