JP2015056766A - Cloud control system, and execution method for control program of the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a cloud control system for making it possible to perform control calculation within a predetermined control period time even when time delay exists in communication in an IP network.SOLUTION: A cloud control system comprises: a detection end part 4 for detecting a state of a plant that is a controlled target; an operation end part 5 for operating the plant on the basis of a detection signal from the detection end part; a management server 1 for executing, in a previously set control period, a control program previously set on the basis of the detection signal of the detection end part to control the operation end part; a plurality of calculation servers 2 for performing parallel processing on a plurality of division programs which are previously set by the management server and are previously divided so as to be capable of performing parallel processing on the control program; an IP network 3 for connecting the management server with the plurality of calculation servers; and a dedicated network 6 for connecting the management server with the detection end part and operation end part.

Description

  Embodiments described herein relate generally to a cloud control system for executing a control program using cloud computing resources, and a method for executing the control program.

  In order to facilitate disaster recovery, attention has been focused on control technology using cloud computing technology that remotely controls a plant via a network. A system that performs control using a cloud computing technology composed of a plurality of remotely located servers is referred to herein as a cloud control system.

  The cloud control system maintains the control program on multiple servers that are physically located in remote areas, and if a server in one area encounters a disaster, the system is controlled by the control program of the server in another area. Since it can be restored, it is possible to reduce the time required for rebuilding the control program and replacing the server equipped with the control program when recovering from a disaster.

  In general, there is a proposal for constructing a field network system by connecting field devices including a controller with a network in a process control system for controlling a plant.

  Conventionally, a fieldbus has been used for such a network, but in recent years, an IP network has been studied as such a network.

  However, since the communication time between each communication device is not secured in the IP network, there is a problem that the control calculation cannot be performed with the latest data. Therefore, there is a technology that can always perform the control calculation using the latest data. (For example, refer to Patent Document 1).

JP 2009-098919 A

  In a system using an IP network, there is a problem that a control calculation cannot be performed within a predetermined control period due to a delay in communication time caused by network communication congestion.

  In the case of Patent Document 1, when the requested time is transmitted in advance and each field device cannot perform communication according to the requested time, it is possible to change the requested time and perform control calculation with the latest data. There is a possibility that the control calculation is not completed within a preset control cycle.

  According to the present invention, in a cloud control system using cloud computing resources communicated over an IP network, even if there is a delay in the communication time of the IP network, control computation can be performed within a predetermined control cycle time. The purpose is to provide a cloud control system.

  In order to achieve the above object, the cloud control system of the present embodiment includes a detection end that detects a state of a plant that is a controlled object, and an operation end that operates the plant based on a detection signal from the detection end. A control program set in advance based on the detection signal of the detection unit and the detection end, and a management server that controls the operation end by executing the control program in a preset control cycle, and is set in advance from the management server, A plurality of divided programs that are divided in advance so that the control program can be processed in parallel, a plurality of operation servers that perform parallel processing, an IP network that connects the management server and the plurality of operation servers, the management server, and the A dedicated network for connecting the detection end and the operation end, and the management server includes the IP network A response time measuring unit that measures a response time with respect to one or a plurality of the divided programs set in advance in the calculation server, including a communication time between the communication server and the calculation server and a calculation time of the calculation server itself; , Storing information for setting a division program executed by a plurality of the calculation servers in advance, a reference response time for determining an abnormality of the response time for each calculation server for each control cycle, and a measured response measured A division program setting table for storing time is determined, and an abnormality in the measurement response time is determined. If there is an abnormality, the division program setting table is updated in the next control cycle by updating the division program setting table. With reference to the table, a divided program setting unit that distributes the divided program to the calculation server in the next control cycle; The control program set in advance based on the detection signal of the leading end is transmitted to the calculation server corresponding to the plurality of preset division programs, and the calculation results of the plurality of calculation servers are received. And an aggregating unit for sending to the operation end, wherein the arithmetic server is configured with a plurality of servers including a plurality of the divided programs and including a spare number, and the management server is configured in advance to perform the division. It is composed of one or a plurality of the divided programs below the maximum calculation time of a program, distributed in advance to the calculation server, receives a detection signal from the detection end, and processes the detection signal. The measurement response time measured at the current calculation cycle is obtained from all the calculation servers, and each response time is predicted. It is determined whether or not the reference response time is set, and for the calculation server that is equal to or longer than the reference response time, the calculation output in the previous control cycle of the calculation server is held and the next time In the control cycle, the calculation server is set to be executed by the backup calculation server. Otherwise, the calculation is continued under the same setting conditions in the same divided program setting table as the previous time, and the management server Even if there is a transmission delay, a transmission delay abnormality of the IP network system connected to the calculation server is detected from the response time of the division program, and the divided program is executed by a spare calculation server. It is characterized in that control can be continued.

  In order to achieve the above object, the execution method of the control program of the cloud control system of the present embodiment is based on the detection end that detects the state of the plant that is the controlled object, and the detection signal from the detection end. An operation end that operates the plant, a control program that is set in advance based on a detection signal from the detection end, and a management server that controls the operation end by executing a control program that is set in advance, and the management A plurality of operation servers configured in parallel to process a plurality of divided programs that are set in advance from a server and previously divided so that the control program can be processed in parallel; an IP network that connects the management server and the plurality of operation servers; A control program for a cloud control system, comprising: a dedicated network for connecting the management server to the detection end and the operation end In the execution method, the management server includes one or more of the divisions set in advance in the calculation server, including a communication time between the IP network and the calculation server and a calculation time of the calculation server itself. A response time measuring unit that measures a response time to a program and information for setting a division program executed by a plurality of the calculation servers are stored in advance, and an abnormality in the response time for each of the calculation servers is determined for each control cycle. A division program setting table for storing the reference response time for the measurement and the measured measurement response time, and determining an abnormality in the measurement response time. The division program is updated at the next control cycle with reference to the division program setting table updated at the control cycle. The division program setting unit distributed to the calculation server and the control program set in advance based on the detection signal of the detection end are transmitted to the calculation server corresponding to a plurality of preset division programs. An aggregation unit that receives calculation results of a plurality of the calculation servers and sends the calculation results to the operation end, and the calculation server is a number of servers including a plurality of the divided programs and including a spare number The management server is configured in advance with one or a plurality of the divided programs not exceeding the maximum calculation time of the divided program, distributed in advance to the calculation server, and detected from the detection end. Receiving a signal, processing the detected signal, transmitting the signal to the arithmetic server including the corresponding division program. The measurement response time measured in the period is obtained, and it is determined whether or not each response time is a preset reference response time. For the calculation server that is equal to or longer than the reference response time, the calculation server The calculation output at the previous control cycle is held, and at the next control cycle, the calculation server is set to be executed by the spare calculation server. Otherwise, the same division program setting as the previous one is set. Continue the calculation under the setting conditions in the table, detect the transmission delay abnormality of the IP network system connecting the management server and the calculation server from the response time of the divided program, The divided program is executed so that control can be continued even if there is a transmission delay.

The block diagram of the cloud control system in an embodiment. The flowchart explaining the operation principle of a cloud control system. An example of a division program setting table. The flowchart explaining operation | movement of a cloud control system.

  Hereinafter, embodiments will be described with reference to the drawings. First, before describing the configuration of each unit, an outline of the operation principle of the present embodiment will be described with reference to the time chart of FIG.

(Operating principle)
The control program executed by the cloud control system 100 and its control cycle Tc are set in advance from the control performance required from the controlled object and its response time.

  In general, the control program executed by the controller that controls the plant is compiled into a plurality of divided programs that can be processed in parallel, and a plurality of divided programs are previously stored. Are set (loaded) as initial values from the management server.

  Next, the management server instructs the execution of the control program, that is, the divided divided program executed by the plurality of calculation servers in parallel, and outputs the calculation output as a response from the calculation server to each operation end. To do.

  At this time, the management server determines, for each control cycle, the time until the computation output including the round-trip communication time between the management server and the computation server is obtained from each computation server in the packet instructing computation. The response time is determined in advance.

  This response time is measured for each control cycle and for each corresponding calculation server in a divided program setting table, which will be described in detail later.

  This divided program setting table designates one divided program or a plurality of divided programs to be executed in advance by the calculation server, and a reference response time indicating whether or not a response within the control cycle time is possible in advance. The communication time abnormality is determined by comparison.

  In communication with the calculation server determined to be abnormal, the calculation output in that cycle holds the previous output, and in the next control cycle, the spare calculation server connected to the IP network executes this divided program. By so commanding, it is intended to avoid a delay in the communication time of the IP network.

  By the way, the control cycle is several milliseconds to several hundred milliseconds in a general plant, and the change rate of the congestion degree of the communication line of the IP network is expected to change at a time later than the control cycle time. Therefore, by setting the margin of the threshold for determining the reference response time as a small value, it is possible to predict a communication time abnormality.

(Embodiment)
Next, an embodiment based on the operation principle of the invention will be described. FIG. 1 is a block diagram of a cloud control system 100 in the embodiment.

  As shown in FIG. 1, the cloud control system 100 according to the embodiment includes a detection end 4 that detects a state of a plant that is a controlled object, and an operation end that operates the plant based on a detection signal from the detection end 4. The control program set in advance based on the detection signal of the unit 5 and the detection end 4 is executed at a preset control cycle, and set in advance from the management server 1 that controls the operation end 5 and the management server 1 A plurality of operation servers 2 that perform parallel processing on a plurality of divided programs that are divided in advance so that the control program can be processed in parallel, an IP network 3 that connects the management server 1 and the plurality of operation servers 2, and a management server 1 and a dedicated network 6 for connecting the detection end 4 and the operation end 5 to each other.

  Next, the configuration of each unit will be described. The management server 1 responds to one or more divided programs set in advance in the computation server 2 including the communication time between the IP network 3 and the plurality of computation servers 2 and the computation time of the computation server 2 itself. A response time measuring unit 11 that measures time is provided.

  Further, information for setting a division program executed by a plurality of calculation servers 2 is stored in advance, a reference response time for determining an abnormality in response time for each calculation server 2 for each control cycle, and a measured response time measured. The divided program setting table 12a is stored, the abnormality of the measurement response time is determined, and if there is an abnormality, the divided program setting table is updated, and the divided program setting table updated in the next control cycle is displayed. A division program setting unit 12 that distributes the division program to the calculation server 2 based on the control program preset based on the detection signal of the detection end 4 as a plurality of preset division programs. And an aggregating unit 13 that receives the calculation results of a plurality of pre-calculation servers and sends them to the operation end unit 5.

  The method of dividing the control program into a plurality of divided programs capable of parallel operation is not the subject of the present invention, and therefore the details are omitted. However, the method is described in source code that requires processing of a large number of input / output signals. The control program, for example, determines the dependency of the input signal or the primary variable on the output signal in advance, divides the source code having the same dependency together, and can execute the divided control program in parallel Can be created by a compiler that converts to

  Next, the plurality of calculation servers 2 are configured by the number of servers including a reserve number equal to or more than the number of divided programs.

  Each of the calculation servers 2 stores all the divided programs in advance, and calculates the divided programs set in advance, but the divided programs are assigned to be executed according to the measurement response time results measured at each control cycle. It is updated at the next control cycle in response to a command from the management server 1.

  By the way, the cloud control system 100 mentioned here is a system in which the management server 1 and the computation server 2 are assumed to use resources provided by the cloud service provider, and communication between these servers is configured by the IP network 3. A computer network interconnected using Internet protocol technology, which is a communication protocol for relaying packets, does not guarantee communication time.

  On the other hand, the dedicated network 6 that connects the management server 1 to the detection end 4 and the operation end 5 is assumed to be, for example, a field bus (IEC 61158), and the communication time between these is set in advance. The communication time is guaranteed.

  Next, a detailed configuration of the management server 1 that controls the operation of the cloud control system 100 will be described.

  The response time measuring unit 11 obtains the packet by instructing the execution of the divided program written in the divided program setting table 12a in advance to each arithmetic server 2 and measuring the time until the reply response is received.

  The measured response time includes a round-trip transmission time during which a packet instructing execution of the division program propagates on the IP network 3 and an operation time of the commanded division program.

  More specifically, the processing included in the control cycle time includes the time for reading (loading) the input data from the detection end 4 and the operation output from one of the operation servers 2 until the operation end 5 is sent. Is included. Therefore, a margin for these values is added to the reference response time for determining an abnormality in the measurement response time.

  Since the divided program setting unit 12 distributes the divided programs to be calculated in parallel so that the divided programs can be completed in the shortest time, the divided program setting unit 12 determines distribution of the other divided programs based on the calculation time of the longest divided program.

  The calculation server 2 may prepare the same number of calculation servers 2 as the number of divided programs to be distributed. However, in order to reduce the number of divided calculation programs, the divided program having the longest time is divided into other divided programs. The programs may be combined and configured by a plurality of divided programs so that a value obtained by adding these calculation times is equal to or shorter than the longest time.

  In addition, the number of arithmetic servers needs to be determined in consideration of the number of divided programs and the number of spare units that are switched when an abnormality occurs in which the response time with any of the arithmetic servers exceeds the reference response time.

  The divided program setting unit 12 distributes the divided program when there is a response time exceeding the reference response time in addition to the configuration of the divided program and the initial setting of the divided program to the calculation server 2 for parallel calculation. Reconfigure (configure)

  Next, an example of such a division program setting table 12a is set with reference to FIGS.

  FIG. 3A shows the distribution of the divided programs POU1 to POU9 distributed to each of the preset calculation servers OS1 to OS5, the reference response time of each divided program, and the measurement response in units of distributed calculation servers. The example of the division | segmentation program setting table 12a which shows time is shown.

  FIG. 3B shows that the measurement response time in the control cycle Ti is abnormal in the response time of the calculation server OS3 (the longest response time is 10 msec of POU1. Therefore), if the reference response time is 10 msec, the calculation server OS3 11 msec, which was determined as above. ) Occurs.

  When an abnormality occurs, an example of a case where the calculation server OS3 is updated to a spare calculation server OS without reconfiguration of the division program is shown in FIG. An example of another updating method is shown in FIG.

  As shown in FIG. 3, the number of divided programs is divided into nine, POU1 to POU9, and the divided number of divided programs is distributed to four calculation servers OS1 to OS4, and the maximum response is shown. The case where the time (reference response time) is set to 10 msec of POU1 is shown.

  In the control cycle Ti, the response time of the calculation server OS3 to which POU3 and POU4 are distributed exceeds 11 msec, which exceeds the reference response time. Therefore, in the next control cycle Ti + 1, the calculation server OS3 is updated to the spare calculation server OS5. Is shown.

  As another update method, a case is shown in which the calculation server OS3 is assigned to the calculation server OS4, and the divided calculation programs POU5 to POU9 that the calculation server OS4 has assigned are assigned to the spare calculation server OS5.

  In the above example, the combination of the division programs is not updated, but the combination of the division programs may be updated.

  Next, the aggregating unit 13 receives the detection signal transmitted from the detection end unit 4 via the dedicated network 6 and sends the detection signal to the arithmetic server 2 that executes the corresponding divided program, and the arithmetic result from the arithmetic server 2 is obtained. It is received and sent via the dedicated network 6 to the corresponding operating end 5.

  That is, the aggregation unit 13 has an interface function with two different networks, the IP network 3 and the dedicated network 6.

  Next, the control operation of the management server 1 which is a controller for controlling such a cloud control system 100 will be described with reference to FIG.

  First, as an initial setting, a split program for all hands is loaded from the management server 1 to the arithmetic server 2 via the IP network 3 (s1).

  Next, input data is read from the detection end part 4 (s2), and the divided program to be shared and the detection data are transmitted to the respective calculation servers 2 (s3).

  Each calculation server receives the data (S11), executes the divided program corresponding to the transmitted input data (s12), and returns the calculation result to the management server 1 (s13).

  The management server 1 transmits the transmitted calculation result to each operation end 5 (s4). Furthermore, the time from when the calculation is requested until the calculation result is obtained is measured to obtain a response time (s5), and the measured response time is written in the divided program setting table 12a, and a preset reference response time and The presence / absence of an abnormality is determined by comparison, and the division program setting unit 12 is notified (s6).

  That is, the management server 1 is composed of one or a plurality of divided programs that are less than or equal to the maximum computation time of the division program in advance, and distributes them to the computation server 2 in advance and receives a detection signal from the detection end 4. Then, the detection signal is processed and transmitted to the arithmetic server 2 having a corresponding division program.

  Further, the measurement response times measured in this calculation cycle are obtained from all the calculation servers 2, and it is determined whether or not each response time is a preset reference response time. For the server 2, the calculation output in the previous control cycle of the calculation server is held, and in the next control cycle, the calculation server is set to be executed by the spare calculation server. Continues the calculation under the setting conditions in the same division program setting table as the previous time.

  Therefore, according to this embodiment, the transmission delay abnormality of the IP network 3 system connecting between the management server and the calculation server is detected from the response time of the division program, and the divided program divided by the spare calculation server is obtained. The control can be continued even if there is a transmission delay.

  Further, by performing parallel operations on the divided programs, it is possible to shorten the control operation time and secure a margin for communication time delay.

  Also, the management server is composed of two management servers, and one management server set in advance is set as a master, and when the management server set as the master fails, the control right is transferred to the other management server. A cloud control system that is configured in such a way that it can not only reliably perform computations within the control period for IP network transmission delays, but also can quickly recover from disasters. Can be provided.

  Although several embodiments of the present invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the scope of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are included in the invention described in the claims and the equivalents thereof.

DESCRIPTION OF SYMBOLS 1 Management server 2, 2a, 2b, 2c, 2i Arithmetic server 3 IP network 4 Detection end part 5 Operation end part 11 Response time measurement part 12 Division program setting part 12a Division program setting table 13 Aggregation part 100 Cloud control system

Claims (3)

A detection end for detecting the state of the plant to be controlled, an operation end for operating the plant based on a detection signal from the detection end, and a control set in advance based on the detection signal of the detection end A management server that executes a program at a preset control cycle and controls the operation end, and a plurality of divided programs that are set in advance from the management server and are divided in advance so that the control program can be processed in parallel A plurality of arithmetic servers that perform parallel processing, an IP network that connects the management server and the plurality of arithmetic servers, and a dedicated network that connects the management server to the detection end and the operation end. A cloud control system comprising:
The management server measures a response time for one or a plurality of the divided programs set in advance in the calculation server, including a communication time between the IP network and the calculation server and a calculation time of the calculation server itself. A response time measuring unit to
Information for setting a division program executed by a plurality of the calculation servers is stored in advance, a reference response time for determining an abnormality in the response time for each calculation server for each control cycle, and a measured response time measured The division program setting table is stored, and when there is an abnormality by determining the abnormality of the measurement response time, the division program setting table is updated in the next control cycle by updating the division program setting table. With reference to, a division program setting unit that distributes the division program to the arithmetic server;
The control program set in advance based on the detection signal of the detection end is transmitted to the calculation server corresponding to the plurality of division programs set in advance, and the calculation results of the plurality of calculation servers are received. A consolidating unit to be sent to the operation end,
With
The arithmetic server is composed of a plurality of servers including a plurality of the divided programs and a spare number,
The management server is configured in advance with one or a plurality of the divided programs not exceeding the maximum calculation time of the divided program, and is distributed in advance to the calculation server and receives a detection signal from the detection end. And processing the detection signal, transmitting to the arithmetic server comprising the corresponding division program,
Furthermore, the measurement response time measured in the current calculation cycle is obtained from all the calculation servers, and it is determined whether or not each response time is a preset reference response time,
For the calculation server that is equal to or longer than the reference response time, the calculation output in the previous control cycle of the calculation server is held, and the calculation server is executed by the spare calculation server in the next control cycle. Set to
If not, continue the calculation with the same setting conditions in the same split program setting table as before,
The transmission delay of the IP network system connecting between the management server and the arithmetic server is detected from the response time of the split program, and the split program is executed by a spare arithmetic server, and the transmission delay is detected. A cloud control system characterized by allowing control to continue even if there is a problem.   The management server is composed of two management servers. One of the management servers set in advance is used as a master, and when the management server set as the master fails, the control right is transferred to the other management server. The cloud control system according to claim 1. A detection end for detecting the state of the plant to be controlled, an operation end for operating the plant based on a detection signal from the detection end, and a control set in advance based on the detection signal of the detection end A management server that executes a program at a preset control cycle and controls the operation end, and a plurality of divided programs that are set in advance from the management server and are divided in advance so that the control program can be processed in parallel A plurality of arithmetic servers that perform parallel processing, an IP network that connects the management server and the plurality of arithmetic servers, and a dedicated network that connects the management server to the detection end and the operation end. A method for executing a control program for a cloud control system comprising:
The management server measures a response time for one or a plurality of the divided programs set in advance in the calculation server, including a communication time between the IP network and the calculation server and a calculation time of the calculation server itself. A response time measuring unit to
Information for setting a division program executed by a plurality of the calculation servers is stored in advance, a reference response time for determining an abnormality in the response time for each calculation server for each control cycle, and a measured response time measured The division program setting table is stored, and when there is an abnormality by determining the abnormality of the measurement response time, the division program setting table is updated in the next control cycle by updating the division program setting table. With reference to, a divided program setting unit that distributes the divided program to the arithmetic server in the next control cycle;
The control program set in advance based on the detection signal of the detection end is transmitted to the calculation server corresponding to the plurality of division programs set in advance, and the calculation results of the plurality of calculation servers are received. A consolidating unit to be sent to the operation end,
With
The arithmetic server is composed of a plurality of servers including a plurality of the divided programs and a spare number,
The management server is configured in advance with one or a plurality of the divided programs not exceeding the maximum calculation time of the divided program, and is distributed in advance to the calculation server and receives a detection signal from the detection end. And processing the detection signal, transmitting to the arithmetic server comprising the corresponding division program,
Furthermore, the measurement response time measured in the current calculation cycle is obtained from all the calculation servers, and it is determined whether or not each response time is a preset reference response time,
For the calculation server that is equal to or longer than the reference response time, the calculation output in the previous control cycle of the calculation server is held, and the calculation server is executed by the spare calculation server in the next control cycle. Set to
If not, continue the calculation with the same setting conditions in the same split program setting table as before,
The transmission delay of the IP network system connecting between the management server and the arithmetic server is detected from the response time of the split program, and the split program is executed by a spare arithmetic server, and the transmission delay is detected. A method for executing a control program of a cloud control system, characterized in that control can be continued even if there is an error.

Images