P.U. Ogeyenko, A.V. Lozovyagin, V.V. Ogeyenko

National Mining University, Dniepropetrovs’k, Ukraine

Decentralized management for problem solution to

minimize specific energy consumption

Modern development tendencies of automation technology are based on the gradual transition from centralized control systems to systems in which the control center does not exist. In such systems each node can be regarded as an intelligent, because all devices itself make a decision for performing actions in management of the actuators. This approach is known as decentralized control. It is based on open information interaction within the group of network modules, in which there is a division of control process for the number of subtasks. Each of them is performed by the device directly connected with a separate actuator. Thus, to achieve specific system goals at the expense of solving local tasks is due to sub team members [1].

Using of decentralized control systems in coal-mining industry will contribute to the objective of improving process efficiency by reducing specific energy consumption in coal mining.

Selection of networking solutions is one of the basic tasks in designing a decentralized management system. Technological objects for which automation of is preferred to a decentralized approach may be generally characterized by a number of factors: distributed in space, the complex topology, a large number of sensors and actuators, interference environment, use of equipment with different I/O interfaces, the need for rapid response in accordance with parameters of sensors, large amounts of requested information. Their presence greatly complicates the selection of universal solution for safe communication channel in the system.

Thus, we can formulate a set of requirements for a decentralized management system and in particular for their transmission system:

1) a large extent;

2) high speed data exchange;

3) a wide range of topologies support;

4) high reliability and noise protection;

5) a large number of workstations support;

6) a wide data field in transmitted messages;

7) multimaster operation mode;

8) ease of design and reconfigure the network.

According to the analysis of several well-known fieldbuses, which characteristics are summarized in Table 1, we can distinguish CAN, Profibus DP, P-NET and LON. These network solutions most suited for the requirements are listed above.

Table 1 - Industrial data communication systems characteristics

Field Bus	Topology	Wire	Stations	Work mode	Segment length (max) (m)	Speed (max) (kb/s)	Data field (bit)
AS-I	bus/tree	2	32	mono	100	167	4
CAN	bus	2	64	multi	5000 (10 kb/s)	1024	to 64
Interbus	ring	2/8	255	mono	400 (total 12800)	500	64x8
Profibus FMS	bus	2	127	multi	19200 (9,6 kb/s)	500	246x8
Profibus DP	bus	2	127	multi	1000 (12 Mb/s)	12288	246x8
Profibus PA	bus	2	32	mono	1900	93,75	246x8
Fip	bus	2	256	multi	2000 (1 Mb/s)	2560	32x8
Modbus plus	bus	2	32	multi	1800	1024	32x8
P-NET	bus/tree	2	32+125	multi	1200	76,8	56x8
FF H1	bus	2	240	multi	2000 (total 9500)	31,25	246x8
FF H2	bus	2	240	multi	2000 (total 9500)	1024	246x8
HART	star/bus	2	15	mono	to 3000	1228	8
LON	bus/tree	2	64 (to 32000)	multi	6100 (5 kb/s)	1228	228x8

However, each of the selected buses for the task has its disadvantages. So there is no universal solution.

CAN-bus is mostly suited for a decentralized approach in the technological objects management. It is interesting from the standpoint of the channel access organization principles, where the struggle for the channel is based on the bus arbitration. The advantages include the growing popularity of the CAN-interface management systems and, consequently, the producing of the entire series of cheaper specialized controllers for this bus. In future, systems based on the RS485 interface, will remain only in the lower price class data devices, because they realize almost all the functions of packet software. For developing data systems such construction can be a bottleneck, so promising solutions based on dedicated bus controllers are justified now.

The main difference between CAN bus and the existing standards is that the transmitted message frame has no address for the receiving device, it only contains the identifier of the data packet. The same package can be both read and used by a number of devices [2].

In addition, CAN-bus has a high noise protection, which is based on a physical level (differential signal). The bus is characterized as resistant to electrical and information overload, and has an internal system of setting priorities.

The disadvantage of CAN in the organization of decentralized control can be its bus topology. However, the use of specialized routers and special organization of the data transfer protocol will allow to use CAN bus for tree topology technological objects [3].

One of the main tasks to be solved in decentralized management of a technological objects is to allocate limited resources [4]. When performing this task it is necessary to provide a broad and reliable channel for step by step exchange of requests for a limited resource, which are initiated by devices. At the end of each step unit involved in the exchange take the decision to stop or continue the distribution based on the values calculated due to requests. Thus, to test the CAN using possibility for the decentralized approach organization, it is necessary to test the performance of the bus at high load communication channel during step by step value exchange and correct sum calculation separately for each node.

Physical model which presented in figure 1 was designed and created for the experiment.

In the information exchange by CAN bus involves six CAN nodes: an industrial controller that implements the functions of line adapter for PC communication, and five CAN-compatible devices of the same type that exchange data. Interaction between the PC and the adapter link is carried out through the Ethernet 100BaseT interface based on TCP.

The software, allowing to monitor information in the CAN segment, run by PC. Remote monitoring application is written in C++ programming language to conduct experiments with the physical model of the system. It is based on dialog application.

CAN bus is based on multimaster approach, whereby each node in the network can transmit data at any time, regardless of other segment devices. Possible conflicts are resolved through the use of dominant-recessive approach for access arbitration. All transmitted frames begin with a unique identifier, which also serves as the message priority. Furthermore, according to the organization, all CAN devices are constantly in receiving mode, thus allowing to monitor the data transmission progress. Synchronization of nodes is based on the interframe bit sequences and on all devices tuning by receiving a start-bit. Thus, for simultaneous transmission of frames by two or more devices a bus will automatically be given to the host, who first set a dominant bit, that is, to whose priority of the message above. Transferring for the remaining devices will be suspended. Such bus access organization itself is decentralized. It opens wide possibilities for performing the limited resources allocating task.

Figure 1. The block diagram model for experiments

Resource allocation process with not optimal initial conditions may require more than a hundred steps, which means that the model develops more than five hundred messages. There is evident need to comply with the requirement of a fieldbus high-speed data exchange.

In accordance with the parameters of the experiments, all five CAN nodes are incrementally transmit its unique request values and at the end of the step to calculate their sum. Therefore, it is necessary to organize a decentralized control over the light of all the values of a step is based on CAN, because each device will not have information about number of nodes involved in this exchange. We must also take into account possible heterogeneity of the combined intelligent modules, which can lead to delays in calculating the final amount before proceeding to the next step.

The graph-scheme (Fig. 2) shows the proposed approach for solving the problem of the experiment. It describes the CAN bus segment node operation algorithm during incremental data exchange.

Node program begins with the initialization of involved modules, variables and flags according to their default values. Then, a timer is switched on for the organization of the delay required to synchronize devices with different intelligent modules (S1). The delay allows all involved in the exchange devices to begin the next step at the same time regardless of the time it takes to calculate the sum on each of them. Thus, the node goes into standby mode for waiting of delay end or receiving data frame (S2). Device, which calculation was made at the least time, exits first from this mode and sends a frame with the current request value (S3), which triggers the other nodes to the next step. In accordance with the specification of CAN, the message is placed in the transmit buffer will be sent at a time when its priority would be the highest on a bus. Through this approach, all nodes involved in the exchange after putting a data frame in transmission buffer can be go into event flags poll mode (S4). Sending / receiving message events are monitored in this mode. In case of the frame sending event, it checks whether it was a request frame or step ending frame (S5). When a step ending frame was sent calculated sum is indicated on the display device (S8). Sending event is marked by appropriate flag. Then receiving of the request frame on the current step is checked (S6). If a frame has been received, then the node re-enters in the flags poll mode. In other case indirect decision as to this node was the first which managed to have sent a request. Thus, the transmission of the current step ending frame is assigned to this unit (S7). Then switching to the flags poll mode takes place. When the frame has been received, its type is checked (S9). If this is a standard frame, the value of its data fields added to the already calculated sum value (S10) and the node enters in the poll mode. When a remote frame has been received, the receiving of the request frame by the current step is controlled (S11). Then the value of the calculated sum is displayed on the indicator (S8). Otherwise, the unit goes into error mode (S12), which means that the unit could not send a request by the current step. When the transition to a poll flags mode events occurred during a frame receiving from the line, the timer is stopped (S13). If, during the poll mode simultaneous sending and receiving frame events are registered, the device goes into error mode (S14), which means that an intelligent device module is too slow to ensure the realization of the task. After the calculated sum has been indicated during the current step, the device goes into the set timer mode (S1).

Figure 2. Graph-scheme of the step by step node data exchange via CAN bus

(ft – delay ending control flag; fT – request frame sending control flag; fTx – frame transmission control flag; fRx – frame receiving control flag; fR – frame receiving for whole step control flag; fD – standard frame control flag)

In the course of the experiment on a model segment nodes incrementally displays the sum of the sended request values. The request values magnitude at each step for different devices was opposed by 10 units. When devices has gone to the next step values have been increased. Thus, each new value of the sum was unique. CAN bus has been tested at all standard speeds.

Obtained in the course of the experiment characteristics depending between the calculated sum and the current step had a linear character. Without using the synchronization delay some nodes has gone into the error mode.

The following conclusions are based on the results of the experiment:

- CAN bus can be used to organize decentralized management of technological objects based on the proposed step by step data exchange approach;

- for the correct limited resources allocating by CAN on the devices it is necessary to use synchronizing software delay.

References:

1. В.И. Варшавский, «Коллективное поведение автоматов», изд. «Наука», Москва, 1973, 408 с.

2. CAN Phisical Layer. – CIA. – 1999. – 43 p.

3. G. Gruhler, G. Pivnjak, V. Tkachev, L. Tsvirkun, D. Poperechnyy, «Very large hierarchical CANopen systems in mining», CAN Newslater 4/2004, 48-54p.

4. Г.Г. Пивняк, С.Н. Проценко, С.М. Стадник, В.В. Ткачев, «Децентрализованное управление: Монография». – Д.: НГУ 2007. – 107 с.