Decentralized Motion Control
Motion control is becoming increasingly complex in order to meet market demands for greater system throughput and improved output quality and precision. These objectives must be accomplished while decreasing overall machine costs. Conventional system designs must be replaced by innovative, high performance solutions. Decentralized motion control is proving to be an attractive alternative by removing the constraints imposed by an expensive and unnecessary centralized controller. Larger numbers of synchronized axes are possible on a single network that provides a high speed, scalable motion solution.
One such project that has benefited from this is an Eagle Manufacturing machine designed for an automotive supplier. The Eaglematic™ notch and cut-to-length fabrication system uses decentralized motion control on a real-time, Ethernet communication network to control 22 servo axes with cycle times of 400 µs.
Decentralized systems divide the motion control responsibilities among the drives. No central motion controller is required. The drives are responsible for their own axis parameters. One or more master axes transmit the master position over the network. Each drive calculates its respective set position internally.
Processor and network load is drastically reduced and system performance is greatly improved in a decentralized system. The job of coordinating the axes has been distributed into small, quick tasks in the drives. Instead of transmitting and receiving set points for each drive, only the master position needs to be broadcast over the network. This broadcast data can be received by all nodes on the network. The motion control performance is not dependent on the processor performance or the number of axes being run by a motion controller. System cost can be greatly reduced by eliminating the central controller. The processing power of currently available, cost-efficient drives can easily process all control loops along with any additional decentralized motion controller features.
Because system performance is no longer dependent on the motion controller, much larger numbers of axes can be configured on a single network. The maximum number of drives is only dependent on the capabilities of the network. Network bandwidth can be maximized by multiplexing time slices for slave drives while master axes are able to broadcast to the network every cycle.
Real-Time Ethernet-Based Communication
Although decentralized motion control allows a network to be used much more efficiently than a centralized design, communication is still a key factor in motion performance. Current, velocity, and position control loops are processed at different cycle times. Typical cycle times are 50-100 µs for the current control loop, 150-200 µs for the velocity control loop, and 300-500 µs for the position control loop.
For systems in which the control loops are closed inside the centralized motion controller, the communication needs to be fast enough to supply sufficient bandwidth to transfer setpoints and feedback values back and forth. As a consequence, compared to a completely decentralized system at the same performance, the centralized system needs to transfer information up to 4 times faster.
This becomes increasingly challenging for digital communication systems such as Sercos, ProfiBus, and DeviceNet, which provide a reliable and economical interface but are only able to communicate in the multiple millisecond cycle range. The EaglematicTM application requires a synchronized position update time of 400 µs, which is more than 10 times faster than the cycle times of the networks mentioned above. When greater numbers of axes are connected to the motion controller, there is increased traffic on the network, since every axis communicates the same amount of information to the motion controller and receives the setpoints in return.
With its peer-to-peer communication capabilities and high communication bandwidth of 100 Mbaud, Ethernet is the perfect medium to transfer vast amounts of information within a minimum amount of time. There is a common familiarity with Ethernet, and standard hardware is available at a minimum cost. However, Ethernet has one limitation due to latency caused by collisions. If a collision is detected on the bus, each sender resends the information after a randomly determined delay. This has prevented Ethernet from being used in motion control applications in its standard form.
Developments in recent years have been successful in providing Ethernet with a high determinism (jitter less than 1 µs) by introducing time slicing mechanisms into the Ethernet protocol. This makes Ethernet capable of real-time communication and suitable for use in motion control applications. The open standard Ethernet Powerlink, developed by B & R and now distributed by the Ethernet Powerlink Standardization Group (EPSG), headed by the University of Zurich in Winterthur, turned out to be the ideal real-time Ethernet solution for the Eagle 22-axis application.
Ethernet Powerlink transfers data in two ways: an asynchronous channel and an isochronous channel. In the asynchronous channel, non-real-time critical information like Internet Protocol (IP) based data (e.g. a video stream) can be transferred. The asynchronous channel is also used to upload and download information to and from the drive (e.g. drive firmware to the drive or motion traces from the drive). Standard IP-based protocols and addressing are used during this phase.
In contrast, the isochronous channel consists of time slices, which are assigned to nodes for broadcasting on the network. All other nodes can receive this data. For the cut-to-length machine, 22 nodes (one drive per node) were linked via Ethernet Powerlink. This would have resulted in a network cycle time of 800 µs. Within the 800 µs cycle time, each drive, represented by one time slice, receives the right to broadcast information, while all the other nodes listen. The network communication manager, represented by the Power Panel, regulates the traffic on the network and initiates each broadcast by a poll to the respective node.
In order to decrease the network cycle time even further, multiple Powerlink nodes were assigned to a single time slot, providing a prescale factor to less important nodes while preserving single-use time slots for nodes that broadcast cycle time critical, master positions.
Each cycle time, another prescaled node per time slot is polled by the network communication master, while broadcast information is received by all nodes on the network. This communication and decentralized motion control concept allows the communication of multiple master positions to up to 240 drives within one Ethernet Powerlink cycle of 400 µs.
Integrated Hardware Architecture
Although a decentralized motion architecture is crucial to the success of the EaglematicTM machine, the system also benefits from several other innovative designs. Traditional automation systems exhibit a clear separation of human machine interface, PLC, motion controller, and communication. Each of these typically requires its own housing, processor, communication ports, and programming software.
For the Eagle project, a B & R Power Panel was chosen that combines these into a single unit with one processor and programming software. Because the highly processor-dependent motion tasks have been decentralized to the drives, this single processor solution becomes not only possible, but ideal. Logic control and visualization tasks can be executed at much lower cycle times. This integrated design also reduces machine cost, space, and installation time because communication is no longer required between separate HMI, PLC, and motion controller components.
Integrated Real-time Operating System
Equally as important as the single processor design is the use of a single programming software. Automation Studio, a real-time multitasking operating system, allows operations to be executed in deterministic cycle classes in which high speed tasks are completed with the shortest possible cycle times while HMI and other low-priority tasks can run at slower cycles. In this way, maximum performance is realized for a given processor.
The use of a single software package also streamlines programming and system integration. A transparent variable database is used for visualization, logic control, motion control, and communication. All drive parameters are accessible from the main logic control. If a drive is exchanged in the field, the control unit can check its compatibility and download the appropriate firmware regardless of the drive''s original firmware version. This real-time operating system and real-time communication network allow motion control, logic tasks, and communication to be highly synchronized for maximum performance.