OBJECT-ORIENTED ROUTE TO INDUSTRIAL AUTOMATION

The term 'Programmable Automation Controller' is becoming increasingly heard in distributed automation. Combining network connection, motion control and I/O functions, this new PLC replacement has become the central device behind industrial mechatronics. Conveyor belt automation - itself a prime application for mechatronics - has been taken to a new level of sophistication. Keith Anderson describes the use of PROFINET RT in component-based automation systems.

Materials handling control systems have evolved from a hierarchal to a decentralized control concept over the last ten years. The first move in this direction was to place I/O remotely around the material handling equipment using fieldbuses. The next development was the arrival of small PLCs with a network connection which began displacing large, centralized controllers. These smaller networked devices enabled greater modularization of subsystems and components together with the ability to use a standard systems interface. Taken together, these architectural changes have reduced costs through control system standardization.

There has also been a parallel trend in the design of control software: the object-oriented approach to programming as typified by IEC61131 has largely ousted sequential programming techniques. This allows for standard objects to be coded with defined interfaces using the IEC 61131 configuration. The same programming trend also eliminates the expensively repetitive hard coding of interfaces.

Siemens Logistics and Assembly has developed a new conveyor product line (CL100) that employs a mechatronics approach. At the lower level of the architecture, a motor with built-in motor rotation sensors now powers the conveyor roller. This design aspect eliminates the need for gearboxes, a major failure point in legacy systems. Each motor is connected to a networked motor controller which also includes I/O for sensors and actuators. These motor controllers also monitor the motor condition and the package flow though the control zone. The motor controllers connect through a self-addressing network to a programmable conveyor bed controller. This supervisory device uses Programmable Automation Controller (PAC) technology. A material handling system can be then created by connecting and configuring these mechatronics conveyor components. This mechatronics component approach also provides standardized performance monitoring, diagnostics and easy system configuration and reconfiguration.

The PAC provides the backbone to the system design. The device itself combines traditional PLC with PC functionality. The PAC used in this particular product is a low cost unit that fits into the side channel of the conveyor. It talks to host systems over Ethernet and uses a self-addressing fieldbus connection to communicate with lower level components.

Peer-to-peer communication

To create a materials handling system with these mechatronics conveyor components requires a standard interface capable of supporting peer-to-peer communication. These requirements are fulfilled by PROFINET Component Based Automation (CBA) and Ethernet. PROFINET CBA defines an automation concept that allows the definition of component and interfaces through a standard engineering model. It also provides a runtime model for the execution of the component interface along with an engineering model that in turn defines configuration of the components used to create a system.

The installation environment for a materials handling system differs from the office environment in several ways, principally with environmental, topology, performance and security requirements. These factors need to be taken into account in during network design and component selection.

First there is the physical layout of a materials handling system. Clearly network topology has to match this. The physical paths on which the conveyed products flow follow the building and production line requirements set by the end user.

The network topology for a mechatronics base material handling system starts below a router which is connected to a customer LAN. Below the router are switches connected together in an high speed redundant ring. The conveyor lane with several PACs appears at the next level down in the overall system architecture. Since each PAC has two exposed Ethernet ports, they can be connected in a row. This differs from the normal office topology where there would be a switch with several leaf nodes.

Network performance requirements for a materials handling system are all about responsiveness. Material Flow Host routing must be done in less than a second to maintain proper material flow rates; these networks are typically isolated from the office network for performance reasons. Discrete control paths must operate in the millisecond time frame with guaranteed network determinism and latency parameters. It requires an isolated network to achieve this.

This isolated materials handling system network - the control network - can typically include 1500 PACs, 250 managed switches, 20 HMIs, 100 identification devices (RFID or barcode), a visualization system (SCADA), OPC servers, a network management system, material flow host, wireless access point, security devices (VPN and Firewall), clients and a router tied to a corporate LAN. Taken together, this represents a network significant size. A Network Management System (NMS) takes the support role in providing configuration management for the network topology and devices. It also assists maintenance tasks.

Handling the network traffic

A 10s system latency occurring in the context of office automation would hardly be noticed. But at conveyor speed of 2m/s, a 10s hesitation would take a package some 20 meters beyond its intended shipping lane. The majority of network traffic within a material flow system takes the form of discrete controls communication generated as packages move through the system. This takes place between adjacent PACs along the materials flow path. This communication is accomplished using PROFINET RT, a Layer 2 protocol using VLAN tagging (IEEE802.1Q) and priority handling (IEEE802.1p) to allow for better determinism. Each PAC has a built in three-port store-forward switch: one port is connected to the internal ASIC while the other two provide the external connections.

Layer 2 switches are typically based on store-forward queuing technology. Once a complete packet is received, it is placed in a queue, and then sent out on the correct port. The amount of traffic flowing through a switch to a specific port will have an effect on the delay in the switch.

IEEE802.1p provides a method for the prioritization of Ethernet traffic. This is accomplished in the Media Access Control (MAC) frame layer of the ISO network model. The 802.1p standard defines eight classes of traffic, with the higher numbered class having the higher priority. Once a Layer 2 switch with IEEE802.1Q and 802.1p sees a tagged packet, it evaluates its priority and, if it is higher than its peers, moves the packet to the front of the queue. This reduces the non-deterministic delay of going through a switch.

PROFINET RT is cyclically framed, i.e., data traffic is generated against a timebase set for 8ms - which makes it the most frequent traffic on the network. However due to the fact that switches route traffic to the port where the end device is located, this traffic is confined to the network topology path between the two physically adjacent PACs. In a mechatronics material handling system this PROFINET RT traffic has a bandwidth use of 1% between the two PACs.

Visualization network traffic

The next major source of communication occurs between the PACs and the visualization systems. Visualization is important for both the operation and maintenance of a mechatronics system. Just as each mechatronics component is standardized, the same standardization can be applied to many aspects of the visualization process. Standard objects can be created for each mechatronics conveyor type, with key performance indicators, status, standard diagnostics, troubling shooting aids, documentation links, part list, and online ordering.

Truly distributed control in automation: a view of the conveyor belt sidechannel showing the PAC (centre left) and associated power supply (right) controlling an individual belt section. The heaviest network traffic takes place between adjacent PACs on the belt line.

The visualization system communication runs on PROFINET DCOM (HMI) over TCP/IP, with OPC PN servers for the SCADA system and HMI panels. If we look at the switch data flow when the OPC PN server is located on the network, bandwidth usage amounts to 24% for a network comprising 1500 PACs.

Material flow host traffic

Package routing information exchanged between PACs and a Material Flow Host (MFH) generates substantial network traffic. The MFH performs the logical routing of packages through a material handling system by communicating with the mechatronics conveyor components that perform the physical routing. When a package arrives at an identification point, it makes its arrival known with an RFID or barcode scanner connected over Ethernet - which communicates with a PAC. The identification and destination requests are then made to the MFH. The MFH responds with a destination for the package which is then used by the conveyor routing components to deliver the package to the correct destination. The routing component also notifies the MFH when it has routed the package. This allows the MFH to keep track of package location within the system.

The volume of generated traffic relates to the quantity of product flowing through the materials handling system at any given time. This communication data uses TCP/IP protocol. The bandwidth requirement would be typically 1.2% at the MFH from the 250 PACs (notification and routing components in a 1500-PAC system) at a material handling system flow rate of 21,000 packages an hour.

There is also network traffic generated by the NMS. There are two types of communication within a running system: SNMP, and periodic pings to network devices. This traffic is well below 1% and of no concern.

Latency

As we can see the bandwidth usage is low and poses little concern for system performance. At the network root we will never see more the 35% peak bandwidth demand. Of greater concern in a controls network is latency and non-deterministic jitter. In the materials handling context latency is the time delay between transmission and reception of signals between PACs. Jitter is the difference of this delay over repeated transmissions.

When multiple conveyor lanes merge into one path all of the PACs needed to supervise this task cannot be connected together directly as each only has two ports. This requires that communication takes a different network route to reach the PACs on the merged conveyor lane. Depending on performance requirements and system layout the communication path between two PACs may occur through several other devices. The average store-forward delay with the current switch in the PAC is 125µs. If for instance the transmission went through 200 PACs, the best case delay would be over 25ms. This would throw up a system error with 8ms PROFINET RT (delay error flagged at RT QoS period x 3). However 20 PACs would result in a delay just over 2.5ms which may be acceptable, as is the case with most material handling systems.

Keith Anderson is principal engineer, Siemens L&A, Grand Rapids, USA

First published in the Industrial Ethernet Book, July 2005

Published by Profibus Trade Organization

16101 N. 82nd Street, Suite 3B

Scottsdale, AZ 85260                                            feedback@profinews.us

©2005 PTO