Machinery safety is often viewed as a market sector where customers are conservative and progress in developing new and innovative products is perhaps slower than elsewhere. Over the last ten years or so, machine builders, system integrators and line builders in Europe have come to terms with CE marking in accordance with the Machinery Directive. At the same time, machinery safety standards have developed, though typically at a slower pace than the corresponding technologies. For example, programmable electronic systems have been successfully and safely used for a number of years, yet it was only in June 2006, with the publication of a new version of EN60204-1 (Safety of machinery. Electrical equipment of machines. General requirements), that the use of such systems has been accepted by the relevant standard.

All types of safety-related input device have seen steady development over the past decade, ranging from mechanical safety gate switches, non-contact switches and pressure-sensitive mats, to light curtains and laser area scanners. The last two are especially beneficial for guarding machinery that needs frequent access by operators, or for parts to be fed into or out of a process. In addition, maintenance is quicker, as the absence of physical barriers means that access is easier.

Nevertheless, such non-physical barriers have their limitations. They cannot, for example, prevent broken parts or tools from being ejected, they do nothing to dampen noise emissions from machinery, and they do not prevent workers being exposed to arc flash, fumes or dust. In addition, light curtains and laser area scanners only monitor a flat plane – though mirrors or multiple devices can be used to create faceted protection zones when using light curtains. As a result, the protected area is often greater than the hazardous zone, which is a waste of floor space and can also mean that a stop is triggered earlier than is absolutely necessary, costing valuable process uptime.

In the world of mass production, where floorspace is very expensive and time is of the essence, it would clearly be advantageous to have a non-physical safety barrier that can protect a three-dimensional zone that closely matches the hazardous zone. And this is exactly what Pilz and DaimlerChrysler have done by developing Safetyeye, a safety-related sensing and control system that is based on machine vision.

The two companies have worked closely together, with Pilz taking overall responsibility for system development and providing the expertise behind the safety functionality, industrial design and programming software. DaimlerChrysler developed the image processing algorithms, specified the practical requirements and supported the comprehensive test programme.

System architecture

Each Safetyeye system comprises a sensing device with three greyscale cameras – which gives three-dimensional coverage – and an analysis unit that contains a

high-performance computer and a programmable safety and control system. The three-camera sensor unit is mounted above the application, enabling a zone to be monitored around a hazard machine (Fig.1). Maximum coverage would be for a three-dimensional envelope that is approximately conical with a base area of 12.8x9.6m and a height of 10m, though the user-friendly configuration software enables zones to be defined within this, both as warning zones and detection (danger) zones. If an operator enters a warning zone, an alarm can be triggered and the speed of the machine reduced; if the detection zone is entered, the machine will be brought to a stop.

Reference marks on the floor or surrounding structure are monitored by the cameras to ensure the sensing device is correctly aligned, and lines would typically be painted on the floor so that operators can see the extent of the warning and detection zones in order to avoid any unintended triggering of the system. Depending on the height at which the sensing device is installed, the system can detect intrusion by body, leg or arm. It is technically feasible to add finger detection, but the initial market requirement is for body, leg and arm detection only.

Signals from the cameras – at a rate of 20 per second – are analysed by two separate image processing systems running on the analysis unit’s two computers, each of which have different operating systems. The outputs from the image processing system are then compared by the programmable safety and control unit, which equates to a dual-channel input being fed to the dual-redundant programmable safety unit. If a breach of the zones is detected by either one of the vision systems, an appropriate output is issued by the analysis unit to signal an alarm, reduce the speed of operation or halt the process, for example. The vision system algorithms are set to respond to violations that extend over three successive images, so the response time for the system is 150milliseconds.

By using the configuration software, zones can be grouped and muting can be programmed to enable parts to be fed into or out of a process. If the application requirements change, then the zones can be quickly and easily reconfigured in software alone – with no hardware changes necessary (Fig.2). Blanking is expected to be added to future versions of the software. Zones do not have to extend from floor to ceiling, they can have their heights specified. However, the zones are fixed, rather than varying dynamically with the position of the robot. While the main emphasis is on the detection of objects that enter the configured zones, it should be noted that the system also detects if a robot or workpiece breaches the programmed working envelope.

Furthermore, the analysis unit is also able to perform some conventional (non-safety) control functions. In one application that has been trialled, a robot was placing components into an oven and then placing them in a collection bin. Safetyeye was used to create a safety barrier around the robot, and the analysis unit also monitored the amount of product in the bin so that it could be emptied once it was full.

An interesting aspect of the system is that any breaches of the warning or detection zones can be recorded. While managers may find this useful, workers’ representative bodies may be less keen to have their members identified and, potentially, disciplined.

Cost-effective alternative

With a price of around E12 500 per system, Safetyeye may seem expensive. But Pilz argues that, compared with conventional safety light curtains, the total cost of an installed and maintained system can be around 70percent lower. Most of the difference is accounted for by the fact that a Safetyeye system can be installed and commissioned in around two hours, whereas a conventional safety system would take a whole day or more.

Nonetheless, it has to be remembered that the non-physical barrier presented by the Safetyeye system shares most of the limitations of light curtains and laser area scanners: it cannot protect workers against broken parts or tools, noise, arc flash, fumes or dust. As such, it is more likely to replace safety light curtains and laser area scanners than conventional physical guarding.

After Safetyeye was launched at DaimlerChrysler’s Sindelfingen site on 15 November 2006, two weeks later it went on to cause a great deal of interest at the SPS/IPC/Drives show at Nuremberg. The system already has concept approval from BG and full product approval is due to be in place by April 2007. After that, trial systems will be offered to prospective customers, and production systems are expected to be available in September 2007.

The first application for Safetyeye will be on the production line for the new Mercedes-Benz C-Class (Fig.3). Meanwhile, the team developing the Safetyeye system are looking ahead to possible future enhancements. One possibility is to distinguish between, say, a human hand and a tool. If this can be achieved, it would pave the work for a human operative and a robot to work in parallel on the same assembly.