Getting more from production systems?

Manufacturers are achieving significant business benefit by exploiting shopfloor sensors and data collection equipment that they already have. These devices, installed to achieve one or several business(es) or manufacturing objective, could now also provide information that if collected and analysed by other, perhaps higher level, systems might, for example, cut waste, improve quality, reduce rework, rein in inventory, or provide real-time factory performance figures and thus improve agility and responsiveness. Factory automation products maker Omron’s business development manager Dean Chespy says: “Systems are put in for a purpose, and by the nature of what they’re addressing they have the inherent capability to do so much more. All the data is there.” While that may be a little hopeful, the fact is that companies’ ability to make imaginative use of whatever existing data they have to achieve new business objectives is likely to be the sticking point. It’s quite often human inhibitors, rather than technology, that are preventing manufacturers from going this step further. Steve Hudd, managing director of system developer and integrator BsquareT, says: “Often the very people that are the recipients of this technology don’t understand what they want themselves. Traditionally, you had a problem, wrote a specification and applied the technology, and the chances are you could solve that problem. However, with a little thought you could go the extra mile to provide a solution that the production guy would have asked for had he known what was available in the first place. It’s very difficult to know what else you want until the product is installed.” Good point. Another human aspect is culture, and in particular the silo mentality frequently found throughout manufacturing industry. “These things are put in by different functions within the company,” says Chespy. “Engineering may be responsible for the operation of a particular machine, and other departments aren’t getting the benefit from it because responsibility for it stays in that domain.” At West Midlands manufacturer Bromford Iron & Steel, chief engineer Bill Venables has a solution to the communication problem. “I go to the production meeting to find out what they want to know. I listen to arguments where they say this caused the problem, no that caused it. I can say, ‘give me 10 minutes and I’ll give you the information on what the real cause was’.” In enterprise management software author Baan’s approach, integration of its Wonderware InTrack MES (manufacturing execution system) layer with Baan ERP business and supply chain software goes some of the way by providing tailored information portals to various departments throughout an organisation. “Different people want different types of information – quality people, sales staff, customers, the boardroom, engineers, even suppliers of raw materials,” says Michel Chouinard, product and solutions marketing manager for iBaan Manufacturing software. Think about improvement The bottom line is that solutions from Baan, Omron and the others can extract data from shopfloor devices to provide a different business context, and help with a user’s business and manufacturing process improvement. Says Chepsy: “How can you improve anything if you don’t know how well it works? You may have a policy of continuous improvement but what are you improving?” And he adds: “You have to understand how it operates now before you can make any decision as to where it wants to go. For example, where are your bottlenecks? You may be trying to improve the efficiency of a machine that is running at 30% but that may make no difference if it’s not the bottleneck.” According to Andrew Ballard, managing director of specialist MES integrator Fortetion (which works particularly with big boy Emerson’s Intellution software), the best place to start on the road to exploiting shop floor devices is simply checking what equipment you’ve got, and what it can do. “At one pharmaceutical company we found two half-million-pound pieces of DCS (distributed control system) equipment in a cupboard that they didn’t even know they had. So do a proper survey of what you’ve got. Then analyse the different ways you can communicate to that stuff, the best way of getting data in and out. Then we advocate putting all that data into a repository – data from PLCs, DCSs, barcode readers, check weighers, SCADA (supervisory control and data acquisition) systems – and then mine that data.” The secret to being able to analyse that data to improve performance is to be able to allocate data sets to batches, as BsquareT’s Hudd explains. “We then synchronise what you’re interested in measuring – it could be taste, in food production, or energy usage in a steelworks – with the different batches.” At Bromford Iron & Steel, this comparison is achieved using BsquareT’s real-time production analyser ICEberg, working on information both from existing and new shop floor sensors (Manufacturing Computer Solutions, December 2001, page 52). “If you are running the same product on a daily basis, you can compare day by day and shift by shift, and look to see why one shift was better,” says Venables. By looking at yield analysis, for example, for every product over every production run, a manufacturer can start to see the anomalies, and identify where improvements might be made. “If you align the best with the worst production run, probably 95% of the data is the same, so you discard all that and investigate the differences,” says Hudd. “If all the data is the same, then you’re probably not recording what you need to be recording.” Snooping on everything At Thales Acoustics, which manufactures professional audio ancillaries, Kronos ShopTrac Pro is the system used to analyse data from the shopfloor and calculate efficiency figures. It can measure individual operators, production cells or departments, piece parts, sub assemblies and final assemblies – everything from entire works orders or contracts, to individual plant or tooling. Analysis like this can often help ameliorate the thorny problem of the real cost of short production runs. “We know that short run lengths cost the company money – what we often don’t know is how much money,” says Hudd. “We have to take the changeover time as a proportion of the run length, and we often find that short runs compared to long runs cost two and a half to three times the manufactured cost.” At Bromford, for example, energy cost is a key variable in determining manufacturing costs. “We can look at both gas and electricity cost per tonne, and thus what sizes we shouldn’t be running at the mill,” explains Venables. “Some sizes were quite uneconomical, but certain sizes that at first looked uneconomical – we found if you increased the tonnage up to a certain amount, all of a sudden they became a very good product.” Many other benefits are achievable from having the ability to analyse all this data. Geoff Kneen, head of business development at system builder WS Atkins, says: “One of the other benefits is asset management. What we’re seeing people do is ensure they are getting the right asset performance out of their shopfloor equipment, as well as ensuring they are using the asset to its optimum – to perform as the manufacturers have told them it will perform.” Again, Bromford is an example. Venables says the firm now uses its data also for maintenance and replacement planning. “Initially we wanted to see where all the stoppages were in the mill, to find out which items of plant were breaking down so we could reorganise it to run more efficiently. But we also wanted a more accurate costing, so we would be able to say if it’s costing so much to maintain over 12 months, maybe we would be better off pulling it out.” Rework is something else that analysis of shopfloor data can help address. At electronics manufacturer Celestica Electronics, extensive use is made of the data available from assembly equipment, including pick and mix machines. First, rework is prevented where possible by testing components before assembly. The component value is known to the system, so it’s a straightforward matter to request that an operator takes a component from a reel to test it before the reel is used in production. Real-time information also helps prevent rework, says Robert Wood, strategic operations director at Celestica: “The machines are being monitored in real-time, so whenever they fail to place a component we know about it. In the past we’ve always had simple reports of what’s going wrong, but with this system it’s reported in real-time. We feed it through a materials review board area in our ERP system and can quickly get engineers to go and look at the issues. If we get a trend, we get immediate feedback and a focus to get our corrective action working.” Cost reduction was also achieved by a reduction in supplier-managed inventory on the shop floor. “We used to have two reels waiting for each machine, one spare and one in case it’s faulty,” says Wood. “But with our Sony TiMMS [production management] software, it counts the placement of every component, so it can predict reel exhaustion. The new reel turns up just before you need it, which gets material off the shopfloor, and improves our cashflow position for materials where we use supplier-managed inventory.” Know where you are? And productivity improvement is another benefit. Wood: “Because the supervisory system knows exactly what’s going on, it records all forms of downtime. In the shift review, we know exactly what the deficiencies are and so we can focus on how to address the issues and not how to collect the data.” Fortunately, achieving these and other benefits is not usually rocket science. Information from existing sensors can be collected by a plant historian (fast database), and then dedicated applications, or a general-purpose data mining and analysis tool used for reporting to help with different business needs. At Bromford Iron & Steel new sensors were installed, but this is unusual, says Ballard: “People seem to think they have to throw away their infrastructure and start again, but we have never done a job ever where we haven’t been able to pull data from the existing equipment.” Kneen confirms this finding: “In many cases they do have the raw data they need. If a photoelectric cell is monitoring goods passing on the line, the current automation layer will just tell you when that has failed, but it could count the products going past.” Thus even the simplest of sensors, installed for the most basic of purposes, has the potential to provide information that can enable a manufacturer to obtain very valuable insights into today’s production performance, which help to yield improvements for tomorrow.