Recipes for success

Here's a question. How do you improve overall equipment effectiveness (OEE) in your plant from 60% to 75%, quality from 98.1% to 99.6%, and dramatically improve flexibility, control, documentation and traceability? Answer: invest in a modest modern batch control system. That's what Graham Hassall, plant manager at Ineos Silicas did at the company's Warrington site, which manufactures liquid and solid silicates, gel and precipitated silicas, and a range of zeolites. As is often the case, necessity was the mother of invention. Hindered by a succession of changes of ownership (the company was formerly under Unilever and ICI) and a lack of investment, Ineos Silicas had a Taylor batch control system dating back to 1981. "It still worked," recalls Hassall, "but there was no longer support." On the hardware side he admits to cannibalising machines from other Ineos sites across the world. The solution, Emerson's DeltaV digital batch control system now controls the plant's 'wet end', including the chemistry of batch making. "It's recipe driven – we make 17 different slurries," affirms Hassall. "It's three streams all making the same product with controlled dilution of raw material, raw material preparation and the batch reactors." More than that, it also controls effluent treatment (previously manual), filtration slurry or drying operations and packing, which the old system could not. This has "given infinitely more control over effluent into the Mersey, plus better control over batch making, and it's widened scope of control." The old system was ripped out and the new one installed in just three weeks, but while it was well planned and executed, Hassall's candour in admitting to his first mistake is refreshing. "We were also doing another project – an automated packing system," he says. "It was a mistake to try and do two projects at once. We did not have the resources [and] both suffered." Neither did the project have a formal project team, and Hassall doubled up as process engineer and plant manager. "That was a mistake too – it needed a dedicated process engineer: it took far too much of my time." Nevertheless, improved flexibility has been achieved by creating a comprehensive recipe portfolio, and new formulas can now be added in 10 minutes. Moreover, the system is object-orientated, meaning the same software can be used on different vessels: hence improvements and modifications can easily be made. Systematic but limited Meanwhile, in terms of hardware, electrical design of the control system was outsourced, with not a single error in 500-plus I/O checks. Ineos also used the opportunity to introduce 'smart' digital instrumentation, including some 20 flowmeters and temperature probes. Success there was down to a "systematic approach to implementation and commissioning." The bottom line: expenditure on the system was in the region of £400,000, but Hassall estimates savings through quality improvements have already reaped £250,000, so the system will pay for itself very quickly. Clearly, this implementation is self-contained and for plant management improvement, with little linking to other business systems, but Hassall doesn't discount that possibility in the future. And that's been the way of it, well, ever since. But there's a host of newer, emerging technologies seeing the light of day, and companies like Ineos form what Ruben Gil, director of MES (manufacturing execution systems) at process management systems specialist Aspentech, sees as the 70% still adopting current or older approaches, while the other 30% rump take advantage of next generation, business-linked technologies. You may take issue with the percentages: our own view is nearer 90% and 10%, but his points are worth attention. So what will be the benefits? First, look at automation scope. Currently process- or product-driven, a newer approach is around managing whole plants, including non-production functions. Indeed, new 'blockbuster' drugs in pharmaceuticals are driving MES standardisation across plants, allowing product manufacturing to be distributed, even outsourced – while maintaining product genealogy across sites and countries – precisely by linking into the business IT infrastructure. It's being done by providing a notional layer on top of scheduling, specification and the rest – and information transfer between sites with their own IT infrastructures. However, Gil concedes that: "There is some technology to do it but the application is still being developed. Integration between MESs is still a problem." Better standards Batch control, currently based on the ISA S88 standard with production areas and process cells, is giving way to ISA S95, focusing on whole plant operations including integration with business processes. But the future, says Gil, will increasingly see more horizontal integration of process development and manufacturing. "No-one has succeeded in doing it," he says, "but pharmaceutical companies are looking at it." And he believes that when it happens, it will provide even more value than integrating plant and business systems. Moving on, look at compliance: traditionally fuelled by ad hoc initiatives, including incidentally 21 CFR 11, the new approach is plant-wide, incorporating systems that standardise production procedures. "We are starting to see enterprise compliance manufacturing," insists Gil, with a general model at the corporate level rolled out for each plant. One MES supplier is chosen, while auditing of all development and implementation methodologies defines standard ways of working in each plant. But to succeed with that, "you need a lot of buy-in from people in the plant," Gil agrees. It's early days, but whereas MES implementation is still considered primarily a process cell level deal, there is a need now for a plant compliance strategy, with manufacturing programme managers responsible for all sites. "The logical end is global MES implementation," says Gil, which could mean internal competition between plants disappearing, although users are currently only piloting on key sites. Looking at integration objectives, the old approach was integration through OPC with control devices. The new approach is increasingly about using XML, Web services and other EAI (enterprise application integration) technologies, focusing on the 'publish and subscribe' model. Integration with customers is a key driver, creating portals to track customer orders, status and quality parameters for example, especially in speciality and fine chemicals. Users are also now starting to concentrate on connecting plant IT to business ERP for real-time status of cost and business unit activity. Customer service now is about increasing flexibility and improving 'capability to promise', so real-time event management and production scheduling have to be more closely integrated and responsive. Making some plant data visible to the customer also helps form special relationships and trust. However, there's still a need to better integrate process development with manufacturing: Gil believes these still exist in parallel universes. "There are good tools for process development to simulate, streamline and scale-up, but when that finishes there's still no link to manufacturing." Recipes are rebuilt from scratch and don't re-use work that's been done before. Gil's battle cry is for information collected on plant to be fed back into development. With the potential now for role-based visualisation, users will want access to applications and data necessary to do their job. At Ineos, for example, team leaders sat down with the IT vendor and designed the graphics for HMI visualisation. Taking that a stage further, role-based workspaces will allow users to log on and get a navigation window to the parts of plant and information for which they're responsible. As on most plants, that's not yet the case at Ineos. "Perhaps in the future there will be a need to do this more at the corporate level now that global MES is here," prophesises Gil. "There is a need to have a product view, but I haven't seen it at the enterprise level." He sees a future of mixed plant and business event management. "Say a new order comes in, or there's a batch delay: this creates an issue that customer service should know about and advise the customer. It is plant and business linked together." And that leads us into KPIs and reporting, which have historically been ad hoc, using applications like Excel. "We see a need for real-time plant performance score cards that you can drill into," says Gil, again role-based. Integration between performance scorecard and MES analysis tools is vital for that. The caveat: "It needs to be [driven] top down to enforce the same measurement system." The danger is plants operating on diverse KPIs and, let's be frank, a tendency to make them look good to observers. "You can still have local KPIs, but also global manufacturing KPIs. The more we automate processes the better – that way no-one can modify the results!" Use of modelling technology will also be instrumental here, he suggests. "This tells you how much you could get out of a manufacturing unit and how much you are actually getting." More realistic KPIs are the result. Gil predicts there will be models that predict and simulate plant behaviour. "You have a model and on-line information, and by connecting the two you can predict the future." He cites Dupont, which is using a modelling/on-line system on a distillation column. "If an operator wants to change the setpoint for flow, he can simulate the effect." Then he can see whether the action is appropriate. BP Oil in the UK is using an Aspentech solution to predict optimum energy usage on a plant – the results of which are used in calculating purchase orders for energy consumption. Modelling is also being used at Ineos to a degree. PCs and monitors adjacent to the control room "virtually commissioned the plant while it was running," says Hassall. "We could almost tune the thing in before putting it to work. It was a huge step forward." Other similar initiatives include real-time and multi-variable statistical process control (SPC), aimed at finding correlations between process variables and quality data to predict product quality based on current process data. There are some very sophisticated SPC tools but most in use remain off-line, analysing what happened after the event. The holy grail is real time SPC, generating alarms as soon as there is deviation, with trend spotting also enabling better process understanding through analysis of historical and real-time batch data. At Ineos, for example, Hassall gets SPC control in real time, and, via his batch management and Aspentech's optimisation systems, he also has instantaneous energy monitoring data and effluent records. More advanced systems What about advanced control? Whereas it's remained largely the preserve of continuous process development, Gil says he's seen advanced non-linear multivariable controllers in polymer production. "The future will see advanced control in pharmaceuticals, thanks to on-line analysis." But he agrees that, despite some prototyping in special chemicals, it's still early days. It's a slight aside, but detailed scheduling at the production level is also a reality, and the near future includes real-time scheduling and capability to promise, with implications mapped and costed, and customers informed. What does it all mean for your people? Staff and operators will increasingly spend less time looking for information, and more on data analysis and prediction. Potentially, the biggest challenge will be finding suitably qualified plant engineers. It will be tough maintaining what will inevitably be more complex systems. It's a changing world, and best advice is don't select a tool for one task, but look first at the whole plant systems picture, including MES and ERP, and what's now possible and desirable. Second, simplify your management and associated IT architecture: you'll need MES systems that complement ERP, and where possible it makes sense to remove islands of automation. Third, take it a bit at a time. "It's modular – you don't need it all at once," says Gil. There's no predefined order either, but it is important to establish priorities. Final word to Hassall at Ineos. He recommends several months of up-front consultative work before tendering out to system vendors. "People don't understand what they want out of their batch systems. Above all, know your process."