The operation of many industrial processes, like those in oil refineries, is frequently complex, involving the monitoring of large numbers of parameters and the control of many different variables. Achieving optimal production while staying within safe operational limits is extremely difficult to achieve with human operators and manual controls. And since safety is obviously paramount, processes are rarely operated at their most efficient. Advanced control systems in the process sector are necessary to help achieve safe optimal plant operation, and can deliver quite substantive business benefit, especially in refining where the stakes are so high. For example, applying software and system vendor/implementer AspenTech’s DMCplus multi-variate controller at Agip Petroli’s Sannazzaro refinery in Italy contributed to such a great increase in utilisation and profitability of the unit that payback on the project was achieved in less than two months! Another example is BP Amoco, which has recently formed a multi-million dollar business and technology alliance with AEA Technology – Hyprotech. Gordon Hutchinson, head of engineering at BP Amoco, says: “We believe the Alliance Management Team will succeed in implementing a more efficient use of simulation technology in existing and grass-root plants, while continuing to reduce plant production costs, increase plant safety, and contribute to BP Amoco’s on-going commitment to environmental improvement.” At the Kuwait National Petroleum Company’s Mina Abdulla refinery, installation of a DMCplus controller on the hydrocracker – which converts vacuum unit residue to valuable products like ATK and diesel – has resulted in an annual benefit of over $500,000. This was attained through an overall yield increase as well as an increase in the production of more valuable products, an improvement in quality control and a reduction of flare gases. But advanced process control is set to deliver even more for the process industries in future, owing to a number of significant developments. Traditionally, advanced process control used empirical approaches, employing data from simulation models that were built up using step tests on a particular unit within a refinery, for example. Within the last few years, one change has been a trend towards process models built up from first principles, based around the fundamental chemical, physical and engineering parameters of the plant. There is also a steady progression from off-line simulations, that model a plant’s operation based on programmed scenarios and then provide data to the human operator to help in decision making, to online open-loop systems that actually monitor the current operating state of the plant. On-line systems provide real-time optimisation, and while most are still open-loop, some manufacturers are now using closed-loop systems that actually control the plant directly without manual intervention. Another change has been to extend the reach of advanced control software. Usually, advanced process control is applied on a process unit (sub-plant) basis, but this approach does not necessarily achieve optimal operation of the entire plant, says Clive Mott, business analysis manager for AEA Technology Engineering Software, a division of AEA Technology. “You would generally apply advanced process control to perhaps the two or three most important units within a refinery, yet you might end up optimising one unit at the expense of the others.” However, even before considering advanced control systems, manufacturers must ensure they fully understand how their plant is performing today, says Neil Holden, training manger with the automatic control technology group, Foxboro Great Britain. “Many customers do not know what happens on a minute-by-minute or hour-by-hour basis – instead they only see figures once a month.” With Foxboro’s technique, plant operators first define those key performance indicators (KPIs) that are relevant to their area of responsibility, business needs and objectives – so that the plant manager is now able to think in terms of business measurements, instead of set points on individual controls. Foxboro decomposes the KPIs into individual variables, which are then measured using a technique called ‘dynamic performance management’. However, advanced control software isn’t always the answer, says Holden: “It could be that the operators require better training, so they need to put in a simulator for operator training, it could be that their alarm management – how they handle alarms – needs improving.” If advanced software control is identified as the best solution, Foxboro first builds a very basic software model of the plant – not entirely going back to first principles – and then runs that on-line for a while. By comparing the model’s predictions with measured plant data, Foxboro can work on the model to improve it before putting the refined version back on-line, in a ‘closed loop’ fully automatic configuration. With Aspen Technology’s systems, data is also collected from live plant measurements, perhaps at one-minute intervals, and then used as the basis for building the model of the plant, explains Sam Dhaliwal, manager – advanced control solutions at Aspen. “We can then build a simulation of the plant, and once that’s working satisfactorily we can go on-site to commission it. First we implement it open loop, and once the customer is happy with the model we can close the loop – plant managers are naturally touchy about having a black box running their plant.” Once again, advanced control software may not be the immediate best solution. “We frequently have to troubleshoot instrumentation issues, and provide a list of repairs that need to be carried out, and then go back once that’s been done,” says Dhaliwal. Both these empirical methods and building models from first principles have their place, believes Tom Fiske, senior consultant with analyst ARC Advisory Group. “Rigorous models are computationally intensive and require a lot of expertise to maintain, while empirical models don’t give you the insight into the plant, but are faster to set up. I do see a trend towards more rigorous modelling, because computers are becoming more powerful.” But Robin Brooks, managing director of independent software vendor Curvaceous Software, believes that the traditional view of processes is fundamentally flawed. “To achieve optimal plant operation we start by looking at the data from the existing plant historian for the last three to six months, and then isolate the best 10% of operation – according to the plant manager’s criteria – over that period. We then define a control scheme that makes the plant operate in the top 10% all the time, instead of just sometimes.” However, by using a breakthrough technique that allows people to visualise as many as 20 to 30 variables on a single screen simultaneously – thus viewing the entire plant’s operating zone as a two-dimensional image – Curvaceous finds that advanced control software isn’t the priority. “For the first time ever, we are letting people see 30 variables at once, and the amount of understanding that comes out of that is astonishing – plant managers have to go back and challenge assumptions that they’ve always relied on,” says Brooks. One of the biggest shocks was the discovery of black holes within processes. “There are bands where you achieve optimal operation and bands where you don’t – traditional control schemes don’t recognise that. Process control theory has always assumed that plants operate in a continuous space, that it would make good material – or operate efficiently – anywhere within a particular range,” says Brooks. “For example,” he continues, “we found a gas turbine that had a black hole in the middle of the ambient temperature range where it was never going to achieve operation within the best 10%. So, once you realise you have these black holes, you then have a lot of work to do to understand and improve your process, rather than just apply better control to it. For those black holes that you can’t get rid of, we can design a control scheme to navigate around.” Currently, Curvaceous is working with clients in the oil, food and pharmaceutical industries to determine best plant operating zones. “We don’t yet have an on-line closed-loop system operating, but expect to achieve that within a year,” says Brooks. Batch processes Although advanced control software is most appropriate for continuous processes, it does have relevance to batch processes too, says Fiske. “The batch-orientated pharmaceutical, food and beverage industries are interested in the same business benefits. A lot of these processes don’t reach a steady state, however – they’re dynamic, and are much tougher to model, simulate, control and optimise.” Nevertheless, Fiske is seeing a concerted effort to apply the same sort of techniques in these areas. “Just as in refineries, they start with the process, and then progress to modelling the whole plant, and then move on up the supply chain. With each step you reap more benefits, but the degree of difficulty increases.” Part of the drive behind this effort has come about because within the pharmaceutical industry in particular, there’s been a change of attitude. “While 10 years ago the thinking was that manufacturing didn’t much matter, that’s no longer the case,” says Phillip Cheng, lead consultant with engineering cosultancy and integrator Eutech. “Now, pharmaceutical manufacturers have to have consistent quality to satisfy regulatory requirements – they cannot be wrong at all. And the edge in the drug industry does come from manufacturing – not necessarily the cost, but the yield. You have to make sufficient quantities for the trial in a very fast time, and you need to launch on the market as quickly as possible. Getting another three months out of the patent lifecycle represents a lot of money, because developing new drugs requires a vast investment, and you want that money back as fast as possible.” Indeed, the value of bringing a product to market one day earlier is estimated at $1m (source: WCI Consulting). Discrete manufacturing Cambashi associate Mike Skidmore, director of engineers AiMS for Industry believes that similar advanced software techniques can also be appropriate to discrete manufacturing. “If there’s no customisation, then there’s little scope for on-line optimisation. To really get the benefits you have to be using the same core capital investment for different products. Where you do see on-line closed-loop systems is in applications like assembling large printed circuit boards populated with surface mount components. Some of the very clever techniques coming in from Japan have learning systems that can be analysed inside the on-line computer, to calculate the optimum path given the capabilities of the high-speed placement computer.” Ultimately, AEA’s Mott believes that advanced software control will extend beyond the enterprise’s boundaries to include the supply chain. “There is no point in plant running the plant optimally if there are no incoming raw materials or you are making the wrong thing. Small mistakes can lead to losses of millions of pounds – if you’re buying the wrong type of crude oil, for example. “You can tie the model of the plant directly into planning and scheduling, and refine the models back to the pipeline and wells producing the crude. Our on-line advanced software control package for the oil industry – Hisys Refinery – can integrate with the upstream systems to achieve that.”