Advanced plant control software moving on up

Agip Petroli, part of the 72,000-employee Eni group, has achieved a feed increase of around 5—7% on cracking units, and other significant gains elsewhere in the company’s plants. All this has been thanks to advanced process control (APC) software, as Santo Biroli, a senior control engineer with Agip, says: “Multi-variable control technology is responsible for increasing unit throughput – it’s a very valuable technology.” Higher throughput such as Agip Petroli has achieved is usually the main driving force for the implementation of APC, and of course leads to increased profitability. In the oil refining and petrochem industries, the large volumes and profits easily justify the time and expense that these schemes can consume. It’s also usually possible to reduce energy consumption, but that doesn’t provide the initial motivation. The Eni group produces more then a million barrels of oil equivalent (BOE) today, and with a goal to increase daily hydrocarbon production to at least 1.5 million BOE by 2003, throughput is crucially important. Initially, some improvement was achieved through offline simulation, which helped Agip to identify and tackle the real constraints of the plant. “Normally, with different panel operators on different shifts – every one operating the unit in their own way – it’s not easy to determine the most significant constraints,” explains Biroli. APC yields its benefits through improved stability of operation. Advanced controls continuously make small adjustments as process variables drift — unlike operators who may not detect subtle changes until much greater corrective action is required – potentially leading to instability. As a result of this more stable operation, product variability is reduced, which means that plants can reliably operate more closely to specification, reducing the margin normally left for error – and thereby saving money. Also, yields are increased as the system drives the process closer to its safe limits. It’s on the basis of this saving that vendors will usually start their pitch, by performing a benefits analysis. However, the foundation that needs to be in place first is a reasonable amount of plant data. “If the plant doesn’t have a plant historian or other means of collecting plant data that does present a challenge to carrying out the benefits study,” observes Simon Angell, sales manager for advanced plant control at process systems’ vendor Honeywell. But in the oil refining and petrochem industries, Angell believes it’s pretty much a done deal. “All the global majors have taken the view that they will implement multi-variable control technology across any process where it shows a return on investment (ROI).” And even once unit-level multi-variable control technology is in place, there are still more benefits to be gained from plant-level optimisation software – although the law of diminishing returns can start to apply, with the ROI being less spectacular. Slow adoption rates In spite of the clear advantages though, APC is not yet universally adopted, especially in the Middle East, Far East and China, according to Sam Dhaliwal, project services manager at APC and process simulation and management software giant AspenTech. “It’s down to how much has been published, what benefits have been reported and how much they believe in the technology.” “There are still those in the oil refining industry who are not yet convinced of the benefits of implementation,” confirms Angell. However, on any new plants, an advanced control package is now fast becoming a de facto requirement, believes Dhaliwal: “It’s almost a necessity from the point of view of being competitive.” One drawback to implementation is quite simply the time it takes. Agip Petroli is now commissioning an on-line closed-loop optimiser running on top of its AspenTech DMC+ advanced controllers, but the project has taken around 18 months. A major contributor to long implementation times, and an inhibitor to deploying the technology in less profitable industries, is building a model of the process unit’s behaviour. “Historically, people have performed step testing, by moving one of the manipulated [process] variables in one direction and observing the effect on the control variables,” explains Angell. “Then, they would build a model with all of these relationships. It’s a man-hour intensive process but we are moving to new technologies that reduce that time.” Improvements in the software used for identifying the model have made this process significantly better over the last few years, so the perturbations carried out during the building of the model are smaller than previously – and well within the changes an operator might make during normal plant operation. However, material feed rate may well have to be reduced during testing, and that’s an ‘investment’ that needs to be taken into account. But time also works on the side of the technophile. “Once the plant management has a benefits study in their hands, they can see that every day that they don’t implement the solution, they are losing money,” says Dhaliwal. Nevertheless, the necessary investment can be considerable, especially on older plants that don’t have the requisite DCS (distributed control system) and modern instrumentation. In that case, the payback period might be as long as two to three years, rather than the usual year or so. “Another inhibiting factor is operator acceptance,” says Angell. With APC, the operator’s responsibilities move to more of a supervisory role, and not everyone will initially be happy to relinquish control to a computer. “But they need to get comfortable with the fact that the algorithm will do things to the plant that they wouldn’t have done – it may push the plant closer to the limit.” Technology transfer There is no doubt that serious investment in training is required – both formal and on-the-job – and this can help to address issues of operator acceptance. In addition, the commissioning procedure can include ‘open-loop’ operation known as ‘warm mode’ (indeed, some plants run like this). “In this mode of operation, the APC software will come up with recommended set point targets but not implement them – that’s down to the operator,” explains Angell. This clearly demonstrates the benefits that APC will deliver. Meanwhile, applying the techniques of APC and simulation to other industries and even discrete manufacturing is happening, but not quickly. “In dairies, cement factories and so on, APC is less prevalent,” says Angell. “It can solve the problem but the economic benefit doesn’t show a good enough return yet. Our objective is to reduce the man-hours taken to implement the solution.” But although controlling multiple variables is frequently a different type of problem in the discrete manufacturing world, there’s no doubt that many companies still face similar challenges related to huge volumes of data, many variables and unknown interdependencies. Bespak, for example, is a manufacturer of drug delivery devices, known for its metering valves for asthma inhalers. These are built in batches and, because the products are for the world’s largest pharmaceutical manufacturers, the requirement for full traceability means there are massive amounts of detailed and comprehensive historical data. Enter Curvaceous Software (see panel), and with no need to build any sort of model of the manufacturing equipment and its interactions (since the data is there to automate that stage), it seems feasible to identify relationships between product quality and specific processes – and to use that. Says Paul Barnes, senior scientist at Bespak: “A lot of our business does generate numerical data and to make sense of that from a troubleshooting or research end can be challenging – sometimes with such a massive amount of data you can’t see the wood from the trees. Investigations can be very time-consuming, and it’s even possible to spend several weeks working on huge numbers of graphs.” Speeding the process Initial work at Bespak has revealed that Curvaceous Software could prove a very useful technique in this environment. Barnes: “A typical metering valve has eight components, of which some are plastic parts moulded on one of several tools… It [the system] can pinpoint whether a particular moulded component is implicated in variations in product quality – and you don’t have to be a scientist or understand graphs to see the implications.” In another example Dr Des Slaughter, food technologist with RHM Technology, uses Curvaceous Software for analysis of data from both production and research batch processes. “As a research establishment, we do a lot of experiments of various natures – typically baking ones – and you have a lot of variables and a lot of outputs. Curvaceous Software is a good way to visualise what the results are, rather than use statistical methods,” says Slaughter. RHM has its own tools to acquire and manage data in its quest for plant optimisation and process efficiency. It uses Curvaceous Software as one means of analysing the data. “We have found relationships in data that we couldn’t find in other ways, and as a result of that we have found out things about our plants that we didn’t know,” insists Slaughter. And there are other potential benefits, as Wim van Wassenhove, regional product manager for HiSys at APC vendor Hyprotech, points out: “Maybe the benefit is not financial but regulatory compliance”. He gives an example of a customer in the beverage industry: “The driving force there was maintaining the same taste and quality of the product. It’s based on natural ingredients, so the product quality is not easy to maintain manually.” And there are parallels across other industries. Advanced process control software has come a long way, but has plenty more to achieve. Only when it becomes even easier and quicker to implement will it make significant progress outside the oil refinery and petrochem industries. And simulation software has some way to go in terms of accuracy too – no one has yet got to the stage of carrying out perturbations on the simulation model instead of on the plant.