Improving performance

Now in its sixth year, the Manufacturing Excellence conference has firmly established itself on the UK process industries' calendar.

With its emphasis on bringing together engineers from the chemical and other industries, with their contractors and suppliers, to share experiences and best practice, the conference has always managed to stay the right side of that fine line between an out and out commercial exercise and a genuine educational event.

Held last month in the pleasant surroundings of the Forest of Arden in rural Warwickshire, ME 2004 proved no exception to this enlightened policy. The whole event - comprising one day devoted to a series of training courses and two days of five parallel conference sessions - was once again organised by Emerson Process Management, with support from Aker Kvaerner, BASF IT Services, Huntsman and the IEE. In between the conference sessions the close on 300 delegates also had the opportunity to visit a 'mini exhibition' put on by equipment and service suppliers.

But few, if any, delegates had come just to be sold to. Their interest lay in the opportunity to hear from their peers and colleagues about their day-to-day experience of running plants under today's constraints of driving down costs while improving performance.

Appropriately enough, given its sponsorship of ME 2004, Huntsman had given free reign to many of its plant operators, allowing them to talk about their experiences in running some the biggest petrochemical and chemical plants in Europe. Since taking over much of the former ICI assets at Wilton on Teesside in 1999, Huntsman Petrochemicals (UK) has invested heavily in the complex, including complete overhauls of the Olefins 6 cracker - the second largest in Europe - and the Butadiene 3 plant.

Project manager for both overhauls, Huntsman's Brian Mitton, presented a case study of the 2003 Butadiene 3 revamp, for which he was tasked with meeting key performance indicators related to environment, health and safety performance, project cost and duration, and overall quality. A key feature in the success of the project, he says, was working in an alliance partnership with main contractor Aker Kvaerner, whose Paul Hardy presented his side of the alliance.

The longer-term goal of the overhaul was to prepare the plant for sustained, reliable, high rate operation until its next planned shutdown and overhaul in 2008. Preparation is one thing, of course, but actually keeping plants running at such required levels devolves to the plant managers, process engineers and plant operators. And here Huntsman speakers were also out in force.

Mike Oglesby, process control consultant to Huntsman Petrochemicals, gave a, quite literally, graphic demonstration of how the performance of the Wilton complex's paraxylene plant had been improved by applying data mining techniques. In this 30-year old, 360 000tpa plant, liquid feed is vaporised and raised to reaction temperature in two parallel trains, each with two vaporisers in parallel followed by a fired heater. The two trains combine to feed the reactor, while reactor effluent returns through the four vaporisers to vaporise the liquid feed.

'Conventional wisdom, passed down over those 30 years,' Oglesby explained, 'is to run the vaporisers at maximum level.' To this end, the plant has an advanced control system -now running on the plant's recently installed DeltaV control system from Emerson - to ensure just that. 'The control scheme works well,' he says, 'but rate varies significantly from day to day for no apparent reason. Why?'

The first step in answering that question was to consider a conventional multi-variable statistical analysis of relevant plant data. But even Oglesby, with a lifetime's experience in process control, blanched at the prospect. Instead, with the help of Robin Brooks, md of Curvaceous Software, he used that company's Visual Explorer package to find out exactly what was going on in the process. This innovative data mining software manages to hide all its underlying maths and statistics from the user, presenting instead a graphical presentation of the interaction between variables.

Importing data from the plant's historian, three months' worth of hourly spot values were fed into Visual Explorer. Oglesby says the crucial factor in this type of analysis is to enter in all the variables. 'Despite what operators might tell you,' he says, 'don't make assumptions about what may or may not be important.'

He found the package very easy to use and was soon able to drill down into the data to show that the strange rate-limiting factor was in fact due to environmental effects. After eliminating most variables, he turned his attention to ambient temperature changes, but these didn't seem to affect reaction rates. But when wind speeds were also factored into the graphical analysis he found the answer. 'Low wind speed seems to result in lower feed rates to the plant,' he says. 'It appears that changes in heat lost from the reactor result in changes in plant feed rate.'

Impressed with the performance of Visual Explorer, he next applied it to the downstream purification section of the plant. The aim here was to maximise paraxylene recovery by balancing the loads on two banks of Escher Wyss centrifuges. 'Again,' he says, 'the data mining gave results showing that what was traditionally considered to be best operating practice did not always give the best overall plant performance.'

The paraxylene purification process - in which this precursor for polyester resins and fibres is separated from its ortho and meta isomers by selective crystallisation and centrifugation - was also addressed by Steve Gill, a process engineering team leader for Huntsman.

Optimum process efficiency here involves controlling the composition of the incoming feed stream to the purification plant — which in turn involves efficient on-line analysis. Early last year, a Rosemount Analytical Raman analyser was installed to replace the previous on-line melting point analysers for monitoring the process stream composition. The performance of the melting point analysers had had to be confirmed with frequent grab samples followed by laboratory analysis. The delays inherent in either of these techniques had been leading to a 2-3% variation in feed composition.

With the Raman analyser, however, process variability has now been dramatically reduced by an order of magnitude to 0.25%, with a consequent improvement in plant stability. The laser-based spectrometer allows full on-line composition monitoring with information updated every minute.

'Considering that we are doing in-line dilution,' says Gill, 'I am very pleased with the performance. While the main benefit of the purification control scheme is to give consistent solids feed to our centrifuges, an additional benefit has been the ability to see the impact of upstream changes on variability. We've never been able to see that in real time before.'

The Raman analyser will now be used for simultaneous measurements at four separate process locations on the plant - feed and recycled material as above, plus two final product streams. The analyser itself is located in a control building, with the laser light transmitted through fibre optic cables to optical probes in the process streams.

The intensity and wavelength of the Raman scattered light - reflected by the sample and returned to the analyser through the fibre optic cables - correlates to qualitative (molecular species) and quantitative (concentration) information about the sample. The derived analysis and concentration data are then transmitted via Modbus to the DeltaV control system, which acts on the signals to control plant feed dilution.

Paraxylene plant manager, Tom Liddle, is certainly pleased to see his plant running more smoothly. 'Without the on-line control provided by the Raman,' he says, 'variability in the process would occasionally lead to excessive solids loading in the centrifuges, resulting in vibration and potential bearing damage. Now we run at maximum output, and have reduced wear on the centrifuges.'

And, as was said earlier, meeting those twin constraints of cutting costs while improving performance.