Unilever takes PRIDE in process diagnostic tools

Every engineer knows that the universe has only one immutable law, and that's the one formulated by Murphy: if it can go wrong, it will. However carefully your system is designed, however many safeguards are installed, failures are inevitable. But as systems become ever more complex, so the methods for identifying and dealing with errors must follow suit.

Advanced process control systems are a case in point. Designed to minimise the variations in process parameters, these systems automate as much of the plant's operation as possible. However, according to Rob Kooijmans of Unilever's research centre in Vlaardingen, the Netherlands, they can lead to operators losing touch with the process and not recognising when the plant's situation slides into the abnormal. When things do begin to go wrong, pollution, safety problems, escalating power usage and wasted materials can result.

Unilever is part of a consortium developing a system known as PRIDE (PRocess Industries Diagnostic Experiments), a computer `toolbox' which uses the output of advanced control techniques to identify abnormal situations before they can send a process off-kilter. The 12 companies involved have dedicated three years and almost E6million to the project; the European Union has chipped in another E3million.

The PRIDE consortium includes Norsk Hydro, which is interested in applying these techniques to its fertiliser processes; BG Transco, Yorkshire Water, BNFL and EDP. Norsk Hydro is also on-board as a technology provider, as are Heriot Watt University, Finnish pulp and paper producer KCL, and the Italian energy research organisation ENEL. The software know-how comes from Cogsys, Elsag Bailey and Hartmann & Braun.

Currently, the consortium is working on three `tools'. ENEL has developed a system called Giotto, which compiles models of the correct sensor output and compares the actual control system data with these. This spots potential problems and identifies the part of the plant which is deviating from steady-state behaviour. KCL, meanwhile, has developed a system called Wedge, for analysing low- and high-speed processes. This looks at disturbances on certain key variables, and traces them back to the problem parts of the plant by analysing patterns of vibrations and oscillations. Norsk Hydro's system, TBD, which was developed to help prevent effluent releases from large plants, traces problems back to sensors which are giving abnormal readings.

These programmes sit within a system such as Gensym G2, which provides interfaces between the control and diagnostics systems and holds the necessary databases.

At Unilever, the system monitors an edible oil refining plant, Kooijmans says. The G2 package interfaces the control system with the PRIDE toolset (in this case Giotto and Wedge). The PRIDE protocol itself runs on an exact copy of the control system interface, also provided by G2.

Giotto and Wedge monitor the process constantly, and G2 `calls' them at intervals to perform diagnostics on the process. Any abnormalities are flagged up as alarms on the duplicate interface. The system tells the operators the possible causes of the alarm, making troubleshooting much more efficient.

PRIDE has turned the plant's control system into an on-line monitoring tool, Kooijmans says. It's also proving useful in training, providing instant tutorials into what can go wrong on a plant. `We're starting to go beyond process automation and into process assurance,' he says. `We can always have the same quality of process - and quality of product. PRIDE is a major part of that.'