Maintenance - What do you know?04 April 2005

Most maintenance improvement philosophies work on the basis that if more time is spent on predictive and preventive maintenance work, equipment will become more reliable: breakdowns will become less frequent, making more time available for other maintenance tasks, and so on in a virtuous circle. That approach certainly works, but there is a hurdle to get over: you have to put in extra time and effort moving to a predictive regime, some time before this pays off in the form of reduced breakdowns. What might be preferable is an approach that freed up time by helping to fix breakdowns more quickly: this would leading to increased levels of uptime for the equipment, and give the maintenance department more time for other activities - such as that predictive maintenance programme they'd have in their sights for a while.

The people at Heathgate Knowledge Engineering believe they have just such as system. It's a software product called Solvatio - which has been produced by a company spun out of the Department of Artificial Intelligence and Applied Computer Sciences at the University of Wuerzburg, Germany. Geoff Walker and Andy Bates, both directors of the company, hesitate to call it an expert system, or even artificial intelligence, because the terms tend to be associated with systems from the 1980s and 1990s that didn't deliver what they promised; but it's certainly the case that using Solvatio is like having your company expert on hand 24 a day, 7 days a week.

Walker and Bates both have established pedigrees in industrial maintenance. Walker set up the maintenance management practice at Coopers & Lybrand in the early 1990s, while Bates spent many years with condition monitoring experts Entek, and moved to Rockwell Automation when it bought that company.

From his days at ICI Grangemouth, Geoff Walker remembers Ken, an expert metallurgist to whom people often turned for help. If you rang him up with a problem, he would ask you a set of questions - what's the pH? what's the temperature? and so on - until he came up with a solution for you. Some problems, though, were too complex to deal with over the phone, and he would ask to look at the metal part in question. The pattern of wear might remind him of a similar problem a couple of years ago; we never did find what caused it, but a particular alloy worked as a solution then, so why don't you try that?

Ken was demonstrating two different types of reasoning: rules-based, in the sequence of questions; and case-based, where he applied his past experience to new problems that had some similarity to previous ones. In the past, most expert systems have been entirely rules-based. This, says Bates, meant that while there was no limit to their complexity, it tended to make them very limited in scope. Also, as the rules changed, it was a major task to amend the software accordingly.

The big change, he continues, is the move away from complete dependence on rules to case-based reasoning. "You don't only have to go through a set of if-then rules - the system can suggest a solution, and if it proves to be correct, it will be more certain next time."

Clearly, then, Solvatio doesn't run straight out of the box - it requires knowledge to be captured from around the organisation. Heathgate's consultants run this process, talking to all the appropriate experts and drawing up a set of flowcharts covering all types of breakdown event, and the associated diagnostics and corrective action. This is then coded into a central knowledge base. "In the software, we've been careful to hide the 'engine' and just show people what they need to know," says Walker.

A key point here is ensuring that the system uses language that is obvious to the user - it doesn't assume that the maintenance engineer knows part numbers, or which light on a display panel corresponds to which function.

Walker and Bates show me how Solvatio has been applied at Royal Mail. The mail sorting machines in its distribution centres are large and complex, and carry out fine mechanical handling on a wide variety of shapes, sizes and weights of package. In the past, when confronted with a breakdown, Royal Mail's field engineers had three ways of finding the solution:

- Relying on their own experience - however, some parts fail rarely, so the particular problem may well have been unfamiliar to them
- Looking through manuals - but these are large and complex, and so it wasn't always easy to locate the appropriate information
- Calling Royal Mail's central helpdesk, which tended to be slow, and so was generally the last option they tried.

Another important point, says Bates, is that engineers tend to look to their own areas of strongest knowledge first: so someone with a background in mechanical engineering will look for a mechanical fault, while an electrical engineer will suspect an electrical problem.

With Solvatio, things are rather different. The system sits in a central server and is accessed on the maintenance engineer's wireless laptop via a web browser. The field engineers don't need any special software on their own PCs.

After logging on, the engineer first selects the part of the machine where the breakdown has taken place, and then chooses from a list the problem he has encountered. He is then told to carry out an action - for instance, checking the voltage between two points - and a labelled photograph helps him to pinpoint exactly where to do it. He is then guided through the rest of the repair process in the same way: he carries out checks, measurements or adjustments and reports the outcome back to Solvatio, until the problem has been fixed.

Links to the appropriate pages of the machine manual are available. "This was what produced the 'wow' factor from the technicians," says Bates. "Their perception was that the technical centre was sitting on this information, when in fact they were regularly updating it and sending it out on CDs. But you can't read those on a regular basis - and when you have an urgent problem to fix, you don't dare go browsing through a gigabyte of information!"

Once the job has finished, the technician completes a record for the job. If any new information has emerged through the course of the repair, it is added to the system's knowledge base.

The system also creates an audit trail of maintenance activity, which is particularly important in safety-critical areas. "In the event of an enquiry, you can show that your people did exactly the same as your best expert would have done in those circumstances," says Walker.

While this example applies to breakdown maintenance, the system can also be used to improve planned and predictive maintenance. It can be linked to other maintenance systems, such as CMMSs (computerised maintenance management systems) and on-condition alarms. "The system doesn't replace anything - it just helps you deploy the different elements of your maintenance regime better," says Walker.

He and Bates believe that the investment of time and money involved in capturing the knowledge and teaching the system means it's not appropriate in all industrial maintenance applications; but where there is complex equipment with complex problems, the cost of downtime is high, or where there are multiple installations of the same equipment, it comes into its own.

SOE

This material is protected by MA Business copyright
See Terms and Conditions.
One-off usage is permitted but bulk copying is not.
For multiple copies contact the sales team.