Saving the knowledge you need

The process of capturing, storing and providing access to their resources of human knowledge is a problem that has tested the ingenuity of manufacturing companies for many years. At one time, it was predicted that all such expertise would be captured by expert systems that relied either on the manual insertion of rules or on a neural net. However, the former approach was time-consuming and problematic if the individuals relied more on experience and intuition than logic, while the logic generated by the latter method can be impenetrable to human programmers. Nonetheless, the importance of retaining knowledge remains clear. Talking to Eureka recently, Bloodhound SSC driver Andy Green revealed that the turbo pump that will be used to pump the hydrogen peroxide into the car is closely based on that originally developed by Rolls-Royce for the Blue Steel standoff missile that entered service in 1961. The drawings had been lost in a fire, but the designer could still be consulted. It was thus possible to reverse engineer parts that still existed and come up with an improved version of the pump without having to develop it from scratch. More chillingly, perhaps, Tom Robertson, a retired engineer from AEA, told Eureka a few years ago that computer test data on the stability of nuclear reactors derived from experiments that nobody would nowadays care to repeat, now only exists in the minds of a few old men and data tapes that can only be read by computers than no longer exist. Clearly, part of the solution lies in meticulous record-keeping processes and effective communication, but only part. There is a considerable gap between retaining knowledge and actually being able to capture it when and where necessary. In a project undertaken between 2000 and 2003 entitled 'Knowledge Capture, Sharing and Re-use in the Design Process', Dr Richard Crowder of Southampton University attempted to address this issue. Funded by the EPSRC (Engineering and Physical Sciences Research Council) this was a multi-university and industrial contract with Rolls-Royce BAE Systems, the Universities of Sheffield Cambridge and Southampton. Says Dr Crowder: "We started with the statement that 90% of industrial design activity was based on variant design, so there needs to be a clear route for the designer to get back that information on to his desktop as soon as possible. A statement we also pulled out was that in design activity, 70% of it basically comes from previous solutions. So the whole object of the exercise was to give the designer tools and approaches to extract information either from colleagues (ie the expertise across companies) or to provide them with tools to search company archives to access the knowledge of those who are no longer with the company." The loss through natural wastage of individuals with expertise is obviously a major obstacle to success for any such system, as Dr Crowder makes clear. "If you go back a number of years, the right people would always be around. But these days, people move so fast that you simply can't rely on that being the case. So you've tended to lose what we always thought of as the wizened old chap at the back of the drawing office who knows everything and who's been there since the year dot. You've lost that kind of resource." The solution was for Southampton University to develop what was called 'Expertise Finder', a system to enable designers to access the wisdom of the entire company past and present to find solutions. However, this approach faced a number of problems, most of which lay in the human aspect of any such system. Says Dr Crowder: "If your approach relies on people to say they have expertise in a particular area, the chances are it will fail. Those who do have the expertise may say 'I don't want to be pestered by 250 other people asking for advice' and so not volunteer. Meanwhile, others will say 'Of course I'm an expert' because they want to get on in the company and they aren't quite as expert as they suggest." Instead, the project chose instead to use existing data to identify expertise. This ranged from examining company reports to noting the positions of individuals within the company and building an information resource from there." We mapped together a number of databases, put in subject areas and it came back with a list of likely people who would be able to help with that particular issue." Of course, this study was undertaken with a view to it being effected by large companies and Dr Crowder does feel there are different requirements, depending on the size of the organisation. "The smaller the organisation, the less of a problem there will be because everyone knows everyone else," he says. "The problem with large companies is that they're often multi-sited and you tend just to work within your group. So you have 'the watercooler effect'. People will ask: 'Do you know anything about x, y or z?' and the answer will either be 'Yes, try Fred in another division' or 'no' and they've got to do all the work all over again." However, certain rules apply regardless of the size of the company. Says Dr Crowder: "The biggest problem is the accuracy of the initial data. Everybody said their initial data was correct, but when you got down to it, there were lots and lots of errors – even down to mistakes in the company telephone directory, which caused a lot of grief at first. However, we proved the concept itself works and therefore I see no problem in rolling it out across any size of company." An example of methodology in approaching this problem can be seen in Procter and Gamble, which employs Siemens' Teamcenter Product Lifecycle Management system for this purpose. Here the problem was particularly acute, as the company's R&D division consists of many separate laboratories. Data was previously fed into vertical applications developed by each laboratory, where it remained accessible only to those that had developed it. No method for either sharing such information across the company or even indexing it it could be located. And, because the company has 130,000 employees in 80 countries, personal networking was impractical to say the least. This problem has been addressed by the introduction of 'electronic notebooks', on which users enter their discoveries and ideas into the software in the same way they did in their laptops or paper notebooks. This information is then available to authorised users in the rest of the company. Again, storage of all data is a critical part of the discipline. For the system to work perfectly, the research and development should begin with capturing the requirements for the project, such as purpose, weight, quietness and cost, leading to a product definition to which all other information can then be linked. It should then be possible for somebody to go back months later and access both the initial requirements and all the design decisions. Legacy data, on the other hand, remains an issue. According to Siemens, materials on laptops could be cut and pasted into the new software module, but material that exists only on paper is still a problem. The emphasis on processes and meticulous recording is reiterated by Dassault Simulia's Tom Bianchi when he says that companies need to have 'a process that requires the recording of design decisions and the reasons for them'. For this reason, Dassault's main PLM offering Enovia enables the linking of engineering models and parts of models to other items held in the database. Says Bianchi: "Because of the data being associated with the model, it is much simpler to access items in the database than it used to be." Examples willing to be cited include Ascari cars, which makes small numbers of supercars at Banbury. Bianchi said that Ascari recently discontinued one of its lines, but recorded everything that had gone into the car in a database so that it could all be retrieved if the company started to sell the car again in the future. Again, though, Bianchi concedes there is a problem with expertise that only exists in the minds of humans, citing the example of steel mills in Yorkshire that have had to bring back 70 year old former employees when they need to make adjustments to their rolling mills. However, in cases where the human expertise no longer exists or is not accessible, Simulia offers a facility to undertake non-linear simulation to rediscover the optimum set-up parameters and this information is then retained electronically in the system so that it cannot be lost a second time. Both Teamcenter and Enovia scale down for use by smaller companies, while for Pro/Engineer users, PTC offers PDMLink and Windchill. Firms accustomed to basing their design on AutoCAD, Inventor or SolidWorks can use the data management packages produced by these companies – Vault in the case of Autodesk and various PDM packages in the case of SolidWorks. In every case, however, it is probably the process of storing all documents, associating them with designs and making them findable and accessible that is more important than which software package is used. For all the advances in knowledge capture techniques, it seems the gap left by Dr Crowder's 'wizened old chap at the back of the drawing office' has still not been filled entirely. However, by use of the available technology and, perhaps more importantly, scrupulous application of the associated disciplines, it should be possible to ensure that, in the future, he doesn't cast quite so long a shadow. Future solutions Because it supports aerospace products that are in some cases decades old, retaining knowledge is critical for BAE Systems. However, while it has an existing process for recording data, it, too, struggles with knowledge lost in the minds of older engineers working on legacy equipment. To deal with the problem, the team at Rochester has been working with the Chatham Campus of Greenwich University and has come up with the idea of creating networks and holding meetings to generate knowledge, and then adapting a proprietary audio-video streaming technology, developed for another purpose, to record, tag and index it. Key to this is a piece of proprietary technology that BAE Systems has that can look through an audio-video stream of data for particular types of sound pattern or event, and then trigger something to happen. In the defence and security arena, this is normally some kind of action, but it could also be to tag to where this happened, with the tag data stored in a meta database. Applied to the knowledge management situation, this would mean looking for key spoken words, and then tagging where they are spoken, allowing appropriate sections of recording to be retrieved and studied subsequently. The project is in its early days – 18 months into a 36 month project. The current aim of the project is to produce an inclusive but secure proof of concept pilot at The Base, which is otherwise known as the Medway Innovation Centre in Rochester. As a result of his presentation, Piorkowski won an award for the best oral presentation at the event which was the "Eighth School of Engineering Research Conference". One of its main themes was, "Innovative Product Development."