posted 27 Jan 2004 in Volume 7 Issue 5
Rolls-Royce: Knowledge acquisition and modelling
As the power behind the recently retired Concorde, Rolls-Royce set about capturing the experiences gained by its engineers for future projects. Michael Moss describes the importance of knowledge capture to the company and how knowledge management as a whole is helping meet the challenges of the future.
Memories of Concorde’s final flight last November will fade, but it is important to Rolls-Royce that our knowledge of the supersonic Olympus engines used to power the aeroplane does not. Our experiences in the design, development, certification, operation and lessons learnt from the past 27 years of Olympus 593 service are worth many millions of pounds in value to future projects. During the past year, Andrew Newman and Brian Lowry have used the knowledge-acquisition and modelling process (Kamp) to complete a project that makes these experiences readily available to future work on supersonic business jets and on defence projects.
Newman, a designer in the new-projects team in the Defence Aerospace division of Rolls-Royce, and Lowry, now retired, but previously chief supersonic propulsion engineer, have successfully gathered and documented the experiences of 46 people. This is a recent example of a project guided by a process that we have been developing and applying over the past five years.
The business context
Rolls-Royce employs 37,000 people globally, and delivers products and services to four distinct market sectors: civil aerospace, defence, marine and energy. The business context holds challenges where the knowledge-acquisition and modelling process, along with other knowledge-management approaches, is in use.
Many of Rolls-Royce’s products have improved through evolution. For example, in civil aerospace the RB211 turbofan engine has evolved over the past 30 years into the Trent and is now powering the latest Airbus and Boeing airliners. This evolution has led to increased thrust, fuel efficiency and reliability, and reduced noise and emissions. The transfer of experience from one project to the next enables greater technical challenges to be met at reduced risk. Capturing and reviewing lessons learnt also enhance this transfer.
The same advanced technologies are applied across projects and sectors. For example, gas turbine designs, originally designed for aerospace applications, are modified for use in ship propulsion and power generation. This requires us to not only transfer the design definitions, but also communicate the rationale behind the original design across organisational and geographical boundaries. Individuals are accountable for technologies and successfully use networks of people, web technologies and software systems to disseminate technical methods and good practice. Kamp is used to capture these methods and practices. Accurate, rapid and repeatable execution of these methods is important to the success of our projects.
Business processes are also common across projects and the organisation. Project management for example is underpinned by measures to build communities of practice in project management and the publication of processes and good practice across Rolls-Royce.
The support of our products in service is important both financially and for customer satisfaction as we have around 54,000 gas turbines in service with 4,500 customers. The products of all the sectors have long service lives, which can often be longer than a working lifetime. To support a product in service we place great importance on both the documentation and transfer of tacit experiences.
People with vast experience are rapidly retiring from the aerospace industry. Indeed, the more experienced Concorde/Olympus engineers are at that age and ten of the experts who contributed to our project had already retired. Capturing their knowledge is valuable both for the future support of a product in service or, as is the case for the Olympus, for the more cost-efficient development of future projects.
There has been a drive over recent years towards collaboration between aerospace companies on projects as a method of risk sharing. As a result, teams are becoming more globally dispersed. IT and web-based technologies, in particular, have become valuable in enabling the flow of experience.
Knowledge-management has become an essential element of our business due to the:
- Evolutionary nature of our products;
- Scope for applying similar technologies to multiple applications;
- Use of the same business processes across the company;
- Long lifespan of our products and services;
- Trends in age of the aerospace workforce;
- Global nature of the aerospace business.
The approach to knowledge management within Rolls-Royce has been to promote a knowledge-sharing environment. Person-to-person based approaches, such as communities of practice and after-action reviews, are used and are driven by the realisation that it is only feasible and economically viable to capture a portion of an expert’s useful knowledge. This is emphasised by the observation that even where reference material is widely used, engineers (people working creatively to solve problems) request information or advice from other people 66 per cent of the time. It is also significant that in a high proportion of the requests, the problem is clarified and the engineer’s request for information is modified.
Where knowledge can be standardised the benefit justifies the extra effort required to capture that knowledge and create knowledge documents that guide people in carrying out processes. These knowledge documents appear as good-practice guides, standards, procedures and training material, which are frequently published on the internal web. The knowledge-acquisition and modelling process is a cost-effective approach to creating these knowledge documents. Where knowledge is stable, and speed or repeatability are of great economic value, then additional knowledge capture and development of knowledge-based systems can be justified. Although knowledge capture is unlikely to fully replace the need for direct contact with experts, it does succeed in making the transferral of expertise more routine.
Knowledge acquisition and modelling
During the 1980s, industry recognised that the loss of key experts was a problem. Projects were undertaken where knowledge engineers would acquire knowledge from experts and computerise it via an ‘expert system’ (see figure 3). The task involved a variety of disciplines, such as computer science, artificial intelligence, psychology and human factors. Significant technical achievements were made at this time. The ‘knowledge base’, where rules and facts acquired from the experts were held, was separate from the ‘inference engine’ that deduced conclusions from the knowledge base. Declarative styles of programming were developed to represent knowledge. Despite these advances, issues still remained. One difficulty was acquiring sufficient knowledge from an expert to make the system useful. Methods for ensuring the quality and maintainability of knowledge were needed. Inference problems also existed, for example, there was no way of deriving results based on alternative assumptions or to maintain truth when an assumption was abandoned. During this period there was also a commercial proliferation of tools, although in some cases and for some tools this was short-lived or they were of poor quality.
During the 1990s tools became available that significantly improve the inference and representation issues faced. The 1990s also saw the development of methods to formalise knowledge acquisition and the subsequent creation of a computer system. These methods of knowledge acquisition aim to improve:
- Maintainability – By structuring knowledge and making it readable by humans, changes could be easily made and systematically incorporated into the system;
- Visibility – By making the system seem less like a ‘black box’ to the user we reduced the risks associated with the loss of skills and retained innovation opportunities;
- Validation – Experts were able to check knowledge directly rather than having to run the system before gaining access to knowledge;
- Cost – As knowledge was more likely to be complete prior to being coded for the computer system the cost of re-work was reduced.
These formalised methods produce a platform-independent knowledge model that is an explicit representation of our knowledge. The knowledge model defines the concepts and their attributes, which are typically organised hierarchically by classification or composition. Additional links give navigation via process steps, for example. Formal methods include CommonKADS (a large general-purpose methodology for expert systems), Moka (aimed at knowledge-based engineering (KBE) systems) and Kamp (where the envisaged product is a document, website or manual rather than a computer system). Project management, knowledge acquisition and knowledge modelling in these methods is similar but modified to suit the planned product.
Expert systems and KBE systems deliver executable software and their creation is resource intensive. Kamp however delivers a human-readable record and is a cost-effective option where the resources or benefits are insufficient to justify an expert or knowledge-based-engineering system. In the case of the Concorde/Olympus knowledge-capture project a human-readable document was the required output.
What is Kamp?
Kamp is a comprehensive method for capturing and publishing knowledge that is suitable for staff members that are not experts in knowledge engineering or knowledge management. The method enables these employees to capture knowledge from experts and publish it as a document. Kamp also provides close support from a coach together with a web definition of what the method is. The web definition is itself presented in Kamp form and therefore provides both guidance and an example.
The method brings together a number of roles and defines what is expected from each one. It also brings to bear techniques from relevant disciplines. Project management (which uses a gated process), psychology (which contributes the knowledge-acquisition methods), information engineering and artificial intelligence (which contributes the knowledge-modelling approaches) have an impact here. The synthesis originated in a collaborative-research project involving computer scientists and psychologists.2 Integration into our processes and organisation is also important. For example, it is beneficial for engineers on the graduate-training programme to undertake Kamp projects in departments they are joining. Similarly, we can use the method to acquire knowledge from experts prior to their retirement.
The benefits of Kamp
There are a number of benefits that Kamp has delivered to Rolls-Royce, such as:
Accelerated learning – Many knowledge engineers are recent graduates or new starters. Estimates from Kamp customers show that the execution of a typical three-month project has increased the knowledge engineer’s learning to levels that might be expected after nine months;
Increased availability of expertise – Expertise that is in high demand, but short supply is a common problem. Knowledge that resides in the heads of experts can be documented and made more readily available. For example, a design standard for an innovative manufacturing technology – called shaped-metal deposition – has been captured from the 30 experts involved in its development. This speeds up the adoption of new technology across the company;
Improved response times – People can obtain information and advice from documentation immediately, whereas an expert may not always be available;
Reduced risk from sudden loss of expertise – We can reduce the risk of loss associated with key areas of the business that rely on only one or two people. These risks include the need to recreate knowledge or the cost of making an error in the absence of knowledge. By publishing knowledge we can improve the availability of key knowledge and experience. For example, the loss of steam-turbine-generator expertise from the closure of a site was a potential risk to a project. By successfully capturing knowledge we were ultimately able to reduce our bid for the project;
Improved quality – Across the board, work is being carried out with greater levels of experience. This includes knowing who and when to ask, what knowledge other business areas have and how it affects the process at hand.
Kamp in more detail
The most active role within Kamp is the knowledge engineer, who is the only full-time member of the project. The remaining roles are part time:
- The customer identifies the business benefit and owns the topic on behalf of the company; the customer is therefore responsible for disseminating the resulting knowledge document and its future maintenance;
- The coach is experienced in knowledge engineering and guides the knowledge engineer throughout the project;
- The knowledge-management lead has overall responsibility for the smooth execution of the project;
- Experts provide and validate the knowledge. Key experts are nominated to arbitrate where disagreements occur and provide final approval of documents;
- The intended users of the document are interviewed early in the project to understand what knowledge should be captured and how it should be delivered.
The scoping phase is vital to the success of the project. This phase determines what knowledge will and will not be captured. This is important because if the project’s scope is too wide it leads to a superficial result, which the intended users will not find useful. The scoping phase identifies the corresponding costs and benefits, and includes normal project planning criteria to ensure that the project is achievable. By this stage we also expect to identify how the knowledge document will be used by consulting with the customer. The document has no value to the company until it is used. Importantly, the scoping phase is also the knowledge engineers’ first experience of knowledge acquisition and modelling, so assistance from the coach is at its most intense at this time.
Experts are in demand, so planning the knowledge-acquisition sessions is important. Knowledge-acquisition techniques are selected in advance. The techniques that we have found useful and that knowledge engineers adopt rapidly are:
- Interviews – Unstructured and, usually later in the project, structured interviews are the most common technique. Written contributions may also be sought;
- Commentating – Experts describe what they are doing as they perform a task;
- Card sorting – Classifying concepts;
- Three-card trick – Comparing and contrasting three concepts to expose the attributes and classification of concepts;
- Twenty questions – The knowledge engineer has a concept in mind and asks the expert to determine what it is. The expert is only allowed to ask ‘yes’ or ‘no’ questions. This technique is used to elicit the attributes of concepts and their relative significance. For example, the knowledge engineer may be thinking of preparing a tender. The expert might ask whether the task occurs before or after a sale, which indicates that the sequence of tasks relative to a sale is a key attribute.
Other knowledge-acquisition techniques are also used when beneficial. Often the coach helps to identify the need for a different technique. Knowledge-acquisition sessions are usually recorded and analysed to identify the concepts and attributes, and model the knowledge. Follow-up sessions with the experts are held to validate the knowledge model. Such modelling combines the knowledge of a number of experts, which can help to identify and resolve differences. It can also be a catalyst for knowledge sharing between experts.
A final presentation is held to close the project, launch the exploitation of the knowledge document, identify lessons learnt during the project and future Kamp or other knowledge-management projects. Kamp has evolved over the course of many projects, of which Concorde was one of the latest. Capturing and incorporating these lessons into the Kamp definition has been important to its success.
Knowledge capture is an important element within a set of knowledge-management methods. It falls midway between people-centred approaches, such as communities of practice, and those that make knowledge explicit, such as manuals and knowledge-based systems. Kamp is a cost effective way of capturing knowledge as the process takes little of the experts’ time and the learning experience is valued by those adopting the role of knowledge engineers, and their leaders. The process reliably delivers explicit, documented knowledge and encourages benefits to be derived from it.
Our most recent Kamp project has been a huge success. It has produced a knowledge asset in the form of a document that allows people to fully understand Concorde’s Olympus engine and its supersonic propulsion system. The document contains the contributions, both small and large, of 46 people with extensive experience in this area. The knowledge asset will allow future supersonic projects to benefit from being able to easily access Rolls-Royce’s experiences in using Olympus technology.
1. Wallace, K. & Aurisicchio, M., Private Communication (Cambridge University Observations 2003)
2. Hammersley, M., Shadbolt, N.R., Golightly, D.A., Cottam, H.D. & Riley, P.H. (1999) ‘Leveraging the dimensions of K: Knowledge engineering for web-based knowledge management’ in Knowledge Management (Ark Group, Volume 3 Issue 2, 1999)
Michael Moss is a specialist in knowledge management at Rolls-Royce. He can be contacted at firstname.lastname@example.org