Design methods optimise cost and performance

A combination of structured design methods and test analysis techniques have been used to optimise the performance and cost of a range of super-precision angular contact and deep groove ball bearings. The project was carried out by the Barden Corporation for a manufacturer of high-speed rotary machines. The two methods used were the Function Analysis System Technique (FAST) and Design of Experiments (DoE), both are popular systems analysis and problem solving techniques which can be applied to any engineered system, mechanical design or production process. Normally, the company would look at the application, designing a bearing solution to suit, then costing the bearing. However, on this project, the customer asked Barden to investigate and develop a range of different high-precision ball bearings and select the most appropriate. To optimise the cost of the bearings, design engineers at Barden adopted FAST, a method for understanding complex systems by converting the activities performed in a system to the functions performed by the system for its customers. Says Nick Dowding, Barden's business development manager: "First, we had to look at what the bearings were required to do... We had to ask ourselves questions such as 'What is it that the bearings have to provide the application or machine to make it operate effectively?'" Once key functions were identified, the chosen method by which the bearing would achieve them had to be listed. Based on this, a list of requirements was drawn up. The completed list included 25 different bearing requirements, including clearances, run-out and vibration harmonics. These bearing functions and requirements were then put into a matrix and ranked accordingly. Next, Barden applied FAST cost optimisation techniques to each function to allocate costs to each function, including raw material costs, inspection and turning of the inner and outer rings. More than 30 separate bearing production steps were considered and scored against the bearing functions. This enabled the engineers to demonstrate to the customer which bearing features were costing the most, but contributing the least to the performance of the bearing in the specific application. Says Dowding: "In effect, you end up with the most important bearing functions at the top of the matrix, which also happen to be the most expensive ones to manufacture. The features at the bottom tend to be the opposite – the least important functions but which are inexpensive to produce. We focused our efforts on the middle range, where the bearing functions were deemed to be relatively important to the application and where the manufacturing costs were relatively expensive." By this method, it was determined that it was unnecessary to carry out 100 per cent functional testing on every bearing, with sample functional testing deemed sufficient. Similarly, it was agreed with the customer not to inspect all bearings after the turning process, reducing manufacturing costs be reduced while still meeting the performance criteria. According to Dowding, the project was an unprecedented success and resulted in a range of bearing solutions that were optimised in terms of their performance for a range of machine variants, but which also, on average, cost the customer 10-20% less. Alongside the adoption of the FAST method, designers also used Design of Experiments (DoE), a series of structured tests in which planned changes are made to the input variables of a process or system. The effects of these changes on a pre-defined output can then be assessed Because multiple design factors can be investigated simultaneously, this allows identification of factors that have a significant effect, as well as the effect of interactions between factors. Mark Pritchard, senior product engineer at Barden, explains: "This meant that we were able to rapidly study multiple bearing design variables simultaneously, enabling us to optimise a set of bearing designs based on performance and cost metrics for multiple customer machine models or variants," he states. For this particular project, Pritchard says he had to conduct 108 separate DoE mathematical models and 72 manual calculations (for validation and verification of optimised settings). The accuracy of the models proved to be in the region of 99 per cent and the result was four optimised bearing designs for each customer machine variant.