Mechanical Design Of An Optimal Transmission

Transmission development is a complex task. Depending on the number of gears and other degrees of freedom, up to 100,000 gearbox structures are possible. With the traditional approach of manually identifying and comparing individual promising system configurations, one may overlook innovative solutions. As part of a research project, scientists at TU Darmstadt have designed special optimisation models that help efficiently find optimal layouts for user-specific objectives, even in the case of large solution spaces.

The development of modern gearboxes is characterised by increasing demands combined with reduced development times. Often, developers are confronted with a multitude of challenging design goals, such as the realisation of more gears, while minimising gearbox dimensions; the realisation of fast gear changes, of the highest possible spread, and of good noise comfort.

A traditional approach to gearbox development is to manually identify and compare individual promising system configurations. However, a gearbox is a highly complex system. A manual comparison of all imaginable layouts is not practical given the multitude of possibilities: depending on the number of gears and other degrees of freedom, up to 100,000 conceivable structures can result [1].

In order to not only examine variations of traditional design variants, but to automatically search the entire solution space for innovative solutions, computer support is required. Using special optimisation algorithms, it is possible to find an optimal layout with regard to individual objectives, even in the case of large solution spaces. These techniques are currently being researched at the TU Darmstadt as part of the Collaborative Research Center 805 “Control of Uncertainty in Load-carrying Structures in Mechanical Engineering.” Thereby mathematicians and engineers work closely together.

THE OPTIMAL GEARBOX FOR EVERY REQUIREMENT

In general, the design of technical systems involves a trade-off between effort and availability for a required predefined performance, 1. The algorithmic design methodology developed at the TU Darmstadt is demonstrated by means of a seven-speed dual-clutch transmission. In contrast to brute-force methods, decision trees are searched efficiently, such that a proven global optimum is found which is “better than all the rest.”