Genetic Algorithms in Search, Optimization and Machine Learning
Genetic Algorithms in Search, Optimization and Machine Learning
| Hi-index | 0.00 |
The central thesis of this paper is that the dynamic performance of machinery can be improved dramatically in certain cases through a systematic and meticulous evolutionary algorithm search through the space of all structural geometries permitted by manufacturing, cost and functional constraints. This is a cheap and elegant approach in scenarios where employing active control elements is impractical for reasons of cost and complexity. From an optimization perspective the challenge lies in the efficient, yet thorough global exploration of the multi-dimensional and multi-modal design spaces often yielded by such problems. Moreover, the designs are often defined by a mixture of continuous and discrete variables--a task that evolutionary algorithms appear to be ideally suited for. In this article we discuss the specific case of the optimization of crop spraying machinery for improved uniformity of spray deposition, subject to structural weight and manufacturing constraints. Using a mixed variable evolutionary algorithm allowed us to optimize both shape and topology. Through this process we have managed to reduce the maximum roll angle of the sprayer by an order of magnitude, whilst allowing only relatively inexpensive changes to the baseline design. Further (though less dramatic) improvements were shown to be possible when we relaxed the cost constraint. We applied the same approach to the inverse problem of reducing the mass while maintaining an acceptable roll angle--a 2% improvement proved possible in this case.