📝 SummarXiv

Abstract

In many industries, the scale and complexity of systems can present significant barriers to the development of accurate digital twin models. This paper introduces a novel methodology and a modular computational tool utilizing machine learning and dimensional analysis to establish a framework for scaling digital twin models. Scaling techniques have not yet been applied to digital twin technology, but they can eliminate the need for repetitive physical calibration of such models in industries where product lines include a variety of sizes of the same or similar products. In many cases, it may be easier or more cost-effective to perform physical calibration of the digital twin model on smaller units of a product line. Scaling techniques can then allow adapting the calibration data from the smaller units to other sizes of the product line without the need for additional data collection and experimentation for calibration. Conventional application of dimensional analysis for scaling in this context introduces several challenges due to distortion of scaling factors. This paper addresses these challenges and introduces a framework for proper scaling of digital twin models. The results are applied to scaling the models between an industrial-size wheel loader vehicle used in construction to a miniaturized system instrumented in a laboratory setting.