Detecting crossovers in polymer fiber using machine learning

点击翻译按钮获取中文摘要

Instrumar Limited's core technology platform is based on the measurement of electrical properties of materials that come in contact with its sensor. The sensor generates a tightly confined electromagnetic field that detects the properties as material enters this field. The sensor is sensitive to the amount, the shape, its density, temperature, conductivity, impurities, moisture content and flow speed of the material. The sensors sample data in real time and provide feedback based on parameters that are determined through extensive testing. This sensor is being used in multiple industries such as aerospace and industrial manufacturing. Two of their more successful applications are in the ice detection and polymer fiber manufacturing. Instrumar receives terabytes of data from its installations worldwide. It is one of their goals to turn this unprocessed high volume data into value-added information.The development process will be to use machine learning methods to identify and determine the top contributors to reliably and accurately determine the outcomes. We want to quantify the contribution (importance) of the original and computationally generated features in the dataset. Also, we want to be able to segregate between signatures, short term and long term trends such as machine harmonics. Once algorithms are successfully coded and models are generated it is important to identify the performance of the models. Using machine learning with extra data means we can develop better algorithms and simplified models. Being able to characterize developed algorithms allows Instrumar to provide better feedback to polymer fiber producers and other operators. This helps the manufacturers meet sustainability and environmental standards. The fiber industry is one of the planets worst polluters. Every efficiency gained has a huge environmental impact.