Intelligent Control of Drying with Machine Vision and Machine Learning

热干燥是一个漫长且耗能的过程，会影响产品的微观结构和质量。干燥的挑战是如何以最小的能源消耗实现最佳的产品质量。目前使用的干燥方法在整个干燥过程中假定温度恒定，导致不必要的能量损失和产品质量的劣化。我的研究计划的长期目标是为干燥的自动控制和优化提供科学基础，以提高工艺效率和产品质量。由于产品质量的核心作用，工艺优化需要更好地了解工艺引起的产品质量变化。在过去的10年里，我的研究一直集中在使用机器视觉自动监测质量属性，如纹理，颜色，密度，孔隙率，收缩率和含水量。结果发现，生物材料的热敏性取决于含水量，这需要智能（自适应）控制。这个为期五年的项目的目标是：（一）确定生物材料在干燥过程中的热敏性变化;（二）建立优化标准和控制策略，以提高产品质量。第一个目标将包括干燥因素对热敏感生物材料质量的数值研究，如水果，蔬菜和蘑菇（1-3年）。因此，将获得有关干燥过程中产品质量变化的新知识。第二个目标将通过数学和推理模型的综合来实现，这些模型将把干燥因素与特定的质量属性联系起来，以建立优化标准（第4年）。采用响应面法确定优化域。这些知识将用于自适应控制系统的设计，该系统将允许根据产品质量连续调整干燥条件（第5年）。这项研究的基本原理是发现自适应智能控制的好处相比，常用的可编程干燥。我们预计机器视觉和机器学习与决策框架的集成将显着提高干燥系统的性能。该项目的科学成果是发现了将机器视觉和机器学习集成到智能控制系统中的基础知识。所提出的研究的新奇是从预先计划的可编程干燥到自适应控制的范式转变，以实现理想的产品特性。在机器视觉，智能控制和优化的接口方面的开创性研究将有利于工程理论，并为许多HQP的培训创造一个刺激的环境。这项新技术的开发将通过改进干燥技术和减少碳足迹，为加拿大的可持续创新组合做出重大贡献。

Thermal drying is a lengthy and energy consuming process that impacts product microstructure and quality. The challenge of the drying is how to achieve the best product quality with minimal energy expenses. Currently used drying methods assume constant temperature during the entire drying process, resulting in unnecessary energy losses and deterioration of product quality. The long-term goal of my research program is to provide a scientific foundation for automated control and optimization of drying to improve process efficiency and product quality. Due to the central role of product quality, the process optimization requires better understanding of the process-induced changes in product quality. Over the past 10 years, my research has been focused on the automated monitoring of quality attributes, such as texture, color, density, porosity, shrinkage and moisture content, using machine vision. As the result, it was discovered that thermal sensitivity of biomaterials depends on moisture content, which requires intelligent (adaptive) control. The objectives of this five-year project are: (i) to identify changes in thermal sensitivity of biomaterials during drying; (ii) to establish optimization criteria and control strategies to improve product quality. The first objective will include numerical study of drying factors on the quality of heat-sensitive biomaterials, such as fruits, vegetables, and mushrooms (Years 1-3). As a result, new knowledge on product quality transformations in drying will be acquired. The second objective will be achieved by synthesis of mathematical and inference models, which will link drying factors to specific quality attributes to establish criteria for optimization (Year 4). Optimization domain will be determined by using response surface methodology. This knowledge will be used in the design of an adaptive control system, which will allow continuous adjustment of drying conditions with respect to product quality (Year 5). The rationale for this research is to discover benefits of adaptive intelligent control compared to commonly used programmable drying. We expect that integration of machine vision and machine learning with a decision-making framework will significantly improve performance of drying system. The scientific outcome of this project is discovering the fundamental knowledge about the integration of machine vision and machine learning into intelligent control system. The novelty of the proposed research is the paradigm shift from pre-scheduled programmable drying to adaptive control to achieve desirable product characteristics. This pioneering research at the interface of machine vision, intelligent control and optimization will benefit engineering theory and create a stimulating environment for the training of many HQPs. The development of this novel technology will significantly contribute to a sustainable innovation portfolio in Canada by improving drying technologies and reducing carbon footprint.