Developing a continuous quality control method for the sawmilling process using sensory features and advanced data mining approaches

锯木厂正在不断调整，以提高他们的回收率和生产率。然而，大多数关于锯木厂使用的周边切割工艺(即切纸机、弯刀和锯子)的研究都是在实验室进行的。虽然这类研究对于更好地了解刀具-木材相互作用中涉及的切割动力学是必不可少的，但将结果应用于或外推到恶劣的锯切条件可能是具有挑战性的。此外，基于现有的文献很难设计出最优的切割系统，因为已发表的结果往往是相互矛盾的。因此，该研究计划的长期目标是开发周边切割(即锯切)的模型，以超越现有的主要专注于正交切割的模型。与研究计划相关的五个具体目标是：1)量化各种切割参数对不同响应变量的影响；2)将不同传感器的感官特征与自变量和因变量相关联；3)在实验室范围内开发新的数据挖掘和人工智能(AI)方法来描述和预测锯切过程；4)在工业背景下验证所开发的方法；5)使用工业大数据集建立多目标模型。拟议的方法由两个阶段组成。第一阶段将在实验室环境中进行，变量受到精确控制，而第二阶段将在工业环境中进行。该项目将由九个不同的步骤组成。步骤1-5将是第一阶段的一部分，旨在评估切割参数对响应变量的影响，并探索新的数据挖掘和人工智能方法。步骤6-9是阶段2的一部分，重点是了解实验室和工业环境之间的基本差异及其对切割动力学的影响，并验证在阶段1中开发的模型。根据结果，可能会编写新的模型。从该项目中获得的知识将与切割动力学以及新的数据挖掘和人工智能方法在木材加工领域的应用有关。实验室和工业环境之间的相互关系也将得到更好的理解，从而能够促进技术进步和研究成果的实施。预期调查结果的相关性与流程优化和产品价值最大化有关，从而鼓励将调查结果转化为实践。同时，开发的模型将有利于木材行业向工业4.0过渡，从而提高质量控制和监测计划的效率。结果将在与研究领域相关的同行评议期刊上传播，并在会议上公布。因此，木材加工界的同事将能够在这些发现的基础上再接再厉，进一步促进我们对周边切割中刀具-木材动力学的理解。

Sawmills are constantly adapting to improve their recovery and productivity. However, most of the research done on peripheral cutting processes used in sawmills (i.e., chippers, canters and saws) has been conducted in a laboratory setting. While this type of research is essential to better understand the cutting dynamics involved in the tool-wood interaction, results can be challenging to apply or extrapolate to the harsh sawmilling conditions. Moreover, it is difficult to design optimal cutting systems based on the available literature since the published findings are often contradictory. The long-term objective of the research program is therefore to develop models for peripheral cutting (i.e., sawing) to move beyond the existing ones that mainly focus on orthogonal cutting. Five specific objectives are related to the research program: 1) quantify the impact of various cutting parameters on different response variables; 2) correlate sensory features of different sensors with the independent and dependent variables; 3) develop novel data mining and artificial intelligence (AI) approaches at the laboratory scale to describe and predict the sawing process; 4) validate the developed methods in an industrial context; 5) build a multi-objective model with an industrial large dataset. The proposed methodology is comprised of two phases. Phase 1 will be conducted in a laboratory environment, where variables are precisely controlled, while Phase 2 will be conducted in an industrial setting. The project will consist of nine distinct steps. Steps 1-5 will be part of Phase 1, aiming to evaluate the impact of cutting parameters on response variables and exploring novel data mining and AI approaches. Steps 6-9 are part of Phase 2, focusing on understanding the fundamental differences between laboratory and industry settings and how it influences the cutting dynamics, and validating the models developed in Phase 1. Depending on the results, new models may be programmed. The knowledge gained from the project will relate to the cutting dynamics as well as the application of novel data mining and AI approaches in the field of wood machining. The interrelation between laboratory and industrial environments will also be better understood so the implementation of technological advancements and findings can be facilitated. The relevance of the expected findings relates to process optimization and product value maximization, thus encouraging translation of findings into practice. In parallel, the developed models will be beneficial to the wood industry in transitioning towards Industry 4.0, thus improving the efficacy of quality control and monitoring programs. Results will be disseminated in peer-reviewed journals relevant to the field of research, and presented in conferences. Consequently, colleagues from the wood machining community will be able to build on these findings, further advancing our understanding of the tool-wood dynamics in peripheral cutting.