Controlling the Properties of Laser-Induced Graphene by Machine Learning

该项目的目标是利用机器学习（ML）改进激光诱导石墨烯（LIG）的制造。LIG是一种新发现的材料，因其制备简单、性能优越而备受关注。LIG可以通过将聚合物（例如柔性塑料箔或3D打印物体）暴露在扫描的激光束下来创建。只要聚合物暴露在激光下，它就会转化为激光。不需要化学合成、石墨烯沉积或光刻等额外步骤。这使得这个过程非常简单、快速和低成本。所得到的LIG是导电的，并且具有多孔微观结构，其特性可以通过改变激光参数或聚合物前驱体的化学特性来调节。为了实现特定应用的最佳LIG特性，由于复杂的底层过程，需要进行大量的实验或复杂的模拟。这限制了LIG设备的性能，以及LIG适应新应用的难易程度。在这里，我们建议训练一个ML模型来自动优化LIG的属性。ML的一个主要优势是能够映射输入和输出之间的复杂关系，即使使用传统的分析模型很难做到这一点。我们将使用我们实验室的实验结果来训练ML模型。项目的实验部分和机器学习部分之间将持续沟通，以确保新收集的数据改善模型中需要进一步训练的区域。这个项目是加拿大的LIG专家和美国的ML专家之间的国际合作，这个项目将在多个方面使加拿大受益。LIG可用于各种设备，从物理和生物医学传感器到超级电容器和储能电池，该项目将有利于加拿大此类产品的制造商。更普遍的是，机器学习在优化先进制造工艺方面变得越来越重要，该项目将在技术成果和学生培训方面增加加拿大在这一领域的能力。

The goal of this project is to improve the fabrication of laser-induced graphene (LIG) using machine learning (ML). LIG is a recently discovered material that has received a lot of attention due to its simple fabrication and favorable properties. LIG can be created by exposing a polymer, for example a flexible plastic foil or a 3D printed object, to a scanned laser beam. Wherever the polymer is exposed to the laser, it is converted to LIG. No additional steps such as chemical synthesis, deposition of graphene, or photolithography are needed. This makes the process very simple, fast, and low-cost. The resulting LIG is electrically conductive and has a porous microstructure with properties that can be tuned by varying the laser parameters or the chemical properties of the polymer precursor. In order to achieve optimal LIG properties for a particular application, extensive experimentation or complex simulations are necessary due to the complex underlying processes. This limits the performance of LIG devices and how easily LIG can be adapted to a new application. Here, we propose to train a ML model to automatically optimize the properties of LIG as desired. A major strength of ML is to be able to map complex relationships between inputs and outputs even if it is difficult to do so using traditional analytical models. We will train the ML model using experimental results in our lab. There will be a constant communication between the experimental and ML parts of the project to ensure that the newly collected data improves regions of the model that require further training. This project is an international collaboration between researchers in Canada who are experts for LIG and researchers in the United States who are experts for ML. This project will benefit Canada in multiple ways. LIG can be employed in a variety of devices ranging from physical and biomedical sensors to supercapacitors and batteries for energy storage and this project will be beneficial for Canadian manufacturers of such products. More generally, ML is becoming increasingly important to optimize advanced manufacturing processes and this project will add to Canada's capacity in this area both in terms of technical results and student training.