A Rapid Design and Fabrication Method for a Capacitive Accelerometer Based on Machine Learning and 3D Printing Techniques
A Rapid Design and Fabrication Method for a Capacitive Accelerometer Based on Machine Learning and 3D Printing Techniques
复制标题
DOI：
10.1109/jsen.2021.3085743
复制
发表时间：
2021-08-15
影响因子：
4.3
通讯作者：
Li, Zhe
中科院分区：
文献类型：
--
作者：
Liu, Guandong;Wang, Changhai;Li, Zhe
MEMS（微机电系统）传感器已经越来越多地用于在健康监测应用中检测人体运动。通常，MEMS传感器（如加速度计）的设计和制造的整个周期需要对器件功能的高度专业理解和微制造工艺的专业知识。然而，物联网（IoT）的出现带来了对低成本和高度可定制的传感器的大量需求，这需要快速制造和灵活设计，即使是对设备本身背景知识有限的客户也是如此。在这项工作中，我们提出了一个快速的设计和制造工作流程的加速度计相结合的人工神经网络（ANN）为基础的逆向设计方法和一步3D打印制造技术的发展。一步3D打印制造方法是基于导电复合材料，聚乳酸（PLA）聚合物与炭黑。在设备设计中，训练的双向人工神经网络被设计为从给定的设计参数预测设备性能，并从设备性能的客户要求中检索设计参数。然后基于检索到的几何参数设计电容式加速度计，并通过集成3D打印工艺制造，而不使用任何额外的金属化和组装工艺。该3D打印加速度计具有75.2 mV/g的灵敏度和良好的动态响应，能够检测和监测人体运动。所提出的快速设计和制造工作流程为适合物联网应用的定制和低成本MEMS器件提供了有效的解决方案。
MEMS (Micro Electromechanical System) sensors have been increasingly used to detect human movements in health monitoring applications. Usually, a full cycle of design and fabrication of a MEMS sensor such as an accelerometer requires highly professional understanding of device functions and expertise in microfabrication process. However, the advent of internet of things (IoT) brings a large demand for low-cost and highly customizable sensors, which requires fast fabrication and flexible design, even by the customers with limited background knowledge in the device itself. In this work, we present the development of a rapid design and fabrication workflow for accelerometers by combining an artificial neural network (ANN) based inverse design method and a one-step 3D printing fabrication technique. The one-step 3D printing fabrication approach is based on a conductive composite material, a polylactic acid (PLA) polymer with carbon black. In device design, trained bidirectional ANNs were designed to predict the device performance from given design parameters and retrieve the design parameters from the customer requirements of the device performance. A capacitive accelerometer was then designed based on the retrieved geometric parameters and fabricated by an integrated 3D printing process without using any additional metallization and assembly processes. With a sensitivity of 75.2 mV/g and a good dynamic response, the 3D printed accelerometer was shown to be capable of detection and monitoring of human movements. The proposed rapid design and fabrication workflow provides an effective solution to customized and low-cost MEMS devices suitable for IoT applications.