Design & Implementation of Automatic Machine Condition Monitoring and Maintenance System in Limited Resource Situations

The different properties of the collected data are given shortly:
5. Data Types: Vibration data of three different types of faults occur in Rotating motor/devices
6. Data length: 1 seconds
7. Rotations/Second: 48.33
8. Data acquisition system: ESP32, gyro acceleration sensor, SD card, OLED display.
9. Initial data dimension: 1600 rows, 4 columns.
Initially, we computed the average of columns containing axis values at different frequencies. Subsequently, to form the dataset, we calculated the partial derivative of the averages concerning frequency, representing changes in data magnitude relative to different frequencies. This data, processed with Machine Learning models, demonstrated that different faults produce varying magnitudes at different frequencies, exhibiting distinct patterns. Structural looseness, misalignment, and bearing problems manifested differing magnitudes at different frequencies, contributing to distinct overall patterns .
Structures:
Folders (Structural_Looseness, Bearing, Misalignment) contains the Raw CSV Files containing vibration signals of faulty machines. The MatLab Files Features and Models containing the features, visualization, spectrum and trained models with the features.

本次采集数据的各项属性简述如下：
5. 数据类型：旋转电机/设备的三类故障振动数据
6. 数据时长：1秒
7. 转速：48.33转/秒
8. 数据采集系统：ESP32、陀螺加速度传感器、SD卡、OLED显示屏
9. 初始数据维度：1600行、4列。
研究初期，我们对包含不同频率下轴分量的列求取平均值。随后，为构建该数据集，我们计算了该平均值关于频率的偏导数，以表征数据幅值随不同频率的变化情况。经机器学习模型处理后，结果表明：不同故障在不同频率下会呈现各异的幅值表现，并具备独特的模式特征。其中，结构松动、不对中以及轴承故障会在不同频率下展现出差异化幅值，进而形成各自专属的整体分布模式。
数据集结构说明：
文件夹Structural_Looseness（结构松动）、Bearing（轴承故障）、Misalignment（不对中）内存储了故障设备的振动信号原始CSV文件。文件夹MatLab Files Features and Models内含基于该数据提取的特征、可视化结果、频谱图以及基于这些特征训练得到的模型文件。