Modeling Smart Multi Module 2-Axis Inclinometer with Machine Learning and Simulink

Extensive research has been conducted on 2-axis inclinometer, covering research of Micro-electromechanical systems (MEMS) accelerometer, calibration studies, performance analysis, data transmission using RS 485, and the investigation of temperature effects on measurement errors. In this research, the focus is on developing a smart inclinometer model using Machine Learning. By utilizing acceleration data (ax, ay, and az), valuable information can be obtained regarding the inclination angle, displacement distance of each inclinometer, total displacement distance, direction of displacement axis, which soil layer is shifting, and triggering danger alarms. Obtaining such information would be challenging and complex if using shallow computations. Modeling was performed using random forest machine learning after experimentation with other machine learning techniques. The parameters used for optimal results were set with max_depth set to 0, min_samples_leaf set to 5, and n_estimators set to 200. Two types of machine learning, namely regressor and classifier, were applied. With a test_ size set to 0.1, the MAE in the first method is 0.9 and the Mean Absolute Error (MAE) in the second method is 0.4. The usage of MATLAB Simulink for simulating a Python program was conducted to observe the characteristics of the created model. The validation model is performed by sending accelerometer data from a mobile phone to the MATLAB Cloud using the MATLAB application on the mobile phone, and comparing it with simulated data.