Many types of sensor have been utilized to monitor milling vibration, and many analysis methods are devoted to the investigation of milling vibration or milling dynamics. In this work, a noncontact sensor and a time-frequency domain analysis method were applied to identify the state of milling vibration. A microphone was employed in practical tests to record the milling dynamics. The Teager-Huang transform (THT) was adopted for the acoustic signal analysis owing to its high resolution in the time-frequency domain. The potential frequency range for the analysis of milling dynamics is reported in this work to improve the recognition accuracy of milling vibration limited by the effect of environmental noise. The THT was used to distinguish the chatter state from the normal milling dynamics. In addition, the statistical index called the coefficient of variation was applied to define the threshold of chatter occurrence. Milling experiments (including dry and wet cuttings) were performed to verify the proposed chatter detection method.

Durians are among the most important fruit products in tropical countries. The environments of durians therefore must support a high yield to meet demand. Sunlight, temperature, and rainfall are all key variables, and any adverse factors will have a negative impact on production. We propose an environmental prediction system for a durian farm on the basis of the concept of the IoT. The system uses multiple machine learning algorithms to analyze collected environmental data and predict the next state of the environmental variables. From numerous experiments, our results show that the support vector machine (SVM) gives the best forecasts for temperature, whereas the convolutional neural network (CNN) performs best for predicting soil humidity. The results of this paper can provide farmers with real-time understanding of their farms and early warning of potential risks. The farm yield rates can hence be increased.