Blending State Identification of a Blender Based on Vibration Signal Analysis

Blenders have become one of the essential kitchen appliances in modern life due to their convenience and simplicity of use. At present, users mostly set the blending time of the blender according to the preset program of the manufacturer or based on personal experience, which often leads to insufficient or excessive process due to the different hardness and volume of the ingredients, and hence influencing the taste and quality of the processed food or resulting in an unnecessary power waste. To address this issue, in this paper, we propose a method to collect the vibration signals of the blender and then extract various features through signal processing to identify the blending state of the blender. Firstly, the vibration signals are collected by mounting a miniature accelerometer on the base of the blender under three types of working conditions, namely, empty cup, full water and ice. Secondly, a short-time Fourier transform is applied to the vibration signals to observe the variation of the main frequency components over the entire blending period. On this basis, a number of time-domain and frequency-domain feature parameters are extracted, and some sensitive feature indicators are compared for different stirred ingredients. Finally, based on the above multiple features, an SVM model is trained to classify the above three types of working conditions. In addition, nine different working conditions were further classified with recognition rates exceeding 99%. This work can serve as an important reference for further optimization of the blender design.