The human foot is easily deformed owing to the innate form of the foot or an incorrect walking posture. Foot deformations not only pose a threat to foot health but also cause fatigue and pain when walking; therefore, accurate diagnoses of foot deformations are required. However, the measurement of foot deformities requires specialized personnel, and the objectivity of the diagnosis may be insufficient for professional medical personnel to assess foot deformations. Thus, it is necessary to develop an objective foot deformation classification model. In this study, a model for classifying foot types is developed using image and numerical foot pressure data. Such heterogeneous data are used to generate a fine-tuned visual geometry group-16 (VGG16) and K−nearest neighbor (k-NN) models, respectively, and a stacking ensemble model is finally generated to improve accuracy and robustness by combining the two models. Through k-fold cross-validation, the accuracy and robustness of the proposed method have been verified by the mean and standard deviation of the f1 scores (0.9255 and 0.0042), which has superior performance compared to single models generated using only numerical or image data. Thus, the proposed model provides the objectivity of diagnosis for foot deformation, and can be used for analysis and design of foot healthcare products.

1. 서 론

Faulty compressors must be detected in advance to speed up the quality control process of the compressor’s performance. Machine learning models have recently been used as fault classification models to distinguish between normal and abnormal compressors, facilitating more sophisticated fault detection methods than those in the past. However, very few studies have been conducted on accurate and efficient feature selection, despite its high importance. Therefore, this study proposes a new hybrid method that combines the merits of existing feature methods, filter and wrapper methods, to obtain a stable, accurate, and efficient fault classification model. For this, three types of filtering methods with different characteristics, such as chi-square, extra tree classifier, and correlation matrix, are used to derive the high-ranked features and then create a powerful feature set consisting of their union sets. Subsequently, using the wrapper method, one combination of features with the highest classification accuracy was selected among all the combinations of features in the union feature set. Using two experimental examples and one numerical example with different types and numbers of data, the robustness and accuracy of the proposed method were verified through comparison with the existing filter methods by combining three classification models: support vector machine, K-nearest neighbor, and multi-layer perceptron.

Most health indicators used to predict bearing performance degradation are calculated through supervised learning of failure data. However, due to the limited availability of failure data during the operation of bearings and the uncertainty of failure occurrences, using a supervised learning method to predict failure thresholds of bearings becomes challenging. In this study, a bearing degradation model is proposed through the development of health indicators using unsupervised learning. The proposed method extracts features from data with high signal amplitudes among multi-time windows data through frequency analysis and develops a health indicator (HI) by fusing these features using principal component analysis (PCA). The six-sigma value for the bearing’s HI data is used as a threshold for determining the first predicted time to failure (FPTF), which is a criterion for an unsupervised method of detecting abnormalities in the performance bearings in real-time. The proposed method accurately detects abnormal states and the degradation model shows improved trendability compared to existing models.

In the fault classification process, filter methods that sequentially remove unnecessary features have long been studied. However, the existing filter methods do not have guidelines on which, and how many, features are needed. This study developed a multi-filter clustering fusion (MFCF) technique, to effectively and efficiently select features. In the MFCF process, a multi-filter method combining existing filter methods is first applied for feature clustering; then, key features are automatically selected. The union of key features is utilized to find all potentially important features, and an exhaustive search is used to obtain the best combination of selected features to maximize the accuracy of the classification model. In the rotating machinery examples, fault classification models using MFCF were generated to classify normal and abnormal conditions of rotational machinery. The obtained results demonstrated that classification models using MFCF provide good accuracy, efficiency, and robustness in the fault classification of rotational machinery.