Anomaly Detection Industrial Items

Abstract

Anomaly detection is one of the machines learning powered approach which identifies the various elements and objects from a big set of the data. The machine learning powered technique checks multiple objects with the help of differentiating properties and detects the objects which are dissimilar to another dataset. Anomaly detection has number of successful applications from security, medical, industrial and virtually all of the tasks which human may fail while identifying the differentiating properties. Machinery elements anomaly detection has been presented which identifies the differentiating objects of the four hundred images. Number of experiments were performed to detect the anomaly among the given 400 images. The training model has used bidirectional LSTM which gets a feature vector of twelve features, selected, and extracted with already defined algorithm. The dataset has been labelled manually and features were stored. While testing of the images, the study achieved 98% of the accuracy while detecting anomaly of the industrial elements. The system can detect anomalies in the presence of the varying colour and other properties. The accuracy can be increased by adding more images to the database and learning model more deeply. The study can be directed to various implementations including multiple learning implementations, applying on various other elements.