Using acoustic impacts and machine learning for safety classification of mine roofs. (November 2021)

• Record Type: Journal Article
• Authors: Wiens, Travis; Islam, Md. Shahriar
• Is Part Of: International journal of rock mechanics and mining sciences. Volume 147(2021)
• Publication Date: 2021-11
• Subjects: Safety -- Mine roof stability -- Machine learning -- Classification -- Acoustics; Rock mechanics -- Periodicals; Soil mechanics -- Periodicals; Mining engineering -- Periodicals; Roches, Mécanique des -- Périodiques; Sols, Mécanique des -- Périodiques; Technique minière -- Périodiques; 624.151305
• DOI: 10.1016/j.ijrmms.2021.104912
• Languages: English
• ISSNs: 1365-1609

Abstract

Abstract: The safety of mine roofs is often evaluated by listening to the sound of a scaling bar impact: well-supported ground sounds "tight" while potentially unsafe ground is described as "drummy". This is a subjective process that requires training and experience before a mine worker can confidently assess the sound. This paper presents a method of removing subjectivity from this process. The authors recorded over 3000 impacts from a variety of potash mines, which were labeled as "tight" or "drummy" by experienced workers. We then used machine learning techniques to classify these recordings and were able to achieve classification errors of less than 1% on new data not used to train the algorithms. Alternately, we could bring false-negative results to zero with a small increase in false-positives. This paper also presents a number of different classification and data dimensionality reduction techniques and compares their effectiveness on this dataset.