Machine learning algorithm for distinguishing microscopic cracking mode based on acoustic emission characteristics

The invention discloses a machine learning algorithm for distinguishing a microscopic cracking mode based on acoustic emission characteristics. The algorithm comprises the following steps: firstly, forming an initial unmarked data set by utilizing acoustic emission average frequency (recorded as AF) and elevation angle cotangent (recorded as RA) monitored by a test; secondly, constructing a weight vector according to a cracking mode category, performing iterative calculation to obtain a central point of data clustering, and marking vectors in the data set; and then, solving a Lagrange multiplier vector corresponding to the data set, calculating to obtain a linear clustering equation of the data set, and finally obtaining a linear discrimination standard of the cracking mode. The algorithm is suitable for judging all materials with cracking mode characteristics which can be established through acoustic emission, is suitable for mechanical tests (including static tests, dynamic tests, fatigue tests, impact test