Automatic crack classification by exploiting statistical event descriptors for Deep Learning
This work addresses crack classification for structural health monitoring systems, representing an incremental improvement with a specific application.
The paper tackled the problem of classifying crack types (tensile, shear, mixed modes) in structural health monitoring by combining deep neural networks with statistical event descriptors, achieving 92% accuracy in classifying incipient damages.
In modern building infrastructures, the chance to devise adaptive and unsupervised data-driven health monitoring systems is gaining in popularity due to the large availability of big data from low-cost sensors with communication capabilities and advanced modeling tools such as Deep Learning. The main purpose of this paper is to combine deep neural networks with Bidirectional Long Short Term Memory and advanced statistical analysis involving Instantaneous Frequency and Spectral Kurtosis to develop an accurate classification tool for tensile, shear and mixed modes originated from acoustic emission events (cracks). We investigated on effective event descriptors to capture the unique characteristics from the different types of modes. Tests on experimental results confirm that this method achieves promising classification among different crack events and can impact on the design of future on structural health monitoring (SHM) technologies. This approach is effective to classify incipient damages with 92% of accuracy, which is advantageous to plan maintenance.