Smart crawfish: A concept of underwater multi-bolt looseness identification using entropy-enhanced active sensing and ensemble learning

Abstract

With the rapid development of the oil industry, more and more subsea pipelines enter into service, and thus we are facing the challenge of safe operation and maintenance of these underwater infrastructures. For instance, as a common component to connect subsea pipelines, the flange is prone to losing sufficient integrity due to severe working conditions and harsh environments under the sea. However, most of the current structural health monitoring (SHM) and nondestructive testing (NDT) techniques are targeted at onshore structures, and we require a more practical method for multi-bolt looseness identification of subsea flanges. In this paper, attempting to solve the above difficulty, we propose the “Smart Crawfish”, which is a concept that employs a subsea robot equipped with a gripper and a pair of Lead Zirconate Titanate (PZT) transducers to implement the active sensing method. Particularly, the main computational contribution of this paper is that we develop two new entropy indexes, i.e., multiscale range entropy (MRangeEn) and multiscale bubble entropy (MbEn) to enhance the current active sensing method (i.e., the entropy-enhanced active sensing method). To detect underwater multi-bolt looseness, we first obtain stress wave signals under different categories (i.e., different cases of multi-bolt looseness) by the active sensing and then compute their MRangeEn and MbEn to construct the feature dataset. Subsequently, via the dataset, a stacking-based ensemble learning classifier is trained to classify different categories. Finally, we conduct a lab-level experiment to demonstrate the effectiveness of the proposed method, which outperforms the current entropy-enhanced active sensing. To the best of our knowledge, this paper is the first attempt to address the detection of underwater multi-bolt looseness, which has great potential for future industrial applications.