PSI - Issue 39

Nabam Teyi et al. / Procedia Structural Integrity 39 (2022) 608–623 Author name / Structural Integrity Procedia 00 (2019) 000–000

610

3

Certain NDTs are carried out on such shafts and beams before they are actually assembled in the machines. NDTs are testing methods which without causing damage to the material, component, or system, evaluate their properties. NDTs like ultrasonic test, radiographic test, electromagnetic test and liquid penetrating test have been used in structures to detect physical flaws. Cheng et al. (2017) detected cracks using a NDT method based on magnetic optical imaging (MOI) on the basis of Faraday Magneto-optical Effect (FMoE) scheme. When the magnetic field moved in the direction of the polarized light, it rotated its polarizing direction. Yusa et al. (2012) used electromagnetic NDT to detect stress corrosion cracking. Hamia et al. (2014) used eddy-current non-destructive testing system to verify of orientation of crack. Sharma et al. (2015) studied identification of a single open crack in a multi-span straight beam using NDT. In this case, the crack was modelled as an analogous elastic rotational spring because it was considered to be transverse and one-dimensional. Such NDTs help the engineers to reject the defective shafts and beams if the flaws are beyond the tolerance and allowance set. Such dismissal of faulty components is very essential so that only healthy links are assembled in machineries. However more often than not, cracks are so small that even NDTs miss their presence at certain times. A faulty motion transfer link like shaft or structural link like beam in any working machines is a big risk. And so, if the engineers have suspicion of presence of cracks in the shaft or beam already in operation in the working machines, then what would they do? Would they stop the operating machineries, disengage the shaft or beam component for the crack diagnosis? If the engineers on site are experienced and their judgment could be trusted, may be, they would actually do that. And let’s say, after putting a halt in production flow and disassembling the component, no crack was detected. This would be a big blow to production time and economy. Therefore, there has to be a robust and a reliable system that identifies presence or absence of cracks while the machines are in operating condition with their shafts and beams performing their intended work. And this is done by signal based methods. This is called as signal monitoring and the signal generated is vibration based signal. Basically, the vibration output of a shaft or beam in operation is recorded as displacements for every cycle of motion. The output of the system, which is the synthetic signal generated, is obtained in the data acquisition system using probes and sensors. The data hence generated as the output of the system as the vibration signals is a data series and a comprehensive dataset. Even yet, the machine’s output signal will have flaws due to the fact that no signal in nature or the universe is free of various noise signals. As a result, the signal generated is a composite of both legitimate and undesirable signals. Data science tools effectively remove the effect of the noise signals from the obtained data to provide the true desirable data. The real data may further be treated with data science tools to perform the required investigation. Data science is the study of data systems and procedures to make sense of vast data volumes. For data clusters, data scientists use tools and algorithms based on statistical inference concepts and principles. Once the presence of crack(s) is ascertained, the next logical step is to locate the exact or approximate probable position of the crack(s). Indentifying crack is called as crack identification, finding the crack location is called crack localization. Data science tools like the AI techniques and ML models have been used by many reputable authors in their crack analysis. Si et al. (2020) published a critical review paper on evolution and application of potential methods in studying crack initiation and propagation. Nasiri et al. (2017) provided an overview of five AI algorithms utilized in fracture mechanics that included Bayesian networks, artificial neural networks, genetic algorithms, fuzzy logic, and case-based reasoning. A similar effort is made in this paper to assemble a reference for future researchers who need to conduct a literature survey in the field of structural crack analysis as part of their research in order to give a reference. 2. Cracks in shafts A crack is a small opening on the shaft’s surface, similar to a slit cut, which can occur owing to faulty shaft material manufacture or during the component’s operation. Even when the shaft is newly made, such cracks are so minute that they are impossible to detect visually. When the shaft is rotating, cracks are virtually impossible to detect. As a result, modelling and analysis employing various methodologies to be confirmed experimentally or vice versa are required for crack identification for its presence, crack localization for its location, and crack assessment for its depth and severity. A single crack, a double crack, or multiple cracks may be present in the models (Fig. 1.). And the cracks could be longitudinal, in which case the crack runs parallel to the shaft axis; slant, in which case the crack runs at an angle to the shaft axis; or transverse, in which case the crack runs perpendicular to the shaft axis (Fig. 2.). The most