Beam Damage Assessment Using Natural Frequency Shift and Machine Learning

Abstract

Damage detection based on modal parameter changes has become popular in the last few decades. Nowadays, there are robust and reliable mathematical relations available to predict natural frequency changes if damage parameters are known. Using these relations, it is possible to create databases containing a large variety of damage scenarios. Damage can be thus assessed by applying an inverse method. The problem is the complexity of the database, especially for structures with more cracks. In this paper, we propose two machine learning methods, namely the random forest (RF), and the artificial neural network (ANN), as search tools. The databases we developed contain damage scenarios for a prismatic cantilever beam with one crack and ideal and non-ideal boundary conditions. The crack assessment was made in two steps. First, a coarse damage location was found from the networks trained for scenarios comprising the whole beam. Afterwards, the assessment was made involving a particular network trained for the segment of the beam on which the crack was previously found. Using the two machine learning methods, we succeeded in estimating the crack location and severity with high accuracy for both simulation and laboratory experiments. Regarding the location of the crack, which was the main goal of the practitioners, the errors were less than 0.6%. Based on these achievements, we concluded that the damage assessment we propose, in conjunction with the machine learning methods, is robust and reliable.

Determining the Severity of Open and Closed Cracks Using the Strain Energy Loss and the Hill-Climbing Method

“…The laboratory setup consists of a rigid structure including a vise in which the beam is fastened, an excitation device, and the data acquisition system. The experimental setup is presented in Figure 8, and described in detail in [27]. The excitation system involves a speaker and amplifier which are controlled using AudioDope software.…”

“…By using the method described in the current paper, by employing Eq. ( 11), we have generated training data for developing a damage detection neural network, similar to the one presented in [13]. The training data consists of the RFS values for the six transverse vibration modes.…”

“…When examining the methods proposed in the literature, we observed that the dependency between the reduction in the size of the crack and the stiffness is rarely considered. In general, the authors limit the study to finding the element in the beam with the lower Young modulus value [13]. When damage detection is applied to trusses, finding the member with the lower Young modulus is usually the target.…”

“…The training data is obtained from simulation or measurements, thus initially it involves a limited number of damage cases. If for a given structure is possible to determine the relationship between the damage parameters and the vibration signal parameters, it is possible to generate a multitude of damage cases [27]. In this way, the training process can be improved, and the AI algorithms provide more accurate prediction results.…”

Determining the Severity of Open and Closed Cracks Using the Strain Energy Loss and the Hill-Climbing Method

“…The training data are obtained from simulation or measurements, thus initially it involves a limited number of damage cases. If for a given structure it is possible to determine the relationship between the damage parameters and the vibration signal parameters, it is possible to generate a multitude of damage cases [27] including the case of imperfect clamping. In this way, the training process can be improved, and the AI algorithms provide more accurate prediction results.…”