Classifying material type and mechanical properties using artificial neural network
04 Mar 2011-pp 207-211
TL;DR: In this paper, the authors focused on experimental data and study for the testing of the material mechanical properties using vibration technique, by applying vibration analysis and testing on the material, they could determine the natural frequencies, the damping ratio and mode shapes of the structure.
Abstract: This paper focused on experimental data and study for the testing of the material mechanical properties using vibration technique. By applying vibration analysis and testing on the material, we could determine the natural frequencies, the damping ratio and mode shapes of the structure. However, in this study, we only considering the natural frequencies of the material as the input data needed for training. As an extension for the study, the system tested with various method of neural network training algorithm. The Levenberg-Marquardt Backpropagation used as the algorithm in an artificial neural network system developed.
Citations
More filters
TL;DR: In this paper, a real-time processing and classification of raw sensor data using a convolutional neural network (CNN) was introduced for shape identification of grasped objects using a pneumatic gripper.
Abstract: This paper introduces a real-time processing and classification of raw sensor data using a convolutional neural network (CNN). The established system is a microcontroller-unit (MCU) implementation of an intelligent gripper for shape identification of grasped objects. The pneumatic gripper has two embedded accelerometers to sense acceleration (in the form of vibration signals) on the jaws for identification. The raw data is firstly transferred into images by short-time Fourier transform (STFT), and then the CNN algorithm is adopted to extract features for classifying objects. In addition, the hyperparameters of the CNN are optimized to ensure hardware implementation. Finally, the proposed artificial intelligent model is implemented on a MCU (Renesas RX65N) from raw data to classification. Experimental results and discussions are introduced to show the performance and effectiveness of our proposed approach.
7 citations
01 Jan 2022
3 citations
01 Jan 2022
TL;DR: In this article , the applicability of a machine learning algorithm known as K-nearest neighbor (KNN) for determining the dynamic fracture toughness of glass-filled polymer composites was investigated.
Abstract: Geometrical features like size and shape of the particles which are used to reinforce the composites affect the mechanical behavior of the resulting particulate polymer composites to a great extent. The aspect ratio of the reinforcing filler is of great importance specially when such composites are subjected to impact loading. Usually, an increase in the aspect ratio results in a significant increase in the energy-absorbing ability which ultimately improves the fracture toughness of the resulting composite. However, the experimental procedure followed for determining the fracture toughness of polymer composites reinforced with particles of varying aspect ratio is very complex and time-consuming. In this view, this chapter investigates the applicability of a machine learning algorithm known as K-nearest neighbor (KNN) for determining the dynamic fracture toughness of glass-filled polymer composites. The proposed methodology aims to predict the fracture toughness in terms of stress intensity factor with limited experimentation and maximum accuracy. The current framework of machine learning utilizes time, dynamic elastic modulus, aspect ratio, and volume fraction of the glass particles as the independent model parameters. The proposed KNN model predicts the fracture behavior of these composites with an accuracy of ~96%.
References
More filters
01 Jan 1999
TL;DR: Experimental modal analysis has been a popular technique for finding the modes of vibration of a machine or structure as discussed by the authors, which has become widespread as a fast and economical means of finding the mode of vibration.
Abstract: Experimental modal analysis has grown steadily in pop ularity since the advent of the digital FFT spectrum analyzer in the early 1970’s. Today, impact testing (or bump testing) has become widespread as a fast and economical means of finding the modes of vibration of a machine or structure. In this paper, we review all of the main topics associated with experimental modal analysis (or modal testing), including making FRF measurements with a FFT analyzer, modal excitation techniques, and modal parameter estimation from a set of FRFs (curve fitting).
346 citations
TL;DR: Investigation into the dynamic characterisation of a two dimensional (2D) flexible structure reveals that the measured parameters associated with the first five resonance modes of the system compare favourably with previously reported results.
Abstract: In this paper, an investigation into the dynamic characterisation of a two dimensional (2D) flexible structure is presented. A thin, flat plate, with all edges clamped, is considered. A simulation algorithm characterising the dynamic behaviour of the plate is developed through a discretisation of the governing partial differential equation formulation of the plate dynamics using finite difference methods. The algorithm is implemented within the Matlab environment, and it allows application and sensing of a disturbance signal at any mesh point on the plate. Such a provision is desirable for the design and development of active vibration control techniques for the system. The performance of the developed algorithm in characterising the dynamic behaviour of the system is assessed in comparison with previously reported results using various other methods. The validation of the algorithm is presented in both the time and frequency domains. Investigations reveal that the measured parameters associated with the ...
14 citations
TL;DR: In this article, a simple impact test method is presented to accurately measure the elastic and shear moduli and Poisson's ratio of a uniform Aluminum 6061-T651 cylindrical specimen with free boundary conditions.
11 citations
Additional excerpts
...H 12 = H 21 (2)...
[...]
27 Jun 2006
TL;DR: The least square and recursive least square are used to obtain linear parametric model of the system and non-parametric models of theSystem are developed using multi-layer perceptron neural networks and Elman neural networks.
Abstract: This paper present an investigation into the development of identification system approaches for dynamic modelling characterization of a two dimensional flexible plate structures. The least square and recursive least square are used to obtain linear parametric model of the system. Furthermore, non-parametric models of the system are developed using multi-layer perceptron neural networks (MLP-NN) and Elman neural networks (ENN). A simulation algorithm of the plate is developed through a discretisation of the governing partial differential equation formulation of the plate dynamics using finite difference methods. The finite duration step input is applied to simulation algorithm of the plate. Finally a comparative performance of the approaches used is presented and discussed.
9 citations
Journal Article•
TL;DR: The contents of four microamount elements of human blood were chosen as recognition index of coronary heart disease patients and normal persons and predicted that this method could be a supplementary tool to diagnose this kind of disease with the determined contents of microamount of elements in human blood.
Abstract: The contents of four microamount elements (Sr, Cu, Mg and Zn) in human blood were chosen as recognition index of coronary heart disease patients and normal persons. The recognition pattern of Levenberg Marquardt Backpropagation algorithm has been established. The first layer transfer function is Tansig function; the second layer transfer function is linear Purelin function. There are four input vectors, eight neurons on hidden layer, and one neuron of output vector. Four samples were chosen as a teat group and 22 samples as a training group. The weights and biases of the neural network were given. The given data could be completely identified, which predicted that this method could be a supplementary tool to diagnose this kind of disease with the determined contents of microamount of elements in human blood.
5 citations