Qualitative evaluation of 3D printed materials' structure using active infrared thermography and signal analysis based on LSTM neural networks

Article Instruments & Instrumentation

Active thermography for quality assurance of 3D-printed polymer structures

Christian Metz et al.

Summary: Test specimens made of PA 12 by Laser Sintering and ABS by Fused Layer Modelling were characterized with active thermography after manufacturing and artificial weathering, using different excitation methods for defect detection. Data processing methods like TSR and Fourier Transformation were applied for improving the quality of thermograms. The long-term stability of these specimens under environmental stress was further investigated with an artificial weathering test over 2000 hours, supplemented by spectral reflectance and UV/VIS spectroscopy monitoring of changes in optical properties.

QUANTITATIVE INFRARED THERMOGRAPHY JOURNAL (2021)

Add to Collection

Article Energy & Fuels

Comparison of Machine Learning Methods for Image Reconstruction Using the LSTM Classifier in Industrial Electrical Tomography

Grzegorz Klosowski et al.

Summary: Electrical tomography is a non-invasive method used in various industries for monitoring objects' interior. Achieving high-resolution tomographic images faces challenges due to the ill-posed inverse problem and noisy input data. The study presents a machine learning concept for intelligent method selection, utilizing long short-term memory networks to classify and choose optimal algorithmic methods for tomographic reconstructions, showing improved efficiency compared to using single homogeneous methods.

ENERGIES (2021)

Add to Collection

Article Optics

Deep learning for defect characterization in composite laminates inspected by step-heating thermography

Roberto Marani et al.

Summary: This study presents a complete procedure for non-destructive analysis of composite laminates using step-heating infrared thermography and deep neural networks. The one-dimensional temperature profiles of the target surface are collected and fed into a compact network of convolutional filters to classify the resulting features for complete three-dimensional characterization of possible subsurface defects. Experimental validation showed successful classification and identification of defects of different depths in glass-fiber-reinforced polymer laminates by varying the duration of heat pulses.

OPTICS AND LASERS IN ENGINEERING (2021)

Add to Collection

Article Computer Science, Information Systems

Multi-Behavior with Bottleneck Features LSTM for Load Forecasting in Building Energy Management System

Van Bui et al.

Summary: This study proposes a novel multi-behavior with bottleneck features long short-term memory (LSTM) model for load forecasting in building energy management systems, combining the predictive behavior of long-term, short-term, and weekly feature models. The model offers improved performance and stability compared to single-model LSTM, displaying excellent adaptability to unexpected situations.

ELECTRONICS (2021)

Add to Collection

Article Chemistry, Physical

Detection and Identification of Defects in 3D-Printed Dielectric Structures via Thermographic Inspection and Deep Neural Networks

Barbara Szymanik et al.

Summary: This paper proposes a new method based on active infrared thermography for assessing the state of 3D-printed structures. The technique involves using an external energy source to create temperature differences between undamaged and defective areas, which may be hard to measure in materials with low thermal conductivity. A dedicated algorithm and deep convolutional neural network are used for signal analysis to enhance contrast between background and defect areas. Experimental results demonstrate the effectiveness of this hybrid signal analysis method in visualizing inner structure and determining defect parameters such as depth and diameter.

MATERIALS (2021)

Add to Collection

Article Chemistry, Physical

Modeling the Average and Instantaneous Friction Coefficient of a Disc Brake on the Basis of Bench Tests

Wojciech Sawczuk et al.

Summary: This study presents test results on the average and instantaneous friction coefficients of railway vehicle disc brakes, with consideration of wear states on friction linings and brake discs, and development of regression models for further evaluation and validation. The coefficients for the models were found to be 0.99 and 0.71 respectively.

MATERIALS (2021)

Add to Collection

Article Mathematics, Applied