The pursuit of lightweight automobiles is accelerated by decreased fuel consumption and CO2 emissions as well as increased battery electric vehicle range (BEVs). The need for lightweight design rises due to investments in new technologies, creating opportunities for structural components to use new, more sophisticated lightweight materials such as carbon-fiber-reinforced polymers (CFRP).
Carbon-Fiber-Reinforced Polymers
CFRP has a high strength-to-weight ratio and exhibits significant fatigue and creep resistance levels. Its qualities are appropriate for use in the commercial vehicle sector, where high-cycle fatigue is prominent and the vehicle service life is steadily rising.
Voids in Carbon-Fiber-Reinforced Polymers Production
Voids are among the most prevalent production flaws, and it takes many costs to prevent the creation of voids in CFRP materials production methods. However, the development of voids is unavoidable for current manufacturing procedures that aim for reduced production costs and quicker production time. Therefore, if the material is to be used in high-volume manufacturing, it is crucial to assess void quality.
Void Characterization Methods
Materials may be characterized using a wide variety of void characterization methods. Several studies have compared common strategies, and they all have benefits and drawbacks. For example, ultrasonic testing and X-ray computed tomography (CT) are more sophisticated methods to quantify vacancies. While X-ray CT is a sophisticated technology gaining popularity, ultrasonic testing is commonly utilized as a quality assurance measuring tool in the industry.
One of the most popular imaging methods for evaluating void content is optical microscopy, widely utilized in institutions, conveniently accessible, and carried out by an engineer. With optical microscopy, the void content may be determined statistically with up to 25 images, and the accuracy can approach 0.2%.
How the Study was Conducted
Pursuing simulation-driven design leads to ongoing advancement and development of increasingly potent modeling techniques. In composite research, neural networks are gaining popularity in mechanical characteristics prediction. In this study, material features are extracted from micrographs using optical microscopy and then statistically defined in a unique and widely applicable way. It is also shown that using a neural network may significantly automate data examination
Material System and Optical Microscopy
The researchers used a unidirectional (UD) CFRP prepreg as material. Heated compression molding was employed to prepare the material system, and three distinct types of plates were produced.
A void fraction analysis was performed on micrograph samples looking towards the fiber direction using a 5x and 50x magnification on an Olympus BX53 microscope to ascertain the void fractions of the produced plates.
Neural Network and Image Analysis
The micrographs were analyzed using several image analysis methods, then contrasted. Two alternative approaches—the selection technique and thresholding, often known as SM and TH, respectively—were used to compute the void fraction.
A convolutional neural network (CNN) with a U-Net architecture was utilized. This kind of neural network is classified as supervised learning, and its main goal is to map input data or an input layer to an output layer. Doing this eliminates the manual work required to employ the SM and TH. The neural network distinguished between the three phases, fiber, matrix, and void, speeding up the analysis of micrograph fractions.
Significant Findings of the Study
A statistical method for characterizing porosity in a UD CFRP material is presented in this study. Micrographs have been effectively segmented using a neural network, identifying the components, including fibers, matrix, and voids.
The study demonstrates that specific micrograph components may be accurately recognized with only a few training photos. For example, only 100 training photos were required for the neural network to recognize voids and void fractions with more than 90% accuracy for the current material and micrographs.
The neural network functions similarly to previous techniques used for micrograph image segmentation, and it may be taught to perform even better and concentrate on solving more complicated problems.
The material's porosity can be determined statistically. This information can draw inferences about how to utilize the material's properties in simulations and explain mechanical behavior during testing.
Future Prospect
The research's conclusions can be used to improve the efficiency of other repetitive manual operations that need a lot of data. The authors have included and shared the neural network implementation script to make it more accessible to the research community.
Reference
Sara Eliasson, Mathilda Karlsson Hagnell, Per Wennhage and Zuheir Barsoum (2022) A Statistical Porosity Characterization Approach of Carbon-Fiber-Reinforced Polymer Material Using Optical Microscopy and Neural Network. Materials. https://www.mdpi.com/1996-1944/15/19/6540/htm
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.