An Indian Institute of Guwahati research team has created new modeling methods for predicting the failure of composite materials. A team led by Dr. Nelson Muthu published the results of their most recent work in the prestigious journal Composite Structures.
- Researchers at the Indian Institutes of Technology, Guwahati, have designed new methods to evaluate the failure probability of different types of biomedical material.
- The team members recently contributed to a paper in the renowned journal Composite Structures.
- Computerized studies like Monte Carlo simulations are often used to predict the failure of composite materials.
- IIT Guwahati researchers developed scalable and highly effective modeling approaches that use machine learning techniques such as support vector machines and sampling tools such as Latin hypercube. The methodology reduces the computational effort.
- A team of researchers from IIT Guwahati developed a method to efficiently model composite materials using multiscale metamodels that combine machine learning and sampling tools like Latin hypercube and equilibria. The method reduced the number of computer programs used by the team by about 95%.
Researchers at the Indian Institutes of Technology in Guwahati have devised novel methods to predict the failure probability of some of the common and useful materials. These include fiber aggregation, matrix cracking, density variation, and impact damage. Scientists have developed an innovative combination of machine learning tools and state-of-the-art sampling techniques that can accurately model the failure probability and mechanism of different materials used in the aerospace and automobile sectors.
This new method of modeling composite materials helps engineers accurately model the materials’ failure probabilities, using various probability techniques, such as Monte Carlo simulation. This is confirmed by the publication of a paper by Dr. Muthu and three other senior researchers: Prof. Aditya More, Puja Rakesh, and Munna Kumar.
Composite structures are materials that are fabricated using more than one component. They are widely used in aerospace, automobile and construction industries because of their good strength-to-weight ratios, durability, long life, etc. Composite materials are usually made up of fiber-reinforced plastic (FRP), which is widely used today.
Specialized fiber-reinforced composite materials are used in the aerospace industry to make specialized aircraft structures. Many parts of aircraft, including the wings, tail, doors, and interiors, are manufactured from these materials. Boeing 747 Dreamliner is 80% composite by volume, saving fuel up to 20-25% versus its predecessors.
Dr. Nelson Muthu, Professor IIT Guwahati, highlighted the critical importance his group of researchers had in their research. They showed that even though most people use them, some problems can happen with them, such as fiber-matrix debonding, matrix cracking, density variation, broken fibers, etc. It is important to know how composite materials can fail and predict them so that the composite and the component can be designed to withstand the failure and help save money.
Composite materials can fail if many different failure mechanisms are not understood or predicted. These modeling studies need a lot of time and memory because many important factors affect the behavior of such materials, such as their fiber and matrix characteristics (microscale) and the design of their products (microscale).
It is unreasonable to expect that computers will become very powerful by relying only on one type of simulation code to predict that all materials will behave as they should.
IIT Guwahati researchers have now devised a computational technique that can predict failures in composite materials with many sources of uncertainty. The technique uses machine learning tools like support vector machines and sampling tools, such as Latin hypercube, to understand the failure risks of composite materials.
Dr. Muthu explained the details of their research by saying, We performed experiments to find out the uncertainties that existed in the microscale and determined the variation that existed at the mesoscale using the numerical method called homogenization. It was possible to estimate the uncertainties on the mesoscale, which we used to drive our macro-scale simulations.
Metamodeling reduced the cost of the simulated components by about 95% versus using Monte Carlo simulation. It was very easy to predict composite properties using this method. The design and development process was made easier by incorporating metamodeling tools.
Dr. Muthu’s group studies the fracture and failure behavior of composite materials. It is also engaged in the development of biomedically useful devices. Researchers from IIT Guwahati regularly work with companies like VSSC-ISRO and TATA Steel (TSAMRC) to research solving problems in the steel industry.