Deep Learning In Product Modeling And Casting Is Creating Circular Product Innovation
During the coronavirus pandemic, high school students designed and produced masks using 3-D printing. Others made parts for respiratory machines used by hospitals. These amateur product designers have inspired us. In the era of reverse engineering, they also make us ask if we are pushing the innovation edge on our own designs.
From prototyping to production, deep learning is improving the ease and speed of innovation. Our limitations are no longer our imagination once deep learning gets a hold of our ideas. Software can simulate any design with any material composition in virtual-reality.
With endless possibilities in a virtual space, researchers are applying deep learning from 3D product modeling through to molding and casting, and inspection.
Smarter 3D CAD Modeling Software
When machine learning and big data team up, CAD software moves beyond design. Designers are simulating material compositions and behaviors, and more complex geometries and features. Now that materials have their own genome project, the Materials Genome Initiative, a renaissance in material discovery is taking place.
This smart learning duo is also creating powerful benchmarking models. The performance of different material compositions and geometries are analyzed across millions of benchmarks. In a circular production model, material composition analysis is then fed back to metal alloy or polyurethane casting material suppliers.
More Realistic Solidification Software
At the molding and casting stage, solidification software is simulating the behavior of popular alloy metals and polyurethane casting materials, as well as novel material compositions. Understanding the shrinking properties of these materials optimizes the casting and molding processes.
Polyurethanes, for example, have been a hotbed of machine learning research owing to their attractive physical properties (e.g, high rigidity, flexibility, tensile strength, tear resistance). Researchers at Carnegie Mellon university have found that machine learning can improve the prediction of the behavior of polyurethane with changes in its chemistry. This predictive technology is currently being commercialized.
Deeper learning allows for more accurate correlations between simulations and real life casting results. This knowledge is expanding opportunities in, for example, bio-based materials—an important frontier for polyurethanes commonly used in coatings, foams, and solids.
Deeper Quality Inspections
At the inspection stage, parts can be analyzed against more critical rejection parameters. Deep learning helps inspection technologies such as ultrasonic and laser scanning learn more about problems and continuously improve defect detection.
These deep learning virtual environments are being designed to work interactively with the real time manufacturing environment. In this way, smart machines are lowering the risks and costs of product innovation.