The strength, flexibility, and durability of plastics have made it ideal for countless applications from energy-efficient electronics and lighter, more fuel-efficient automobiles to single-use medical devices that improve patient safety. In all cases, quality control is crucial to ensure that the plastic part performs as intended. In 2020, this challenge has been compounded by the effects of the pandemic. Plastics processors have had to deal with fluctuations in consumer demand, workforce scarcity, supply-chain volatility, and new workplace safety mandates. This perfect storm has put an unprecedented burden on manufacturers trying to maintain their machine utilization rates and production line efficiency while producing quality products.
Fluctuating consumer demand
When COVID-19 hit, many injection molders saw decreased demand for large volume orders of some parts because of shrinking consumer activity. Meanwhile, demand for single-use plastic items surged. These products had been in the crosshairs of environmental groups for years, but consumers staring down a pandemic are much more focused on the hygiene and safety offered by these single-use products. At the same time, manufacturers of plastic products saw an increase in requests for prototypes and small batches of medical-grade equipment and supplies for essential workers.
Ramping production up and down for various types of products means adjusting quality control requirements and processes. For large batches and volume runs, visual inspection is typically done on a small percentage of the products by a human operator. Sometimes, individual part inspection is even left to the customers to perform before they incorporate the part in an assembly. Alternatively, small batches and new products require quick, increased quality-inspection protocols, i.e. the ability to inspect each part and track that inspection process, sometimes with minimal infrastructure or expertise on site.
Automating quality inspection on the production line
Injection molders and other plastics processors facing a higher product mix than before need to rapidly adjust their quality inspection protocols. The human operator who used to perform the inspections may not be available, and medical device parts, in particular, typically require stringent quality protocols. Given the new normal of unpredictable consumer demand, the key to success is flexibility, and an added “pair of eyes.”
Consider, for example, the global automotive plastics market, valued at more than $35 billion last year. Many injection molders are running dozens of production lines, each with their its own machine vision set up for quality control. They are being challenged to efficiently handle line changeovers and modifications. And many surface inspections of the plastic parts are subjective in nature, an assessment that is very difficult to accomplish using machine vision alone.
Artificial intelligence (AI) allows manufacturers to quantify defects that were previously qualified or subjective at the discretion of an operator. Surface inspections might have had guidelines, but they were identified by sight and not standardized because of the complexity involved in customizing the programming of a machine vision system. AI and deep learning facilitate this standardization and are fueling automated inspection that is faster, cheaper, and more accurate than ever before. Plastics manufacturers can quickly integrate Vision AI software with common automation protocols, such as GigE vision standard cameras and Modbus TCP. By connecting to manufacturers’ existing hardware, Vision AI provides tremendous flexibility to handle production line volume changes, workforce constraints, higher inspection rates, and more stringent quality requirements.
As the world re-discovers the unique versatility of plastics to meet new challenges, AI and machine learning can help manufacturers speed up inspection processes, decrease errors and false positives, and perform quality control inspections when humans are unavailable.
About the author
Max Versace is CEO and co-founder of Neurala, which has developed award-winning, patented AI technology based on advanced research work conducted for NASA, DARPA, and the Air Force Research Labs.