Using AI to catch visual defects in plastic injection molding

Scrap rates in injection molding operations are often reported at 2-3%, but industry surveys suggest the true figure is closer to 5-10% when you include downstream defects and customer returns. At $1-5 per pound for engineering resins, those losses add up fast. 

Using AI-based quality control to catch visual defects, one of the most common, can improve customer satisfaction and profit margins. In this post, we cover the unique defects for plastic injection molding, and how AI solutions can be integrated to solve them. 

Knowing the real cost of visual defects and manual inspection

Every defective part costs more than just the material. Industry estimates suggest that it takes seven good parts to recover the loss from one rejected one, and production costs are four times as much to break even. Even more costly than direct scrap is the cost of an unidentified root cause. Today, the best solution is using AI-powered automated quality control paired with human operators. 

Research suggests that manual visual inspection accuracy in manufacturing environments is around 80%, meaning that one in five defects get missed. This increases in sampling-based inspection, where sporadic defects can be missed for longer, creating a lag time before patterns and root causes are found.

For mechanical tasks, such as scanning for defects with high throughput, machine vision systems perform much more efficiently. This enables every item passing the cameras to be scanned and focuses human operators on the edge cases to inspect. 

With modern-day solutions, AI models and intuitive interfaces also makes it easier for staff to label new defects, add samples for existing ones, and update descriptions that can help team members look into root causes. For example, AI quality control can set acceptability thresholds for defects, add a label, such as “discolouration” as well as an actionable description: “Could be due to contamination.”

Catching visual defects from short shots and flashes in plastic injection molding

Incomplete fills (short shots) and excess material at parting lines (flashes) are highly visible defects that are eventually caught during quality inspection. The problem with current processes is how costly they are before they are detected and corrected before more scrap is produced. Automated solutions like Enao Vision are installed on the production line to scan without interference. The quality control solution detects short shots and flash in real time, alerting operators within seconds of the first defective part rather than after a bin of rejects has accumulated.

For plastic injection molding, incomplete features, thin walls, or missing sections are common issues. The root causes such as insufficient injection pressure, low melt temperature, or inadequate venting can also be addressed. Enao Vision’s system logs every defect by type and location, giving you the data to calculate true quality costs and identify which defects and root causes are actually hitting your bottom line.

To calculate whether an automated quality control system is worth the investment, it is important to consider AI solutions as an operational expenditure, rather than depreciating capital expenditure historically used for industrial machinery. Track your true scrap rate—including parts rejected downstream and customer returns. Understanding the real cost and using that as a basis for new investments in optimisation will ensure your competitive advantage.

Using lighting and machine vision systems to catch surface defects 

Surface defects can be trickier for visual inspection, but the right setup can create robust monitoring. With the right lighting and camera angles for scanning, the surface defects can become visible for machine vision systems, in contrast to manual inspection where accuracy drops significantly when operators need to constantly adjust lighting and angles.

Instead of replacing all manual inspection with automated solutions, we recommend testing a solution and only expanding once it’s proven to work for your runs. You can get details on our post on closing the quality gap for manual inspection to understand where you might place your quality control.

Common surface defects for plastic injection molding include sink marks, flow lines, and splay. If you are trying an AI quality control solution for the first time, increase your chances of success by reading our blog post on setup. If you’re unsure, ask your provider to make recommendations for lighting, mounts for their cameras, and support for labelling the defects so that you can integrate root causes, whether it is pack pressure, variations in cooling rate or moisture and trapped gas. This will help your team quickly see the value of having an AI partner in quality checks.

How AI quality control can catch contamination and color variation

Contamination is notoriously difficult to prevent, and color variation is just as difficult to detect. Using AI quality control solutions helps to reduce cost and risk by replacing sampling-based inspection with 100% coverage, but we also caution manufacturers against immediately replacing manual checks.

Choose your AI provider based on the ones that have the right features, usability, and camera capabilities to catch your most common defects. To start, you can speak to these leading machine vision system solutions to understand how they differ, whether it’s scan speed, AI model, or camera capabilities to catch the specks and foreign materials or colour variations.  

Getting Started

Adding AI solutions to your production line should follow the principles of lean production. By understanding where your biggest costs are, you can decide where AI quality control can be added to make measurable improvements, whether it’s optimising costs, throughput, or processes.

Unlike in the past, AI quality control solutions like Enao Vision can be tested before you commit to multi-year contracts. We recommend speaking to different providers and understanding how long it takes before you can be live-testing a solution on your shopfloor. To ensure success, we suggest focusing on one production run and only expanding it once your staff feel comfortable with it. To ensure a successful setup for your automated quality control, make sure you cover critical steps such as ensuring stable WiFi for the AI model to process properly, getting the right lighting and mounts, and requesting support for setup where needed. Enao Vision makes it easy by using iPhone cameras, which are not costly to replace, and sending a starter kit with 5G connectivity, lights, and mounts included.

Visual defects in injection molding are inevitable, but shipping them doesn't have to be. The technology exists today to inspect every part, classify defects by type, and give your team the data they need to fix root causes—not just sort rejects.

Try Enao Vision for the first month free to see how it works on injection molded parts.