Optimizing Quality Control with Visual Inspection for Manufacturing

Editor's Note: This post was originally published in May 2023 and updated with new information in September 2023. 

Key takeaways

• Integrating AI and AR technology into inspection processes leads to faster and more accurate quality control, reducing human error and improving overall production efficiency.

• With Vuforia’s Step Check, companies can accelerate and streamline end-of-line inspection with augmented reality (AR) by providing an integrated mobile solution that combines automated assessments, real-time feedback, and automated reporting, leading to better decision-making and increased customer satisfaction.

• Vuforia’s Step Check offers a comprehensive, competitive advantage for today's manufacturing inspection needs, benefiting both front-line and quality control workers.

Quality control is one of the most important steps in the manufacturing process. It's the last chance to ensure that a product meets customer needs. No matter how good a production line is, a failure in quality control can lead to product failure and damage confidence in the supplier.

Visual inspection, a preferred quality control method in many industries, is difficult to automate. Most manufacturers still prefer to trust human judgment over automated solutions. But human traits like stress and fatigue can take their toll on even the best and most experienced workers.

Now, with Vuforia's Step Check, you can help these workers be more efficient while feeling more engaged, thanks to advances in AR and AI. And they're asking for it: 77% of frontline workers surveyed by Forrester suggested they don't have access to the technology they need to be productive.

What is visual inspection in manufacturing?

Visual inspection is the process of verifying, traditionally with the unaided eye, whether a product, component, or system is in a fit state to be shipped.

In the case of a finished product, visual inspection ensures that the product can be used safely and effectively by a customer. In the case of component manufacturing, it can help identify if a component is of the quality required to be used within a larger system.

The Evolution of Visual Inspection

Technologically basic forms of visual inspection might be so simple as an expert confirming that a piece is satisfactory without the aid of reference models or images. This is an intuitive process, but one that puts a great deal of pressure on a small number of inspection engineers with a high level of expertise.

Visual inspection can also be conducted by comparing a unit to a reference model – a unit of the same type that is known to be in perfect working order. In the case of large units, this can be unmanageable as it requires inspection engineers to navigate in and around multiple large units. In the case of smaller units, it still burdens the inspection engineer with holding onto a reference model.

Reference models have largely been replaced by reference images. The benefit of reference images is that they are lighter than small units and easier to work with than a large reference model. The downside is that, because reference images are two-dimensional, any set of worthwhile reference material will include numerous images from multiple different angles. This can great a lengthy document that it is difficult for inspection engineers to navigate efficiently.

Until very recently, the peak of visual inspection evolution was reference to a CAD model or even a virtual reproduction of a physical unit. Visual inspection with reference to such a model combines all of the benefits of reference models and reference images with few of the drawbacks. Inspection engineers can easily rotate the virtual model, which can be viewed on a handheld smart-device no matter what the size of the actual unit may be. This means that one virtual model takes the place of pages of reference images.

The current heights of visual inspection technology incorporate the emerging fields of augmented reality and artificial intelligence. AR overlays a virtual model over the inspection engineer’s view of their physical environment, enabling more efficient comparison while maintaining better situational awareness. AI can help inspection engineers access large quantities of actionable data, including helping them identify and report issues during the visual inspection process.

When is visual inspection needed?

Visual inspection can be performed any number of times, by any number of people, at various stages of a product's lifecycle. Multiply that by the thousands or millions of parts and products a manufacturer might put out each year. Consider an airplane as an example. Each component and system may be inspected before assembly. The entire airplane is inspected again before delivery, and then again before each flight.

In manufacturing, the process is carried out before a unit is rejected due to a flaw, sent on to the next stage of assembly, or shipped to a customer.

The primary purpose of visual inspection is to check the quality of the units themselves. It also serves the secondary purpose of identifying potential problems in the production process. A well-performed visual inspection will not only identify issues with the units but, ideally, improves the overall manufacturing process by identifying the human error or machine malfunction that caused the issue.

Visual Inspection Methods in Quality Control

Visual inspection can take a number of forms, often depending on the type of product and the volume of products that need to be inspected.

Random Sampling

Random sampling involves only visually inspecting occasional units rather than every unit that comes off the line. This approach can save time for simple units but can be a risky choice for units that are more complex or that might prove dangerous to an end user were they to ship in less-than-perfect order.

One issue that can occur with random sampling is when visual inspection discovers a fault caused by machine wear-and-tear or a procedural error. In these events, it is likely that the unit that was detected is not the first unit that was affected. In this case, the production line may need to be stopped while the issue is addressed and other units are inspected – or worse, recalled.

Full Manual Sampling

Full manual sampling is the detailed inspection of every piece that comes off the line. This is time and resource intensive and can be taxing for the inspection engineers charged with carrying it out. However, it prevents some of the shortcomings of random sampling and can be necessary for complex or sensitive products, or for safety reasons.

Remote Visual Inspection

Remote visual inspection involves using a remotely controlled device to visually inspect units, or even buildings, from a distance using cameras and other sensors. This method of visual inspection is typically reserved for hazardous situations and environments.

There are some benefits, including a more comfortable work environment for the inspection engineer as well as increasingly powerful sensors and more capable computer vision resources, but it remains an awkward method that involves additional expenses including hardware like drones.

Automated Visual Inspection

Automated visual inspection means turning over the entire visual inspection process to machines with advanced cameras and sensor capabilities. This method is increasingly promising as experienced workers are increasingly hard to come by, as it could replace human inspection engineers and potentially save cost and consideration for companies. However, many companies are wary of giving complete decision-making power to computers.

Challenges with existing visual inspection approaches

Despite significant advances in areas such as machine learning, automation technology has yet to fully augment human expertise in visual inspection. This is a missed opportunity, as leveraging automation can significantly reduce the cognitive load on frontline workers and increase their efficiency during the inspection process.

Human attention is a valuable resource that companies need to make the most of

High output and high quality are goals of all manufacturers. While optimizing and automating manufacturing and assembly processes can speed up production times, it can also mean faster cycles for human workers – including inspectors. This can lead to strain and fatigue, especially over long shifts.

Inspection worker stress and fatigue not only take a toll on the individual, but can also lead to errors in this critical final step of the production process. This can lead to worker burnout as well as lost revenue, shipping delays, and reduced customer satisfaction. This is especially true when multiple items must be inspected on each unit.

Companies can and do try to mitigate these issues by providing materials like reference images. However well-intentioned, such steps often require a high level of abstraction for the inspector and can slow down the process – whether or not they contribute to more accurate inspections.

The Limitations of Manual Inspection

While many are rightfully wary of replacing human judgment, the unaided eye can and does make mistakes. This can be particularly true when an inspection worker repeats an intricate task or series of tasks seemingly countless times per shift, typically while juggling reference images and manually filling out reports.

What's the state of artificial intelligence and AR in manufacturing today?

Artificial Intelligence (AI) and Augmented Reality (AR) have their own parallel history in manufacturing. AI might be used to optimize the manufacturing floor or ensure machines don't overheat. Meanwhile, AR could help a worker get assistance from a remote expert or allow them to follow step by step 3D digital instructions, keeping their hands free.

While these technologies are powerful tools on their own, using AR and AI together is even more impactful. Combined, these two technologies make computers more cognitive, supporting human perception and enhancing attention.

The power of combining AI with AR

Envision AR (computer vision) as the computing system's visual interface, akin to eyes, while AI serves as its cognitive processing hub, comparable to a brain.

Augmented reality displays virtual information over a view of the physical environment. In most situations, that is information that is either entered by other human users, or provided by connected smart devices that can provide their data to the program.

Artificial intelligence allows a device to generate or provide its own information about the environment. For example, it may recognize and contextualize items in a user’s surroundings.

Augmented reality devices equipped with AI-powered software are already utilized on many manufacturing floors. These devices can count inventory items for a worker, improve their situational awareness, or direct them across a floor or campus to do their jobs more efficiently and safely. AI can also increasingly automate the filing of reports – one of the necessary clerical steps that can take an inspection worker's mind and eyes off the job.

The greatest strength of AI-enhanced AR is the human component. There's a reason that humans still make the final decision in so many visual inspection situations. AI-enhanced visual inspection solutions finally give them the tools they need to do their jobs more accurately and at the speed needed to keep up with increasingly automated production lines.

What are the benefits of using AI-enhanced AR inspection for quality control?

Strictly in terms of output, the use of AI-enhanced AR quality assurance measures results in faster cycle times, increased throughput, and a reduction in inspection errors and omissions.

Accelerated Inspection Cycle Times

AI-enhanced AR inspection helps to accelerate inspection time by helping human operators – including those with little or no experience with a product – navigate around a unit and check key points, even helping them check for and address common points of failure. These tools are also increasingly capable of helping inspection engineers to generate reports, saving them additional time and stress.

Decreased Ramp Up Times

Ramp up times, the time that it takes for a new hire to “learn the ropes,” can be drastically reduced by AI-enhanced AR inspection solutions. These tools don’t just have the information that they are programmed with, they also learn as they are used. That means that the program is an ever-deepening trove of company experience.

This is also where the impact of industrial AR comes in. All of that information has been available to companies for ages, but it has not been available in a format that could be efficiently conveyed. Through AR, even the freshest hire can easily access that information.

Reduced Test Failures

AI-enhanced AR helps to reduce test failures by helping to identify common issues, and even troubleshoot them. In this way, some issues that might have resulted in a test failure in previous methods can be addressed on the floor by the inspection engineer.

Reduced Tests Omitted

AI-enhanced AR solutions can recognize the physical piece being viewed to automatically summon the correct process. One of the ways that it can identify the physical unit is by an identification number. This can also help the program to automatically catalog that specific piece when generating documentation.

This is a handy tool that helps inspection engineers efficiently keep accurate records. However, it also helps to ensure that all units required for testing have in fact been tested.

Increased Throughput

Whether working with experienced inspection engineers with intimate knowledge of a product, or new hires fresh on the floor, AI-enhanced AR solutions increase the rate at which inspection engineers can effectively conduct inspections. While increasing throughput, these solutions also decrease the stress on inspection engineers by guiding them through the process in an engaging way, removing the difficulty of dealing with physical reference materials, and automating the documentation process.

Vuforia’s Step Check is a best-in-class, fully customizable, AI-enhanced AR inspection tool

Vuforia’s Step Check, an AI-enhanced AR visual inspection solution, empowers frontline workers by streamlining their tasks. By seamlessly integrating AR and AI technologies, this solution improves accuracy, efficiency, and overall effectiveness.

Step Check uses an AI training process that leverages image data and operator feedback. Integrated with AR, it provides users with enhanced visual support for quality control tasks. With the ability to track and evaluate components beyond the limitations of traditional CAD-based systems, Step Check is a holistic visual inspection solution.

Step Check is part of Vuforia Expert Capture, PTC's SaaS AR solution providing a collaborative environment to create, share and view AR inspections and work instructions. Its Vuforia computer vision technology offers the best accuracy on the market and multiple ways to recognize objects, images and environments.

Vuforia Expert Capture also streamlines the documentation of Step Check inspections, simplifying reporting and increasing efficiency while providing verifiable proof of quality.