AI has become more prevalent in every industry, including manufacturing. More plants are adopting AI technologies to improve product quality, reduce downtime and recalls, support maintenance efforts, and more. Still, there is so much to the discourse on AI that it can be overwhelming for even the experts to stay updated as new advancements and breakthroughs happen seemingly daily.
To help our customers gather information on these advancements, we have broken down three considerations about AI: vision, predictive maintenance, and training. By learning how these systems work and their capabilities, we hope you will better understand whether AI makes the most sense for your packaging operations now or later.
One of the most exciting areas where AI can support production goals is vision systems and detection. AI could detect subtle things, like minor scratches, coloration anomalies, and packaging defects. With traditional tools and visions, AI can process things more efficiently and clearly.
That’s because of how current AI and deep learning technologies work, which we touched on in a past Short Talk on vision systems. These systems combine the flexibility of human visual inspection with the speed and robustness of a computerized system. They use neural networks that mimic human intelligence, which enables them to distinguish defects and anomalies.
As AI constantly learns and evolves, programmers don’t need to add specifics when training the system on good and bad examples of products to monitor for defects. The beauty of AI is that it will figure it out if you provide consistent inputs. However, consistency is crucial, as inconsistent examples could confuse the system and cause it to hallucinate.
Since there’s no one set of criteria for human inspection, quality issues can be very subjective, which could also increase programming complexity. Also, remember that what you’re inspecting for will differ from country to country. If you invest in AI software for your vision system, you must have skilled programmers with experience in AI and data science who can help manage and program the AI correctly to avoid inconsistencies and hallucinations.
Machine learning has to be conducted by an engineer who can train on what’s good and bad and even retrain when a training set needs to be changed, like when introducing a new recipe or SKU to your line.
Sometimes, when a system alerts the operator of a defect, it isn’t done to reject the product but to identify trends to report back upstream so they can make the necessary adjustments.
Vision systems with AI could help floor operators catch upstream issues earlier and avoid wasted production and downtime that impact EOL operations. This is one of AI's most important potential benefits: it could improve end-to-end decision-making. By offering real-time alerts and troubleshooting when subtle defects and other quality issues occur, AI can help operators and management respond faster and more effectively.
As AI increases its capability to detect subtle defects, it should help improve product quality and waste. It could also avoid packaging issues that sometimes lead to recalls by ensuring the packaging is closed and sealed, contains the correct amount of products, and includes accurate labeling and bar codes.
Advancements in technology have helped support and improve vision capabilities in the past, but each advancement has its challenges. AI is the next step in helping operators and management increase visibility, efficiency, and throughput, but it isn’t a fix-all. Consider some past technological advancements in vision:
BluePrint Automation is developing more robust vision systems with AI capabilities that minimize the importance of camera position and reduce the need for a vision box or special lighting. They combine the best of these vision technologies with AI's learning capabilities. In the very near future, vision systems should be able to see whatever your eyes can see.
Conversely, anything difficult for an operator to spot would also make it more difficult for a vision system combined with AI to see. However, the next possible step could be to use 3D vision with two color cameras to pinpoint the exact position of a product in relation to the camera, much like your eyes give you depth perception. AI can make this much more viable as the technology advances.
Additionally, as cameras become less expensive yet more precise with AI, line managers can place them at critical points upstream and downstream to monitor what is happening, for instance, above each delta cell. We also see cameras being used more as sensors in the future to not only see if a product or case is present but also to monitor its orientation, position, and current state.
As we consider the potential to increase visibility and monitor quality with AI, we can also envision AI vision systems used to grade products and the filled cases. For example, a combined system could help identify if each product is placed neatly in the case or if missing products or products are sticking out of the packaging. These systems could also make bin-picking and shallow pile-picking much simpler.
But, as with any advanced technology, education, and partnerships are essential to ensure you invest in the correct system and technology, whether you invest in a fully automated system combined with AI or start slow to determine your needs and comfort with automation. We advise you always to choose the technology you’re comfortable with, as you can upgrade later.
Partnering with a reputable partner who can help you identify the best solution to meet your needs is one way to ensure you don’t invest in an AI-based system that’s unnecessary or too complex for your plant and people. We will discuss several other considerations for AI in packaging first but will later explain why it’s beneficial to partner with an OEM equipment supplier to help you consider AI without all the noise and hype that has saturated the conversation.
One challenge with AI is that these systems generate a lot of data, and companies must have a strategy for how it is gathered, collated, presented, stored, managed, and presented to users. This is especially true for Industry 4.0 and predictive maintenance, where it’s pretty straightforward to gather a vast amount of data, but it is much more difficult to assign any tangible meaning to the collected data without a robust data management system and strategy.
AI can be used to separate the wheat from the chaff. Internally, you must be able to distill the vast amounts of data coming in into something useful, recognize patterns in the data, and turn this into usable information to aid decision-making. AI and data science can make this much more accessible.
A starting point could be determining what kinds of data your people need to complete their tasks efficiently.
A programmer or data scientist could then apply a methodology for collecting and transforming the raw data from AI into actionable insights for optimizing throughput, enhancing quality, and ensuring timely maintenance of your secondary packaging equipment. Of course, it’s essential to have skilled programmers with experience in data management and AI systems, as these can become exponentially complex and overwhelming quickly.
Various sensors embedded in everything from the robotic arm to a simple bearing can monitor essential parameters, such as temperature, vibration, pressure, and operational speed, to identify anomalies and inconsistencies as they occur. AI systems can identify trends using historical data and other sources associated with adverse events or failures and alert operators when repairing or replacing a component may be necessary.
In the future, AI systems could assist with predictive maintenance, prioritizing urgent repairs and optimizing maintenance schedules for reduced downtime. Resource allocation could improve as AI helps predict upcoming needs more accurately. It could also be an asset for root cause analysis, using past data leading up to the incident to offer possible corrective actions or insights for a maintenance manager to make an informed decision.
It’s essential to continuously educate and train your team on operating these AI systems correctly and effectively. First, you should involve the floor operators as early as possible to identify what the AI system needs to do. They’re the ones who see the products all day and know their subtle defects. Management must support their operators and ensure they have the training to champion these systems.
AI requires a skilled, highly technical team to operate, manage, and troubleshoot it, and your team must buy into its adoption completely. If one critical operator or floor person doesn’t see the added value and is opposed to AI, this could lead to system failure due to deliberate or accidental sabotage.
AI can provide significant benefits if the correct system is chosen and implemented. Ample training and support from management can ensure your most essential people understand its value, and, more importantly, it helps them do their jobs safer and better by aligning needs with capabilities.
As we’ve laid out some considerations for AI, we hinted at the need to find expertise to help you identify your production needs and educate you on the emerging technologies and concepts surrounding AI. A lot of wishful thinking and false claims have fueled the AI narrative. We fear that some companies are buying into the hype of AI without having the need, capability, budget, or desire to manage these state-of-the-art systems.
This is where a partnership can and should come into play to provide guidance and help you. Producers must become comfortable with the technology and understand how it works, which requires expert guidance as it is complex to operate and maintain.
While they will undoubtedly become more affordable as the technology matures and becomes more prevalent, current AI systems are costly. Remember, you’re only as modern as your oldest piece of equipment. You don’t have to do all upgrades at once, but smartly choosing which ones will help you reach where you want to be and where the industry will be. This is where the expertise from BPA comes in.
We can help you find the best application for AI-based technologies in secondary packaging or identify the best solution to impact throughput now.
AI is not a magic quick fix for every problem, and there’s no one-size-fits-all approach possible. Yet, it’s impossible to deny the potential AI can bring to vision systems and detection, predictive maintenance, scheduling, resource allocation, and so much more in packaging.
By partnering with experts who work with these technologies and support customers daily, you can ensure that AI is used effectively in your operations.
Keep an open mind, stay flexible, avoid trendy buzzwords, and start with low-hanging fruit. AI requires a different way of working and a different mindset. Let us know if we can help!