Does The Future of Quality Manufacturing Lie in Predictive Analytics?
Metrology News recently sat down with David Isaacson, Vice President of Product Marketing at ETQ, where he develops market strategies and product positioning for the company’s cloud-based quality management solutions.
ETQ, part of Hexagon, is a leading provider of integrated quality management, health, safety, and environmental solutions for manufacturers. Firms around the world rely on ETQ to ensure optimal quality at scale, reduce costs and improve the velocity of data-driven decisions.
Q: Can you provide an overview of how predictive analytics is transforming quality management in manufacturing industry?
A: Predictive analytics involves using data analysis and machine learning techniques to forecast future quality issues in products. By analyzing historical data, trends patterns, issues and resolutions, manufacturers can identify potential quality problems before they occur, allowing for proactive measures to be taken. In addition to early detection and proactive resolution to manufacturing quality problems, predictive analytics can limit production variation, speed decision making, deliver prescriptive solutions to supply chain issues and boost overall quality.
Importantly, you can’t just focus on automation. You still need the human element – and so you’ll see our approach as a melding of both the artificial and human elements to solve complex quality issues.
Q: What are the main challenges manufacturing enterprises face when transitioning from reactive to predictive quality management?
A: Addressing quality issues – such as product defects or recalls, safety incidents in the plant or equipment failures – after they have occurred, can be extremely costly and resource intensive. Predictive quality, on the other hand, anticipates issues before they have a chance to wreak havoc on the enterprise. The problem with transitioning to predictive quality analytics is that it can be difficult to justify the cost when the problem has not yet occurred. Yet, according to a recent ETQ study, the cost of product recalls can be overwhelming for manufacturers. In fact, the study found that financial costs can reach up to $99.9M in the U.S. alone for each recall.
Another key aspect of transitioning from reactive to predictive analytics involves data. Predictive quality analytics requires organizing what could be massive amounts of quality data gathered from sensors, machinery, production lines and employees. Most modern manufacturing facilities have some data, but it may not be easily accessible. Older factories may have limited data, but this is actually not an issue – you can start with whatever data is available. This data trains AI algorithms to perform better. It’s important for manufacturers to first get their data house in order so that they can create the most intelligent solutions possible.
Q: What role does ETQ play in helping manufacturers adopt predictive analytics for quality control, and how do your solutions differ from traditional approaches?
A: For a long time, quality control in manufacturing has typically been carried out by human workers who have inspected products for physical defects. Yet, when an issue is discovered, trying to understand the root cause has been difficult due to the complex nature of many manufacturing processes.
While visual inspection has always been an effective way of detecting and resolving surface-level defects, it falls short when it comes to understanding more subtle defects that are harder to find, or when there are complex products or large amounts of data. And it is next to impossible to take what was discovered and apply this knowledge to predicting future failures.
Advances in AI and machine learning, however, have led to the development of new and increasingly sophisticated tools that do a much better job of identifying and preventing defects in manufacturing – including flaws that can be difficult for humans to detect. ETQ has a growing portfolio of shop floor tools that allow issues to not only be identified but also prevented before they occur. By partnering with Acerta, an innovator in AI-driven predictive analytics for manufacturers, ETQ has accelerated its ability to deliver AI and machine learning to customers of our ETQ Reliance quality management system. Customers now benefit from a predictive, closed-loop quality system that marries real-time data analysis with advanced quality workflows for the first time.
Q: How do you ensure the accuracy and reliability of predictive models in manufacturing, given the complexity and variability of production environments?
A: As machine learning models encounter new data, they adapt and provide increasingly better results. With each production cycle, the AI gains an increased understanding of the variables required to optimize quality as the outcomes from previous cycles are fed back into the model. The models then use the data to uncover new patterns and relationships between variables that are difficult to spot and may not have been uncovered before. Manufacturers use this information to continuously fine-tune their processes and identify new opportunities to enhance quality.
Q: What are some best practices for integrating predictive analytics with existing quality management systems (QMS) and processes in a manufacturing enterprise?
A: Quality managers should consider several key factors before integrating AI and machine learning into their quality management systems. The first step is to evaluate your current QMS and processes to identify existing problems and see how predictive quality analytics could be implemented to offer the most immediate improvements across the organization.
Predictive quality analytics (PQA) moves quality management to the shop floor. Therefore, it is critical for quality managers to work with their counterparts in the factory to identify where PQA could be most effective.
In addition, since data is essential for AI to work effectively, quality managers must determine which data streams are the most relevant to their quality objectives, so that they can map data streams to specific processes. The organization can then focus on data that is most likely to impact quality outcomes and structure it so that machine models can analyze it effectively. Understanding the factory infrastructure and where data will come from to inform the AI models is critical to a successful rollout.
It’s also important to collaborate with data scientists to ensure there is alignment to the business goals and AI models can be developed to meet manufacturers’ needs.
AI systems require consistent oversight to ensure they are delivering the expected results. Manufacturers can implement a feedback loop designed to regularly review performance data from AI systems and make relevant adjustments to the algorithms or processes. Managers should also invest in training teams to interpret AI-driven insights and integrate them into decision-making processes.
Q: In your opinion, how can manufacturers foster a data-driven culture to fully leverage the potential of predictive analytics for quality improvements?
A: Fostering a data-driven culture requires an enterprise-wide commitment to the power of real-time data and its role in continuous improvement. Quality professionals, manufacturing executives and business leaders should demonstrate a commitment to data-driven decision-making by using data in their own decisions and prioritizing data initiatives. Communicating the goals and approaches will be necessary to ensure that everyone is aligned.
Additionally, there should be data literacy training across the enterprise and at all levels – from the connected worker in the plant to the c-suite. Not only is data literacy a necessity, but access to data is vital, along with user-friendly dashboards and analytics tools that allow teams to visualize and analyze data easily and in teal time.
Since the deployment of predictive quality analytics can be a big undertaking, it’s important to start small, with projects that demonstrate the value of data-driven decisions with less risk.
Finally, it’s important to integrate data-driven goals not only into specific projects, but into corporate goals as well.
Q: How do you see regulatory requirements evolving in the context of predictive quality management, and what should manufacturers be aware of to remain compliant?
A: Automated predictive quality management requires the data to be accurate. Any time data is collected and leveraged, however, there is the risk of sensitive information being exposed. It’s important for enterprises to establish their own data governance policies to ensure data quality, consistency and security, making it easier for employees to trust and rely on that data. Additionally, any time AI is being used to make business decisions, it’s important to ensure that the AI algorithm is explainable, so it’s understood what data and metrics informed the decision in a very transparent way. Finally, you need to understand the regulatory environment in the jurisdictions where you do business. AI regulations are evolving rapidly and will likely change over the next several years as the impacts of AI are felt across society.
Q: Looking ahead, how do you envision the future of quality manufacturing with the further integration of AI, machine learning, and predictive analytics? What are the next big trends?
A: The use of AI in the manufacturing process for predictive decision-making is still in its infancy. As these technologies become more advanced, we will continue to see greater adoption of AI across manufacturing operations, not only predicting the likelihood of a product or part failing, but also to predict supply chain issues – such as weather delays or political issues – that can impact the delivery of shipments. It’s going to be a journey; I can’t wait to explore future potential use cases
We also will see the greater integration between different types of AI, such as machine learning, predictive analytics, generative AI and computer vision, for example. This is known as multimodal AI and will bring together all of these solutions to provide deeper insights that will ultimately result in greater product quality. Despite all of this artificial intelligence, however, we need to keep in mind that human intelligence will always be at the center of decision-making.
For more information: www.etq.com
