Fluctuating market conditions, supply chain constraints, labor shortages and a fast-paced global industry are forcing manufacturers of all sizes to reevaluate how they operate. Many have begun to embrace technology to maintain a competitive edge and address long-standing business challenges. From automation to digital technologies, industrial IoT and more, companies can leverage these innovations to finally unlock data from disparate systems, processes and people to provide the strategic insight needed to make better decisions.
And make no mistake, these organizations have a lot of data to leverage. According to a study from McKinsey, the manufacturing industry creates 1.9 petabytes—or 1,900,000 terabytes—of data every year. The catch is that they need a better way to capture and analyze the data and turn it into usable information—and they need to do it quickly. As a result, many are turning to artificial intelligence (AI) to utilize their data to spot opportunities to enhance their operations.
Why AI is perfectly suited for data analysis
From increasing manufacturing yields and uptime, to accurately forecasting demand and remotely monitoring machines, even controlling assets and improving product quality, AI can be leveraged to make significant improvements to overall efficiency and productivity metrics.
It’s not magic, but rather a set of complex algorithms that are used to analyze massive amounts of data, correlate or learn patterns in a broad range of variables, and apply that knowledge to current conditions to help predict the future state. It’s not that humans can’t perform these tasks, but rather Al can do it much faster and handle a much larger amount of data, with greater precision, to improve business outcomes.
For example, in any manufacturing environment, there are traditionally several different working groups and machines all gathering their own data. Each piece of equipment’s information may vary in quality, formatting and timing, which can create barriers and make it difficult to analyze and glean any meaningful insights from the data.
AI can handle the data tsunami and enable companies to combine operational information quickly and accurately to predict outcomes based on alternative scenarios, allowing manufacturers to make agile, well-informed decisions and drive bottom line impact. This ability to predict issues before they become a problem is where AI especially shines, and that can make a big difference on product yield.
By identifying the root cause of product quality issues, AI can help reduce product defects and scrap rates and increase manufacturing yields. Armed with detailed information and analysis, manufacturers can address quality control issues before they directly impact the company’s bottom line. Let’s take a look at one such example.
Using AI to improve engine quality
A global engine manufacturer produces large diesel engines that are used in generator sets, naval and marine applications, and military vehicles. Once assembled, each engine is subjected to rigorous testing. During this testing process, subtle indications of a pending problem often went unnoticed by even the most experienced operators, leading to a catastrophic failure during the test or after the engine was put into service. These failures caused extensive damage, delayed shipments and created backlogs in the testing area and upstream production, costing the company millions of dollars annually and negatively impacting on-time delivery.
The problem wasn’t lack of data, but in how it was used. In fact, the plant had been collecting process data for years, but used it solely for follow-up after a failure already occurred. By looking at the data in such a reactive way, the team was unable to understand why these failures were occurring or proactively address them. Eventually, these issues were viewed as a cost of doing business—until the company looked into using AI on existing data to predict critical asset failure before it happened.
The manufacturer started out with a pilot program that established the data foundation needed for AI to be impactful. Given the need to use historical data, the company first went through a data cleansing and analysis process—20 billion data points from 100 engines were reduced to 6 billion of the most impactful data points, all within 48 hours with the help of AI. Next, multiple model sets were joined by time and model number, the data was visualized and any data gaps were identified. Based on the gap analysis, an adjustment was made to pull certain data more frequently to improve modeling and, by utilizing an AI platform, this entire analysis was completed in a low-risk environment with no impact on current production.
From that data the manufacturer was able to establish baselines, identify trends and anomalies, and develop a plan to operationalize the information. In a matter of weeks, they produced a report which identified a group of at-risk engines by serial number. Based on that information, the manufacturer suspected that these engines had a higher probability of experiencing a problem during quality control testing or in the field. By correlating the test data to actual product failures, the report accurately identified more than 80% of the engine problems over a period of several years.
It’s important to note that this project was an iterative process, as the AI model was constantly learning. Within approximately 45 days, the model was able to predict failures 30 minutes in advance with zero false positives.
Driving business outcomes
During the official rollout, the Al solution was connected to the live data generated by the testing control systems and human machine interfaces (HMI). There was no impact on normal operations. In fact, the model was integrated with the company’s standard test software and operators were not even aware of the implementation. All they needed to know was that their HMI screens would now advise them of any potential imminent problems and how to react. Within the first 90 days, the AI application detected 20 real-time events, prevented more than $4.5 million in engine damage and delivered a 10x project ROI (return on investment).
As this use case illustrates, utilizing artificial intelligence can provide manufacturers with a way to proactively reduce quality defects, save money and improve their on-time delivery, all with minimal disruption to operations. By starting with a solid data foundation and working with experienced partners, AI can deliver the insights needed to drive business outcomes and help manufacturers compete in today’s fast-moving business landscape.
But AI isn’t necessarily a one-size-fits-all solution. Depending on your needs, use case and situation, different solutions may be more appropriate. That’s why it’s important to have a trusted partner by your side. When it comes to AI, they can assess where you are on your digital transformation journey, understand your goals or challenges, and identify solutions from top suppliers that are best suited to your environment.
Scott Dowell is senior vice president and general manager, U.S. industrial and CIG, Wesco.