P&G enlists IoT, predictive analytics to perfect Pampers diapers
HomeHome > Blog > P&G enlists IoT, predictive analytics to perfect Pampers diapers

P&G enlists IoT, predictive analytics to perfect Pampers diapers

Aug 24, 2023

The consumer packaged goods giant has turned to Microsoft IoT and edge analytics to capture real-time data about manufacturing processes to anticipate impending failures before they damage diapers.

If there are everyday items you want to be failsafe, diapers are surely among them. That’s why The Procter & Gamble Co. goes to great lengths to ensure the fidelity of its Pampers products.

But when tossing away thousands of diapers damaged during the manufacturing process becomes an everyday occurrence, something has to be done to provide relief for the bottom line. That’s when P&G decided to put data to work to improve its diaper-making business.

“We’re always looking at what are the biggest sources of our losses are and where things could run better,” says Jeff Krietemeyer, IT senior director of Global Baby Care Services & Solutions at Procter & Gamble, whose team began planning a solution in late 2021 to fix the costliest manufacturing glitches, particularly those that impacted diapers.

Diapers are made of fluff pulp, plastics, absorbent granules, and elastics, and different processes — such as streaming hot glue and heat binding — are used during various aspects of the highly-mechanized manufacturing process.

But things go awry and when they do, Procter & Gamble now employs its Hot Melt Optimization platform to catch snags and get the process back on track. The project, which earned Procter & Gamble a 2023 CIO 100 Award for IT innovation and leadership, has had a profoundly material impact on the manufacturing floor.

Hot Melt Optimization employs a proprietary data collection method using proprietary sensors on the assembly line, which, when combined with Microsoft’s predictive analytics and Azure cloud for manufacturing, enables P&G to produce perfect diapers by reducing loss due to damage during the manufacturing process.

Since deploying the solution in 11 plants, P&G estimates it has eliminated 70% of the flawed diapers that have to be scrapped. Executives would not disclose the exact amount saved every week but it’s well into the seven-figure range.

During the diaper-making process, hot glue stream is released from an automated solenoid valve in a highly precise manner to ensure the layers of the diaper congeal properly.

“The diapers are flying through the manufacturing line at high speeds during the assembly process, so you need this super precise application of glues that are hot melt adhesive,” Krietemeyer says, adding that the glues are safe for human skin.

If the temperature and pressure of the glue stream is inaccurate, however, or the glue gets clogged in the valve and is not corrected in time, the resulting diapers must be scrapped.

To address these issues, Procter & Gamble worked closely with Microsoft to deploy Microsoft’s IoT and Edge analytics platform, its Azure cloud for manufacturing, and its IoT sensors, edge analytics, and machine learning models.

The resulting platform was pilot tested for nine months at one P&G plant before being rolled out half of P&G’s Pampers manufacturing plants across the US.

En route to one of those plants in Missouri, Kietermeyer explained to CIO.com that the combination IoT and edge platform, sensors, and edge analytics rules engine have been successfully employed to address pressure and temperature anomalies and the valve hardware issues that can occur in the diaper-making process.

“The project team explored several algorithms, including training neural network models, and found that the Microsoft AI Rules Engine achieved the best results,” Kietermeyer added.

On the assembly line, P&G employs Rockwell programmable logic industrial controllers and other sensors to closely monitor and record glue stream temperature and pressure data. The data is fed into analytics platforms and in-house developed code to identify errors or anomalies that must be corrected in real-time — while not taking the manufacturing offline. This ensures that the output of each facility exceeds what was achieved before Hot Melt Optimization was launched.

The data streaming measurement was configured using an industrial control database dubbed Influx Historian. Data is streamed to Microsoft’s Edge analytics model using a broadcasting system and Grafana pre-visualization. The sensor and software can detect whether something is going awry and, in several hours, makes the fix automatically.

“These industrial microcontrollers run at super high speed and are very finicky,” Kietermeyer says. “Getting them to run very precisely to manufacture the perfect diaper every time takes a lot of effort and inspection and there’s not a highly expert person available 24 by 7 to watch the line. Even if there were, they would need break time. So that’s where the project idea came from.”

Here, predictive analytics are key. P&G’s manufacturing specs are continually tested against the incoming data in a rules-based manner via Microsoft’s edge analytics engine, which helps spot necessary corrections several hours in advance. “If the data is trending in a bad way, you can see in six to eight hours if it would fail [in manufacturing],” Kietermeyer says. “We can predict it in time to stop, and do the maintenance before it actually goes outside the spec.”

Procter & Gamble, which as one of the world’s largest consumer product companies generates more than $75 billion annually, emphasizes how important this use of data collection and predictive analytics has been to the company’s bottom line.

“Business demand for baby care products is extremely high, and the production lines needed to create these products are asset-intensive,” the company reports. “P&G’s ability to keep the lines running has a significant business impact, including supporting our ability to maintain and increase production capacity, reduce unplanned downtime, and reduce the amount of scrap generated during production.”

Hot Melt Optimization comes on the heels of broader commitments P&G has undertaken to its evolve its manufacturing business using digital technologies and AI.

One analyst who follows the use of digital technologies in manufacturing notes that it is critical for vendors to know their processes inside and out to benefit from advanced manufacturing technology.

“Digital transformation uses advanced sensing, data analytics, and the latest in artificial intelligence to gather insight into production processes,” says Carlos Gonzalez, research manager of IoT Ecosystem & Trends at IDC. “The drive of digital commerce is driving organizations to be flexible and produce goods efficiently and quickly. To do so, organizations must deeply understand their industrial processes. IoT platforms and advanced data gathering are necessary to ensure successful and resilient industrial operations.”