IBM Uses Machine Learning to Lower Bottling Costs
You're probably familiar with the thick plastic wrapping around a 24-pack of water bottles. The plastic is usually strong enough to prevent bottles from slipping through when carried. Water bottle manufacturers try to stretch this plastic during production to save money. However, there's a risk of breakage, which can cost a company the money it's trying to save. Therefore, water bottle manufacturers—such as bottled water company Niagara Bottling—are turning to IBM machine learning (ML) and artificial intelligence (AI) to help optimize the right amount of plastic that automatic pallet stretch-wrapping machines use. Ontario, California-based Niagara Bottling operates hundreds of these machines across more than 24 plants.Niagara Bottling packages its own namesake bottles as well as others' bottles, including PepsiCo's Aquafina, Arizona Beverage Company's Arizona Iced Tea, and The Coca-Cola Company's Glacéau Vitaminwater. About 90 to 95 percent of Niagara Bottling's business comprises manufacturing water bottles, and the other 5 to 10 percent involves specialty beverages, according to Sreesha Rao, Senior IT Manager at Niagara Bottling.
With data analytics, Niagara Bottling factors in data points such as humidity amount, package size, pallet size, weight, and the number of wraps to determine the best, most efficient way to wrap. If Niagara Bottling can find the right amount of plastic to use, then it can avoid breakage while reducing production costs on the amount of plastic.
"There's always a balance between cost of the plastic that's going to be used [for] wrapping versus stopping production," said Carlo Appugliese, Program Director of ML and AI for IBM Analytics. "You really don't want to stop production."
Avoiding Stretch-Wrap Breakage
Niagara Bottling implemented the IBM Data Science Experience (DSX) platform and IBM Watson Studio to build data models using open-source tools such as Python. The goal was to study how to reduce the amount of plastic used. The process involved not only data modeling but data cleansing as well. "It takes a four-step process to extract data from source systems: cleanse it, do the modeling, and then test and deploy the model," Rao said.
ML from IBM Watson Studio can help Niagara Bottling create data models to determine the most the company can stretch the plastic (without breaking it) in order to save on materials. The four components of IBM Watson Studio are: Decision Optimizer, IBM SPSS Statistics, IBM's statistical modeling toolset, and the open-source tools RStudio and Jupyter Notebook. Open-source tools form the foundation for ML and data science, according to Rao. Niagara Bottling uses IBM Cognos for reporting, dashboarding, and data visualization. In addition, it uses Informatica and Tableau for extracting data, cleansing it, and importing it into the company's data warehouse.
Although the research is still ongoing, the bottling company found that ML was a secondary tool in gaining the insights. "We thought we could use DSX or [IBM] Watson Studio to help us reduce that plastic. But we found that tuning the machines and the quality of the stretch-wrap material that we have are the primary determinants," Rao said. "ML is more of a secondary [benefit] or a nice-to-have. Our models are not sophisticated enough to incorporate stretch-wrap material characteristics, which is hard to track from each supplier and associate with each roll on each equipment."
As part of the manufacturing process, Niagara Bottling uses two pallets to hold six packages of bottles per pallet. Niagara Bottling must build the water bottle plastic to handle that type of pressure, Rao noted. "Then, stretch wrap is applied around the pallets for transportation. There's some containment force that keeps all of the cases together so, when these pallets are transported to our customers, they all stay in place," Rao said.
Transporting the bottles requires about 200-250 grams of stretch-wrap plastic to hold the cases of water together. Usually, Niagara Bottling rotates the plastic 15 to 20 times around the water bottle pallets, according to Rao. "The amount of plastic is directly proportional to the number of wraps and how much stretch you apply on that stretch wrapper machine," Rao said.
Predictive Maintenance in the Factory
In addition to ML, the Internet of Things (IoT) and predictive maintenance will come into play as Niagara Bottling will collect data from machine sensors. Sensory data includes humidity within the building and the amount of pressure and speed when factory workers pull the plastic stretch wrap. "Getting into predictive equipment maintenance is going to be an area where we could use data science and ML," Rao said. "We have just started down that path of being able to instrument different manufacturing equipment with the right sensors, and being able to measure and gain insights from the measurements." In addition to water bottle manufacturing plants, IoT data from predictive maintenance is valuable in power plants and oil refineries.
This year, Niagara Bottling would like to use data analytics to gain more operating cycles out of the automatic pallet stretch-wrapping machines and more closely identify when they would need to go out of operation for maintenance. "Especially with some of the critical motors, we want to know what is the most appropriate time to take them down," Rao said.
Going forward, Niagara Bottling will take a look at how it can use ML for condition-based maintenance, which involves gaining data from machine sensors to track how the machine is doing. This step will require watching the machines for variables and creating signature patterns to determine which is normal operations and which isn't. "We will know exactly based on the signature pattern when a machine condition deteriorates," Rao said.
For three months (beginning mid-year 2018), Niagara Bottling worked with the IBM Data Science Elite team to learn how to build data models and cleanse the data. The collaborative environment was helpful in gaining insight about the tools, according to Rao. "The important thing is that we were able to have a discussion about the business problem," Rao said. "We tried to leverage the tools and then experience the capabilities of modeling, apply that to the business problem, and then try to be narrowly focused on results and outcomes."
Niagara Bottling is still deciding on which technology to use for its cloud platform and possibly an edge-computing architecture.
Using Data Analytics to Improve Quality
Potentially, Niagara Bottling could connect the predictive analytics model to the automatic stretch-wrapping machine and adjust to environmental changes in real time. In September 2018, IBM held its "Winning With AI" event in New York City. At the event, Rao told ESPN host Hannah Storm that it's too early to glean results from the data. However, Niagara Bottling has learned that the primary indicators are how Niagara Bottling tunes its automatic pallet stretch-wrapping machines and the quality of the stretch wrap material.
Niagara Bottling is awaiting results on the quality of the plastic, and will study the plastic's characteristics and source minerals in the next phase of the research. "It validated the approach: that ML is the driving technology that will dynamically predict machine settings to influence plastic usage in the production line," IBM Analytics' Appugliese said. "The next phase is to collect plastic characteristics and retrain the model."
This article originally appeared on PCMag.com.