In 1999, Abbott Nutrition began working with OSIsoft’s PI System to integrate, collect and contextualize data at the company’s manufacturing plant in Columbus, Ohio. Based on its success at this plant, the company entered into an OSIsoft Enterprise Agreement (EA) in 2012 to include all Abbott Nutrition manufacturing sites globally. This created a huge volume of data, which presented the challenge of extracting maximum value from the data.
About two years ago, Abbott began focusing on ways to obtain more value from the data. The company used OSIsoft’s Asset Framework (AF) to contextualize and organize the data using AF’s asset-centric models. The company also started using Seeq’s advanced analytics applications to reduce the time and effort needed to connect to AF, create models and find insights quicker.
Seeq can be used to contextualize time series data and create models to help engineers quickly derive insights and value from industrial process data. Abbott decided to use this approach to help reduce clean-in-place (CIP) rinse time while maintaining or improving product quality.
In 2016, Abbott started a modest pilot project on one OSIsoft PI Server workstation to prove the return on investment of this idea. “By the end of the pilot, we felt confident in the value we could obtain and increased the project scope to include additional PI Servers,” says James Li, an engineering manager in the Abbott Nutrition division.
Abbott launched a formal analytics program globally at the beginning of 2017, with near-term plans to connect Seeq to all OSIsoft PI Servers globally.
Creating the model
Seeq’s three-stage model includes: 1) data wrangling, 2) engineering investigation and 3) sharing insights.
第一步,数据角力,取代了小时of manual effort typically required to aggregate, cleanse and contextualize data. Data wrangling turns specific asset data into useful information “capsules.”
Capsules, a Seeq innovation, consists of a user-defined slice of time that can be visually correlated with a unique identifier and then compared with other assets. Once defined, capsules can be overlaid on top of each other or compared side by side for analysis and visual comparison.
The second phase, engineering investigation, pertains to developing the model. Seeq’s analytics tools helped Abbott target the problem by searching and annotating the sensor data, pattern matching and historical benchmarking. This enabled the company to quickly identify and target previously unidentified problems.
The third phase, sharing insights, includes performance monitoring and reporting, dashboards and real-time collaboration. “Our engineers love their own spreadsheets with their own formatting,” Li says. “However, to improve collaboration and transfer process knowledge to newer workers, the information needs to be shared.”
Seeq Worksheets support the sharing of displays with a common URL; the software also exports data to Excel and their business tools for sharing and role-based dashboards. Both approaches improve collaboration globally.
Abbott Nutrition operates 14 manufacturing sites (with more than 70 manufacturing sites globally for all of Abbott). The company’s goals include:
- Improving asset utilization for CIP processing.
- Maintaining or improving product quality.
- Reducing overfilling of products.
- Increasing production of saleable products, without increasing variable costs.
实现这些目标,公司使用Seeq analytics technology in two use case pilots. The first involved CIP processes; the second focused on reducing overfills in packaging operations.
Optimizing CIP systems
To reduce CIP cycle times while maintaining or improving product quality, the company focused on asset utilization. Li explained that he wanted to be able to use Seeq to review the data stored in the PI Server to quickly identify periods of interest.
CIP is commonly used to clean process equipment between batches or production runs. While a process is in the CIP phase, the equipment is not available to produce product.
Typically, CIP is a centralized system that can connect to and be used to clean multiple process units. CIP systems typically comprise multiple circuits or steps that involve circulating caustic, acid and water to clean the process equipment. To determine the endpoint for each CIP step, Abbott uses conductivity sensors to measure the water/acid/caustic parameters.
While Li had previously created trend reports to identify and isolate specific CIP issues, trend data alone does not always tell the complete story. For this pilot, he created a model using Seeq to validate and improve the way the available data was used.
“To be able to capture the most important data and assimilate the intelligence, we used Seeq capsules to break down the data into phases specific to the CIP process and equipment,” Li said. “This information allowed us to quickly create a phase capsule with information that compares circuits and equipment.”
Seeq enabled Li to quickly identify patterns, see the duration time for each circuit and phase, identify the longest CIP run, and drill down to determine the reason for the longer run. He was also able to analyze the incremental costs accrued. The capsules helped identify excessively long CIP runs, over-rinsing and/or when and in which CIP circuits the acid or caustic was overused. Then, he could drill down to quickly identify the root causes, such as valve leakage or another instrument failure.
The ability to quickly identify the root cause enabled the company to fix the issues to improve operational efficiencies through more consistent CIP duration. In some cases, Abbott was also able to decrease CIP cycles to increase production availability.
Another benefit was the ability to reduce product contamination by quickly identifying conductivity peaks. For example, if it was revealed that a valve leak caused a conductivity spike, an operator could put the batch on hold to prevent product contamination.
For the CIP cycle use case, the tool helped Abbott identify key areas to focus on for production improvements.
Reducing product giveaway and scrap
The second Seeq pilot use case focused on Abbott powder filling equipment. Here, it’s important to reduce product filling errors to enable the company to meet its filling label claims for product weight, while minimizing costly product giveaway. In the past, it was time-consuming to examine the filling data and obtain batch information.
使用Seeq,李能说明连续数据streams using a model that calculates product giveaway and aggregates the information over the time periods of interest. He did this by creating capsules that enabled the company to review the weight data vs. the target and filling state and compare filling operations, analyzing loss-per-fill and aggregating by product and shift. This enabled the company to determine if the target was met in about five minutes, providing a resolution to the overfilling problem.
To assess total giveaway across shifts and find variability, Abbott uses dashboards in PI Coresight (now called PI Vision) to visualize the data. Also, by monitoring the filler head actual weight distribution, the company can determine when filler maintenance is required.
For more information, visit Seeq atwww.seeq.com.
