COMPANY CONFIDENTIAL. DO NOT DISCLOSE.
Please note that this was first published as an internal communication within ExxonMobil.
JMP USER QUARTERLY
SUMMER 2021 | ISSUE 1
EDITORS’ NOTE
Over the course of its many decades-long history, ExxonMobil has grown from a small regional outfit to one of the world’s largest publicly traded companies and pioneering early advocate of digital transformation. One of the tools that has accelerated that transformation – by empowering domain experts to make more data-driven decisions – is JMP® statistical discovery software.
Thanks to our company’s investment in an enterprise JMP license, any and everyone who deals with data at Exxon can now access JMP from their laptop or desktop, at home, on site or in the field. So whether you’ve been using the software for years – or just installed it yesterday – keep reading. Regardless of your background and level of statistical expertise, we hope you’ll be inspired by this new quarterly newsletter to get more business value from your data.
With each issue, we’ll highlight how JMP users from Houston to Kuala Lumpur have introduced actionable improvements that affect our bottom line. You’ll read examples of how, in the hands of talented engineers, scientists and analysts, JMP is enabling Exxon as an organization to work more reproducibly, accelerate production, increase yield, lower costs, improve processes, communicate new ideas, explore our data, find opportunities and most importantly – continue to innovate and push the boundaries in the global energy sector.
Read on and reach out to your colleagues, even those on the other side of the world. And find new ways to share best practices for data management, analysis and reporting. After all, discovery is what Exxon – and JMP – are all about.
Ekenechukwu Casey Ofoma, DAS Analytics Advisor, ExxonMobil North America
Meg Eberle, Global Customer Reference Manager, JMP
_____________________________________________________________________________________
1 USER SPOTLIGHT: EMRE FPO KLTC PMA
Customizations in JMP created a user-friendly, scalable tool for extreme value analysis that helped optimize inspection processes and eliminate human error
Amar Sufian Mohd Sapihie, Business Support Centre, EMRE FPO KLTC PMA, Kuala Lumpur, Malaysia
Marco Salfi, Advanced Analytics Advisor, Milan, Italy
Challenge
In the face of increasingly complex inspection data siloed into different data storage systems, in-house inspectors at ExxonMobil’s NDC facilities in Malaysia sought a user-friendly tool that could provide high-impact quality analysis, reduce human error and overcome data access barriers.
Methodology
Engineer Amar Sufian Mohd Sapihie, in consultation with Advanced Analytics Advisor Marco Salfi, used a non-linear equation in JMP to determine the best values for a model’s constant variables. To evaluate the model’s accuracy and variability, they then compared the resulting equation with theoretical figures. life distribution application in JMP enabled Sapihie and Salfi to perform an in-depth analysis of extreme values based on historical data. Finally, a comparison of theoretical and existing EVA equations provided insight into their efficiency and skewness.
Results
By automating extreme value analysis (EVA) with JMP scripts, Sapihie explains, the team created a new standard process for repeat analysis that can be done many times faster than the previous manual approach. With its customizable, in-built analytics workflow, JMP made the NDC site’s standard inspection process more reproducible and less prone to error.
ROI
$200,000 in annual savings with strong scale-up potential for any application involving extreme value analysis.
_____________________________________________________________________________________
2 USER SPOTLIGHT: EMCC Operations Asia Pacific
A multivariate process monitoring approach in JMP leads to more than $5 million in annual savings and a 30% reduction in analysis time
Masoud Golshan, Manufacturing Analytics Advisor, Houston, Texas
Yew Boon Teo, Chemical Operations, EMCC Operations Asia Pacific, Singapore
Challenge
At ExxonMobil, each process engineer typically supports between 3 and 6 plants with as many as 1,000 tags each spanning up to 100 days of operation. Tasked with multiple objectives from identifying and resolving reliability issues to evaluating measurement systems and optimizing operating conditions, process engineers required a better standard of practice for actioning the data from extremely complex physical systems. Given the quantity and complexity of data generated by advanced sensor technology, data management had become one of the most significant – if not the primary – challenge for process monitoring and analysis.
Methodology
Working together with Manufacturing Analytics Advisor Masoud Golshan, process engineer Yew Boon Teo outlined a multivariate process monitoring and analysis system in JMP that would enable engineering teams to execute on their vision of moving to more predictive or even prescriptive approaches to process monitoring. JMP, Golshan says, offers an in-built platform that allows users to visualize complexity in much reduced dimensions .
Results
The team’s multivariate approach to process monitoring proved to be both efficient and effective; a solution that could be run from the central office with low investment and high return in both the short and long term. While only one pilot has been completed to date, Golshan and Yew Boon Teo report that there is an extended pilot underway involving 20 operations and the creation of more than 30 multivariate monitoring systems. Ongoing tech extensions are expected to further advance the full scale-up.
ROI
A combined value of more than $5 million per year and 30% reduction in process monitoring analysis time.
_____________________________________________________________________________________
_____________________________________________________________________________________
DID YOU KNOW?
RTO data is available to be pulled out of the data lake into JMP DLC/XOMMFG/SEEQ/GetPlantData!
Contact: Charles Fryer
JMP DLC was enhanced to version 2.0
Contact: Luiz Padilha
Our timeseries API can now provide straight average results, not just weighted average
Contact: Marcel Mueller
We are looking to innersource JMP DLC, XOMMFG and more for rapid updates
Contact Marcel Mueller if interested to help
_____________________________________________________________________________________
_____________________________________________________________________________________
3 USER SPOTLIGHT: EMCC Global Technology
When a step-change quality improvement was needed, the Global Surface Spot Team used the Functional Data Explorer platform in JMP along with principal components and outlier analysis to introduce an annual improvement of $200,000
Urtzi Apraiz, Chemical Process Engineer, Advanced Elastomer Systems, EMCC Global Technology, Newport, NJ
Marco Salfi, Advanced Analytics Advisor, Milan, Italy
Challenge
After observing persistent performance issues with Santoprene, a thermoplastic elastomer polymer, the Global Surface Spot Team sought a more long-term solution to chemical process quality optimization. First, they would need to identify and understand the main process drivers causing a high number of surface spots.
Methodology
Chemical process engineer Urtzi Apraiz, together with Advanced Analytics Advisor Marco Salfi, developed a methodological approach that could be undertaken entirely within JMP. The team first used a distribution report for outlier analysis followed by predictor screening to identify the contribution of all response variables on surface spot outcomes. Exploratory data analysis in the software’s Functional Data Explorer (FDE) combined with principal component analysis (PCA) helped to address functional, signal and series data. Ultimately, they applied a predictive modeling technique known as bootstrap forest to create a model that could predict a response value by averaging the predicted response values across many decision trees.
Results
Apraiz and Salfi identified two root causes of Santoprene surface spots. First, they identified the downstream extruder high melt temperatures, particularly for oil and rubber, as having the most significant impact on FPC1. Second, they discovered that the RB powders ratio significantly affected performance .
ROI
An estimated $200,000 in annual improvement, with more expected as the analysis is extended to the Pensacola Chemical Plant.
_____________________________________________________________________________________
4 USER SPOTLIGHT: GSC PROC Downstream and Chemical
A customized take-or-pay contract optimization tool developed in JMP provides ExxonMobil contract negotiators with a competitive advantage
Bethannie Faith Canter, GSC PROC Downstream and Chemical, Spring, Texas
Ekenechukwu Casey Ofoma, DAS Analytics Advisor, Houston, Texas
Challenge
Take-or-pay contracts guide industrial gas negotiations between ExxonMobil and its suppliers, with contract negotiators agreeing to either take the supplier’s product at a predetermined price or pay a penalty. Pricing for industrial gas volumes is generated from a combination of product quantity, capital, operations and maintenance costs and labor costs. Existing contracts base their volumetric tier thresholds on historical consumption, while new projects are based on theoretical consumption. With excess consumption charges, the price typically exceeds base volume tiers by 4-13 times.
Given the complexity of contract negotiations, it is a challenge to identify thresholds of volumetric tiers that optimize the minimum take-or-pay (MTOP) while minimizing cost incurred due to excess consumption. Analysts must further monitor site consumption rates over time to track performance.
Methodology
Bethannie Faith Canter and DAS Analytics Advisor Ekenechukwu Casey Ofoma saw an opportunity to leverage the JMP platform to develop an end-to-end optimization tool. By using the JMP Scripting Language (JSL), they were able to write a script that applies a Monte Carlo simulation of daily production to evaluate a range of pricing and volume parameters on take-or-pay contracts. The model uses penalties and total spend as the response variable and pricing and volume parameters as the input variables and is fit in JMP using a neural network. The JMP Prediction Profiler is the used to visualize changes to the response variable as input variables change. The desirability functions help find optimal input settings to minimize penalties and total spending.
Results
Canter and Ofoma believe they have developed an optimization tool that will provide ExxonMobil with a competitive advantage in industrial gas contract negotiations.
ROI
Potential cost savings of as much as $10 million and a competitive advantage during contract negotiations.
REMINDER: COMPANY CONFIDENTIAL. DO NOT DISCLOSE.
