Questions answered: What is just enough industrial data analysis?
Extra questions from the data analytics webcast are answered here, from an April 21 webcast, archived for a year.
Audience members participating live in the April 21 webcast on “What is just enough industrial data analysis?” have their additional questions for the speakers answered below to help learn if “just-enough” industrial data analysis working for operations.
Speakers for the Registered Continuing Education Program (RCEP) webcast are Laurie Cavanaugh, director of business development, E Tech Group, and Matt Ruth, president, Avanceon. The webcast, archived for a year, was moderated by Mark T. Hoske, content manager, Control Engineering.
Who is typically the driving force to engage on analytics in a manufacturing plant?
Matt Ruth: There are two paths where analytics starts in manufacturing.
First is a top-down approach where a program is sponsored from executive leadership and is designed to mine the data holistically across the enterprise. These are often large programs that have set scope, structure and human element change management functions to ensure uniformity.
Second is the ground-up path where the champions are from operations (because they need specific context about a specific problem) or from operational technology (OT) resources (because they see the need/problems in operations and know the data exists to help). Results from this type of project is usually a key first win that drive the second project and perhaps an ultimate rollout (and maybe even feed the bigger initiatives listed above).
Laurie Cavanaugh: I agree with Matt’s comments regarding the paths. Both can and may happen in parallel. A top-down approach to data analytics is typically driven by corporate and involves corporate IT and the use of accessible data in the form of financials and some operational data that has been consolidated (transactional data) based on enterprise resource planning (ERP) and business intelligence software. A bottom-up approach is typically driven by manufacturing departments or sites and starts with granular detail (time-series data) using historian databases and extraction tools offered by automation and controls manufacturers. Where true value and insight can be experienced is in the convergence of these two analytics’ efforts to connect the corporate financial and operational data with the more granular process execution data to provide a more thorough and holistic view of key factors that drive organizational decisions from top to bottom.
What do you see as the future of analytics in manufacturing?
Ruth: The future of analytics and its real value in manufacturing comes when the confidence in the models and problems solved by analytics become so reliable that manufacturers feel comfortable with closing the loop from the analytics directly to the process. This occurs today in most control systems with process feedback loops. Manufacturers will eventually look at analytics as another control platform.
Cavanaugh: The future is now. User expectations and demand for contextualized and intelligent information, driven by the tools used outside of the workplace, are countering old excuses or reasons for the failure to deliver fact-based analytics, directed decision making, and immediate access to critical information. Technical tools and platforms are not lacking. What is lagging is the complete understanding and response to the non-technical bottlenecks and barriers. Greater awareness of how analytics applies to processes and proper context for critical decision factors produce the value that addresses those non-technical barriers.
Do you realistically see the wall between information technology (IT) and OT disappearing?
Ruth: I do. As it becomes more challenging to find a wide array of resources to target specialty platforms and the platforms become more and more similar, IT and OT will converge. I believe the common languages/common platforms need will be driven by security protocols and a focus on ensuring the cyber safety of the supply chain. At Avanceon, we have a phrase to decide if someone will be successful as a developer for controls and/or IT – “a true programmer can program anything.” With the needed convergence of IT/OT, we will see this reality first hand.
Cavanaugh: The IT/OT wall or divergence must be overcome for the sake of shared responsibility: the health and wellbeing of the company for which they both work. While there is acknowledgement and acceptance of this fact (2 of the 4 As discussed in IT/OT convergence), the “heavier lift” part is action and accountability. We have found that engaging IT and OT in a shared goal and corporate need, such as cybersecurity protection or analytics and business intelligence, provides a focus and path to the final two As, action and accountability.
What percent of your customers are actively involved with analytics?
Ruth: Nearly 100% of our customers do some form of Analytics. Descriptive reporting is prevalent. As is diagnostic reporting. A handful (10%) of those are committing to predictive and prescriptive problem-solving models. Those same 10% are looking to get their data into a self-service model by creating a data factory to make “advanced analytics” more feasible. It is a logical step for many manufactures to take the value of the historical info that feeds descriptive and diagnostic, remove the human element/risk from the root cause analysis and allow the cloud to lean and crunch all that data (right tool for the job).
Cavanaugh: 100% of our customers engage in analytics – from Microsoft Excel spreadsheet graphs posted weekly on a bulletin board to real-time dashboards displaying key metrics on mounted screens throughout the plant and consolidated views at the corporate level. Differences are the breadth and depth of the analytics, the source of and timeliness of data, and the sophistication of the analysis being done on the data sets and in the user interface/user experience (UI/UX) delivery.
What return on investment (ROI) is achieved with analytics?
Ruth: This is a “it depends” question. ROI requires an action or change to be realized. Most analytics examples we shared show operations or engineering where problems exist but changes aren’t made. Typically, it takes people to act on that suggestion and close the loop to the process. In the proportional-integral-derivative (PID) loop tuning example, when the “loop gets closed” there is 5% improvement in the time it takes to clean-in-place (CIP), resulting in an ROI (of the analytics project + the improvement project that follows) of less than 3 months. Those are great results, but it takes a commitment to act on what the data tells you.
Cavanaugh: ROI categories can be grouped into revenue generation, cost reduction, and risk avoidance. Some examples are noted:
Revenue generation is measured in terms of increased yield/throughput, more turns on inventory, higher return on asset performance. If analytics and the feedback and information provided allows for active decision making, or eventually closed-loop decision making for auto-adjustments or in-line statistical process control, the ROI is definitely tangible and measurable. Notification of alarms, reducing short stops with more informed maintenance response or predictive failures also are ways to meet these revenue increase goals.
Cost reduction can be achieved by increasing mean time between failure (MTBF), reducing mean time to repair (MTTR), identifying scrap upstream or with more immediate notification to allow for corrective action. In doing some of these things, it also may be possible to analyze and reduce spare inventories and hire skill sets in operations, maintenance, and engineering to meet highest need requirements.
Risk avoidance is sometimes more difficult to quantify because it requires either a past event, like a security breach, recall, governing agency penalties (OSHA, FDA, USDA), or obsolescence failure, to provide tangible financial impact. However, more recently with ransomware attacks on process industries and corresponding rises in costs for cyber-insurance premiums, higher deductibles, and reduced ransomware coverage, there is now a case to be made for ROI by reducing the costs in these areas. Cost reductions can be made through analytics to expose vulnerabilities and post-remediation security and with safety audits to reduce insurance costs.
Please review how to eliminate industrial manufacturing bottlenecks.
Ruth: I categorize the “bottlenecks” as the barriers to use of analytics. The main issues with the adoption and acceptance of analytics are:
- Cloud “willingness”: Accept the use of the tools that we leverage every day in our lives inside the factory
- Connection to OT and business data: Breakdown the silos of the past and focus on the business value as the goal
- IT allocating resources for OT data: Develop trust and take an IT slice of time for OT
- Addressing security concerns/access: Follow cyber policies but understand the value in allowing connections for analytics.
- Warehouse/factory your data: Organize, cleanse and filter your data separate from their original data sources to make it possible to break down barriers to achieving self-service data use and developing advanced analytics.
Cavanaugh: I love to focus on the non-technical barriers having spent my career on the up-front business development process with our customers and their teams. Without the people and the passion for change to make a positive impact for the organization, it won’t matter how technically “cool” a solution is. Stakeholders across all levels of the organization need to be engaged early and updated often. While many won’t have a “vote,” they can certainly have a “voice.”
Do you have related research on analytics?
Ruth: Our experience and info comes from the value provided in our applications and the opportunities they exposed.
Cavanaugh: Case studies and post-project audits provide a retrospective to identify what went well, where adjustments are needed, and next steps on the path.
What data analytics are appropriate for ongoing commissioning (OCx)?
Ruth: As our PID process analytics showed, there is ROI in baselining a system, performing a change to the process and then measuring the result. Most plants do not look across the spectrum of time in their operations. The concept of baselining before the commissioning and then running the same analysis after (or even during) the commissioning process will shed light on the improvement and resulting “J curve” of the investment.
Do you have other real-world examples and applications of analytics improving a manufacturing process?
Ruth: Yes. Quite a few. Additional examples include leverage Internet of Things (IoT) sensors, cameras as data sources and real cost calculation of production. For more info, please contact firstname.lastname@example.org.
Cavanaugh: Yes, we have quite a few and as noted previously, we share those on our social media channels and website through blogs, articles, and white papers. www.etechgroup.com
How do you think the notion of the retirement of long tenured staff and their tribal knowledge is expediting the importance of diagnostic analytics for the next generation of plant engineers?
Ruth: As we discussed in the session, analytics exposes and presents operational issues in three time frames.
- After they happened
- As they happen
- Before they happen.
This type of insight is invaluable when considering the gap in info that a new plant engineer has to do the job. Helping plant engineers look in the past to predict and react to the future will make them better equipped to add value sooner in their tenure.
Cavanaugh: The “great resignation” has offered 2 sides to the coin. We are losing a lot of insight and analytical expertise that was not captured within systems as engineers and technicians as young as 55 are choosing to retire. On the flip side of that coin, though, we are seeing those senior level resources feel an “itch” and want to return to the work, but perhaps in a less conventional way. E Tech Group has seen a return of these key individuals in roles as part-time consultants working three days a week, or even technical members who want to work four or eight months and are excited to travel and assist with our commissioning work at customer sites across North America.
One thing the pandemic assured all of us is that the talent can be anywhere… as long as there is Internet access.
Once the decision has been made to use data analytics, is it difficult to get agreement on goals for the project (what success looks like)?
Ruth: Depending on the approach (top down or bottom up as per question 1), it changes the way analytics goals are developed. If “top down,” the goals are enterprise focused and often the plants will get their versions of the enterprise application and (hopefully) will be able to develop additional analytic models on the data to solve plant specific issues. If “bottom up” from a plant or operational focus, the ability to understand the needs of the plant operators will be considered holistically and addressed directly. The risk exists, without discipline and overall understanding of best practices for structure and standards, that one analytics system will not leverage forward as easy to the next facility and will need to be customized to be ported from site to site.
Cavanaugh: The presentation noted that it is important to consider when navigating the inside sale for any type of project: What’s in it for me (WIIFM)? Know your audience and what their pain points are and how the analytics output will help address that pain. Revisit some of the ROI examples. Increasing revenue, reducing cost, and avoiding risk are music to any management person’s ears. Making operations’ users jobs easier, helping them be heroes, and alleviating their stresses will be music to their ears.
Laurie Cavanaugh is director of business development, E Tech Group; and Matt Ruth is president, Avanceon. Edited by Mark T. Hoske, content manager, Control Engineering, CFE Media and Technology, email@example.com.
KEYWORDS: More answers about industrial data analytics
Review the return on investment for industrial data analysis.
Examine barriers and how to overcome obstacles to implementing industrial data analytics.
Eliminate industrial manufacturing bottlenecks with information here and from an associated webcast.
Have your operations applied industrial data analytics in more optimal ways?