Overview of TrendMiner`s practical use cases, from root cause analysis to automation and monitoring.
As a project engineer at TrendMiner, I support our customers in the use of our software. In this role, I’ve seen the great results that the software delivers and how much it helps people like process and production engineers. To inspire new and existing customers, this post will provide a general overview of some common customer use cases. The quotes you read here are from customers who have experienced the benefits of using TrendMiner in their own roles.
In addition to being a fast and intuitive trend viewer, TrendMiner’s advanced search functionality instantly unlocks the information in the process data to our end users, who are generally not data scientists, by providing fast and objective answers to specific questions, like:
- How often do we see these pressure drops in our hydrogenation process?
- How many times did the isomerization temperature raise above the safety limit in the last five years?
- What is the duration distribution of the cooling phase?
- Does dryer fouling actually occur faster after the most recent shut-down?
The answers to these questions assist the users in setting priorities for process optimization and building business cases for improvement projects.
Instead of linearly scrolling through the process data, various search dimensions allow batch process engineers to quickly retrieve batches of a specific product or particular process phases. This assists grade plant end users in finding back periods in which a distinct product was produced or transitions between particular product grades. Likewise, it helps engineers in continuous plants to acquire data at comparable production set points.
The resulting periods can be easily overlaid for further analysis and comparison, for instance to determine the impact of process changes, for example:
- How does the installation of new sieves impact production?
- What is the effect of using new mixers in our reactors?
- How do these second generation batches compare to historical batches of the same product?
“I needed to retrieve process data for batches of 10 different products, about 40 batches per product. Without TrendMiner, it would have been a nightmare.”
Turning data into action with TrendMiner
TrendMiner assists process and production engineers in their day-to-day work by allowing fast hypothesis testing and analyzing process upsets. Instead of working from assumptions, engineers can make quick comparisons between a current situation and a historical time period with comparable set points to identify what action is needed.
The results of the analysis can be easily shared with maintenance and can be used to determine alarm limits that can prevent the problem from happening again in the future.
“I assumed that a broken valve caused problems in our production line. A quick comparison between the current situation and a historical time period with comparable set points instantly revealed a large difference in the valve opening. I only had to mail the trends to our maintenance department to have the valve replaced.”
Root cause analysis
Root cause analysis is probably the most widely spread use case of TrendMiner at our current customers. The lack of overlay and compare functionality in most existing trend viewers forces engineers to revert to quite unorthodox methods when searching for the cause of more elaborate process upsets. Stories about printing trends on A3 sheets of paper and pasting them against the window to compare different periods, or exporting data to Excel, (which takes over two hours every time a new variable is added) are not one-time encounters. Therefore, time restrictions often lead to early abortion of the analysis, leaving cold cases and many frustrations in the engineers’ heads.
By overlaying periods of normal and abnormal behavior and using the overlay comparison functionality, users can quickly identify tags that exhibit deviating behavior and need further investigation, revealing answers to questions such as:
- What causes this unexpected pressure rise in the conveying system, forcing us to abort production of this specific product in 90% of the runs?
- Why has the duration of the filtration step in our silver process increased compared to two years ago? Can we find the cause of this problem in the preceding leaching phase?
- How can we explain this violent foaming in our flush tank, leading to blockage of piping and resin deposition on measurement instruments?
- Where is the origin of low pH values in this pumping station, causing violations of the aggressive CO2 norms?
- Since last year, we lose tons of product due to the formation of folds in the paper when the tambour reel is replaced. Can we find the root cause of this problem?
Unlike other trend viewers, TrendMiner contains a logbook back end that enables users to add context information as annotations on trends and assets. Several companies use TrendMiner as their day-to-day logbook in the control room, generating insightful reports for efficient shift handovers. Even when the operators aren’t involved in this process, information sharing right on top of the trends can lead to shortcuts in analyses and decisions. Imagine yourself analyzing a situation that a colleague of predecessor already came across a couple of months, or even years, ago. If you had instant access to their findings when performing a similarity search, it could save you hours of unnecessary work.
Other use cases of this functionality include labeling maintenance periods to filter them out in future analyses, categorization of upset situations with known and unknown causes via keywords to focus on the latter ones, and annotations of test runs.
Finally, when combined with the monitoring power of TrendMiner, losses or periods of high energy consumption can be detected automatically and annotated with their cause to generate daily/weekly/monthly reports on overall equipment efficiency (OEE) or energy consumption.
“Annotating my test runs makes it possible to easily find them back for analysis and comparison later on.”
Equipment fouling, catalyst degradation, energy consumption, start-up of spare equipment, water usage… Many process engineers have a long list of items they should be following up on a daily or even more regular basis via process trends or data export tools. However, most engineers’ to do lists are even longer and it is not unusual to lose sight of these follow-up tasks for days or even weeks. TrendMiner offers the possibility to monitor every saved search, sending you an email when new results are detected.
In addition to monitoring on saved searches, TrendMiner introduces the concept of fingerprints for common process situations. Using fingerprints based on historical good operation, process start-ups and shut-downs, grade transitions or batches can be followed-up live and action can be taken when deviations from the fingerprint are detected. Furthermore, fingerprints can be used to analyze and optimize, e.g., historical transitions and batches. Creating fingerprints for typical process upsets allows (early) detection and alleviation of these problems.
“If we can build monitoring schemes around all individual parts of our process, we should be able to ensure the stability of the overall process. That’s what Industry 4.0 means to me.”
Finally, TrendMiner’s (search-based) predictive analytics mode gives you a glimpse of the future. Based on the assumption that historical data is the best predictor for future behavior, you get an idea of how your process will evolve in the next few minutes or hours. When operators feel a process is degrading, they can use the predictive mode to check how similar conditions have evolved in the past. That way, unnecessary maintenance can be avoided.
“We recover energy from our wastewater to heat the main process stream. When both flows are equal, no extra steam should be needed. TrendMiner warns me when the steam valve is opened in such situations, indicating fouling of the heat exchanger.”
Automation, optimization and modeling
TrendMiner use cases can be found in any aspect of modern engineering. The development and tuning of new or existing controllers, process optimization projects and the construction of process models for simulation and monitoring purposes all require high quality process data, be it from normal operation, steady-state production, process transitions or step changes. The combination of search-based filters and fast data export get you that data in just a couple of clicks. Moreover, the layer comparison functionality can assist users in the selection of meaningful parameters for their analyses or variables.
“After the plant tests for the design of a new APC controller, an additional manipulated variable was added. With the help of TrendMiner, I was able to find the data to include this variable in my model without the need for new plant tests.”
Do some of these use cases apply to you?
If you saw an example in this post that you’d like to address, please contact us – we’ll be glad to discuss it with you.
Discover the real-life applications of self-service process analytics. Looking at real use cases, we will show 4 practical examples of making data actionable to inform decision-making and improve equipment availability.