Manufacturing analytics in action
Our methodology for the requirements analysis makes all the difference
There are various ways to get a data analytics project up and running. The starting point is usually a kickoff workshop at the customer’s production facility, where we meet with the customer team to discuss and understand the project objectives and underlying issues based on visual impressions of the product and the production process.
Production engineers tend to focus on and deep-dive into the specific process step that seems to be causing the problem, and apply common problem-solving approaches.
But data analytics can deliver a wealth of further benefits because it doesn’t focus purely on one specific process or machine. Instead, it also takes into account machine, process, and material data from up- and downstream production steps so it can identify previously hidden cause-and-effect relationships, correlations, and patterns.
So besides introducing the potential and power of data analytics in manufacturing at the beginning of a project, one of our main goals in the kickoff workshop is to understand the customer’s problem from a business perspective. Even more important, we try to view the problem in terms of the physical principles behind it. This allows us to focus on the right data sources so we can leverage the full power of data analytics.
Spending time on business understanding pays off
In the first part of the workshop, the goal is to identify the core problem and ensure that we align both our understanding of that problem and our data analysis with it. We call this first part of the workshop “business understanding.”
Those in the workshop are asked targeted questions to help them achieve a deeper understanding of the product, processes, and overall conditions. This, in turn, yields the initial signs pointing to possible root causes. The targeted questions include:
What is the exact sequence of process steps? At which station do the problems first occur? What nearby stations might be having an effect on that? Are there any special characteristics of the value stream, such as parallel steps, rework, or repeated steps? How many product variants are there? How many suppliers are involved? And so on…
Data understanding comes next
In the second part of the kickoff workshop on analyzing and recording the project requirements, we start talking about data:
What sources are providing data? Do we first need to integrate, or even generate, data? What data from other processes is important? What time frames are we looking at? When working with various sources of data, is it possible to clearly trace data back to its origin (e.g. using unique IDs)?
Iterative analysis instead of a massive project
The first analytics phase usually lasts no more than a week, depending on how long it takes to prepare the data. After this week, our manufacturing analytics experts present their initial results to the customer’s expert team with the aim of demonstrating the feasibility of solving the customer’s problem based on the given data.
Ideally (and in fact, this happens very often), the results of the first cycle of analysis already provide actionable ideas on how to achieve the project objectives (e.g. reduce the scrap rate in a particular value stream).
At the same time, based on customer input and their reaction to the intermediate results, it’s possible to readjust the analysis strategy to make it fit with the newly gained insights. This is crucial for the success of data analytics projects! Why? The team rules out incorrect conclusions, the immediate value added by data analytics becomes visible and usable, and the next steps are jointly defined:
What else is needed to verify results? And ultimately to automate them? How exactly should we define the expanded data scope for this?
Maintenance & support for predictive models once the project is over
Is there such a thing as post-project maintenance and support? many customers ask us. The answer is a resounding YES! This aspect is crucial for all customers who want to apply the predictive model to their real-time data, for instance to schedule wearing-parts replacement at the optimum point in time or to predict test results.
That’s why we don’t focus solely on providing maintenance and support for the installed software solutions, but also on supplying the appropriate technical support for training and monitoring predictive models.
Next level: standardized tools for standard problems
Web-based analytics tools yield immediate insights and put data analytics to use in the day-to-day work of engineers without having to engage data scientists. This next level is exciting. Learn more about it in the video.