In manufacturing, it’s true that a picture is worth a thousand words - and often hours or days. At Instrumental, we’re building an intelligent factory quality system to modernize manufacturing. We learned from speaking with engineers and leaders at both product companies and manufacturers that simply looking at dashboards of numerical data hasn’t lived up to its promise of enabling more efficient development and production. We believe it’s because the data that’s typically collected and stored is too low-resolution to be actionable -- manufacturing needs better data sources. Our first product incorporates multiple data types, but is designed around images, one of the few types of “measurements” you can take where you don’t need to know what you want to measure beforehand. Below, we’ll cover best practices to follow when using such images to keep manufacturing schedules on track.
Make them accessible from anywhere
Unfortunately, assembly line data is too often trapped on individual machines and simply overwritten over time (if it’s collected at all). However, when images are available and other best practices are followed, “using [assembly line images] to review issues is more efficient than being in the factory” according to John Brock, Director of Product Design at Pearl Automation. That’s because images can provide a complete historical record and can be compared easily, whereas an engineer on the assembly line can only see a subset of units in real time.
Make them searchable
Finding the right data quickly and easily is the key to problem-solving efficiency, and that means you’ll need a fast search system that supports unit and inspection metadata. Here’s an example of a search for unit data in Instrumental:
Instrumental customers often search for images of units that have failed functional tests in the factory in addition to more casual perusal of incoming units. In fact, 44% of searches use a serial number filter, showing the importance of pinpointing specific units. (The other 56% of searches use criteria like “units that went through Station A yesterday” and then users browse the results to keep tabs on incoming unit quality.) We keep things fast, too: 100% of searches in the last 30 days were processed in under 50 milliseconds.
Images are an incredibly rich medium, and much can be learned from zooming in on high resolution images to see features that are not visible at native resolution. This is most useful when similar units can be compared to help identify trends and root causes. For example, Instrumental’s image viewer has a feature we call “window boxing” that makes it easy to compare the same zoomed-in feature across multiple units. To illustrate, compare these two potted conductors from a consumer product - an engineer reviewing the images would be able to easily see the bubbles in the second one:
Similarly, the ability to see “more like this” makes comparison faster. In practice, this typically means seeing data from other units that went through the same station around the same time. Users do this in about 57% of Instrumental sessions.
We've also written about data management best practices as a whole. There’s much more to say about data derived from images -- but we’ll save that for another post.
Request a demo to see how Instrumental images can streamline your issue discovery and failure analysis process.