Preserving design data, ensuring long-term access

The recent developments in data analysis and computation-enabling technologies mean nothing if the datasets you need become like buried treasure. Most businesses assume that the progress in high-performance computing and big data analytics has been matched in the data access realm. But the reality for data access is grim: the simple task of finding some of the most important unstructured datasets related to industrial product lines remains elusive.

PLM databases and other systems may contain formal information about a design, but they don't account for the collective intellectual property about a product or system. Much of this information resides in unstructured form and is much harder to manage because it could live just about anywhere. During periodic tech refresh cycles, such file-based assets often "go dark," leaving engineers with a very fragmented picture.

The challenge of preserving access to such assets is not going to get easier with time. It's well known that, as predicted by IDC's Ashish Nadkarni, the growth of unstructured assets has now outpaced the growth of structured data by a wide margin. Even formal information such as geometry files, simulations, test and telemetry data—critical to informing design, testing, maintenance and redesign—is often rendered inaccessible to engineers after a few years. The inability of current systems to make accessible unstructured information such as informal notes and calculations leads to more problems. Asking engineers to work with an incomplete picture doesn't make sense. Engineers' final products are only as good as their ability to understand the decisions and "thinking behind the thinking" of a design.

Recreating the thinking behind the thinking
Trade-offs are made at every stage of manufacture and design. Unfortunately, informal notes and meeting discussions between members of engineering teams are rarely, if ever, preserved for even a portion of a product's lifecycle. This has serious implications for the overall success of the product.

Consider the automotive microcontroller. Designs are produced, schematics drawn and prototypes created. Test data is collected, and design changes are integrated into a production model. The design team may be fully aware that, during the test phase, bit error rate test X led to design change Y. Or they may have chosen to trade higher leakage for the maximum performance boost that comes with a shorter channel length. But these insights will gradually deteriorate over time as engineers and designers forget their reasoning or the data on which it is founded—or simply leave the design team. Months, years or even decades later their successors may be tasked with making changes to their design. But in the absence of the paper trail of notes, e-mails and other historical documentation, engineers are left with an unfortunate archaeological puzzle.

Too frequently, this critical "homework" is easily lost—buried in numerous laptops, desktops and other siloed computers. While conventional wisdom says that PLM systems contain comprehensive lifecycle data, in reality these systems are only marginally useful for repurposing designs or understanding the thinking behind design trade-offs. Typically these systems do not have sufficient unstructured and informal data.