Smart Engineering
The OCR Solution to Data Entry
A top-down approach for reducing error-prone and time-intensive manual operations.
Last month we talked about the often ambiguous, unstructured design data packages running rampant in the PCB industry, which drive non-value-add administrative tasks across all phases of our data exchange and processes, and we underscored the urgency to integrate “smart engineering” data-driven processes, as becoming more efficient as an industry in reducing cost and NPI cycles should be a critical objective to all organizations. What exactly do we mean when we talk about smart engineering or data-driven processes? Buzzwords and acronyms are all around us, such as digital transformation, RPA (repetitive or robotic process automation), BPM (business process management), SaaS (software as a service), etc. All encompass a similar objective: optimizing our processes throughout the enterprise.

In the PCB manufacturing facility, some classic examples of duplicated data entry when receiving a new design package are in the front-end engineering process steps (FIGURE 1). Several generic steps occur across the industry, and all of these must occur, with the sequence varying based on the company or manufacturing facility. In many cases, each of these process steps are completely segregated software applications, which in essence results in non-value-added administrative tasks.

diagram of PCB process flow
Figure 1. Most PCB operations have redundant manual data entry processes.
All the software applications we use today, from quote to engineering and CAM, require some form of data entry to complete a process step. An engineer must read the fabrication drawings, specifications (industry and customer), and manually populate the applicable defined fields. Examples are surface finish, solder mask type and color, drill and tolerances, product dimensions, materials and thickness, impedance and SI requirements, and inspection criteria, along with many other product attributes. In a typical scenario, these manual steps occur in quote, then are duplicated in engineering/planning upon order receipt, and in many instances occur once again in CAM. Multiple personnel read the same documentation and manually enter the same information into their respective software applications. The inefficiency of all this could not be more evident.
diagram of downstream processes
Figure 2. By entering all information into a central master database, downstream processes acquire their required data autonomously.
Data entry is tedious and error-prone, so how can we enter data one time only, or even make it semiautonomous? A solution is getting this information entered once at the entry point into the operation and into a central master database. All other related applications downstream in the process then acquire their required information autonomously (FIGURE 2). Reading the fabrication drawing is certainly still a necessity, but we eliminate manual and duplicated data entry, and simple validation checks replace the monotonous administrative data entry tasks.
“OCR exceeded my expectations, and the results amazed me.
Another fascinating approach I have researched the past few years provides the methodology for semiautonomous data entry. The data package arrives to the designated portal; the fabrication drawing(s) is extracted and downloaded into an optical character recognition (OCR) application; and within minutes all critical product descriptions and attributes are extracted with a quick validation by an engineer, exported into an XML file, uploaded to a central master database, and are easily accessible to all software applications in the operation.

OCR has been around for years, and frankly I was reluctant when initially investigating. It exceeded my expectations, and the results amazed me. All data package documentation, specifications (industry and customer), purchase orders and beyond would no longer require an engineer to read, interpret and manually enter in another application. Many companies in multiple industries are adopting this technology to drive efficiency and reduce costs. The potential for utilizing this in electronics industry processes is real and attainable. My next column will provide real-world examples of how this works and its benefits.

The industry as a whole must become more efficient. While we wait patiently for structured data packages to be widely adopted, there are opportunities within our operational environments to drive productivity and improve quality. Taking the first step toward smart engineering is often the most difficult, but the rewards can be bountiful.

Portrait photo of Kent Balius
Kent Balius
is senior consultant at EPIC Front-End Engineering (epicfee.com); kent.balius@epicfee.com. Listen to his PCB Chat podcast on smart engineering at pcbchat.com.