Automated Inspection as a Poka-Yoke

How a novel approach cut takt time from 2 minutes to 2 seconds.

Poka-yoke is a Japanese term meaning mistake proofing. It was introduced by Shigeo Shingo as part of the Toyota Production System (TPS) and refers to simple fixes that address small variances in the manufacturing process.

Even production operations that focus on best practices in new product introduction (NPI) require poka-yokes. For example, SigmaTron International’s Tijuana, Mexico facility applies Advanced Product Quality Planning (APQP) and conducts failure mode and effects analysis (FMEA) to establish the most efficient, defect-free process. The facility utilizes a product part approval process (PPAP) for automotive projects to validate the process, while customer-specific validation processes apply to projects in other industries. Once production begins, the team monitors processes and tracks defects using statistical process control (SPC) and other forms of quality data collection. When defects occur, the team schedules a kaizen event, employing tools such as 8D problem solving, Six Sigma’s define-measure-analyze-improve-control (DMAIC) and poka-yoke to analyze and correct the root cause.

In this example, the issue occurred on an automotive product with four different models of right-left printed circuit board assemblies with multiple color wires. During a portion of the production operation, operators manually placed wires, but defects arose when they inverted wires or mounted the wrong color wire. An electrical test at the end of the production line successfully caught defects; however, operators spent from 20 seconds to 2 minutes manually inspecting and debugging failed assemblies.

A Gemba walk revealed this issue needed improvement, and the test engineering team felt a poka-yoke could address the problem. The challenge was designing a simple, cost-effective fix. The automated visual inspection systems utilized in other areas of the facility offered more functions than necessary and were cost-prohibitive. The test engineering team determined that only color and object detection were required. This enabled it to significantly reduce the cost of visual inspection by utilizing web cameras. While these cameras lacked the necessary embedded functionality, their USB interface enabled porting to a serial interface board. The team utilized C++ programming language to create inspection libraries that compared images captured via webcam with known-good images.

The prototype inspection system consists of two independent cameras placed over the PCBAs. If the system detects good PCBAs, the screen turns green, and the conveyor moves the PCBA to the next process. If the system identifies an issue with a PCBA, the screen shows a red square in the position with the NG (no good) reference, and the system moves the PCBA to a repair queue.

The takt time for the inspection is now under two seconds. The current system has been approved by the QA department for inspection of the eight PCBA configurations. The system is undergoing validation testing for propagation to other programs. During this phase, it is standalone. Following validation, the team will link the system to the proprietary manufacturing execution system (MES) to integrate inspection results into quality trends analysis.

Poka-yoke focuses on three types of solutions:

• Contact type – Eliminate defects caused by variation by making variation impossible.
• Constant number/visual management type – An error signal is triggered if a specific number of movements, or steps, are not completed.
• Performance sequence/failsafe – If steps are not completed in a specific order, the operator is locked out.

The aforementioned example represents the contact type of poka-yoke.

This example also highlights the value of analyzing readily available consumer technology for poka-yoke purposes. High-volume manual assembly operations experience variation that drives defects because production operators are not machines. Manual inspection operations also incur high costs and have inherent variation. While inspection never serves as the best solution, in the electronics manufacturing services (EMS) environment, where OEM customers typically dictate design and processes, it often becomes a necessary non-value-added activity. When manual assembly operations cannot automate, proprietary automated inspection systems that use low-cost hardware may be the best solution for ensuring defects do not escape the factory.

Developing a proprietary system requires test engineering and programming resources. Once the team validates the system, it becomes a poka-yoke in similar processes throughout the company. The approach improves customer satisfaction and reduces costs associated with returning defective products.

Alan Salmeron is a test engineer in SigmaTron International’s Tijuana, Mexico facility (sigmatronintl.com); alan.salmeron@sigmatronintl.com.