How to avoid assembling the wrong parts

Recommendations on the prevention of picking errors in mixed-flow lines

Version 2, 10/31/98

The full text of this document can be purchased from the MMTI store.


Part confusion in assembly is the #1 cause of defects in manufacturing today, and the magnitude of this problem grows as manufacturers implement more and more mixed-flow assembly lines. There are many low-technology solutions to prevent, make less likely, or quickly detect picking errors but these techniques are not massively applied. There are also high-technology solutions developed for warehouse distribution applications that can be applied on assembly lines. These high-technology solutions fall far short of full automation and are orders of magnitude cheaper.

We work our way backwards from the assembly station to the supplier, and examine means of preventing or reducing the likelihood of errors at each point where they may occur. The only way to make an assembly mistake impossible is to (1) standardize product designs around as many common parts as possible and (2) design each product so that only the right parts will fit in the right place. Most plant managers do not have the authority to do this and must live with product designs they have had no input to. They can, however, modify workholding and part presentation devices to make errors less likely, taking advantage of automatic identification (auto ID) technology based on bar codes, ID squares, or radio-frequency ID (RFID) chips.

At the assembly station, kits of parts for different products are more difficult to confuse than individual parts, particularly if the goes-into product ID is clearly marked both on the kit pallet and on the assembly fixture. When a model comes with a multitude of options, a large options tag tells assemblers how to configure each unit. This visual check can be strengthened by electronic validation of auto IDs on both sides. The mistake-proofing of lineside picking, for parts not included in kits, involves a variety of flip-lids and carrousels programmed to expose the right tray, based on an auto ID read on the product fixture.
Upstream from the assembly station, we also need to prevent kitting errors. In “supermarkets” near the assembly line, we can apply the “pick-to-light” concept and have an LED display under each bin showing the number of pieces needed for the next product unit.

Upstream again is the material delivery system that delivers parts from Receiving and Stores to supermarkets and to the line side. There, mistake-proofing can be addressed through replenishment policies and visual management. “Push” policies make transporters drop parts near the line regardless of whether the storage space intended for these parts is full, which makes sequencing parts on the line side all but impossible. A pull system, on the other hand, ensures that parts are delivered when the line is ready for them, and first-in-first-out can be enforced.

For visual management, plant stores should be benchmarked against a neighborhood supermarket. These grocery stores are laid out for customer convenience, and new customers find thousands of products with no training and without mistakes. Products are identified by mnemonic names rather than 15-digit item numbers, and all markings and labels are kept up to date.

Markings in stores should be designed to make picking easy and error-free. Labels should be at the right height and orientation for the drivers of forklifts or marshalling trains to read without turning their heads of taking their eyes off the road. Formal item numbers should be used for pick validation using auto ID rather than as retrieval keys for people. Where people have to read item numbers, font and size should be used to emphasize the differences between similar items.
None of these approaches will prevent a supplier mislabeling parts, unless he uses them as well. Once the user organization has achieved success in preventing errors in-house, it can invite suppliers to see it and be inspired to implement their own versions.