A robotics-department reset at a metal-fabrication shop

This entry describes the first of an eighteen-month sequence of kaizens at a metal-fabrication shop in the western United States that makes the steel and aluminium covers, hatches, doors, and grates that sit on top of underground utility vaults. The Robotics Department was the largest single piece of capital on the floor and the binding constraint on what the firm could ship. Orders queued at the robots for almost four weeks; the operators tending the machines were spending more of each shift walking to fetch material than they were spending watching the robots work. The client’s name and the names of individuals have been altered at the client’s request.

The shop’s product is the metal access hardware — hatches, covers, doors, grates — that, fenced from view at street level, carries vehicle and pedestrian load on top of underground utility vaults for thirty years at a time. The Robotics Department welds and finishes the steel and aluminium parts the rest of the plant has cut and formed; it contains the firm’s largest piece of capital and is the constraint on what the firm can ship. The robots themselves were the right machines for the work and were operating well within their rated capacities. The problem the plant manager presented was that the orders waiting for them were waiting for far too long, and that the orders coming out the other side were not coming out fast enough to meet the firm’s growing book of business. Three-and-a-half weeks of order queue is inventory waste in its purest form — work paid for, materials committed, no value added.

On the Tuesday morning of the event the kaizen team mapped the Robotics line in the configuration it had been left in when the equipment was installed: workstations laid out for the convenience of the equipment vendor, materials staged in a small-parts cabinet across the bay from the robots, the Forming Department upstream pushing parts to Robotics in large, infrequent batches that piled up at the line and waited for someone to walk them in.

The Tuesday afternoon was the moment. Derek, the Robotics shift lead, agreed to walk a single shift with a pedometer clipped to his belt. The result, presented at the close of day: one thousand five hundred and thirty-four feet. The operators were not watching the robots. They were walking — to the small-parts cabinet for fasteners, back to the line, across the bay to retrieve material from the Forming staging area, back to the line again. The robots, when their operators were elsewhere, ran out of consumables and stopped. The Lean canon has a name for what Derek was measuring: motion waste, on a continuous-shift basis. The robots stopping for the want of a fastener was waiting waste. Both were symptoms of the same upstream condition. The constraint was not the robots. It was the upstream feed and the operator-material relationship. Three previous attempts to push throughput had read the line at the level of the robots themselves, and had asked the wrong question.

From Wednesday the team walked the cycle back. The redesign had three principal moves.

Pull the layout in. The Robotics line was reorganised so that material came to the operator at the workstation, rather than asking the operator to retrieve it. Workstations were repositioned, the small-parts cabinet was brought to the line, and the floor was redrawn for one-piece flow rather than the batch-and-wait pattern that had developed under the original layout. The transportation waste between the Forming staging area and the Robotics line was eliminated by removing the need for the operator to make the journey at all.

Install Kanban for small parts. The fasteners, fittings, and small consumables the robots needed were pulled out of the across-the-bay cabinet and put on the line in a two-bin Kanban arrangement. An empty bin returned to a holding spot triggers replenishment automatically; the operator never has to leave the station to keep the robot fed. Andy, the senior operator at the second station, designed the bin sizes during the Wednesday afternoon and tested them during Thursday’s run.

Install Kanban between Forming and Robotics. Vic, who ran the Forming Department upstream, agreed during the Wednesday-morning mapping session to feed Robotics in a cadence Robotics could actually use, in exchange for clear forward-visibility of what Robotics was about to need. The mechanism was a wall-mounted shelf of WIP and Complete slots installed at the press brake itself, holding the travellers for the work in progress and the work just completed for each downstream destination — the robots and the manual-weld stations alike. The press-brake operator, in front of the rack, could see at a glance which downstream cell was running short and route the next traveller accordingly. The Forming-to-Robotics handoff, which had been a queue-and-batch arrangement that left some robots overloaded while others sat idle, became a level-loaded pull. Standard work for changeover on the robots was developed during the event by Chase and posted at the line.

Press-brake-side WIP and Complete travellers — the level-loading mechanism for downstream robots and manual-weld stations. — **Press-brake-side WIP and Complete travellers, photographed shortly after installation.** The press-brake operator, standing in front of the rack, could read the state of the downstream cells at a glance and pull the next traveller from the slot most in need of work. Photograph: house archive.

By the close of the five days, the team had drawn a Robotics Department in which the operator stayed at the station, the small parts arrived at the station, the upstream parts arrived at the station in a cadence the robots could consume, and the changeover discipline was written down rather than left in the heads of the senior operators.

The before-and-after, as the team measured it on the closing day:

Order queue at the robots: 3.6 weeks → 4 days. A 75% reduction.
Changeover time: 26 minutes 37 seconds → 6 minutes 54 seconds.
Operator travel during a shift: 1,534 feet → 0 feet.
Robot utilisation: Up 22%, against a 15% target.
Product velocity through the Robotics Department: Doubled.
Standard work: Developed and posted at the line.

None of the changes required new equipment. The robots that had been the binding constraint at the start of the week were the same robots at the end of it, running on the same fasteners, welding the same parts. What had changed was what arrived at the robots, in what cadence, and what the operators did while the robots ran. Masaaki Imai, in Kaizen: The Key to Japan’s Competitive Success (McGraw-Hill, 1986), put the discipline’s operating premise as plainly as anyone has:

The starting point for improvement is to recognize the need. This comes from recognition of a problem. If no problem is recognized, there is no recognition of the need for improvement. Complacency is the archenemy of kaizen.

The corollary — implicit throughout Imai’s book and obvious to every kaizen team the practice has worked with since — is that once a problem has been correctly recognised, the solution tends to follow as a natural reaction to it rather than as the product of clever invention. Three previous attempts at the Robotics line had failed because they had not correctly recognised the problem. The five-day event succeeded because Derek’s pedometer made the right problem visible, and the redesign that followed was scarcely more than a careful response to what the line itself had begun to say. The eighteen months of kaizen work that followed extended the same discipline across the rest of the site.

· · ·