Improving how lawn mowers are made at John Deere

[In photo above, Carolyn Elliott stands in front of the John Deere sign at their facility in Greeneville, Tennessee. Through an innovation that Carolyn and her fellow intern developed while at the company, a process which once took two weeks to perform was reduced to two days.]

Carolyn Elliott served as an intern at Vitech in the summer of 2017. Elliott is currently a junior at Virginia Tech, majoring in Industrial and Systems Engineering and planning to graduate in May of 2019.

Not every engineering undergrad gets a chance to help improve a process at a major, well-regarded company before they’re even out of college. I guess I was lucky. From June, 2016 to June, 2017, I worked as a co-op at John Deere in Greeneville, Tennessee. My job was to help in streamlining the assembly line for making lawn mowers.

This was a great opportunity for me to learn about manufacturing engineering, a field I’m considering entering after I complete my degree. Although manufacturing engineering is different from systems engineering in many ways, there are also instances in which the disciplines overlap. This is a story about how I worked with engineers and production line associates to apply model-based systems engineering (MBSE) to a manufacturing problem.

The John Deere plant in Greeneville manufactured lawn mowers—to be specific, 60 different kinds of them. Model changes occurred several times per day on the assembly line when production switched from one style of lawn mower to another. Although the lawn mowers shared many common parts, there were also many parts or combinations of parts unique to each model. To help manage all this, “changeover sheets” were born. Each sheet displayed a list of the key components in the order of assembly for a particular model. This sheet traveled down the assembly line with the first mower after a model change to ensure that the right parts were put on the mower.

Prior to my project, the changeover sheets were produced manually every year. Each changeover sheet contained a list of about 50 parts, and there was a changeover sheet for each of the approximately 60 models produced at the facility. If you do the math, that’s 3,000 parts that a group of manufacturing engineers needed to research in a database, copy into an Excel workbook, and then format for ease of use on the assembly line. As you can imagine, doing this by hand was a tedious, time-consuming, and error-prone method.

The key to improving this process was MBSE. I, along with one other co-op, was tasked with improving the system of creating the changeover sheets. The goal was to save time, money, and manpower, all while improving accuracy. This required us to reverse engineer the existing system.

Once we had completed that, we turned to designing a new system, starting by addressing the first domain of MBSE, the requirements. For our new system to be implemented, we needed to accommodate our customers—that is, the manufacturing engineers and production line associates. The engineers needed to be comfortable with the format and software used to create future changeover sheets, and the associates on the assembly line had to be able to gather the information quickly and without additional training. Lastly, we also needed to improve accuracy while reducing time spent creating the changeover sheets.

These requirements drove the behavior—what our system was going to do. During the first iteration of our layered approach to MBSE, we did not make many changes to the behavior of the system. The bill of materials in the database provided the part names and numbers to the changeover sheet. The changeover sheet provided a concise list of components to the associates assembling the mowers. Although the functions of the system remained relatively the same, how these behaviors would be accomplished needed a serious update.

To this end, we tackled the third domain of MBSE, the physical architecture. During this step, we determined how we were going to accomplish the necessary behaviors. Keeping the requirements in mind, we chose Excel to create the changeover sheets. The manufacturing engineers and associates on the assembly line were already familiar with Excel, thus making the transition to the new changeover sheets much simpler. Through the use of formulas and macros, the workbook would import the bill of materials from the database, search for the key components of each model in the bill of materials, and finally pull the part numbers onto the corresponding changeover sheet.

At this point, we were ready for the final step in the MBSE process—verification and validation. Unfortunately, verification failed during our first iteration of the process. The parts had been given different names on the bill of materials (which was created by design engineers at a different facility) and on the existing changeover sheets (which were created by our manufacturing engineers). We needed to maintain the existing names on the changeover sheets, since these were the names with which the associates on the assembly line were familiar.

It was apparent that we needed some sort of electronic translator to create an association between the two names for the same part. This brought us back to the behavior step in the MBSE process. After some work, we had the bill of materials in the database provide the part names and numbers to a translator we had developed, which then read the colloquial part names on the changeover sheet and output the correct part number back onto the changeover sheet. The changeover sheet then presented a concise list of key components for the associate assembling the mowers.

The physical architecture was then modified to reflect this change in behavior. A new sheet was added to the workbook to serve as the translator. This changed the architecture and added an intermediate step between pulling parts from the bill of materials and inputting them onto the changeover sheets.

Finally, this brought us back to the verification and validation step. Alas, during this iteration, validation again failed. We did not meet one of our requirements—improving accuracy. Our system still required a high level of human interaction, which allowed errors to occur. We solved this problem during the next iteration by creating a control panel that allowed for user input while protecting the other sheets from editing. This created a sleeker user interface that was much simpler to use, as well as making the process less error-prone by limiting direct interaction with the data.

By implementing a layered approach to problem solving, which required several iterations of the MBSE process, we were able to identify the big problems early on, saving time in the long run. In the end, we satisfied the requirements and successfully implemented a solution. The changeover sheets now take about two days to create instead of two weeks. And, the new changeover sheets have actually helped identify errors in the bill of materials—an unforeseen benefit and a testament to the accuracy of the new system. I hope that this example has demonstrated how systems engineering and MBSE can provide a method to quickly and effectively improve almost any complex process. For myself, I was pleased to see that our efforts had made a lasting impact. I was proud to know that even in my short time at the company, I had been able to help improve how something was done.

Leave a Reply