When Automation Results in Lower Overall Performance

When Automation Results in Lower Overall Performance

Machine Designers Sometimes Build Great Labor Reducing Machines That Slow Down Assembly Lines

Many times, automation design is conducted in a limited way by a company that focuses on design and fabrication of machinery. They do so to add features and benefits that reduce or eliminate direct labor for previous manual operations. This enables the machine manufacturer to sell more machinery and in and of itself is an important strategy for the machinery company. If the machine is stand alone this is typically great for the customer. If the machinery is part of a production line, however, this may result in a decrease in overall production rates despite the localized improvement of this particular operation.

Over-Automating (Too Much at Once)

Elon Musk of Tesla has recently stated that current inability of Tesla the promised production levels for the Model 3 at Fremont is the result of excessive automation. On April 17th, production was suspended to “improve automation and systematically address bottlenecks to increase production rates.” This issue occurs from time to time when a company attempts to achieve too much automation at one time. This article is not about the complexity of integrating several different systems and their short-term issues. Tesla will sort all of this out and achieve their objectives over time. This article is more about the permanent downside of looking at automation on a localized basis rather looking at the overall system effects.

When a Machine Is Designed to Perform Too Many Tasks

On several occasions, I have reduced the number of tasks that a machine performs. This has been when the tasks have been performed in a time sequence rather than simultaneously. The example that comes to mind is for a large vinyl window company. When a sash is machined (the sash is the top part of the lower window upon which a lock is mounted), it requires drilling of several holes to allow for mounting of the lock mechanism and the routing of the ends of the sash to allow for the installation of shock absorber mechanisms on both ends. Previously, these were done manually by operators positioning the sash and operating a drill press. The automated machine would position the drill automatically for the two holes that needed to be drilled and this was quicker than an operator manually positioning the sash, hence a significant reduction in time, but more significantly, the elimination of an operator. The routing operation had been previously done the same way by an operator positioning the material and then operating the router to create an opening at each end. The automated machine also performed this function and eliminated an operator as designed. However, instead of these tasks being simultaneous, they were sequential. The result is that while two operators were eliminated, the overall theoretical assembly line speed was slowed down by 20%. This caused no concern at the time the machine was designed and installed and fully eliminated the two positions as intended. The issue was identified several years later when I was hired and started a project to dramatically increase production rates throughout the manufacturing plant.
In reviewing metrics, the only two measures were headcount on the production line and the total quantity of windows produced. In looking in to how to increase the production rate, an analysis of machine by machine and operator by operator cycle time was conducted. All other individual tasks could be performed by operators much more quickly than this automated machine. The overall reason that the company did not see this as an issue in that productivity was not measured, production supervisors were not held accountable, and the company had been growing at double digit rates in both revenues and profits. When the production teams and supervisors were measured against theoretical line performance and rewarded with celebrations in the form of pizza for setting new records, the obstacle to even further improvement became obvious as it was the limiting factor. Most individual tasks could be performed in much shorter time with the time required for the longest of these tasks still being 30% less than the cycle time of the automated machine. If we could identify a way to reduce the cycle time of this machine, then we would be able to increase line throughput by 42% until this next constraint would restrict the rate of production.
Since the two functions of the machine were performed linearly and the latter could be easily replaced by a manually operated machine punch and by adding an operator, the second function was disabled. On the surface, this looks like a bad idea in that the number of operators on the line was increased by one. However, the number of operators was increased from a total of 11 to 12 or an increase of 9.1% in labor. Overall output did not increase by 42% due to other disturbances to the production line, however, it did increase by an average of 30%. Labor productivity then went from 250 units per shift divided by 11 operators or 22.7 units per operator to 325 units per shift divided by 12 operators or 27.1 units per operator for an overall labor content decrease of 19.3%. With this small change plus posting of key metrics and holding the team and supervisor accountable for results, in-place capacity increased by 30% per shift. Positive secondary effects included a reduction of routine overtime expense.
The machine designer had looked only at the reduction in headcount that was possible. The company was somewhat unsophisticated and lacking metrics but for several years benefited greatly by installing the automation. However, with addition of performance expectations and looking at the overall production line as a complete system, greater productivity was experienced by reducing the level of automation.

When Automation Results in Lower Performance Due to Set Ups

Many times, automation achieves its objectives by reducing labor and sometimes by increasing line speed. The downside is when time is required to change the machine from running one item to a different one. Sometimes that time is minimal like for a color change. When sizes change, greater complexity occurs. Many companies will focus on Single Minute Exchange of Dies (SMED) to reduce drastically the changeover time and should be complemented for doing so. Other times, the changeover time is reduced to the least time possible, but the complexity of the changeover and the frequency of changeovers due to requirements for a high number of shorter run products greatly reduces production output.
Automation works best for high volume production runs. In the world of frequent changeovers, consideration should be given either to reducing the product mix (which Sales and Customers will resist) or removing some of the products from the production line and building them in an off-line or manual operation. Unfortunately, many managers resist this and exert ever greater effort to increase the efficiency and output of the automated production line when the simpler solution is to produce off-line or even outsource the production of the lower volume products. The result of either of these efforts is a higher cost per unit for these products. However, the benefit is a lower cost per unit of the products produced on the automated line. The management team should evaluate the impact on total production cost for all products. Many times, it is a much lower cost to implement one of these strategies.
If there are several automated production lines, then dedicated one or more to higher volume products or to a size range may also reduce the complexity and result in higher throughput by not having every line produce every type of product.

Looking Ahead

Over the last few years, there have been a plethora of articles concerning the return of manufacturing to the United States. In nearly every article, there is the premise that automation will be the principal reason that will allow this to occur. I strongly believe in automation and in necessity to offer customer greater choice through customization. In addition to the above considerations that sometimes limit output by attempting to eliminate too many manual tasks or to automate every aspect of manufacturing, there remains a work force limitation. The factory worker that has successfully performed in a repetitive manufacturing environment for years and the skilled tradesman may not be trained and indeed may be incapable of learning the new skills such as electronics or programming that may be required to manufacture the same product in an automated environment. Some planning is required to ensure that the equipment does not outstrip the capability to staff and operate the new machinery.
Deploying lean manufacturing or the Theory of Constraints would be recommended in most operations. However, simple observation of where inventory builds up on a production line or where there is somewhat of a vacuum is often enough to facilitate the discovery of major production limitations. These disciplines would be of great value in assessing the potential benefit of additional automation.