Roger Richardson | 03.15.16
Today we will look at some examples of applying Augmented Reality (AR) to real-world assembly applications. In this paper, our focus is on assemblies that we consider “complex”. That is, they take some hours to complete and require a person with significant skill to perform the tasks. Assemblies that are serialized to the point that each task only takes a few seconds or minutes, and is thus sufficiently simple that years of experience is not a relevant factor in the quality of the finished product, are not the kinds of assembly tasks discussed here. The multipliers that are used in this paper are based on testing conducted by DSC and reported in a paper entitled “The Effects of AUGMENTED REALITY on HUMAN PERFORMANCE”.
The first factor to consider is the proportion of the work day spent doing value-added tasks versus non-value-added tasks. How AR may be applied to these task types varies, and how much it may improve the task varies. However, the value-added tasks are much more consistent in regards to the effect of AR on the process. Although, these are fairly generic terms, we will start with a definition of value-added and non-value added tasks as used in our context:
Again, it is not that NVA tasks are non-essential, it is just that they are not categorized the same as VA work, and the best ways to optimize VA and NVA work are typically not the same. Dealing with both types of tasks – things that touch the assembly and things that do not – independently will provide the best outcome with the least effort regarding the efficiency of the end process.
Our initial focus will be limited to value-added work, and we will focus on the three key characteristics needed to create a highly efficient assembly process for complex objects:
In the opening paragraph of this paper, we referenced another paper which describes the effects of AR on human performance. We will not reiterate the contents of that document here; it is available at the link provided for those who want to understand the source of the values below. Those values are:
Let’s start by talking about a key benefit of AR – the ability to bring a very high level of expertise to a relative novice. When the AR environment has the expertise built-in, it requires less expertise of the user. This does not take away from the craftsmanship of the user – in fact, quite the opposite. A craftsman is not really applying his trade by reading drawings or manuals – craftsmanship is applied by “doing it”. If the AR system is the container for “where and when” in the build sequence, the craftsman is still the possessor of the “how” (how to drill a straight hole, how to apply sealant to a panel or fastener, how to torque a fastener, etc.). If, in an instant, everyone knew equally well “when and where”, then the focus becomes exclusively on “how”, and that can and will directly improve the quality of the craftsmanship.
As stated previously, the standard deviation is about 1/3 that of traditional assembly techniques. Suppose we had a task that had an average build-time of 80 hours, but ranged between 48 and 150 hours, depending on who built it. Across 85 assemblies we determined a standard deviation of 22.6 hours. With over a half workweek of standard variance and almost 3 weeks of potential variance, keeping an assembly line balanced can be very difficult when all of the key players are not in place every day. In real life, as we well know, they are not. But, if an AR system is providing the “when and where” information, and a craftsman that already knows “how” can step into a work station and instantly perform at the same level as the person that regularly works at this station – then that time variance drops drastically, in fact it drops from over 22 hours to about 7 hours.
Here is what that would look like in an AR environment:
Diamonds represent temporary fasteners (Cleco’s)
Circles represent fasteners to be installed
Bands around the perimeter of a circle represent a process to be applied
The assembly technician sequences through the steps of the task as the AR system delivers the “when and where” information to exactly the right place at exactly the right time. The technician needs to know how to insert and tighten a Cleco, how to apply the right amount of sealant to the fastener, and how to set and use the torque wrench. But significantly, the technician does not need to know when or where to do any of these things – that information is provided at the moment of need – so, the technician can freely move between work stations working at maximum effectiveness in all cases. The net of all of this is that standard deviation is reduced by about 67% compared to people reading drawings and manuals to do the exact same work.
Build-time and error rates need to be considered together because of their inherent relationship. It is natural to assume that going faster results in lower direct build-costs, but also means more errors, and errors mean rework, and rework costs money. So, there is a trade-off where the maximum cost benefit is achieved by balancing the two. Using an AR system we are going to see that our technicians are going to easily be able to produce much more product with less effort than they expend today using traditional assembly techniques, and at the same time come pretty close to eliminating errors entirely.
Here we will look at the cost implications related to the application of AR to VA tasks of complex assemblies.
CASE 1: We will first look at an aircraft wing assembly. The single wing surface is 60 feet (18.3m) long. There are three assembly technicians that work in this station full time. Each of them makes, on average, three errors per week that each add $500 in rework costs.
The facility has 10 holidays per year where the plant is closed, so there are 50 workweeks available for production. In our example, we will use a fully burdened labor rate of $100 per hour, and a base workweek of 40 hours. During the course of a week, the technicians will have 2 breaks of 15 minutes each every day (-2.5 hours of productivity), a 20 minute meeting each start of shift to plan the day (-1.5 hours), 25 hours per year for training of various kinds (-0.5 hrs per week), 30 minutes of clean-up at the end of shift each day (-2.5 hours). Factoring in the two weeks the plant is closed, this leaves an average of only 31.73 hours per week of production availability. Now we are going to estimate that of the 31.73 hours where productivity actually happens, 40% of that time is spent doing non-value-added tasks, leaving only 19.04 hours per week of available time for value-added work.
In our ROI calculations, we will only use the 19.04 hours as the available time during which we can take gains in productivity for the type of AR system under discussion. The 8+ hours of non-productive time and 12+ hours of NVA time, will remain at the status quo for this discussion.
Next, we will look at an approximation of the costs to implement an AR system on this one wing surface. In this case we have 1 controller, 6 projection heads, and $35,000 in other start-up costs, for a total of $177,000. So, that concludes our inputs.
Now, let’s review the outputs. Just in the 19 hours per week where we get to apply the savings generated by AR, we will net out $79,519 per person in direct build time, and another $58,344 in rework savings. We have 3 people in this station, which will result in $413,589 per year in total cost savings, generated from this $177,000 investment. The key numbers here are:
Full Cost Recovery: 5.14 months
For an investment of: $177,000
1 Year net ROI: $236,589 134%
5 Year net ROI: $1,890,945 1,068%
10 Year net ROI: $3,958,891 2,237%
CASE 2: So, the chart above is to build one wing surface – what if we expand the numbers to build all 4 surfaces:
Full Cost Recovery: 4.56 months ($1,654,356/year cost reduction)
For an investment of: $628,000
1 Year net ROI: $1,026,356 163%
5 Year net ROI: $7,643,782 1,217%
10 Year net ROI: $15,915,564 2,534%
If any of the following statements are true, then your savings will be greater than shown:
CASE 3: Next we will look at a much smaller application – one that can be done using a single projector and only uses one technician to build the assembly. Also, we are now considering a simpler process so our technician now only makes one error per week. All of the costs and time take-aways are the same. So now we get:
Full Cost Recovery: 6.39 months ($103,335 per year cost reduction)
For an investment of: $55,000
1 Year net ROI: $48,335 88%
5 Year net ROI: $461,675 839%
10 Year net ROI: $978,350 1,779%
These are realistic numbers. This is for a system that has minimal or no adverse impact to your existing production line during installation. This is for a system that poses no hazards to your product during the start-up phase. This is for a system that can often be installed and providing these results to your operation in about 8 weeks. These gains come from a system that makes people more efficient via human performance enhancement. But, it is the same people performing the same tasks, using the same tools the same way: thus, there are no process changes that affect regulatory compliance.
The best articles, conveniently delivered to your inbox once a quarter.