The Working Anthroengineer

Automation has made it easier than ever before to collect large amounts of data, and with the advent of Industry 4.0, this data has become a major player in the manufacturing industry. Data collected from machines and sensors can help manufacturers better understand their processes, reduce downtime, and get a real-time look at efficiency. Although this data is invaluable, it is still limited and requires us to integrate human insight and data in order to get the full picture. However, qualitative data, especially in relation to all the objective quantitative data, can appear quite messy and its usefulness is often overlooked. Leveraging the ubiquitous hour-by-hour production control chart and giving it a meaningful language can help organizations share a richer accounting of how equipment and processes perform.

Data Visualization & Storytelling

Data visualization allows for complex stories to be told quickly with minimal effort on the part of the reader. By creating dynamic visuals that depict relevant information in an easily digestible way, data visualizations can help manufacturers make sense of large amounts of data and tell compelling stories about their processes efficiency and equipment performance. For example, a company may create visuals that demonstrate how many units flow through a machine in a given day and how much time a unit spends at a specific point in the process or any errors that may have arisen throughout a shift. This provides an easy way to understand the scope of an entire manufacturing day without having to wade through large sets of numbers. Although incredibly powerful, this still doesn’t give a full accounting of some factors which we may deem critical, such as an undetectable or unmeasured quality issue that might have occurred during the day. It is important to be aware that not all problems are easily predicted in the design of automated equipment and some sensors simply cannot be trained for every possible combination of anticipated or unanticipated problems.

The Need For Human Insight

In order for us to make use of all the data collected from automated manufacturing processes, we need to supplement it with human insight. This means gathering additional information from workers so that we can understand their experiences with our processes better to make improvements accordingly. Additionally, having experts on staff who can review the collected data can provide valuable insight into areas where improvements could be made or changes should be implemented—for example, finding ways to reduce waste or improve quality control monitoring in the production process. It is particularly useful to look at this human collected data as a form of community memory (also: collective memory) as discussed by Julian E. Orr in “Talking About Machines: An Ethnography of a Modern Job.” This perspective recognizes that systems of workers interact with machines and processes and while doing so collect an incredibly rich amount of data. In Orr’s example, he describes community memory as something “in which they [repair technicians] preserve and circulate their hard-won knowledge of machine arcana.” And he goes on to say that “the technician’s own memory of these stories [is not] all that is available; other technicians called for purposes of consultation will bring their own recollections to bear.” It is easy to imagine how invaluable this data and community memory is for both troubleshooting and for giving a thorough accounting for a days production gaps and inefficiencies. The problem that naturally arises is how exactly we may collect this qualitative data in a way that is concise, useful, and aligns to automated data collection.

Restructuring the Production Control Chart

One of the most used tools in manufacturing to document performance has been the production control chart or the hour-by-hour by chart. As the name suggests, it is meant to track the hourly progress of a manufacturing line against a predetermined hourly output rate. When the output goal is not met, for any kind of reason, workers on the line or a line-lead are expected to write why in a blank box next to the output. For example, if a wire at an automated test station is damaged and maintenance is required for the equipment to continue working a technician may write, “15 minutes downtime for maintenance to replace wire.” Now imagine later in the shift, the same issue occurs and causes a large string of test failures but someone only writes, “maintenance issue.”

Hour	Rate	Actual	Notes
7:00am-8:00am	60 Units	44 Units	15 minutes downtime for maintenance to replace wire
8:00am- 9:00am	60 Units	40 Units	maintenance issue

What this highlights is a problem of non-standardized and disparate data, a problem which is all too real. Although we may ask this worker what the actual quality issue was the next day, this poses a systemic inefficiency and it is difficult to align this with the concise and direct data collected by the tester which may just show a run of failing units. In this system then, one can see how the test data collected automatically by the station could very easily be supplemented by collective memory if it was collected well. A particular solution I have proposed is the restructuring of the chart which prompts an expected level of information and detail. If an organization is highly concerned with test data, downtime, then at any point in the process via either machine or person, this data collection should be streamlined.

Hour	Rate	Actual	Downtime (mins)	Tester	Cause	Resolution
7:00am- 8:00am	6o Units	44 Units	15	#4	machine: broken wire	called maintenance to replace

In this chart, it is clear to see how the amount of downtime, its cause, and subsequent resolution at a specific tester is something the organization deems necessary to collect. With sufficient training and clear categories such as “machine vs. part quality,” this rich data which was not be collected by the equipment can be systematically used to support automated data.

Reflecting, Strategizing, and Moving Forward

Anecdotally, this kind of problem is something I have faced before when reviewing test data. Even though I have been able to show a string of test failures from weeks prior, it has required much more time to figure out or remember exactly what caused it by finding someone who may have been present and does remember. This kind of problem is what caused me to recognize the gap in my organization’s automated data and to see the value of the collective memory of technicians and other engineers in my workplace. In the interim, it has allowed my team to rethink old tools and manual process to better streamline our qualitative data and pair it with quantitative data to overcome a uniquely modern problem.

Moving forward, as we think about automation and the new kinds of problems we are bound to face in manufacturing, it is necessary to understand that these environments stress the people/technology and culture/technology relationships outlined in sociotechnical theory. So even though there is more novel technology being implemented, it is crucial to remember the importance of all the data us people knowingly (and unknowingly) collect as well as how organizational culture impacts our relationship to this technology. At the end of the day, people currently have one of the largest impacts on how well automation runs and how well the stories about it are told.