Archaeological Washing Machine
Designed an automated process for washing and drying archaeological finds from the Vedi Fortress site in Armenia; will resume prototyping, building, and testing in the Winter
Overview
The archeological find washing machine is a two-term engineering capstone project that I am completing in a team of six people. Our sponsor, Dr. Peter Cobb, is a Professor at Hong Kong University and a field archeologist who is currently leading a dig at the Vedi Fortress site in Armenia. He presented us the challenge that the manual cleaning of archaeological artifacts is time-consuming, tedious, and increases the risk of human-introduced errors, such as breaking or losing objects. Our task is to design an automated process for the washing of such archeological artifacts at his dig site that minimizes the amount of time it takes to wash and dry artifacts while maintaining artifact organization and preserving sherd integrity.
Design Considerations
Our first step was to conduct interviews with potential users (mainly undergraduates working at the dig site) and stakeholders (expert archeologists and conservators) to understand the challenge, risks, user needs and the solution space better. I personally conducted 3 out of 5 of these interviews. Based on these user interviews, conversations with our sponsor Dr. Peter Cobb and scientific research we have done, we fleshed out our metrics of success for the washing and drying solutions. Our main objectives were to reduce human labor, clean effectively, protect artifacts, clean a range of sherd sizes, maintain sherd organization, dry sherds effectively, and reduce resource utilization.
After understanding what our solution should accomplish, we brainstormed cleaning and drying techniques that could meet the specifications we have listed. For washing, the tools we thought of were various mechanical brushes, ultrasonic cleaning mechanisms, and fluid pressure systems (including air and water). We decided to test the efficiency of the aforementioned techniques.To mimic the conditions we are designing for as closely as possible, we researched the soil composition at the Vedi Fortress site in Armenia and made our own soil sample that was similar to that. We then gathered sherds and broke them up into pieces that roughly had the same size. In order to quantitatively compare them against one another, we came up with standardized testing procedures mainly looking at the percent mass difference of a sherd piece before and after being subjected to the cleaning method in question. We wrote one test report per test performed and ended up with over 10 reports. We found ultrasonic cleaning to outperform the other cleaning methods we tested for. Although ultrasonic cleaning was the winner of our cleaning testing, we wanted to conduct further in-depth ultrasonic cleaning tests to see how the conditions of ultrasonic cleaning can be optimized. We found that ultrasonic cleaners are more effective on one side of the sherd than the other. The solutions we found to be effective in tackling this were flipping the sherds during the cycle, minimal brushing, or incorporating rinse cycles in that respective order.
We also made sure to talk to industry experts and conduct a thorough literature review to find out why ultrasonic cleaning has not previously been utilized for this application. Our research did not yield any concerning information.
We then proposed a preliminary design and workflow solution to our Professors, sponsor, and faculty advisor. They approved our solution for prototyping and building.
Before we started building the device, we came up with contingency plans. Some of our alternative paths included pursuing a brushing solution, rinse cycling, and trying other frequencies, powers, or transducer configurations for the ultrasonsic cleaner. Another de-risking measure we took was to ship a commercially available ultrasonic cleaner to Armenia for them to replicate our testing procedure on site and ensure that it works for their sherds without causing any damage
We simultaneously worked on a solution for the drying portion of the challenge. We tested heat and air flow as drying mechanisms and compared the drying time data collected during these tests to the benchmark tests of ai-drying time. Although heat combined with air flow ended up being the most promising, the concern that heat could damage the fragile architectural finds and result in a loss of information made us eliminate the heat alternative. Thus, we were constrained to searching for energy-efficient air flow alternatives and finally landed on the usage of fans. Our preliminary design sketches proposed building a brand-new product but for ease-of-manufacturing on site in Armenia, our team later decided to purchase and modify off-the-shelf materials.
The second-term of the project was focused on implementing our proposed solution. As we had foreseen, we ran into many road-blocks and ended up making small pivots.
Our initial proposal included a plumbing system used to filter out the sediment and recycle the clean water in the ultrasonic cleaner to streamline the process of running cycles continuously without the need to replace the water repeatedly. However, we found that neither the mesh filter nor the cartridge filter collected an amount of sediment significant relative to the amount put into the cleaner (each filter collected less than 25% of the total sediment). Moreover, we realized that pumps are a critical failure point and it may be difficult to quickly replace a pump at remote excavation sites. To help develop an alternative solution, we revisited which types of sediment we are removing from the cleaner and why. First, we determined that we can split up sediment behavior into two categories: settled or suspended. We realized that when settled sediment covered the transducers of the ultrasonic cleaner, the process of cavitation was impeded. Thus, our system must prevent too much settled sediment from covering the transducers. After many cycles of brainstorming and prototyping, we concluded that siphoning the settled sediment out of the ultrasonic cleaner and into a settling tank would be the easiest workflow. To remove the suspended sediment clinging to the sherds, we also implemented a post-rinse basin. We were able to effectively remove the sediment clinging to the surface of the sherds with a 5 to 10 second rinse in a bath of clean water.
Regarding mechanical flipping halfway through the cleaning cycle, we explored prototypes utilizing cranks or levers that would allow the user to seamlessly flip the sherds with minimal manual input. The integration of mechanical parts that would require maintenance and would be more likely to break during use posed risks towards the feasibility and usability of our system. We then reframed the problem and explored ways to keep the sherds slanted while cleaning in order to minimize the amount of flat surface area for sediment to settle. We were able to address this challenge by designing a slanted stainless steel mesh insert, and we observed that the slanted sherd orientation allowed by our design ensured the effective cleaning of both sides of the sherds to get cleaned.
We simultaneously worked on implementing our drying solution using off-the shelf products. This endeavor featured less challenges, so we spent the bulk of our time optimizing fan size, settings, numbers as well as tray orientation and numbers to make our solution as energy-efficient as possible while keeping the drying time at a minimum.
Insights from thorough testing and prototype iteration led us to our final system design. In broad terms, our final system is made up of three key components: (1) an ultrasonic cleaner, (2) settling tank, and (3) drying rack system. Each of the key system components contain sub-components that contribute to the overall effectiveness of our delivered design. As part of our handoff of the project, we shipped all components of our final design, as well as an extensive manual-style handoff document (that includes information about the materials needed, instructions for assembly, operation, and maintenance, links to 4 different tutorial videos we filmed, troubleshooting tips, product manuals, and an appendix that includes the key findings from our experimentation for future reference) to Armenia.
