Optical Coffee Bean Sorting System with Robotic Arm

I. Introduction

Our optical coffee bean sorting system integrates computer vision, machine learning, and a robotic arm to isolate contaminants, serving as a case study for the broader engineering trend of low-cost rapid prototyping. PLA fused deposition modeling was selected to build the physical prototype because it functions as an iterative design methodology rather than a mere material choice.

This approach allowed the team to rapidly adjust mechanical parameters, correct camera regions of interest, and resolve lighting reflections caused by part coloration. However, the project exposes a critical contemporary engineering issue: the technical gap between a successful functional prototype and a scalable commercial machine. While FDM excels at fast design verification, its inherent surface roughness and layer lines present significant cleanability and hygiene challenges for food-contact equipment [1].

II. Contemporary Engineering Trend

One contemporary engineering trend related to this project is the growing use of rapid prototyping through low-cost additive manufacturing. Traditional machining, molding, or outsourced fabrication can produce strong parts, but these methods are slow and expensive when a design is still changing.

Modern 3D printing changes this process by making it easier for engineers to move from a digital model to a physical part in a much shorter amount of time. FDM printing is especially useful in early development because it has low tooling cost, short production time, and high design flexibility. PLA is commonly used because it is inexpensive, easy to print, and widely available [2]. Its low melting temperature, about 150-180 degrees C, supports layer adhesion for intricate designs [3].

III. Connection to the Project

This trend directly applied to the optical coffee bean sorting system because the physical design required multiple iterations before it worked reliably. The project did not only depend on the camera, machine learning model, or software. It also depended on whether the beans and rocks moved through the machine in a controlled and predictable way.

One important example was the flipper mechanism. The flipper had to be large enough to redirect a rejected object, but not so large that it blocked the normal flow of coffee beans. With 3D printing, the team could adjust the flap size, print a new version, install it, and test whether movement improved.

The base of the system also required iteration because it affected the camera's region of interest. The vision system needed a consistent viewing area where the bean or rock could be clearly detected. Part color also mattered: black printed material reflected light in a way that caused camera errors and affected classification reliability. Rapid prototyping therefore helped test how physical materials interacted with sensors, lighting, motion, and the environment.

IV. Challenges, Responsibilities, and Implications

Although 3D printing was useful for prototyping, the same design method creates challenges when the system is considered as a possible commercial food-related product. In industrial food systems, scalability and economic viability remain primary concerns because current additive manufacturing methods are significantly slower and less efficient than traditional food-processing methods [4].

A part that works well in a student prototype is not automatically suitable for long-term use around coffee beans. Standard FDM parts are built layer by layer, creating layer lines, small gaps, surface roughness, and possible spaces where dust, coffee residue, or microorganisms could remain after cleaning. Because coffee beans move through the printed parts, the bean path would have to be treated as a food-contact area in a real product.

The larger implication is that rapid prototyping can make engineering faster, but it can also create a false sense that a working prototype is close to a finished product. The project showed that the prototype could be improved quickly through PLA printing, but manufacturing decisions become more serious when the device is used around food.

V. Lifelong Learning

This project helped the team understand that a working prototype is not the same as a finished product. Initially, material choice was mainly about cost, strength, and whether the part was easy to print. Because the project involves coffee beans, the team realized that material choice also includes cleanability, surface finish, durability, and food-contact safety.

The project also showed that engineering areas are connected. The software and camera system depended on the physical machine. The flipper size affected sorting, the base affected the camera's region of interest, and printed-part color affected lighting and classification.

VI. Conclusion

Rapid prototyping with 3D printing was important to the optical coffee bean sorting system because it allowed the team to test and revise parts quickly. PLA was useful for the prototype because it was cheap, easy to print, and allowed multiple design iterations for the flipper, base, and other parts.

The project also showed that a prototype material is not always suitable for a final product. If the system were used commercially, the bean-contact parts would need to be more durable, cleanable, and appropriate for food-related use. The main lesson is that 3D printing is a powerful tool for prototyping, but engineers must understand when a design needs better materials and manufacturing methods before it becomes a real product.

References

1. T. Hanisch et al., "Influence of surface texture on the cleanability of 3D-printed stainless-steel components," Heat and Mass Transfer, vol. 61, Art. no. 84, Aug. 2025.
2. J. M. Chacon, M. A. Caminero, E. Garcia-Plaza, and P. J. Nunez, "Additive manufacturing of PLA structures using fused deposition modelling," Materials & Design, vol. 124, pp. 143-157, Jun. 2017.
3. M. A. Cuiffo et al., "Impact of the fused deposition (FDM) printing process on polylactic acid (PLA) chemistry and structure," Applied Sciences, vol. 7, no. 6, Art. no. 579, Jun. 2017.
4. E. C. A. Trindade, M. Ouichka, O. Rayan, J.-S. Deschenes, and J. Brousseau, "Evaluation of the chemical resistance of selective laser-sintered PA12 pieces for food industry machinery applications," Progress in Additive Manufacturing, vol. 10, pp. 11697-11709, Aug. 2025.
5. K. Muchtar et al., "Edge AI-based detection for defective coffee beans using deep learning and Streamlit framework," IEEE Access, vol. 13, pp. 67977-67992, 2025.