Automatic Trash Sorter

This project was completed during my junior year for the ‘Product Design’ course. The aim of the project was to create a product that could automate the sorting process of recyclable plastic, glass, and aluminium containers. The project arose as current, human-based sorting causes the contamination of waste, thus making it unrecyclable, so an automated approach would be beneficial to reduce this issue.

Our four-person group composed of myself, Emily Wilczewski, Victoria Thomas, and Tess Ravick decided to achieve this goal by implementing the use of various sensors alongside a four-bar linkage system and belt and pulley mechanism to sort the waste.

I primarily worked on the material identification part of the project, this being sensor usage and code. Inductive proximity sensors were used to detect the aluminium waste, while load cell sensors were used to differentiate between plastic and glass, all coded using the Arduino IDE and Arduino UNO boards. In addition, the motors for the mechanical system were also controlled using an Arduino UNO board, having a connection between the sensor and motor boards to share data.

Although our project was successful in recognising the desired materials 92% of the time (tested 100 times), it did not always sort it properly due to issues with the four-bar mechanism as parts would slip through. In addition to this, the load cells were quite sensitive and sometimes would produce erroneous results due to inadequate calibration.

The written report for this project can be found here, and the presentation can be found here.

Product Goals and Design

The main goals our team aimed to achieve during the semester were as follows:

Material Sensing

The critical uncertainty for this product was its ability to differentiate correctly between the materials. To do this, the team decided to use a variety of sensors. We decided aluminium and metal containers would be the easiest to differentiate from the other waste as they have distinct conductive properties. We then decided to differentiate between glass and plastic through weight, plastic 2 fluid ounce bottle weighs 0.023 lbs while a glass one weighs 0.177 lbs.

We decided to use inductive proximity sensors to check for metal, and a load-cell setup to measure weight and check for plastic and glass. In our code, we tested for metal first to prevent erroneous sorting due to aluminium cans being so lightweight. The inductive proximity sensors we used only worked if the material is within a range of about 8 mm, so to ensure detection we used multiple within the chute.

Material Sorting

To sort the materials, the team divided a trash can into three sections and used a four-bar mechanism along with a belt-pulley system to sort the waste into these. The four-bar mechanism worked to hold the waste in place during the material detection phase and then release it once the belt-pulley mechanism rotated the trash into the correct position.

The motors used for the four-bar and belt-pulley were controlled using an Arduino board, as were the sensors. The sensors were connected to one board while the motors were connected to a separate one, and an I2C protocol was used to connect both and transmit data. The sensor board was the ‘Master’ and the motor board was the ‘Slave.’

Product Demonstration

Project Takeaways

Assessment

Although the sensors were able to work at least 80% accuracy to detect the material, sorting was an issue due to the conjoined setup of the mechanical and sensor systems. The Arduino boards were also able to successfully control the sensors and the motors, but issues arose when these were positioned in their final locations.

Issues with the setup of the sensors on the four-bar linkage caused issues with the material detection. The load cells were particularly difficult to manage as changes in the angle of the four-bar linkage required the team to recalibrate the sensors often to work in new orientations, as it was not at the desired 90º the system was designed at. Due to the issues with sensing and the usage of the four-bar linkage, the system did not consistently achieve the 30 second or less requirement for this item to be considered fully achieved, especially when measuring the weight of the waste.

Future Improvements

Our main sources of error were the mounting of our individual components in addition to the materials used. The four-bar linkage was made using K’NEX, a construction toy system. This was due to our limitations in accessing workshops due to COVID-19, but improving the materials used on the system would be beneficial in terms of improving the sturdiness of the system. The K’NEX bent in ways we did not expect and lacked the strength to hold itself and the load-cell setup up, thus causing it to be at an angle larger than the desired 90º.

Moreover, stronger motors could be implemented to better the motion of the four-bar linkage as sometimes the opening and closing were jittery, causing erroneous sorting. The proximity sensors were also a limiting factor due to the small range they function at, so small containers would sometimes be missed by these. Repositioning the sensors in a different setup could have been beneficial to ensure proper detection.

In addition to this, this project demonstrated the need for iterating designs and working at smaller scales before implementing systems in a full-scale setup. If our team had iterated on smaller models we may have been able to catch issues with our design sooner and correct these, thus preventing larger issues with our final design.