Conveyor Belt Sorting System

Overview

#

A university project demonstrating industrial sorting automation using a PiXtend PLC (Raspberry Pi-based programmable logic controller). The system sorts objects on a conveyor belt using three increasingly sophisticated approaches: inductive sensing, color classification, and machine learning-based image recognition.

Hardware Setup

#

The conveyor belt system integrates multiple industrial sensors and actuators:

• Capacitive sensor — Detects the presence of any object (metal, plastic, wood)
• Inductive sensor — Distinguishes metal from non-metal parts
• Color sensor — Classifies objects by color (4 trainable classes)
• Raspberry Pi Camera — Captures images for ML-based classification
• Pneumatic cylinders — Eject sorted objects at defined positions

The Controller: PiXtend

#

The PiXtend is a PLC based on the Raspberry Pi with industrial-grade digital and analog I/O. It supports RS232, RS485, CAN, Ethernet, WiFi, and Bluetooth — making it a versatile controller for automation projects.

Test Setup 1 — Material Sorting

#

The simplest approach: distinguish metal from non-metal using the inductive sensor.

• Capacitive sensor detects object presence
• Inductive sensor checks for metal (HIGH = metal)
• Non-metal → ejected at cylinder 1
• Metal → ejected at cylinder 2

Test Setup 2 — Color Sorting

#

Using the color sensor with 4 trained color classes:

• Green → ejected at cylinder 1
• Blue → ejected at cylinder 2
• Red → passes through to the end

Test Setup 3 — ML-Based Object Classification

#

The most advanced approach: a trained neural network classifies objects from camera images.

The flow: capacitive sensor detects object → belt stops → camera captures image → ML model predicts class → pneumatic actuator sorts accordingly.

Key Learnings

#

• Industrial sensor integration with PLCs
• PLC programming on Raspberry Pi-based controllers
• Training and deploying ML models for real-time classification on embedded hardware
• Pneumatic actuator control and timing