AI Autonomous Monitoring of Milling Machine Tool

As AI is advancing and manufacturing demand increases, there is a need to incorporate AI into manufacturing. Our hand drill press currently requires an operator to manually move the drill press to create a hole. However, the process of detecting contact is inaccurate with manual control. In order to automate workpiece contact detection, a machine learning can be developed and a mechanical retrofit can replace the manual Z-axis movement. The machine learning model can be uploaded on a raspberry pi as well as a sensor to read live vibrational sensor data. The Raspberry Pi can simultaneously control the Z-axis movement using a stepper motor that powers the mechanical retrofit encompassing a belt-pulley drive. In turn, an AI autonomous electro-mechanical system is developed. In order to record data to train into a machine learning model, an appropriate sensor was required to be chosen to accurately measure vibrational spikes when the drill bit is in contact with the workpiece. An acoustic emission sensor (AE), power amplifier, and DAQ was used to capture and store the data. 100 trials were conducted and the data was appropriately trimmed and formatted to be trained on a machine learning model. A contact and no contact label (0/1) was created with each associated voltage value for training. Additionally, the standard deviation, rolling max, and rolling mean were other features the Machine Learning Model was developed on. From this, a supervised Machine Learning model was trained and refined. The model was able to achieve an accuracy of 97%. This model was used to validate live data by using the model logic to manually bring the drill press down and display Contact/No Contact. When the drill press was touching, Contact detected lit up on the screen. Now, the AE sensor data will be read using a Raspberry Pi rather than the large DAQ by integrating a DAQ 118 HAT to latch on to the Raspberry Pi. The stepper motor is being wired appropriately and powered with a stepper motor driver and power supply. A button switch is also used to start the whole process. A code is being developed to begin reading sensor values and referencing the Machine Learning Model after pushing the button. Once the sensor is reading data and the model has been successfully uploaded using tensorflow Lite, the stepper motor will activate to bring the drill press down. The DAQ will be reading live AE sensor values and the machine learning model will be interpreting the values as contact or no contact. Once the machine learning model detects contact, the stepper motor will retract the drill bit up in the opposite direction. The stepper motor has been able to be powered safely and the machine learning model is able to be opened on raspberry pi. The principle of automating machinery functions can be applied from this project. Although the project, focuses on contact expansion, this principle of using AI can be applied to things like tool wear, workpiece damage, and xy axis sweeping. This aligns with the trend of industry 4.0 where automation and AI will boost manufacturing production.