Smart Monitoring for SLA-type 3D Printer using Artificial Neural Network Danielle Jaye S. Agron * , Cosmas Ifeanyi Nwakanma † , Jae-Min Lee ‡ , and Dong-Seong Kim § Networked Systems Laboratory, Department of IT Convergence Engineering, Kumoh National Institute of Technology, South Korea. {danielleagron * ,cosmas.ifeanyi † , ijmpaul ‡ , dskim § }@kumoh.ac.kr Abstract—In additive manufacturing, a relatively high oxygen concentration inside a Stereolithography Apparatus (SLA-type) of three dimensional (3D)-printer during the operation have high risk of degrading the quality of the product. Monitoring such level of oxygen is still an open challenge to solve in the additive manufacturing industry. This paper presents a design of a collaborative sensor system that monitors the normal oxygen level during the printing process. Furthermore, a self organizing map (SOM) which is a form of artificial neural network(ANN) to cluster the oxygen concentration to reach the critical level was proposed. Results of the ANN performance is 96%. Index Terms—ANN, machine learning, real time monitoring, remaining useful life, 3D Printing. I. I NTRODUCTION Nowadays, 3D printing has been implemented in many industry [1]. One type of a 3D printer is a metal laser additive manufacturing printer and one variation of it is the StereoLithography Apparatus (SLA), this technology uses an ultraviolet (UV) laser to turn light-sensitive resin into solid 3D objects layer-by-layer from bottom to top. However, SLA printer may be the fastest additive manufacturing technology for functional, durable prototypes and end-use parts but there are a lot of wasted materials, time and money because of some printing failures. Some of the existing problem includes the high concentration level of oxygen inside the 3D-printer pro- duced during operation. The allowable oxygen concentration to preserve the highest quality of the printed product is 0.2% if it violates the threshold value then, it escalates the risk on having high porosity on the printed product [2]. Recent trend in condition monitoring is the application of digital twin and machine learning approaches. Such approach involve real time monitoring and classification of remaining useful life (RUL) of applications and devices or what is rightly called prognostics and health management (PHM). Chief among the machine learning approaches are support vector machine (SVM), artificial neural network (ANN) [3]. In [3], an echo state network (ESN) was adopted and shown to have superior performance over other machine learning algorithms like SVM. The drawback was also acknowledged in the work of [4] where in SVM was only used in classifying a good or bad print of work done. The need for more robust means of remaining useful life monitoring can not be overemphasised. This paper adopted the use of ANN to help in real time prediction of the threshold of the oxygen and prompting action if threshold for refill is detected. Most works in the area of combining neural networks with 3D printing monitoring appears to be faced with challenges and look promising as a research direction due to numerous challenges identified such as dataset limitations, adoption of unified application programming interfaces (APIs), hardware limitations and expertise [5]. To the best of our knowledge, this is the first time a work is done in the area of using ANN to predict of the level of oxygen inside the 3D printer within acceptable threshold. Motivated by the works of [6], this paper attempts to develop an ANN to monitor the threshold level of oxygen in SLA-type 3D printer. The rest of the paper is organized as follows. Section II discuss the system architecture of the 3D printer and the ANN of the proposed system. Section III demonstrated the preliminary results while Section IV is the conclusion of the paper. II. SYSTEM ARCHITECTURE In this paper, the effectiveness of using Artificial Neural Networks (ANN) to predict the future errors that may occur during the printing process was presented. The error that we are referring to is the abundance of oxygen present inside the 3D printer that leads to the degradation of quality of the product. In order to do that, an oxygen sensor is strategically placed inside the SLA 3D printer and it is connected to a data acquisition device (DAQ) to collect the data from the sensor and feeds it on a computer to process it (See Fig. 1). The data are filtered and logged in the local database, a comparison of the acquired data and the threshold value that indicates whether the condition of the printing environment whether it is normal or in critical/alarming condition. If the data collected is found to have an error or is strained, the system activates an alarm that sends a warning to the administrator about the irregularities on the values acquired by the sensors. These values may lead to conclusions that (1) Fig. 1: Monitoring System Test Design