A Fundamental Study on Estimating the Concentration of Carbon Monoxide (CO) By Fabric-Based Colorimetric Analysis Using an Optical RGB Color Sensor

Abstract

We propose a sensitive colorimetric sensing system for estimating the concentration of Carbon Monoxide (CO) using an optical RGB color sensor. Tasteless and odorless hazardous gas, such as carbon monoxide (CO), can be life-threatening, even at prolonged exposure at low concentrations. Early detection of these gases under low concentration conditions is a difficult problem. Thus, studying environmental health, monitoring of air quality, and the safety of people from harmful gas exposures have motivated increasing efforts to develop low cost and high-performance various sensors[1-2]. A detection method using the visual and optical device is relatively easy to apply because they are widely used and cheaper than in the past. This method can effectively detect harmful substances with reactive dyes. Potassium disulfide dye is suitable as a reactive dye because it has a property of discoloration in response to CO. Therefore, we used this to detect the optical response of the fabric discoloration process and perform quantitative gas concentration estimation. The chemical reaction of CO gas and disulfide potassium dyes is as shown in Formula 1, and the fabric exposed to the reaction turns black. The discoloration information for the fabric was collected from chambers that can control the gas concentration, and the concentration of CO gas considered in the experiment is 0-600 ppm. The time to reach full discoloration was recorded according to each gas concentration condition, and an RGB sensor was used to detect discoloration information. We used Moving-median filtering to remove noise caused by sensor characteristics of the collected data. To estimate the gas concentration, a curve fitting model using a least-squares method was applied. This method of converting to a power law scale and then use poly fitted to fit a linear curve to data. In the process of discoloration, we observed the difference in the discoloration progression rate according to the gas level and time. Also, under each gas concentration condition, the spending times from 20% progression to 80% progression in the whole discoloration process have a significant difference. These characteristics were applied to approach algorithms for estimate concentrations. Results show that the spending time to discolor from 20 % to 80 % progression at 200, 400, and 600 ppm was 275, 163.8 and 151.4 seconds, respectively in Figure 1. In the low concentration section, a significant time difference was observed compared to the high concentration section. The time characteristics of each section were extracted and applied to the concentration estimation algorithm. Except for outliers, the actual measurement data applied to the estimation algorithm, the concentration estimated were about 95%, 77% and 72% accuracy at 200, 400 and 600 ppm. As a result, we could estimate the concentration of CO gas by analyzing some of the discolored sections. This result is since the response time of discoloration varies depending on the gas concentration level. As future work, the gas concentration estimation method in this study could be improved by machine learning and data processing methods. The early detect system on dyed fabric color by gases exposes can be applied to for the next generation of the smart industrial safety system and intelligent clothing.AcknowledgementThis study has been conducted with the support of the Korea Institute of Industrial Technology.