Carbon dioxide (CO₂) is one of the main greenhouse gases, and its massive emissions can lead to an increase in the concentration of greenhouse gases in the atmosphere, thereby causing global temperature rise. The frequent occurrence of extreme weather events, rising sea levels, and glacier melting caused by global warming pose a serious threat to the environment and human society. Industrial activities are one of the main sources of carbon dioxide emissions, so it is necessary to strictly control carbon dioxide emissions in industry. By installing and using carbon dioxide concentration sensors, real-time monitoring of carbon dioxide emissions in industrial processes can be achieved, accurately grasping the emission amount and providing scientific basis for formulating carbon reduction measures. This article introduces the design of a small dual channel non dispersive infrared (NDIR) carbon dioxide concentration sensor with a long optical path chamber, aimed at improving detection accuracy and flexibility.This article adopts a dual channel infrared differential detection method to optimize the sensor design from three aspects: chamber structure, hardware circuit, and signal processing software. Firstly, a multi-stage folded gas chamber was designed to achieve a long optical path within a small volume, in order to enhance detection sensitivity. Using SolidWorks and TracePro for the structural design and luminous flux simulation of gas chambers, continuously optimizing the optical path design based on simulation results, and ultimately determining the optimal gas chamber structure (Fig.2). Secondly, a dual channel signal readout and processing circuit based on a second-order bandpass filter was developed (Fig.3), and amplifiers and filters were reasonably configured to ensure the stability and accuracy of the system. In addition, a data acquisition and preprocessing program was developed, and a neural network model was constructed based on the processed data on the upper computer. The initialization parameters of the backpropagation neural network were optimized using genetic algorithm, and the optimal relationship model between the detector dual channel output values, chamber temperature, and carbon dioxide concentration was established (Fig.4). Finally, a gas sensitivity testing platform was built to comprehensively test the performance indicators of the sensor, such as response time, repeatability, stability, and accuracy, verifying the feasibility of the system.According to the simulation results of the gas chamber, an effective optical path of 208.47 mm was achieved in a small volume gas chamber of 30 mm×30 mm×12 mm, significantly improving the sensitivity of the sensor. The dual channel output voltage of the sensor indicates that temperature drift and nonlinear factors have a certain impact on the measurement results, which need to be corrected through compensation algorithms (Fig.6). The performance indicator test results show that under the condition of 20 ℃, the response time of the sensor is 34 s, demonstrating good repeatability and stability (Fig.7). The backpropagation neural network model optimized by genetic algorithm was used for concentration prediction, resulting in an R² of 0.9999, an average absolute error (MAE) of 0.606 1, a root mean square error (RMSE) of 0.994 7 (Fig.9), an average error of less than 0.043%, and a relative error of less than 0.015% (Tab.1). These results indicate that the designed sensor has high measurement accuracy and excellent gas sensing performance.This article successfully designs a small dual channel NDIR carbon dioxide concentration sensor for monitoring carbon dioxide emissions in industrial production processes. This sensor has the characteristics of small size, simple structure, good stability, and high accuracy. It can achieve high-precision measurement at a concentration of 0%-3% in the temperature range of -10 ℃ to 40 ℃, with an average error of less than 0.043% and a relative error of less than 0.015%. This study provides valuable reference for the development of miniaturized NDIR gas sensors.