HeteroSense-VOC Suite

This dataset suite comprises three gas-sensing datasets derived from our prior studies, organized as one source domain and two target domains, and is intended to support cross-domain modeling and transfer-learning evaluation across heterogeneous sensor platforms. The three datasets mainly differ in sensor array type and the set of target gases. The source-domain dataset was collected using commercial gas sensors and covers twelve VOCs, with fifteen independent trials conducted for each gas. Target domain A was acquired using a MEMS gas sensor array; compared with the source domain, it excludes certain gases and adds two additional gases, with six independent trials for each gas. For both the source domain and target domain A, gas concentrations range from 10 to 100 ppm in 10-ppm increments, and the sampling rate is 2 Hz. Target domain B was collected using a self-developed intelligent gas sensor array based on NiO\/ZnO nanocomposites and includes six VOCs. For each gas, five independent trials were conducted, and each trial includes an increasing concentration ramp from 10 to 100 ppm followed by a decreasing ramp back to 10 ppm; additional measurements below 10 ppm were also collected. The sampling rate for target domain B is 1 Hz. All experiments were performed on a fully automated, in-house test system to improve throughput and reduce environmental interference. The suite provides both raw resistance curves and response curves defined by the ratio (R_a\/R_g), where (R_a) and (R_g) denote the sensor resistance in air and under exposure to the target gas, respectively. These data enable studies of domain shift, sampling-rate mismatch, low-concentration sensing, and cross-sensor generalization.