Study on the Oxidative Deterioration Process of Diesel Based on Dielectric Spectroscopy

Conventional methods for assessing diesel oxidation stability often lack real-time monitoring capability and quantitative precision. This study investigates the evolution of dielectric spectra during diesel oxidation and the predictive efficacy of dielectric difference spectra for oxidation stability indicators. Using a self-developed dual-channel differential dielectric spectrometer based on the AD5933 chip and an oxidation stability tester, the oxidation of 19 diesel samples from various Chinese refineries was accelerated under controlled conditions (140 °C, 700 kPa oxygen). The dynamic changes in the real (ε') and imaginary (ε″) parts of the dielectric spectra during oxidation were tracked. The results reveal that the low-frequency response (<20 kHz) is governed by ionic conduction and interfacial polarization, whereas the high-frequency region (>50 kHz) primarily reflects dipole relaxation behavior. The dielectric parameters (ε' and ε″) of all oil samples exhibited continuous and regular changes with oxidation progression, without abrupt transitions. Dielectric responses showed significant variations in the low-frequency region, a finding consistently supported by the characteristic relaxation patterns in Cole-Cole plots. The introduction of exogenous polar substances (e.g., water) markedly interfered with relaxation characteristics. Although the current measurements were conducted offline, the methodology demonstrates potential for future adaptation to online monitoring systems. As a key advancement over previous work, this study integrated dielectric difference spectroscopy with dynamic principal component-based partial least-squares regression to predict key oxidation stability indicators (filterable insolubles at 140 °C, induction period, and oxidation inflection point). The use of difference spectrum (Δε) enhances the sensitivity to spectral changes induced by oxidation. Predictive models based on the imaginary part of the difference spectrum (Δε″) achieved outstanding performance, with correlation coefficients (R) of 0.9778, 0.9786, and 0.9581 for the three indicators, respectively. This research provides apromising methodological foundationfor thefuture development of real-time, nondestructive monitoring of fuel oxidation and holds significant potential for the online assessment of fuel quality.