Materials

Hydrogen peroxide (H₂O₂, 30%, w/w) was purchased from Sinopharm Chemical Reagent Co. (China). The DURSET cleaning agent (PSC-002) was purchased from the NCH HUYANG Co. (Dalian, China). The physicochemical properties of the DURSET are listed in Table 1. The measured value of COD was 6.20×10⁵ ± 3.7% mg·L^-1, which was used to calculate the required amount of hydrogen peroxide oxidizer for fully oxidizing DURSET using stoichiometric method. The density, pH, flash point, and saturated aliphatic hydrocarbons were obtained from the product details. The GC-MS chromatogram and main components of DURSET are shown in Fig. 1 and Table 2, respectively. It can be seen that the main components of the stock solution are monocyclic alkanes (1,2,4-trimethylcyclohexane, 1,2,3-trimethylcyclohexane, 2-cyclohexylbutane, etc.) and bicyclic alkanes (decalin, 2-methyldecalin, etc.). In addition, it contains a small number of refractory monocyclic aromatic hydrocarbons.

Fig. 1Full-screen View

GC-MS chromatogram of DURSET

Apparatus and procedure

The SCWO-1000 system was constructed based on the SCWO-250 system used in previous studies [13, 27], as shown in Fig. 2. The reaction system mainly included one tank reactor (1000 mL, Inconel 625) and two preheaters (200 mL and 250 mL, Inconel 625). The maximum operating temperature and pressure of the equipment were 600℃ and 28.4 MPa, respectively. Considering that pressure had no significant effect on organic matter removal, all experiments were performed at 24 MPa. Before the start of each SCWO experiment, deionized water was pumped into two preheaters and one tank reactor, and the temperature was gradually increased using the PID meter (proportional integral derivative) meter. When the internal temperature exceeded 100℃, the pressure should be slowly increased to the expected value. Deionized water and a hydrogen peroxide solution were jointly pumped into the preheaters via a pressure metering pump (pump 1). Then, the feedstock solution (DURSET) was pumped directly into the tank reactor with the assistance of another pressure metering pump (pump 2). After the exothermic SCWO reaction, the outflowing reaction mixture was cooled to room temperature using a tubular cooling device. The condensed reaction mixture was adjusted to atmospheric pressure using a backpressure regulator (BPR). The mixture was sent to a gas-liquid separator, and the required liquid product sample was obtained after separation of the gas and liquid. The organic matter content in the liquid effluents was measured and evaluated using a HACH chemical oxygen demand digestion reactor and colorimeter.

Fig. 2Full-screen View

Schematic diagram of SCWO system (T: Temperature meter, P: Pressure gage)

Analysis methods

The main ingredients of DURSET and liquid effluent were analyzed by gas chromatography/mass spectrometry (GC7890A/MS5975C, Agilent, USA) with a HP-INNOWAX capillary column (60 m × 0.25 mm I.D., 0.25 μm film thickness). Helium (1 mL/min) was used as carrier gas. The temperature program of the oven was as follows: isothermal at 40 ℃ for 5 min, ramped up to 100℃ at 5 ℃/min, and ramped up to 230℃ at 10 ℃/min, and held for 15 min. The working voltage of the MS ion source (EI) was 70 eV, and its working temperature was 230℃. 150℃ was used as the operating temperature of the MS quadrupole.

The COD values of DURSET and the liquid products were determined by rapid digestion spectrophotometry (HJ/T399–2007). The measuring equipment used was the digestion device (DRB 200, HACH, USA) and portable spectrophotometer (HACH, DR 890, USA). First, 2 mL of the liquid product sample was injected into a digestion tube (in the range of 0-1500 mg·L^-1). Then the digestion tubes were placed into the digestion reactor for the digestion reaction for two hours at 150°C. Finally, the COD values were measured using a portable spectrophotometer. To ensure the credibility and accuracy of the COD measured by DURSET, a dilution measurement method was adopted to conduct six parallel experiments.

The COD removal efficiency (CRE, %) of the liquid effluent was calculated using the following formula: $CRE=\frac{{\left[COD\right]}_0-{\left[COD\right]}_{\text{L}}}{{\left[COD\right]}_0}\times 100\text{\%,}$ (1) where $［COD{］}_{0 }$ represents the concentration of chemical oxygen demand in the feedstock after considering the dilution influence of deionized water and hydrogen peroxide solution and $［COD{］}_{\text{L} }$ represents the concentration of chemical oxygen demand in the liquid effluent.

The excess oxygen coefficient (α, %) was calculated using the following equation: $\alpha =\frac{{\left[O_2\right]}_r-{\left[O_2\right]}_0}{{\left[O_2\right]}_0}\times 100\%,$ (2) where $［O_2{］}_{0 }$ is the concentration of oxidant needed to completely oxidize the organic substances calculated in accordance with the theoretical stoichiometric ratio, $［O_2{］}_{r }$ is the actual concentration of oxidant during each reaction test.

The reaction time (t, s) refers to the residence time in the reactor, and is defined as follows: $t=60\times \frac{V_0}{Q}\times \frac{V}{V_r},$ (3) where $V_{0 }$=1000 mL is the volume of the tank reactor, $V_{ }$ and $Q_{ }$ imply the specific volume and rate of volumetric flow from the liquid product at room temperature and atmospheric pressure, respectively, and $V_r$ implies the specific volume of feedstock under the conditions of the oxidation reaction.

Experimental model design

Response surface methodology (RSM) was used to design the experimental scheme for SCWO of DURSET, which was based on the CCD principle (central composite design). Design Expert software version 8.0.6 (Stat-Ease, Inc., Minneapolis, MN, USA) was used for the design. The application of statistical auxiliary tools can reduce material consumption and expenses. This is mainly due to the reduction in working hours and number of experiments. In addition, the scheme design can establish a multiple equation model between each independent factor and the response value. It can also optimize the model to reveal the multiple, interactive, and linear influences between the independent factors and response values.

Before designing the response surface scheme, it was necessary to determine the main influencing factors and range through single-factor experiments. In this study, the feed concentration (C, 1.5–3.0 wt.%), temperature (T, 440–580℃), reaction time (t, 30–105 s) and excess oxygen coefficient (α, 50–300%) were selected as influencing factors, which affected COD removal of DURSET via SCWO. Next, the influence of reaction parameters was analyzed through single-factor experiments. Finally, the central selection points of the main influencing factors, such as the temperature, reaction time, and excess oxygen coefficient, were determined (see Sect. 3.1).

Subsequently, the set pressure and feed concentration were set to 24 MPa and 2 wt.%, respectively. The influences of temperature, reaction time, and excess oxygen coefficient were studied using the experimental method of central composite design (CCD) of three factors and five levels. The levels and ranges of the independent factors are tabulated in Table 3. The CCD is mainly suitable for multi-factor and multi-level experiments, usually with five levels for each factor. The interaction between these significant factors can be intuitively determined, and the optimal reaction conditions can be obtained using a high-precision quadratic regression equation. The total number of experimental tests is expressed by the formula [30] $N_{\text{T}}=2^k+2k+c,$ (4) where k is the number of influencing factors and c is the number of center points. A total of 20 tests were applied to this RSM design, of which there were six axial points, eight factorial points, and six repeated experiments at the center point. The relationship between the independent variable and the response value was regressed using a multiple model equation [31]. This is expressed by the following equation: $Y={\beta }_0+\sum {\beta }_i{X}_i+\sum {\beta }_{ii}{X}_i^2+\sum {\beta }_{ij}{X}_i{X}_j+\epsilon ,$ (5) where $Y$ is the expected response value, $X_i$ is the independent variable, ${\beta }_0$ is a constant value of the model, ${\beta }_i$ is the coefficient of linear variables, ${\beta }_{ii}$ is the coefficient of quadratic variables, ${\beta }_{ij}$ is the coefficient of interactive variables, and $\epsilon$ is the residual random error in the test.

Finally, the reliability of the response surface model was assessed by analyzing R² (coefficient of determination) and $R_{adj}^2$ (adjusted value). Subsequently, the significance of the quadratic model equation was verified by analysis of variance (including the P-value, F-value, and coefficient of variation).

Single-factor experiment on COD removal efficiency

All the experiments were performed at approximately 24 MPa. In this section, the effects of feed concentration, temperature, reaction time, and excess oxygen coefficient on the CRE were studied. The optimal range of the main factors in the response surface experiments was selected through single-factor experiments.

3.1.1

Effect of feed concentration on CRE

Figure 3 presents the influence of feed concentration of DURSET on COD removal in SCWO at 540℃, a reaction time of 80 s, and excess oxygen coefficient 300%. As shown in Fig. 3, when the feed concentration increased from 1.5 to 2.0 wt.%, the CRE increased from 98.66% to 99.67%. However, when the content exceeded 2 wt.%, there was a relative downward trend. When the feed concentration increased and the organic removal rate decreased, the COD of the liquid effluent tended to increase. Moreover, a high feed concentration may lead to an increase in second-order polymerization reactions, resulting in char and tar byproducts [32]. In general, the feed concentration of the reactants had no significant effect on the COD removal. Consequently, the effect of feed concentration was not considered in subsequent response surface experiments.

Fig. 3Full-screen View

Effect of feed concentration on COD removal efficiency

3.1.2

Effect of temperature on CRE

The CRE of DURSET was significantly affected by the temperature, as shown in Fig. 4. The CRE gradually increased with increasing temperature at a feed concentration of 2 wt.%, reaction time of 55 s and excess oxygen coefficient of 300%. When the temperature increased from 440 to 540℃, the CRE clearly increased from 88.75 to 99.41%. However, the CRE only increased by approximately 0.40% when the temperature increased from 540 to 580℃. Accordingly, when the temperature was relatively low, an increase in temperature was an effective measure to quickly increase the COD removal of organic matter. This may be due to the increase in temperature, which helped trigger molecular activity in the supercritical water reaction system. Yang et al. [33] studied the decomposition of 15 aromatic compounds in supercritical water oxidation at 350–550℃. The results indicate that temperature played an important role in the removal of various organic matter in the SCWO. Al-Duri et al. [34] conducted an experimental study on the SCWO degradation of nitrogenous hydrocarbons 1,8-diazobicy-clo [5.4.0] undec-7-ene (DBU). It was found that, at the conditions of oxidant stoichiometric ratio of 200%, pressure of 25 MPa, and residence time of 7 s, the TOC removal rate increased from about 17% to 98% when the temperature increased from 400℃ to 525℃, which showed that the effect of temperature was very significant. In this work, to obtain a fitted quadratic model with a wide temperature range, a temperature range of 430–550℃ was used of the response surface experiment, and a temperature of 490℃ was selected as the center point.

Fig. 4Full-screen View

Effect of temperature on COD removal efficiency

3.1.3

Effect of reaction time on CRE

The effect of reaction time from 30 to 105 s on the COD removal of the SCWO process at a feed concentration of 2 wt.%, temperature of 540℃, and excess oxygen coefficient of 300% was demonstrated in Fig. 5. It can be seen from Fig. 5 that under the given conditions, the CRE obviously grew from 91.79 to 99.90%. The CRE increased rapidly with increasing reaction time in the range of 30–55 s, after which the increasing trend slowed with depleted reactants. This result may be explained by the free-radical reaction mechanism in a supercritical water environment [35]. Bühler et al. [36] reported that OH· and H· free radicals play a key role in the degradation of organic molecules in SCW. Yu et al. [37] studied the treatment of oily wastewater using SCWO and concluded that when the reaction time increased from 1 to 9 min, the COD removal rate approximately increased from 69% to 78%. At the same time, with the use of Cu²⁺ catalyst, the COD removal rate increased from 83% to 93%. Chen et al. [38] investigated the effect of the reaction time on SCWO degradation in oil-based drilling cuttings. At the condition of the reaction temperature 500℃ and oxidation coefficient 250%, the TOC removal rate increased by 25% when the reaction time increased from 30 s to 120 s. These results show that the effect of reaction time plays a significant role in SCWO. Hence, a reaction time of 55 s was selected as the response surface center design point, while the reaction time ranged from 25 to 85 s.

Fig. 5Full-screen View

Effect of reaction time on COD removal efficiency

3.1.4

Effect of excess oxygen coefficient on CRE

The influence of excess oxygen coefficient in the range of 50–300% and at the conditions of a feed concentration of 2 wt.%, 540℃, and 80 s is given in Fig. 6. When the excess oxygen coefficient increased from 50 to 200%, the CRE clearly grew from 92.01 to 99.43%. When the excess oxygen coefficient exceeded 200%, the amount of oxidant had little effect on the CRE. This may be due to an excess of oxidant. Gong et al. [39] found that the TOC removal rate increased rapidly when the oxidation coefficient increased from 0% to 200% during SCWO degradation of quinazoline. When it exceeded 200%, the TOC removal rate plateaued to a constant value, which also confirmed our results to some extent. In addition, we found that if the excess oxygen coefficient was lower than 50%, a lot of insoluble black tar appeared in the liquid effluent, which could easily cause the reactor pipe to block. Related studies [32, 40] have pointed out that the precursor molecules of tars are polycyclic compounds, including naphthalene and phenanthrene. Besides, Wahyudiono et al. [41] studied the liquefaction of tar in sub- and supercritical water and found that the main products from the liquefaction of tar were phenol, biphenyl, diphenylether, and diphenylmethane, respectively. This phenomenon may be due to the formation of intermediate compounds and their polymerization processes, which resulted in the production of tar and char [42]. Therefore, to reduce the production of tar, the range of excess oxygen coefficient was selected to be 50–250% in the response surface experiment.

Fig. 6Full-screen View

Effect of excess oxygen coefficient on COD removal efficiency

Model fitting and ANOVA

The response surface method based on CCD was used to probe the effect of the process parameters on the SCWO of the oil cleaning agent. Table 4 shows the experimental results of the CCD assays for SCWO of DURSET. The results showed that there were significant differences in the CRE of the SCWO systems under different reaction conditions. Fig. 7(a) shows a favorable linear correlation between the experimental and predicted values. Fig. 7(b) presents the normal plot of residuals that shows that the points follow a linear regularity, which means that the error terms are normally distributed [43]. Therefore, the proposed model is credible.

Fig. 7Full-screen View

(Color online) Comparison of (a) CRE obtained by the experiment values and model predicted. (b) Normal probability of internally studentized residuals for CRE

Combined with the results in Table 4, which shows the experimental and predicted values of the three responses, an empirical relationship between the experimental variables and response was obtained using Eq. (6). In Eq. (6), the CRE was used as the response value, which was associated with the coded independent variables, including temperature, reaction time, and excess oxygen coefficient. $\begin{array}{c}CRE \left(\%\right) = 95.56 + 3.33T+1.84t+1.46\alpha +0.72Tt+0.36T\alpha +0.026t\alpha –\\ 1.31T^2- 1.12t^2- 0.57{\alpha }^2\end{array}$ (6)

Analysis of variance (ANOVA) was used to verify the reliability of the model. The ANOVA of the fitted quadratic model for the SCWO process of DURSET is presented in Table 5. As shown in Table 5, a model P-value of less than 0.0001 indicated that the model was statistically significant. P-values less than 0.05 implied that the model terms were stable and significant [44]. In this case, the significant model terms were T, t, α, Tt, T², t², and α², with corresponding P-values of <0.0001, <0.0001, <0.0001, 0.0294, <0.0001, <0.0001, and 0.0052, respectively. In other words, temperature, reaction time, and excess oxygen coefficient, as well as the interaction between temperature and reaction time, significantly affected the CRE response. The model F-value of 57.84 meant the model was significant. It can be observed that the temperature with an F-value of 277.44 was a determinant of the reaction time and excess oxygen coefficient, with F-values of 84.49 and 53.37, respectively. Therefore, the order of factors affecting the CRE was temperature > reaction time > excess oxygen coefficient. Compared with the pure error, the “lack of fit F-value” of 3.43 and its P-value of 0.1014, imply that the lack of fit was insignificant. In other words, there was a 10.14% possibility of noise influence caused by such a lack of fit value. The coefficient of determination ($R^2$) and its adjusted value ($R_{\text{adj}}^2$) were 0.9812 and 0.9642, respectively, proving that the correlation between the experimental and predicted values was very high. The signal-to-noise ratio (Adeq. Precision) was 29.42. Usually, the signal-to-noise ratio is required to be greater than four, indicating that this multiple model can accurately reflect the experimental results [45]. Smaller coefficients of variation (CV) of multiple models lead to higher accuracy. The value of CV was 0.86% (< 5%), implying that the model was accurate and reliable [40].

Interactive effects of process parameters on responses

To better understand the three independent variables and their interactions, response plots provide a statistical method to visually predict the relationship between the response and experimental level [46]. Figure 8 presents the response surface (3D) plots and contour plots that signify the interactive effect of multiple variables on the CRE obtained by the SCWO treatment of DURSET.

Fig. 8Full-screen View

(Color online) The 3D and contour plots showing the interactive effects of (a, b) temperature and reaction time at excess oxygen coefficient of 150%, (c, d) temperature and excess oxygen coefficient at reaction time of 55 s, (e, f) reaction time and excess oxygen coefficient at temperature of 490℃ on COD removal efficiency of DURSET

Optimization and verification of process conditions

The optimal combination of various influencing factors in the reaction process can be obtained using RSM. Usually, in the optimization and selection of a combination of independent variables that meet the objective experimental conditions, the maximum CRE will be obtained in the supercritical water oxidation treatment of DURSET. Therefore, the CRE was selected to be “maximized”, while the independent factors were set to be “in range”. Table 6 presents the solution optimized via Design Expert 8.0.6 software, where the optimal process conditions were 515.03℃, 66.33 s and an excess oxygen coefficient 211.38%. However, considering the operability of the experiment, the optimum process conditions were 515℃, 66 s, and excess oxygen coefficient 211%. These levels of process conditions led to a maximum removal rate of approximately 100%. To verify the agreement between predicted values and experimental values, three parallel experiments were conducted applying the optimum process conditions: the CRE of DURSET was 99.49 ± 0.23%, and the COD concentration of liquid effluent was 84.32 ± 38.03 mg·L^-1. The error between the experimental and the predicted values was less than 1%, which indicates that the parameters of SCWO treatment of DURSET optimized by response surface analysis method are reliable, and the prediction is feasible according to the regression model.

Product analysis for SCWO of DURSET

In this section, the reaction intermediate products formed by the decomposition of DURSET in SCW are analyzed. The compounds extracted from the liquid effluent were separated and identified using GC/MS. The identification of reaction products at different retention times (RT) was achieved by comparing with the standard MS library (NIST 11). Table 7 lists major reaction products identified from the SCWO of DURSET conducted at 440℃, 70 s an excess oxygen coefficient 50%. As shown in Table 7, the main intermediate products were ketones, ethyl ether, benzene compounds, ethyl propionate, and others. Among these products, cyclic ketones and benzenes were considered refractory products, which mainly hindered the decomposition of organic matter and COD removal. It was worth mentioning that the newly produced naphthalene may be produced by dehydrogenation of decalin [50]. In addition, under special conditions, the phenyl radical of SCWO may generate dimers or polycyclic compounds such as naphthalene and phenanthrene through coupling reactions [17]. The mechanism of oxidative degradation is very complex and cannot be explained in detail through simple experiments. Cyclohexane in the stock solution can be oxidized to cyclic ketones under certain circumstances [51, 52]. Previous studies have shown that aromatic compounds (such as benzene, phenol, and naphthalene) are often regarded as dominant intermediates that hinder the degradation reaction in SCW [26, 53]. Therefore, ketone and benzene compounds may be important intermediate products in the degradation process. In other words, the work in this section provides a basis for further study on the degradation mechanism of oil cleaning agents treated by SCWO.