Membrane processes for the reuse of car washing wastewater

This study investigates alternative treatments of car wash ef ﬂ uents. The car wash wastewater was treated by settling, ﬁ ltration, and membrane ﬁ ltration processes. During settling, total solid concentration decreased rapidly within the ﬁ rst 2 hours and then remained constant. Chemical oxygen demand (COD) and conductivity were decreased by 10% and 4%, respectively. After settling, wastewater was ﬁ ltered throughout a 100 μ m ﬁ lter. It was found that ﬁ ltration had a negligible effect on COD removal. Finally, wastewater was ﬁ ltered by four ultra ﬁ ltration membranes of varying molecular weight cutoff (MWCO) (1, 5, 10 and 50 kDa) and one nano ﬁ ltration membrane (NF270, MWCO ¼ 200 – 400 Da). The permeate COD concentrations varied between 64.5 ± 3.2 and 85.5 ± 4.3 mg L (cid:2) 1 depending on UF pore size. When the NF270 nano ﬁ ltration membrane was used, the permeate COD concentration was 8.1 ± 0.4 mg L (cid:2) 1 corresponding to 97% removal. FeCl 3 precipitation and activated carbon adsorption techniques were also applied to the retentate and 60 – 76% COD removals were obtained for activated carbon adsorption and FeCl 3 precipitation, respectively. This study investigates alternative treatments of car wash effluents. This study investigates alternative treatments of car wash


INTRODUCTION
The car wash industry is one of the leading consumers of large volumes of clean water. Water used per car varies between 150 and 600 L depending on the size of the car and equipment used (Panizza & Cerisola a). Therefore, there is a growing interest in wastewater treatment and reuse in this sector, in addition to the recognition of the environmental impacts (Zaneti et al. ). In Queensland, Australia, water is limited to 70 L per car. Similarly, some European countries restrict water consumption to 60-70 L per car and/or impose reclamation percentages (70-80%) (Boussu et al. ; Zaneti et al. ).
Often, characteristics of the wastewater depend on the socioeconomic structure of the country. In a study conducted in Malaysia, chemical oxygen demand (COD) ranged from 75 to 738 mg L À1 ; conductivity 150.7-260.7 μS cm À1 and turbidity 34.7-86 NTU were reported (Lau et al. ). Another study in Brazil reported COD at 259 ± 40 mg L À1 ; conductivity at 446 ± 55 μS cm À1 , and turbidity at 139 ± 45 NTU (Rubio & Zaneti ).
Although there is a growing interest in recycling car wash effluents and implementing different technologies, there is no comprehensive standard for recycled water conditions. Metcalf & Eddy () describe reclaimed water as water that has been brought to specific criteria and suitable for its intended use. According to Brown (), no dust, oil, or grease should be present in the water that is to be recycled in car wash units. Any additional process for the treatment of dust, oil, and grease increases the quality of recycled water and allows the water to be used in different washing stages (pre-soak, wash, rocker panel/undercarriage, first rinse, and final rinse) (Zaneti et al. ).
There are many treatment alternatives, including reverse osmosis-nanofiltration ultrafiltration (Jönsson & Jönsson ), ultrafiltration-activated carbon adsorption (Hamada & Miyazaki ), electrochemical oxidation (Panizza & Cerisola b), biological treatment, and flocculation (Rubio et al. ). In one study, high turbidity and color removal (higher than 90% and 75%, respectively) were obtained by flocculation and the column floating method (Rubio & Zaneti   For that purpose, car wash effluents were precipitated and filtered. Finally, four various sized ultrafiltration and one nanofiltration membrane were used to treat the wastewater. The results were evaluated with respect to COD, conductivity, and total solid rejections.

EXPERIMENTAL Car wash effluent
Wastewater was collected from a car wash station in Atasehir/Istanbul, Turkey; its characteristics can be seen in Table 1. Wastewater was collected during three washing periodsrinsing, foaming, and final rinsingand subsequently mixed. Wastewater was collected in proportion to water used in each period. In order to also focus on detergent removal, detergent was also taken from the station.

Treatment process
The treatment process included three steps: settling, filtration, and membrane filtration. In the settling tests, 1 L of wastewater was taken into a 1,000 ml graduated cylinder and sampled for every 30 minutes from the 300 ml level of the cylinder. After 11 h settling, the test wastewater was siphoned up to the 300 ml level. The siphoned wastewater then passed through a ∼10 μm filter (coarse filtration -Chmlab). The newly filtered water was then used in several membrane filtration tests in which four ultrafiltration membrane (1, 5, 10 and 50 kDa) and one nanofiltration membrane (NF270, 200-400 Da) were used. The properties of membranes are presented in Table 2. Membranes used were left in pure water for 1 day prior to the dead-end analysis. At the end of each process, wastewater was sampled for color, COD, total solid, pH, conductivity, and PO 4 3À analysis.
In order to investigate high concentrations of detergent removal from the retentate, detergent from the car wash station was added to the water. This water was then treated by adsorption and chemical precipitation.
Granular activated carbon and FeCl 3 were used for the adsorption and chemical precipitation tests, respectively.
The results were evaluated based on the COD and conductivity analysis.

Analytical methods
Total solids were measured according to the evaporation methods as reported by another study (Bhattarai et al.  Stannous Chloride method with maximum detection limit of 0.05 ppm). Color was measured using the Pt-Co method by a spectrophotometer (HACH, DR/5000). The dead-end filtration unit was used for both coarse and membrane filtrations ( Figure 1). In the dead-end unit, wastewater was forced by nitrogen gas pressure to pass throughout a membrane lying in the bottom of the unit. Wastewater mixing (230 rpm) was provided by the magnetic stirrer within the dead-end mechanism (Figure 1). The weight of the filtrate was measured on a scale and the data was processed by a computer connected to the scale. The volume of the filtrate was calculated based on the assumption that filtrate density is 1 g cm À3 . These dead-end filtration tests were performed -P (mg L À1 ) 9.05 ± 0.4

Pre-treatment of wastewater
The pre-treatment stage consisted of settling and coarse filtration. In settling tests, wastewater was sampled every 30 minutes during the first 150 min of the experiment. Then two more samples were taken after 210 minutes and again after 630 minutes. Results showed that most of the settleable solids were precipitated in 1 hour and then the concentration of total solids remained the same (see supplementary material, Figure S1, available with the online version of this paper). Initial total solid concentration was 1,054 ± 21 mg L À1 and decreased to 609 ± 14.2 mg L À1 at the end of 1 hour corresponding to 42% solid removal. At the end of 10.5 hours, removal efficiency was 47% (Figure 2).
Wastewater was then filtered throughout a coarse filter To specify that the detergent used in this study can be measured as COD, 2 ml and 4 ml of detergent were added to 1 L of distilled water and the COD was analyzed. The COD concentrations were 228.3 ± 3.5 and 427.6 ± 10.6 mg L À1 for 2 and 4 ml L À1 detergents, respectively (COD ml À1 detergent ratio was calculated as 110 ± 4 mg COD ml À1 detergent).

Membrane filtration -flux
In the membrane filtration tests, four 1,000-50,000 Da UF membranes and one NF (NF270) membrane were used.
The flux variations were investigated for a single 1 hour period and decreases were monitored. The highest flux decrease was observed for the 50,000 Da UF membrane.
While the initial flux was 177.9 L m À2 h À1 , it rapidly decreased to 68.9 ± 1.9 L m À2 h À1 in 10 min. After the 1 hour filtration, the final flux was 35.2 ± 2.1 L m À2 h À1 .
As MWCOs decreased from 50,000 to 1,000 Da, the flux  Figure 1 | The dead-end filtration unit used in filtration tests: 1, nitrogen gas, which is decline also decreased. For the 10,000 Da UF membrane, the initial flux of 136.5 ± 2.05 L m À2 ·h À1 decreased to 24.8 ± 3 L m À2 ·h À1 corresponding to 82% flux loss. The 1,000 Da UF membrane showed the lowest flux with an average of 13.97 L m À2 ·h À1 (Figure 3). The flux decrease for NF270 membrane was 35% at the end of the experiment.
Pore size is one of the most important parameters affecting the flux in membrane filtrations. Flux is also affected by the membrane structure, constituents in the water, and wastewater pH. In some studies, it has been reported that NF membranes (especially NF270) can provide a higher flux than UF membranes (Lau et al. ). This is primarily due to electrostatic interactions between the membrane surface and wastewater constituents as reported (Chidambaram et al. ).
In this study, the initial flux of the NF membrane was 57 L m À2 h À1 and decreased to 37 L m À2 h À1 (35% decrease).
The NF270 membrane has a hydrophilic nature and is composed of a piperazine and benzenetricarbonyl trichloridebased polyamide layer on top of a polysulphone microporous support reinforced with a polyester non-woven backing layer (Gryta et al. ). With this structure, the NF270 membrane can provide good rejections at high flux values. In addition to the hydrophilic layer, the highly negatively charged surface of NF270 reduces the clogging effects and prevents flux drops in wastewater treatment (Ong et al. ). The reason for the low flux in the ultrafiltration membranes may be due to oil, grease, and other petroleum derivatives resulting from the engine washing process. Often the composition of oil and grease originating from the engine is quite complex and treating these materials during the process of membrane filtration can be difficult. Generally, benzene, lead, zinc, chromium, arsenic, pesticides, nitrates, and different concentrations of heavy metals are found in oil and grease (Lau et al. ).
For all UF membranes tested, effluent PO 4 3À concentrations were less than 1 mg L À1 . Variations in water characteristics during the settling, coarse filtration, and membrane filtration are presented in Table 3.
As expected, better results were obtained with the NF270 membrane. With the NF membrane, COD, conductivity, and PO 4 3À -P removal efficiencies were 98%, 47%, and 100%, respectively (Table 3). Images showing the changes in wastewater appearance are presented in the supplementary material ( Figure S2, available with the online version of this paper).

Fate of concentrate
Although high effluent quality can be obtained by membrane Adsorption of detergent on activated carbon was studied by preparing 2 ml L À1 detergent solution. Then 2 g of activated