Research into natural organic matter (NOM) removal in drinking water treatment processes is mostly independent, distributed, disconnected and unable to meet the needs of technology application; therefore, an assessment of the value of NOM treatment processes is necessary. In this paper, a hybrid evaluation model based on rough set theory and a matter-element model was used to evaluate the value of eight NOM removal processes. The counting process of the weighting factor did not include any subjective information which avoided the artificial factor deviation and made the evaluation more objective. The result indicated that in addition to the MIEX + coagulation + sedimentation + filtration (MCSF) treatment process, the rest of the NOM treatment processes had a certain value; the coagulation + sedimentation + filtration + adsorption (CSFA) and coagulation + sedimentation + filtration + membrane (CSFM) treatment processes had the highest values which meant that these treatment processes could remove the NOM in drinking water effectively. It also illustrated that the coagulation + sedimentation + adsorption + membrane (CSAM) treatment process had high feasibility, which has important significance for guaranteeing the safety of drinking water.

INTRODUCTION

Natural organic matter (NOM) is defined as a complex mixture of organic materials present in all natural waters, particularly surface waters (Dias et al. 2013). Although NOM has no side effects, it is the primary source of colour, odour and taste problems and, to a large extent, affects the sensory index of water quality in water treatment processes (Zhang & Zhao 2014). Meanwhile, during drinking water treatment, it is responsible for the increase of coagulant dosage affecting the coagulation effect, the acceleration of filter bed clogging and membrane fouling (Roccaro et al. 2011). Moreover, NOM has been noted to contribute to the production of disinfection by-products and has even caused bacterial growth in the water conveyance system (Zhang et al. 2015).

Currently, the primary and most feasible drinking water processes for the removal of NOM include coagulation, adsorption, ozone oxidization and various combined technologies. Research into NOM removal in drinking water treatment processes is mostly independent, distributed, disconnected and unable to meet the needs of technology application; therefore, an assessment of the value of NOM treatment process is necessary (Burchard-Levine et al. 2014). At present, there are many methods to evaluate the drinking water process, such as water quality index evaluation, grey correlation, principal component analysis, fuzzy theory and fuzzy synthetic evaluation, which are widely used in hydrology, economics, medicine and other fields (Gong et al. 2012). Every method has its weakness, for example the water quality index evaluation method is subjective, and the grey clustering and fuzzy theory methods neglect some important information in the process of calculation and need special programs to finish the complicated calculation. The matter-element method can solve the realistic contradiction qualitatively and quantitatively (Islam et al. 2013). It has been applied to many fields, such as environmental quality assessment, land suitability evaluation, and risk assessment of urban network planning, but less work has been done on applying it to evaluating drinking water treatment processes (Liu et al. 2013).

The evaluation of NOM treatment processes is affected by many factors and when using the conventional classical mathematical analysis method of dealing with this kind of problem it is often difficult to obtain good results, but a matter-element analysis model can transform complex problems into a pictorial mathematical model and quantify evaluation results. Meanwhile, the rough set theory is used to determine the weighting coefficient of attributes which can eliminate the weight calculation of anthropogenic interference and has a certain theoretical basis. In this paper, a hybrid evaluation model based on a matter-element model and rough set theory was used to evaluate the NOM treatment processes, and the counting process of the weighting factor was obtained by calculating the comprehensive correlation degrees and extension indexes. Meanwhile the counting process did not include any subjective information which avoided the artificial factor deviation and made the evaluation more objective.

METHODOLOGY

NOM treatment process evaluation index system

There are many factors that influence the value of NOM treatment processes, not only the social, economic and natural factors, but also the process itself (Roccaro et al. 2013). However, during the evaluation, the higher value is not suited to the index system. Not only do the indicators make a lot of work, difficulties and poor operational values, but also complex relationships between indexes can easily produce a correlation between each other and reduce the impact of key indicators (Houtman et al. 2014). In addition, too large an index system makes the weight of each index small and also cannot fully embody the role of the main indicators (Shahid et al. 2014). Therefore, the complex characteristics of the NOM treatment process evaluation can be well represented by strongly relevant and easy to quantify comprehensive parameters which reflect the NOM process evaluation and establish the index system.

According to the above principles, we selected a water synthesis pollution index, a drinking water treatment synthetic index and a drinking water quality index to set up the evaluation index system.

Water integrated pollution index

The water integrated pollution index (WPI) can not only produce the quantitative analysis and level of pollution for raw water, but it is also easy to compare and evaluate the pollution levels of different areas (Lavie et al. 2014). The determination of the pollution index method is as follows:

If P represents the WPI, it contains multiple pollutant evaluation factors that can be represented as: 
formula
1
where is the weight of a single pollution factor; is the single factor pollution index, ; is the pollutant concentrations measured; is the pollutant standard values of appropriate categories; n is the number of pollutants in the evaluation. The pollution index of the grading range is determined by water quality standards according to environmental quality standards for surface water (Zhang & Zhao 2014). The NOM weight is higher than others and set to 0.5.

Drinking water treatment process integrated index

The drinking water treatment process integrated index (DWTPI) is a comprehensive indicator which can evaluate different treatment processes quantitatively (Chen et al. 2007). It has seven aspects including project investment, operation cost, treatment area, running stability, process maturity, operational difficulty level and environmental benefits. The index calculation steps are as follows:

  • (1)

    The normalized processing of variable values. The bigger optimal type and the smaller optimal type are adopted for different processing methods.

  • The bigger optimal type: 
    formula
    2
    The smaller optimal type: 
    formula
    3
    where , and are respectively the value of variables, and the maximum and minimum values of the data series; are the values normalized by variables, between 0 and 1.
  • (2)

    In order to avoid the problem of boundary values of 0 and 1 as well as the difference in the evaluation of regional gaps, Equations (2) and (3) can be improved and the highest and lowest of the variable data series increased and reduced by 5%, respectively.

  • (3)
    The DWTPI expression is showed as follows in Equation (4). The higher indexes calculated illustrate that the process comprehensive efficiency is higher, which means that the process is suitable for NOM removal. 
    formula
    4

Drinking water quality integrated index

Drinking water quality depends on the composition of pollution variables and as each variable reflects the water quality situation from one aspect, evaluating the water quality reasonably and objectively is an important step in the evaluation process (Xiao et al. 2014). The drinking water quality integrated index (DWQI) can produce a quantitative description for the effluent of a drinking water treatment process which can compare and evaluate the effluent quality of different water treatment processes. DWQI contains multiple water quality evaluation factors: 
formula
5
where is the weight of a single water quality factor; is a single factor water quality index, ; is measured drinking water quality indexes concentrations; is the drinking water quality index standard values of appropriate categories; n is the number of drinking water quality indexes that are included in the evaluation. The drinking water quality index of the grading range is determined by water quality standards according to environmental quality standards for drinking water (Zhang & Zhao 2014).

The index weight determination based on rough set theory

The rough set theory is a kind of mathematical tool which depicts imperfection and uncertainty. The rough set theory considers that ‘knowledge’ is a kind of ability, which humans and other species have to classify different features of things (Pai & Lee 2010). In order to ascertain the index weight, we use the rough set theory to determine the important degree of each index attribute which affect the NOM treatment process value.

Establish the knowledge expression system

Four tuples is a knowledge expression system; U is the universe showing the non-empty finite set of the object; , , the subsets C and D are regarded as condition attributes and decision attributes, respectively; , is the range of attribute a; means one information function of the attribute value of each sample, that is ,, .

Determine information quantity

Because , , so the information quantity of p is denoted by Equation (6): 
formula
6
where represents the cardinal number of a set.

Determine the weight

Because , the significance of condition attribute () in C can be defined as Equation (7): 
formula
7
If , the weight of condition attribute is Equation (8): 
formula
8

The matter-element analysis model

Matter-element theory was first introduced for solving incompatible problems in the 1980s by the Chinese mathematician Cai (1999). Since then matter-element theory has been applied not only to mathematics but also to system theory and a number of other disciplines. New methods of system matter-element analysis were created and adapted to suit its applications in system theory. Systems were considered as a set of matter-elements, with each element consisting of objects, characteristics and values which participate in a range of processes and transformations (Tang et al. 2009). Following on from this idea, matter-element analysis now includes the following basic steps: firstly, the system is divided into matter-elements (objects). Analysis or evaluation factors are then selected and classes are defined. Class intervals for each factor are then also defined. For the classes the range of values is called the classical domain, while the whole range of values for all classes is called the segmented domain. Thirdly, the correlation degree for each single factor (in other words how well each factor matches the criteria for the category) is calculated. Finally, the integrated correlation degree of matter-elements for each class is calculated through model integration methods such as the weighted average method. The class (which includes the maximum integrated correlation degree) defines the grade the matter-element falls within.

The matter-element analysis theory is a crossing edge subject of the systems of science, cognitive science and mathematics. It has been widely used in the comprehensive evaluation of many fields (Jing et al. 2012). In this paper, the matter-element analysis theory is applied to NOM process evaluation in order to establish an assessment model of treatment process multi-index parameters. With the assessment results described by quantitative values, the value of the treatment process is fully reflected.

The definition of the matter-element

The name of a given thing is N; the quantity value of a thing's characteristics (c) is v; ordered three tuple R= (N, c, v) as the basic element of a thing is composed from v. If the thing N has multiple characteristics, its n characteristics (c1,c2, ···, cn) and corresponding quantity values (v1,v2, ···, vn) constitutes an array R as follows: 
formula
9
where R is n-dimensional matter-element, it is denoted as R = (N, C, V).

Classical domain and segmented domain

If the matter-element awaiting assessment is divided into j grades, the value range of each eigenvalue is the classical domain. The value range of all grade characteristics (ci) is the segmented domain.

Correlation function and correlation degree

The relevancy formula of each quality level which belongs to each individual characteristic is as follows: 
formula
10
where 
formula
If is the weight coefficient of , the correlation degree of the jth level about is as shown in Equation (11): 
formula
11
The evaluation grade of the matter-element awaiting assessment is as shown in Equation (12): 
formula
12

RESULTS AND DISCUSSION

Build the evaluation index system

We have selected three comprehensive indexes, WPI, DWTPI and DWQI, to set up an evaluation index system. According to the survey data in the literature and the existing engineering and social economic data, eight types of NOM treatment processes mentioned above have been selected as the research object in order to analyze the applicability of the NOM treatment process. The eight treatment processes specifically include coagulation + sedimentation + filtration (CSF), coagulation + sedimentation + filtration + adsorption (CSFA), coagulation + sedimentation + filtration + ozone (CSFO), coagulation + sedimentation + ozone + biological activated carbon + filtration (CSOBF), MIEX + coagulation + sedimentation + filtration (MCSF), coagulation + sedimentation + filtration + membrane (CSFM), coagulation + sedimentation + adsorption + membrane (CSAM) and coagulation + sedimentation + ozone + membrane (CSOM). Combining the current situation of water quality and the NOM removal efficiency of each treatment process, three comprehensive indexes, which influence the NOM treatment process value have been selected as indicator elements. The indicator elements of WPI (X1) include NOM, COD (chemical oxygen demand), TP (total phosphorus), TN (total nitrogen), NH3-N (ammonia nitrogen) and DO (dissolved oxygen); the indicator elements of DWTPI (X2) include project investment, operation cost, treatment area, running stability, process maturity, operational difficulty level, environmental benefits and removal efficiency of NOM; and the indicator elements of DWQI (X3) include COD, Fe, Mn, NH3-N, total number of bacteria communities (TNBC), nitrate, trihalomethanes (THMs) and turbidity. The basic data used in the calculation are derived from the above cited literature (Zhang & Zhao 2014).

WPI can not only produce the quantitative analysis and the level of pollution for raw water, but it is also easy to compare and evaluate the pollution levels of different areas (Lavie et al. 2014). According to the weight of a single pollution factor, the measured pollutant concentrations, and the pollutants standard values of appropriate categories shown in Table 1, the X1 is calculated by Equation (1) and the results are presented in Table 5. Seen from Table 1, there are mainly six elements in the raw water (NOM, COD, TP, TN, NH3-N, and DO). They have indirect influence on other parameters of groundwater, such as TDS, total hardness and conductivity, thus they can affect groundwater quality to some extent. NOM is closely related to daily life; water with high NOM is not suitable for drinking, washing and industry. The maximum NOM is 4.00 mg/L in CSFM; the minimum is 3.20 mg/L in CSF. COD reflects the organic matter's concentration; whether water is polluted by organics can be judged from the value of COD. The highest value of COD is 6.00 mg/L in CSFM; the lowest value is 4.50 mg/L in CSF. COD is slightly higher, which may be due to the organic pollution. The TP, TN and NH3-N reflect the eutrophication level of groundwater. The minimum values of TP, TN and NH3-N are 0.180, 0.74 and 0.581 mg/L in CSFO, CSFO and CSFA, respectively. The DO is an indicator of the self-purification ability of water.

Table 1

The data for calculating the evaluation index X1 (WPI)

 NOM (mg/L)COD (mg/L)TP (mg/L)TN (mg/L)NH3-N (mg/L)DO (mg/L)
CCSF 3.20 4.50 0.190 0.76 0.600 4.20 
CCSFA 3.41 5.00 0.200 0.75 0.581 4.00 
CCSFO 3.92 5.50 0.180 0.74 0.640 4.12 
CCSOBF 3.85 5.40 0.192 0.78 0.609 4.21 
CMCSF 3.41 4.80 0.194 0.79 0.613 4.09 
CCSFM 4.00 6.00 0.191 0.80 0.650 4.23 
CCSAM 3.59 5.20 0.189 0.81 0.630 4.18 
CCSOM 3.87 5.19 0.187 0.79 0.625 4.15 
Si 1.00 6.00 0.200 1.00 1.000 5.00 
 0.50 0.10 0.10 0.10 0.10 0.10 
 NOM (mg/L)COD (mg/L)TP (mg/L)TN (mg/L)NH3-N (mg/L)DO (mg/L)
CCSF 3.20 4.50 0.190 0.76 0.600 4.20 
CCSFA 3.41 5.00 0.200 0.75 0.581 4.00 
CCSFO 3.92 5.50 0.180 0.74 0.640 4.12 
CCSOBF 3.85 5.40 0.192 0.78 0.609 4.21 
CMCSF 3.41 4.80 0.194 0.79 0.613 4.09 
CCSFM 4.00 6.00 0.191 0.80 0.650 4.23 
CCSAM 3.59 5.20 0.189 0.81 0.630 4.18 
CCSOM 3.87 5.19 0.187 0.79 0.625 4.15 
Si 1.00 6.00 0.200 1.00 1.000 5.00 
 0.50 0.10 0.10 0.10 0.10 0.10 

DWTPI is a comprehensive indicator which can evaluate different treatment processes quantitatively. It has eight aspects including project investment, operation cost, treatment area, running stability, process maturity, operational difficulty level, environmental benefits and NOM removal efficiency. According to the value of variables , the maximum and the minimum values of the data series and the values normalized by variables shown in Table 2, the X2 is calculated by Equation (4), and the results are presented in Table 3.

Table 2

The data for calculating the evaluation index X2 (DWTPI)

xmaxxjxmin
Project investment (×104 RMB)Operation cost (×104 RMB)Treatment area (103 m2)Running stability (%)Process maturity (%)Operational difficulty level (%)Environmental benefits (%)NOM removal efficiency (%)
CSF 52.5 21 22 10.5 52.5 10.5 10.5 21 
12.8 11 18 48 12 
9.5 4.75 9.5 6.65 38 4.75 4.75 9.5 
CSFA 52.5 30.45 21 15.75 47.25 10.5 10.5 21 
16.9 15 17.5 35 8.5 5.5 14 
9.5 7.6 9.5 4.75 28.5 4.75 4.75 9.5 
CSFO 84 42 30.45 21 42 8.4 15.75 30.45 
35 25.9 20 12 25 13 20 
19 19 9.5 9.5 19 2.85 4.75 14.25 
CSOBF 52.5 36.75 21 21 42 10.5 15.75 36.75 
28 13 16.78 16 37 14 22 
9.5 7.6 14.25 14.25 28.5 1.9 9.5 14.25 
MCSF 105 52.5 31.5 26.25 42 9.45 21 36.75 
50 20 20 22 30 18.5 28 
38 12.35 14.25 14.25 14.25 2.85 9.5 14.25 
CSFM 94.5 94.5 21 31.5 26.25 10.5 31.5 42 
69 60 15 26 20 8.5 25 25 
28.5 47.5 4.75 14.25 9.5 1.9 9.5 19 
CSAM 105 84 31.5 36.75 36.75 8.4 31.5 52.5 
60 50 27 35 27 27 45 
38 28.5 14.25 23.75 23.75 1.9 9.5 19 
CSOM 157.5 126 15.75 52.5 31.5 6.3 42 52.5 
95 101 11 45 28 35 42 
66.5 57 4.75 28.5 17.1 1.9 19 19 
xmaxxjxmin
Project investment (×104 RMB)Operation cost (×104 RMB)Treatment area (103 m2)Running stability (%)Process maturity (%)Operational difficulty level (%)Environmental benefits (%)NOM removal efficiency (%)
CSF 52.5 21 22 10.5 52.5 10.5 10.5 21 
12.8 11 18 48 12 
9.5 4.75 9.5 6.65 38 4.75 4.75 9.5 
CSFA 52.5 30.45 21 15.75 47.25 10.5 10.5 21 
16.9 15 17.5 35 8.5 5.5 14 
9.5 7.6 9.5 4.75 28.5 4.75 4.75 9.5 
CSFO 84 42 30.45 21 42 8.4 15.75 30.45 
35 25.9 20 12 25 13 20 
19 19 9.5 9.5 19 2.85 4.75 14.25 
CSOBF 52.5 36.75 21 21 42 10.5 15.75 36.75 
28 13 16.78 16 37 14 22 
9.5 7.6 14.25 14.25 28.5 1.9 9.5 14.25 
MCSF 105 52.5 31.5 26.25 42 9.45 21 36.75 
50 20 20 22 30 18.5 28 
38 12.35 14.25 14.25 14.25 2.85 9.5 14.25 
CSFM 94.5 94.5 21 31.5 26.25 10.5 31.5 42 
69 60 15 26 20 8.5 25 25 
28.5 47.5 4.75 14.25 9.5 1.9 9.5 19 
CSAM 105 84 31.5 36.75 36.75 8.4 31.5 52.5 
60 50 27 35 27 27 45 
38 28.5 14.25 23.75 23.75 1.9 9.5 19 
CSOM 157.5 126 15.75 52.5 31.5 6.3 42 52.5 
95 101 11 45 28 35 42 
66.5 57 4.75 28.5 17.1 1.9 19 19 
Table 3

The index X2 value of NOM treatment process evaluation

yjProject investmentOperation costTreatment areaRunning stabilityProcess maturityOperational difficulty levelEnvironmental benefitsNOM removal efficiencyX2 = Σyi
CSF 0.923 0.615 0.261 0.390 0.690 0.739 0.217 0.217 4.053 
CSFA 0.828 0.676 0.304 0.795 0.347 0.652 0.131 0.391 4.124 
CSFO 0.754 0.700 0.499 0.783 0.261 0.748 0.750 0.355 4.848 
CSOBF 0.918 0.815 0.625 0.741 0.630 0.709 0.720 0.344 5.503 
MCSF 0.821 0.810 0.667 0.354 0.568 0.326 0.783 0.611 4.938 
CSFM 0.386 0.734 0.369 0.319 0.627 0.767 0.705 0.261 4.168 
CSAM 0.821 0.613 0.261 0.135 0.250 0.323 0.795 0.776 3.974 
CSOM 0.687 0.362 0.432 0.313 0.757 0.250 0.696 0.687 4.183 
yjProject investmentOperation costTreatment areaRunning stabilityProcess maturityOperational difficulty levelEnvironmental benefitsNOM removal efficiencyX2 = Σyi
CSF 0.923 0.615 0.261 0.390 0.690 0.739 0.217 0.217 4.053 
CSFA 0.828 0.676 0.304 0.795 0.347 0.652 0.131 0.391 4.124 
CSFO 0.754 0.700 0.499 0.783 0.261 0.748 0.750 0.355 4.848 
CSOBF 0.918 0.815 0.625 0.741 0.630 0.709 0.720 0.344 5.503 
MCSF 0.821 0.810 0.667 0.354 0.568 0.326 0.783 0.611 4.938 
CSFM 0.386 0.734 0.369 0.319 0.627 0.767 0.705 0.261 4.168 
CSAM 0.821 0.613 0.261 0.135 0.250 0.323 0.795 0.776 3.974 
CSOM 0.687 0.362 0.432 0.313 0.757 0.250 0.696 0.687 4.183 

The DWTPI of eight NOM treatment process is shown in Table 3. The higher indices illustrate that the process comprehensive efficiency is higher which means that the process is suitable for NOM removal. As seen from the values (X2) in Table 3, the maximum value is 5.503 in CSOBF which means that this drinking water treatment process is most suitable for NOM removal from raw water.

DWQI can produce a quantitative description for the effluent of a drinking water treatment process which can compare and evaluate the effluent quality of different water treatment processes. The drinking water quality index of the grading range is determined by water quality standards according to environmental quality standards for drinking water. According to the measured drinking water quality indexes concentrations , the drinking water quality index standard values of appropriate categories , the single factor water quality index and the weight of the single water quality factor shown in Table 4, the X3 is calculated by the formula in Equation (5) and the results are presented in Table 5. As seen from Table 5, through comparing the values of different water treatment processes, the maximum DWQI (X3) is 0.744 in CSF which means that the effluent quality of CSF is better than other treatment processes.

Table 4

The data for calculating the evaluation index X3 (DWQI)

aiCOD (mg/L)Fe (mg/L)Mn (mg/L)NH3-N (mg/L)TNBC (CFU/100 mL)Nitrate (mg/L)THMs (mg/L)Turbidity (NTU)X3
CSF 2.0 0.28 0.08 0.31 75 0.82 0.75 0.744 
CSFA 1.5 0.25 0.075 0.3 76 7.5 0.76 0.81 0.698 
CSFO 1.2 0.19 0.072 0.21 90 5.2 0.51 0.5 0.524 
CSOBF 1.4 0.22 0.071 0.25 80 0.80 0.71 0.639 
MCSF 0.98 0.15 0.056 0.17 86 4.1 0.60 0.41 0.448 
CSFM 0.52 0.10 0.021 0.11 64 4.5 0.51 0.32 0.305 
CSAM 0.11 0.11 0.051 0.09 56 5.2 0.88 0.39 0.354 
CSOM 0.57 0.05 0.02 0.05 69 3.8 0.35 0.07 0.200 
 3.00 0.30 0.10 0.50 100 10 1.00 1.00  
 0.25 0.10 0.10 0.10 0.05 0.05 0.10 0.25  
 0.67 0.93 0.80 0.62 0.75 0.70 0.82 0.75  
aiCOD (mg/L)Fe (mg/L)Mn (mg/L)NH3-N (mg/L)TNBC (CFU/100 mL)Nitrate (mg/L)THMs (mg/L)Turbidity (NTU)X3
CSF 2.0 0.28 0.08 0.31 75 0.82 0.75 0.744 
CSFA 1.5 0.25 0.075 0.3 76 7.5 0.76 0.81 0.698 
CSFO 1.2 0.19 0.072 0.21 90 5.2 0.51 0.5 0.524 
CSOBF 1.4 0.22 0.071 0.25 80 0.80 0.71 0.639 
MCSF 0.98 0.15 0.056 0.17 86 4.1 0.60 0.41 0.448 
CSFM 0.52 0.10 0.021 0.11 64 4.5 0.51 0.32 0.305 
CSAM 0.11 0.11 0.051 0.09 56 5.2 0.88 0.39 0.354 
CSOM 0.57 0.05 0.02 0.05 69 3.8 0.35 0.07 0.200 
 3.00 0.30 0.10 0.50 100 10 1.00 1.00  
 0.25 0.10 0.10 0.10 0.05 0.05 0.10 0.25  
 0.67 0.93 0.80 0.62 0.75 0.70 0.82 0.75  

According to the above analysis, the NOM treatment process evaluation weights are calculated and are shown in Table 5.

Table 5

The index values of NOM treatment process evaluation

ProcessWPIDWTPIDWQI
CSF 1.489 4.053 0.744 
CSFA 1.546 4.124 0.698 
CSFO 1.607 4.848 0.524 
CSOBF 2.127 5.503 0.639 
MCSF 1.751 4.938 0.448 
CSFM 1.669 4.168 0.305 
CSAM 1.707 3.974 0.354 
CSOM 1.896 4.183 0.200 
ProcessWPIDWTPIDWQI
CSF 1.489 4.053 0.744 
CSFA 1.546 4.124 0.698 
CSFO 1.607 4.848 0.524 
CSOBF 2.127 5.503 0.639 
MCSF 1.751 4.938 0.448 
CSFM 1.669 4.168 0.305 
CSAM 1.707 3.974 0.354 
CSOM 1.896 4.183 0.200 

The calculation of index weight

The attribute set (U) and attribute feature set (P = C ∪ D) are established based on the data in Table 5. Condition attribute subset (C) and decision attribute set (D) are the set of the evaluation index and the results of fuzzy synthetic evaluation, respectively. After combining environmental quality standards for surface water and drinking water, the data are analysed using the method of discrete analysis, and the discrete standard is as follows:

  • Index X1:0 - (X1 < 1.7), 1 - (1.7 < X1 < 1.8), 2 - (X1 > 1.8);

  • Index X2:0 - (X2 < 3.8), 1 - (3.8 <X2 < 4.3), 2 - (X2 > 4.3);

  • Index X3:0 - (X3 > 0.8), 1 - (0.4 < X3 < 0.8), 2 - (X3 < 0.4);

Decision attribute set (D): 0 - low value, 1- middle value, 2 - high value.

The attribute feature and decision feature of the decision attribute set of each NOM treatment process after discrete analysis are calculated and classified. Then substituting the classified results into Equations (7) and (8) produces the importance degree and weights of each condition attribute which are as follows: 
formula

The established evaluation matrix and the evaluation of the matter-element analysis

The classical domain (RI, RII, RIII, RIV, RV) and the segmented domain (Rp) of the analytical model of the NOM treatment process value assessment are shown in Table 6. Then substituting the correlation degree calculated by Equation (10) and the weights of each condition attribute into Equation (11) produces the values of the eight NOM treatment processes. The results are shown in Table 7.

Table 6

Classical domain and segmented domain

IndexHigh RIRelatively high RIIMiddle RIIIRelatively low RIVLow RVRp
X1 ‹2.0, 3.0› ‹1.8, 2.0› ‹1.7, 1.8› ‹1.6, 1.7› ‹0, 1.6› ‹0, 3› 
X2 ‹4.8, 5.2› ‹4.2, 4.8› ‹3.8, 4.2› ‹3.4, 3.8› ‹0, 3.4› ‹0, 5.2› 
X3 ‹0, 0.3› ‹0.3, 0.5› ‹0.5, 0.7› ‹0.7, 0.9› ‹0.9, 1.0› ‹0, 1.0› 
IndexHigh RIRelatively high RIIMiddle RIIIRelatively low RIVLow RVRp
X1 ‹2.0, 3.0› ‹1.8, 2.0› ‹1.7, 1.8› ‹1.6, 1.7› ‹0, 1.6› ‹0, 3› 
X2 ‹4.8, 5.2› ‹4.2, 4.8› ‹3.8, 4.2› ‹3.4, 3.8› ‹0, 3.4› ‹0, 5.2› 
X3 ‹0, 0.3› ‹0.3, 0.5› ‹0.5, 0.7› ‹0.7, 0.9› ‹0.9, 1.0› ‹0, 1.0› 
Table 7

The comprehensive evaluation results of NOM treatment processes

ProcessHigh(a)Relatively high(b)Middle(c)Relatively low(d)Low(e)Evaluation grade
CSF −0.525 −0.298 0.024 −0.399 −0.289 
CSFA 0.127 −0.401 −0.255 −0.464 −0.169 
CSFO −0.271 −0.114 0.034 −0.163 −0.388 
CSOBF −0.086 −0.255 −0.408 0.354 −0.291 
MCSF −0.406 −0.151 −0.392 −0.206 0.127 
CSFM 0.024 −0.129 −0.307 −0.144 −0.296 
CSAM −0.185 −0.011 −0.375 −0.297 −0.151 
CSOM −0.102 −0.207 0.171 −0.212 −0.200 
ProcessHigh(a)Relatively high(b)Middle(c)Relatively low(d)Low(e)Evaluation grade
CSF −0.525 −0.298 0.024 −0.399 −0.289 
CSFA 0.127 −0.401 −0.255 −0.464 −0.169 
CSFO −0.271 −0.114 0.034 −0.163 −0.388 
CSOBF −0.086 −0.255 −0.408 0.354 −0.291 
MCSF −0.406 −0.151 −0.392 −0.206 0.127 
CSFM 0.024 −0.129 −0.307 −0.144 −0.296 
CSAM −0.185 −0.011 −0.375 −0.297 −0.151 
CSOM −0.102 −0.207 0.171 −0.212 −0.200 

Table 7 shows that in addition to the MCSF treatment process, the rest of the NOM treatment processes have a certain value; the CSFA and CSFM treatment processes have the highest values which means that these treatment processes can remove the NOM in drinking water effectively. It also illustrates that CSAM has high feasibility, which has important significance for guaranteeing the safety of drinking water.

CONCLUSION

In order to avoid the influence of artificial subjective factors on the evaluation results, a hybrid evaluation model based on a matter-element model and rough set theory was used to evaluate eight NOM removal processes. Based on social and economic data and the routine monitoring data of water treatment process effluent, the evaluation index system of NOM treatment processes was constructed by WPI, DWTPI and DWQI. The research evaluated the eight NOM treatment processes through the combination of rough set theory and a matter-element analysis model and the result indicated that, in addition to the MCSF treatment process, the rest of the NOM treatment processes had a certain value and the CSFA and CSFM treatment processes had the highest values which meant that these treatment processes could remove the NOM in drinking water effectively. It also illustrated that the coagulation + sedimentation + adsorption + membrane (CSAM) treatment process had high feasibility, which has important significance for guaranteeing the safety of drinking water.

ACKNOWLEDGEMENTS

We are very grateful for the financial support of the National Natural Science Foundation of China (NSFC) under Grant No. 51308373 and No. 51308385.

REFERENCES

REFERENCES
Burchard-Levine
A.
Liu
S.
Vince
F.
Li
M.
Ostfeld
A.
2014
A hybrid evolutionary data driven model for river water quality early warning
.
Journal of Environmental Management
143
,
8
16
.
Cai
W.
1999
Extension theory and its application
.
Chinese Science Bulletin
44
(
17
),
1538
1548
.
Dias
J. M.
Oliveira
R.
Semmens
M.
2013
Modeling NOM removal by softening in a surface water treatment plant
.
Water Science and Technology
67
(
5
),
1008
1016
.
Houtman
C. J.
Kroesbergen
J.
Lekkerkerker-Teunissen
K.
van der Hoek
J. P.
2014
Human health risk assessment of the mixture of pharmaceuticals in Dutch drinking water and its sources based on frequent monitoring data
.
Science of the Total Environment
496
,
54
62
.
Islam
N.
Sadiq
R.
Rodriguez
M. J.
Francisque
A.
2013
Evaluation of source water protection strategies: a fuzzy-based model
.
Journal of Environmental Management
121
,
191
201
.
Jing
J.
Hui
Q.
Yu-Fei
C.
Wen-Juan
X.
2012
Assessment of groundwater quality based on matter element extension model
.
Journal of Chemistry
37
,
78
85
.
Lavie
E.
Morabito
J. A.
Salatino
S. E.
2014
Application of an integrated water quality index in the Argentine Andean Piedmont
.
Tecnologia y Ciencias del Agua
5
(
4
),
23
37
.
Pai
P.
Lee
F.
2010
A rough set based model in water quality analysis
.
Water Resources Management
24
(
11
),
2405
2418
.
Tang
J.
Wang
C.
Lin
N.
Li
Z.
Li
H.
Mao
Z.
2009
Application of matter-element model in soil nutrient evaluation of ecological fragile region
.
Chinese Geographical Science
19
(
2
),
168
176
.