![]() |
Citation: | Zhihua Cai, Xincai Wu. Association Rule Discovery and Its Applications. Journal of Earth Science, 2001, 12(3): 279-282. |
Data mining, i.e., mining knowledge from large amounts of data, is a demanding field since huge amounts of data have been collected in various applications. The collected data far exceed people's ability to analyze it. Thus, some new and efficient methods are needed to discover knowledge from large database. Association rule discovery is an important problem in knowledge discovery and data mining. The association mining task consists of identifying the frequent item sets and then forming conditional implication rules among them. In this paper, we describe and summarize recent work on association rule discovery, offer a new method to association rule mining and point out that association rule discovery can be applied in spatial data mining. It is useful to discover knowledge from remote sensing and geographical information system.
Diosgenin (CAS number 512-04-9) is an important steroidal metabolite used as a starting material for the synthesis of steroidal drugs. It is mainly used as the starting material for partial synthesis of oral contraceptives, sex hormones and other steroids. Possible sources of plant material for the isolation of diosgenin are the plants of Dioscorea, Costus and Trigonella. In China, Dioscorea zingiberensis is the dominant source for the production of disogenin. The main producing areas are Shiyan and Enshi of Hubei Province, where more than 1 500 t of diosgenin is produced annually. Diosgenin has become the main export product of these areas.
In the traditional process of diosgenin production, the tuber of Dioscorea zingiberensi was first washed, ground and then hydrolyzed by HCl. The acidic slurry is then centrifuged. The supernatant fluid is discharged as wastewater. The residue is washed by water and centrifuged, while the supernatant fluid is discharged as wastewater and the residue is extracted by petroleum ether and diosgenin is then obtained by crystallization. The residue after extraction is disposed of as solid waste. For the whole traditional process of diosgenin production, more than 500 t of mixed wastewater per ton of diosgenin is generated. The chemical oxygen demand (COD) concentration of wastewater centrifuged out from the original acidic slurry and that from the first wash are up to 130 000 and 110 000 mg/L, respectively. Also, the COD of mixed wastewater is in the range of 30 000–50 000 mg/L. This wastewater shows relatively low biodegradability, with the biochemical oxygen demand (BOD)/COD ratio of about 0.27.
Biological processes, including anaerobic process using upflow anaerobic sludge bed (UASB) (Shan et al., 2003) and yeast (Song et al., 2004), and integrated anaerobic and aerobic processes (Li et al., 2004; Liu et al., 2004; Zhang et al., 2003) have been applied in the treatment of diosgenin wastewater and up to 92% COD elimination has been reported. However, due to high loads of pollutants in diosgenin wastewater, the mixed wastewater is too strong to be directly fed into a biological reactor. In most cases, diosgenin wastewater needs dilution before bio-treatment, which makes it a less viable and not economical approach. Dan et al. (2003) reported microeletrolysis as a pretreatment process of diosgenin wastewater; this method, however, is associated with operational difficulty and potential hazard of intermediate reaction products. Therefore, effective pretreatment alternatives are urgently needed prior to the bio-degradation process.
The coagulation-flocculation process has advantages of simple design and equipment, low capital costs as well as high efficiency in removing dissolved and colloidal particles (O'Melia et al., 1999; Vilge'-Ritter et al., 1999). This process has been extensively used in the pretreatment of high strength wastewater, such as landfill leachate (Tatsi et al., 2003; Amokrane et al., 1997), oily wastewater from petroleum refinery (Demírcí et al., 1998) and beverage industrial wastewater (Amuda and Amoo, 2007).
Aluminum sulfate and ferric chloride are the most commonly used conventional coagulants. Once added into water, metal ions instantaneously undergo a series of hydrolytic reaction to produce species that play an essential role in coagulation. Coagulation efficiency using metal salts largely depends on physiochemical factors such as the coagulation conditions and wastewater characteristics. The disadvantage of conventional coagulants is that it is extremely difficult to control the degree of hydrolysis. In contrast, polymeric coagulants are pre-polymerized metal salts that contain a range of relatively stable hydrolysis species. By optimizing preparation conditions, the most favorable species for coagulation, e.g., species carrying highly positive charges, can be produced and thus coagulation efficiency improved. Pre-polymerized coagulants have been widely used in recent years because they are more effective than monomeric salts under a variety of water quality conditions, especially at lower temperatures and in broader pH range, with less coagulant dosage and metal residuals. Polyferric sulfate (PFS) is an alternative polymeric iron-based coagulant with molecular formula of [Fe2(OH)n(SO4)3–n/2]m (n < 2, m > 10) (Cheng, 2001). PFS contains a range of pre-formed hydrolysis iron (Ⅲ) species, including monomeric and polymeric species. The higher surface activity and charge neutralizing capacity of PFS make it more effective than conventional coagulants (Jiang and Graham, 1998). Jiang et al. (1993) had proved the superiority of PFS to the conventional iron salts in removing turbidity, algae, color and natural organic matter (NOM). Synthetic polyelectrolytes have received far more attention in water treatment in recent years (Ahmad et al., 2008). The application of organic polymers in the drinking water treatment process can increase the settling rate, reduce costs, improve the finished water quality, and provide better dewatering characteristics or sludge and reduce sludge volume (Kawarnura, 1991). Although coagulation-flocculation is widely used in water and wastewater treatment, up to date, little work has been done on the coagulation-flocculation pretreatment of diosgenin wastewater.
The objective of this study is to investigate the feasibility of coagulation-flocculation as a pretreatment process of diosgenin wastewater using PFS as primary coagulant and a cationic polyacrylamide (CPAM) as coagulant aid. The effects of main influential factors, such as effluent pH and coagulant dose on coagulation performance, were examined and operational conditions were optimized to achieve better effluent quality. The effect of CPAM dose on coagula-tion efficiency was also tested.
Diosgenin wastewater under investigation was mixed liquor collected from a diosgenin manufacturing facility in Hubei Province, China. All samples were kept in a refrigerator at 5 ℃ before use.
Polyferric sulfate used in this study was lab-prepared and brown-yellow in color. The final product was in liquid form. The main properties of polyferric sulfate are listed in Table 1. Cationic polyacrylamide with molecular weight of more than 3×106 was of reagent grade and prepared as 0.1% (w/v) solu-tion by dissolving into deionized water.
![]() |
Coagulation/flocculation tests were conducted using a conventional jar test apparatus (TA-6, China). In each run, 1 L of molasses effluent sample was poured into an acrylic plastic beaker. The pH values were adjusted to desire levels using 30% (w/v) lime solution. PFS was added and the coagulation began with rapid mixing of 300 revolutions per minute (rpm) for 2 min, followed by slow stirring of 60 rpm for 10 min. Finally, the water samples were allowed to settle for 50 min. Following sedimentation, approximately 150 mL samples were withdrawn with a plastic syringe from near 2 cm below the liquid-air interface for chemical analysis. In the case of coagulant aid, CPAM was added prior to flocculation stage. All the experiments were carried out at ambient temperature of 25–28 ℃.
The pH value was measured with a pH meter (PHS-25, China). COD was determined in accordance with the standard methods (APHA, 1998). Turbidity was directly read on a turbidity meter (TDT-2, China) in a nephelometric turbidity unit, NTU. FeT (total iron) was measured by inductively coupled plasma-mass spectroscopy (IRIS Intrepid Ⅱ XSP, Thermo Elemen-tal, USA).
The acidic diosgenin wastewater was dark brown in color and had high COD content, as listed in Table 2.
![]() |
Residual turbidity of diosgenin wastewater varied dramatically with pH, as shown in Fig. 1. Although the initial turbidity of raw effluent is rather low, as shown in Table 2, it reached a high level of more than 90 NTU at pH 5. Turbidity then gradually decreased with a lower level occurring at neutral pH. Turbidity increased to a second peak of around 62 NTU at pH 8, but decreased with further increase in pH.
In addition to neutralizing agent, lime can be used as coagulant or coagulant aid. Georgiou et al. (2003) reported desirable COD reduction of 50%–60% in the treatment of cotton textile wastewater using lime alone. In this study, however, lime precipitation performed poorly in organic removal since COD reduction was less than 2% (data not shown). This suggests that lime has little effect on organic removal from diosgenin wastewater.
In order to determine the dissolved organic fraction in diosgenin wastewater, effluent samples were filtered through a 0.45 μm filter membrane since 0.45 μm is the size operationally defined to isolate dissolved organic matter.
Figure 2 shows filterable COD and residual turbidity at varying pH. Insoluble COD varied in a narrow margin of 8.0%–11.5% over a broad pH range of 3–12, indicating that the major fractions of organic compounds in diosgenin wastewater are in dissolved forms. Residual turbidity was rather low (approximately 1 NTU) over the tested pH range, with the only exception of 56 NTU at pH 12.
Experiments of influential variables were performed by changing one variable while keeping the other one constant. In this run, polyferric sulfate dose was fixed at 100 mg/L while pH varied in the range of 5.0–9.0.
Figure 3 shows the pH effect on the coagulation of diosgenin wastewater using polyferric sulfate. COD removal efficiency increased slowly with effluent pH, reaching the highest level (14%–15%) in the pH range of 6.5–8.0. At pH 9.0, however, COD reduction dropped sharply to approximately 6%.
Lower turbidity removal was observed in the pH range of 5.5–7.0 while higher removal efficiency was obtained in the pH range of 7.5–8.5. At pH 9.0, turbidity removal sharply reduced to less than 40% (Fig. 3a). Residual turbidity initially decreased with increasing pH and the lowest turbidity occurred in neutralweakly alkaline region (pH=7.0–8.5). Further increase in pH led to increased turbidity, i.e., 25 NTU at pH 9.0 (Fig. 3b). The difference between the turbidity removal efficiency and residual turbidity lies in different initial turbidity levels. At pH 7.0, for instance, effluent turbidity was the lowest over the pH range tested. Therefore, even residual turbidity was in the lowest level, the percentage of turbidity removal at pH 7.0 was relatively lower than that under other pH conditions.
The curve of residual iron concentration against pH is similar to that of residual turbidity with the lowest levels occurring in the pH range of 7.0–8.5, which favors the formation of ferric hydroxide that readily precipitates.
In the experiment of testing the effect of polyferric sulfate dose, effluent pH selected was 7.0 to reduce lime consumption and sludge production. The dosage of polyferric sulfate varied in the range of 10–500 mg/L.
The effect of polyferric sulfate dose on coagulation is presented in Fig. 4. Coagulation efficiency was rather low at the lowest coagulant dosage with less than 5% COD removal. There was a sharp increase in organic removal as the dose of polyferric sulfate increased from 10 to 100 mg/L; and then COD reduction reached a plateau regardless of further coagulant addition. Residual turbidity of treated effluent initially decreased with increasing coagulant dose until the lowest level occurring at 100 mg/L polyferric sulfate. Further coagulant addition did not result in better supernatant clarity but elevated turbidity. The maximum dose (500 mg/L) gave rise to turbidity higher than that without coagulant addition. The increase in turbidity indicates coagulant overdosing, which leads to the formation of a colloidal suspension. Clearly, the optimal dose of polyferric sulfate was 100 mg/L on the basis of coagulation efficiency.
Measurement of the residual metal ion content in coagulated effluent is one way to determine the most effective coagulant demand. Iron concentration in the coagulated effluent continued to rise with coagulant dosage over the dose range tested. Despite this, at the optimal coagulant dosage, residual iron dosage was negligible since the iron residue accounted for less than 1% of Fe(Ⅲ) added.
Although the degree of hydrolysis of polymeric salts can be controlled during manufacturing (Cheng, 2002), polyferric sulfate still, to a certain extent, undergoes hydrolytic reaction, as indicated by a drop in pH with coagulant dosage. On the other hand, the slope of the pH curve was rather low, suggesting that polyferric sulfate consumes alkalinity at a relatively lower rate. This confirms slow hydrolysis of polyferric sulfate compared to conventional coagulants.
Basically, the mechanisms of organic removal involve charge neutralization and adsorption. It is clear that PFS carries highly positive charge while organic compounds are negatively charged under the tested conditions. Charge neutralization occurs and the reaction products aggregate into settleable flocs in the subsequent flocculation process. Besides, PFS contains more highly polymerized iron species (gels) which enmesh organic colloids in the settling process. At pH 9.0, however, Fe(Ⅲ) hydrolysis produces increasing fraction of negatively charged Fe(OH)4-, which is ineffective in coagulation, leading to inferior coagulation efficiency.
Randtke (1988) suggested that organic removal by coagulation varies widely between 10% and 90% since removal efficiency is influenced by many factors such as coagulation conditions and the characteristics of organic matter. COD removal efficiency in this study is comparatively low, presumably due to the nature of organic compounds present in diosgenin wastewater. Coagulation is the most effective in removing macromolecular and hydrophobic organic matters but less effective in removing hydrophilic organic fractions. It is reasonable to hypothesize that filterable COD is amenable to removal through coagulation while the majority of soluble COD fraction in diosgenin wastewater is of non-hydrophobic nature and not readily subject to this process. On the other hand, up to 4 000 mg/L COD was removed by coagulation alone, which greatly reduces the organic load of the subsequent treatment processes. This suggests that coagulation can be used as a feasible pretreatment of diosgenin wastewater.
A cationic polyacrylamide was used as coagulant aid in this study. The selected initial pH and polyferric sulfate dose were fixed at 7 and 100 mg/L, respectively. CPAM was dosed in the range of 0–20 mg/L.
The effect of coagulant aid on coagulation efficiency is presented in Fig. 5. It can be seen that CPAM had little effect on COD reduction since the curve was leveled off over the CPAM dosage range investigated. On the other hand, polymer addition significantly improved turbidity removal. Removal efficiency was increased from 53% (without CPAM addition) to 82% even at the lowest CPAM dosage (1 mg/L). Maximum turbidity reduction of 93% was achieved in the dose range of 2–5 mg/L. Further increase in CPAM dosage led to reduced removal efficiency when more than 10 mg/L CPAM was dosed, suggesting an overdosing phenomenon of the polyelectrolyte.
The effect of polymers on coagulation depends upon their concentrations; at low concentration they promote coagulation by forming an interparticle bridge. At high concentration they envelope particle surfaces, hence hindering floc formation and causing suspended particles. The effective CPAM dose range was determined as 2–5 mg/L when used as coagulant aid in this study. Visual observation reveals rapid settling of flocs formed under the optimum dosage of CPAM, as compared to that without CPAM addition. This confirms that polyelectrolytes enhance floc settling rate when used as coagulant aid.
The following conclusions can be obtained from the present study.
(1) The optimal pH range was 6.5–8.0 for the pretreatment of diosgenin wastewater using polyferric sulfate, and the optimal coagulant dose was 100 mg/L.
(2) Under the optimum conditions, polyferric sulfate was capable of removing up to 15% COD and residual turbidity was in the range of 10–20 NTU.
(3) The combination of cationic polyacrylamide with polyferric sulfate does not result in significant improvement in organic removal, but substantially reduced residual turbidity. The appropriate CPAM dosage was 2–5 mg/L.
(4) Soluble organic fraction constitutes the major portion of the organic compounds in diosgenin wastewater.
(5) The turbidity of diosgenin wastewater changed sharply with effluent pH and lime addition contributes little to organic removal.
ACKNOWLEDGMENTS: This study was financially supported by the Na-tional Natural Science Foundation of China (No. 40830748), and China Postdoctoral Science Founda-tion (No. 20080440976).Agrawal R, Tmielinski T, Swami A, 1993. Mining Association Rules between Set of Items in Large Database. Proc of the ACM SIGMOD Conf on Management of Data, Washington D.C. . http://www.almaden.ibm.com |
Agrawal R, Srikant R, 1994. Fast Algorithm for Mining Association Rules. Int Conf on Very Large Data Bases, Zurich, Switzerland |
Brin S, 1997. Dynamic Itemset Counting & Implication Rules for Market Basket Data. Proc of the ACM SIGMOD int'1Conf on Management of Data. http://www -db.stanford.edu/Ullman |
Egenhofer M, 1991. Reasoning about Binary Topological Relations. Proc Sympp. SSD '91, Zurich, Switzerland, http://spatialodyssey.ursus.maine.edu |
Lu W, Han J, 1993. Discovery of General Knowledge in Large Spatial Databases. Proc Far East Workshop on GIS, Singapore. http://dbs.cs.sfu.ca/han |
Ng R T, Han J, 1994. Efficient & Effective Clustering Method for Spatial Data Mining. Proc1994Int Conf VLDB, Santiago, Chile http://dbs.cs.sfu.ca/han |
Park J S, Chen M S, Yu P S, 1995. An Effective Hash-Based Algorithm for Mining Association Rules. Proc of ACM, SIGMOD, 24(2): 175-186 |
Savasere A, 1995. An Efficient Algorithm for Mining Association Rule in Large Database. T he 21st Conf on Very Large Databases. (VLDB'95), Zurich, Switzerland |
Yen S J, Chen A, 1996. An Effective Approach to Discovering Knowledge from Large Databases. Conf on Parallel and Distributed Information System. http://www.cs.nthu.tw |
![]() |
![]() |