Abstract

The internal corrosion of cast iron and steel pipes is one of the main issues that drinking water distribution operators are facing. This study evaluated the relevance of 10 known corrosion indices according to their estimate of corrosion rate and iron particle release for 20 different water qualities. Pilot-scale contact trials were run over 45 days using cast iron and steel coupons. Corrosion rate was measured by coupon weight-loss and by an online linear polarization rate probe. Particle release was monitored by an online turbidimeter. The results showed that none of the indices properly predicted the level of risk associated with each water and that corrosion and particle release were not correlated. Two novel indices were developed to predict the corrosion and particle release risks independently of each other. The corrosion index showed a strong linear correlation with the corrosion rate of cast iron and slightly less reliable results for steel. The Particle Emission Index presented good correlation with turbidity in waters following contact with cast iron. These two indices thus showed interesting potential as tools to limit internal corrosion risks for metal pipes in water distribution networks.

INTRODUCTION

The condition of water distribution networks is a major concern for drinking water treatment suppliers. The American Water Works Association (AWWA) State of the Water Industry Report (AWWA 2015) highlighted it as the top issue facing the profession ahead of the ability to find financing for capital improvements, long-term water supply availability and public understanding of the value of water resources, systems and services.

This urgent need for the rehabilitation of water distribution networks was well-documented by the AWWA report on reinvesting in drinking water infrastructure (AWWA 2001). At the time, the AWWA estimated that re-placement of pipes could reach upwards of $250 billion over 30 years in the USA alone.

Most of the buried drinking water infrastructure in the USA was built 50 or more years ago, in the post-World War II era of rapid demographic change and economic growth (AWWA 2011). In Australia, distribution networks consist predominantly of pipes with ages on average greater than 60 years (Rajeev et al. 2014). The average age of pipes in French urban networks such as Orléans, Dunkerque and the Paris metropolitan area tend to be higher than 40 years. The same is true of many other urban areas around the world.

Corrosion, and in particular the corrosion of cast iron pipes, has been underlined as the most common problem in water distribution networks (McNeill & Edwards 2001; Rajeev et al. 2014). Estimating what proportion of distribution networks is at risk is difficult due to scarcity of data on the length, material and the internal/external environment of pipes in place. Data, collated by the USEPA (2009), estimate the total pipe mileage in the USA to over 2,000,000 miles. According to an AWWA (AWWA 2002; Grigg 2010) survey (Table 1), 22.5% of these pipes are directly concerned by corrosion: unlined cast iron (14.4%), unlined ductile iron (4.3%) and steel (3.8%). An additional 34.1% of pipes could be indirectly concerned as their linings deteriorate: cement lined ductile iron (19.7%) and cement lined cast iron (14.4%). It is expected that similar proportions are to be found in other countries, which underscores the potential impact of this issue for drinking water operators worldwide.

Table 1

Pipe material in place in the USA (Grigg 2010)

Pipe Material% of total miles of pipe
Ductile iron, cement lined 20 
Ductile iron, unlined 
Cast iron, cement lined 14 
Cast iron, unlined 14 
Steel 
PVC 17 
Asbestos cement 15 
Concrete pressure 
Polyethylene 
Other 
Misc/unknown 
Pipe Material% of total miles of pipe
Ductile iron, cement lined 20 
Ductile iron, unlined 
Cast iron, cement lined 14 
Cast iron, unlined 14 
Steel 
PVC 17 
Asbestos cement 15 
Concrete pressure 
Polyethylene 
Other 
Misc/unknown 

In the water industry, pipe degradation phenomena are often examined in terms of external factors but several studies concur that the nature of the soil and pipe environment cannot explain a large number of pipe failures observed in actual networks (Rajani & Kleiner 2012; Rezaei et al. 2015). A survey of pipes from 568 distinct locations of a UK utility, indicated that internal corrosion appeared as a larger issue than external corrosion (66% of the samples had high internal corrosion versus 34% with high external corrosion) (Rajani & Kleiner 2012). While the negative impacts of internal corrosion on water quality are widely recognized (particle release, metal leaching, disinfection control), the impacts on the condition of the pipes can be neglected at times (obstruction, leaks, bursts).

As distribution networks age, it becomes ever more important to acknowledge that internal corrosion control plays a key role, not only in preserving the quality of distributed water but also in extending the useful lifetime of pipes. Adjusting water chemistry at production sites is one of the few mitigation actions that can be effective in reducing degradation rates on the full extent of an existing system. These actions can include pH control, adjustment of calcium and alkalinity, application of inhibitors or the elaboration of suitable water blending strategies to ensure water compatibility with network materials.

A great deal of literature on internal corrosion is dedicated to indices that correlate different water quality parameters to the corrosion potential of the water. Alkalinity, or more generally dissolved inorganic carbon, is one of the main parameters integrated in the definition of such indices (Merrill & Sanks 1977; Legrand & Leroy 1995). Another parameter often taken into account is the pH and, more specifically, the saturation pH (Langelier 1936; Ryznar 1944). Other approaches incorporate the levels of chloride and sulphate ions (Larson & Skold 1958; Imran et al. 2005) or the buffer intensity (Stumm 1960; AwwaRF 1996).

Several industry guidelines suggest drinking water quality targets aimed at mitigating corrosion in distribution systems. The Langelier saturation index, for instance, is widely referenced in countries such as the USA, France, and Australia, as well as the World Health Organization (AwwaRF 1996; NHMRC Australia 2004; French Drinking Water Standards 2007; WHO 2011). In practice, however, there is no consensus on the validity of any single index and few regulations actually incorporate these recommendations into compulsory treatment requirements.

This article presents the results of a pilot-scale study of the relevance of eleven different indices for the prediction of cast iron and steel pipe corrosion effects on two levels: iron particle release and material degradation/mass loss. Twenty different water qualities were examined with varying levels of alkalinity, hardness, pH and salinity (chloride and sulphate concentrations). Corrosion rates and water turbidity were monitored online and complemented with data on coupon mass loss. The results prompted the development of two new indices, building on existing ones, to anticipate corrosion effects on particle release on the one hand and pipe wall degradation on the other.

MATERIAL AND METHODS

Water distribution system simulation

Corrosion behaviour in water distribution networks was simulated by pilot units (Figure 1) each composed of an individual 1 m3 water tank and two water circuits connected in series. The first circuit (front side of the pilot unit) is designed to study the impact of corrosion on cast iron and steel material degradation. The second circuit (back side) evaluates the impact of corrosion on water quality by particle release. Five of these pilot units ran in parallel over four testing campaigns, each testing the impact of a specific water quality.

Figure 1

Schematic of water distribution network simulation pilot.

Figure 1

Schematic of water distribution network simulation pilot.

Coupons of cast iron and steel were installed in the pilots (ASTM 2005) and put in contact with flowing water over the course of 45 days. The water was monitored daily to ensure constant quality. The flow through the pilot was maintained at 12 L/min, such that the water velocity through the online corrosion sensor was of 0.88 m/s while the water velocity through the online turbidimeter was of 0.2 m/s.

Water quality

The water qualities of study were selected to be representative of the expected range found in drinking water systems relying on various resources, from soft surface water reservoirs to high mineral content groundwater and desalinated waters represented by two samples (numbers 17 and 19). Alkalinity values were set between 50 and 300 mg/L CaCO3 with pH levels ranging from 6.6 to 8.9. Conductivities varied from 130 to 1,800 μS/cm. Table 2 summarizes the water quality parameters for the 20 waters that were tested.

Table 2

Water quality for the different trial runs

Trial run #Temperature (°C)pHAlkalinity (mg/L CaCO3)Hardness (mg/L CaCO3)Ca2+ (mg/L)Mg2+ (mg/L)Na+ (mg/L)K+ (mg/L)Cl (mg/L) (mg/L) (mg/L)CO2 (mg/L)Conductivity (μS/cm)
20 8.1 50 86 20 28 43 38 61 0.7 343 
20 7.4 300 332 120 25 38 35 365 21 730 
20 7.5 300 513 120 51 166 23 256 230 365 17 1,766 
20 8.9 50 55 20 0.6 58 0.1 138 
20 6.7 300 513 120 51 166 23 256 230 365 95 1,775 
20 6.6 300 332 120 25 38 35 365 127 745 
20 8.9 50 85 20 28 43 38 61 0.1 378 
20 8.1 50 55 20 0.6 58 0.8 133 
20 7.5 175 246 70 17 56 86 77 213 11 768 
10 20 7.5 175 246 70 17 56 86 77 213 11 763 
11 20 8.9 50 55 20 0.6 58 0.1 135 
12 20 7.4 200 276 98 27 58 244 16 584 
13 20 6.7 50 55 20 0.6 58 19 135 
14 20 8.2 80 80 32 18 28 97 1.0 251 
15 20 7.5 80 80 32 18 28 97 250 
16 20 8.3 60 60 24 11 71 0.6 158 
17 20 8.4 80 80 32 227 350 95 0.8 1,282 
18 20 7.4 50 55 20 0.6 61 135 
19 20 7.4 50 55 20 227 0.6 300 61 1,246 
20 20 8.3 80 80 32 11 96 0.9 195 
Trial run #Temperature (°C)pHAlkalinity (mg/L CaCO3)Hardness (mg/L CaCO3)Ca2+ (mg/L)Mg2+ (mg/L)Na+ (mg/L)K+ (mg/L)Cl (mg/L) (mg/L) (mg/L)CO2 (mg/L)Conductivity (μS/cm)
20 8.1 50 86 20 28 43 38 61 0.7 343 
20 7.4 300 332 120 25 38 35 365 21 730 
20 7.5 300 513 120 51 166 23 256 230 365 17 1,766 
20 8.9 50 55 20 0.6 58 0.1 138 
20 6.7 300 513 120 51 166 23 256 230 365 95 1,775 
20 6.6 300 332 120 25 38 35 365 127 745 
20 8.9 50 85 20 28 43 38 61 0.1 378 
20 8.1 50 55 20 0.6 58 0.8 133 
20 7.5 175 246 70 17 56 86 77 213 11 768 
10 20 7.5 175 246 70 17 56 86 77 213 11 763 
11 20 8.9 50 55 20 0.6 58 0.1 135 
12 20 7.4 200 276 98 27 58 244 16 584 
13 20 6.7 50 55 20 0.6 58 19 135 
14 20 8.2 80 80 32 18 28 97 1.0 251 
15 20 7.5 80 80 32 18 28 97 250 
16 20 8.3 60 60 24 11 71 0.6 158 
17 20 8.4 80 80 32 227 350 95 0.8 1,282 
18 20 7.4 50 55 20 0.6 61 135 
19 20 7.4 50 55 20 227 0.6 300 61 1,246 
20 20 8.3 80 80 32 11 96 0.9 195 

Water samples were all synthesized from reverse osmosis filtered water to which was added a mix of CaCl2, MgSO4, KCl and lime in varying concentrations to attain the desired composition. For each trial, a fresh batch of 1 m3 was prepared and run in a recirculation loop through the pilot. The pH was kept constant throughout each experiment by correcting the concentration of dissolved CO2. Temperature was controlled and maintained at approximately 20 °C.

Monitoring of corrosion impact on material degradation

Two different methods were applied to monitor corrosion effects on cast iron and steel materials. The first method is based on standard cast iron and steel coupon mass loss measurements (ASTM 1999) after 45 days of testing which represents the cumulative corrosion rate of the material. Mass was measured with an analytical balance (ED124S, Sartorius, Göttingen, Germany). The second method uses a linear polarization rate (LPR) probe (Corrater, Rohrback Cosasco, Santa Fe Spring, CA, USA), for online monitoring and can be interpreted as the instantaneous corrosion rate of the material (ASTM 2013).

Monitoring of corrosion impact on particle release

Online turbidity measurement was monitored with a turbidimeter (Turbicube 1000, Bamo, Argenteuil, France) positioned downstream of two 67.5 cm long, 50 mm diameter PVC pipes, each equipped with a 77 mm × 25 mm × 13 mm cast iron slab designed to ensure a surface to volume ratio equivalent to that of a 125 mm diameter pipe.

Corrosion indices

Eleven referenced corrosion indices were evaluated during this study. The data provided by coupon mass loss, the LPR probe and turbidity levels were used to evaluate their relevance. These indices were calculated for each water quality and correlated with the corrosion rates and turbidity measured at pilot-scale. Table 3 summarizes these indices and their associated interpretation thresholds to evaluate corrosion risk. Based on the data produced during this study, two novel indices were developed to evaluate the risks of material and water quality degradation (iron particle release).

Table 3

Conventional corrosion indices

Index nameFormulaInterpretation for the evaluation of risk of corrosionReference
Langelier Index (LSI) LSI = pH – pHs
pHs is the pH at CaCO3 equilibrium 
LSI<0 – Aggressive water for CaCO3 – Risk of corrosion
LSI=0 – Water at equilibrium
LSI>0 – Scale-forming water – can limit corrosion 
Langelier (1936)  
Calcium Carbonate Precipitation Potential (CCPP) 
a = 2, b = −2[Ca2+]–Alk c = Alk[Ca2+] − 0.5(K′S)–[Ca2+
Quantity of CaCO3 that can precipitate (CCPP > 0) or be dissolved (CCPP < 0).
To limit corrosion 4 < CCPP < 10 mg/L 
Rossum & Merrill (1983)  
Driving Force Index (DFI)  Concentrations in mol/L DFI<1 – Aggressive water– Risk of corrosion
DFI=1 – Water at equilibrium
DFI>1 – Scale-forming water 
Rossum & Merrill (1983)  
Ryznar Index (RI)  RI < 5 – Highly scale-forming
5 > RI > 6 – Mildly scale-forming
6 > RI > 7 – Equilibrium
7 > RI > 7.5 – Mildly corrosive
RI > 7.5 – Very corrosive 
Ryznar (1944)  
Larson Index (La)  Concentrations in meq/L La > 0.5 – Risk of corrosion (according to Larson)
La < 2/3 – Recommendation in Finland
La < 1 – Recommendation in Germany 
Larson & Skold (1958)  
Modified Larson Index (MLI)  Concentration in mg/L, T in °C, Hydraulic Residence Time in days MLI > 0.5 – Risk of corrosion Imran et al. (2005)  
Feigenbaum Index (Y)  A = 3.5 × 10−4, B = 0.34, C = 19, H = ([Ca2+] × /[CO2])
concentrations expressed in mg/L
expressed in mg/L of CaCO3 
Y > 500 – Mildly corrosive water
500 < Y < 200 – Moderately corrosive water
Y < 200 – Corrosive water 
Feigenbaum et al. (1978)  
Riddick Index (RCI)  Alk and TH expressed in mg/L of CaCO3
The other concentrations expressed in mg/L 
RCI < 5 – Non-corrosive
6 > RCI > 25 – Very mildly corrosive
25 > RCI > 50 – Moderately corrosive
50 > RCI > 100 – Very corrosive
RCI > 100 – Extremely corrosive 
Singley (1981)  
Pisigan Index (IPisigan
β is the buffer intensity Cond is the conductivity 
No known threshold
An increase in the index value leads to a decrease in water corrosiveness

 
Pisigan & Singley (1987)  
Leroy Index (ILeroy Alk is alkalinity and TH is hardness in °f 0.7 < ILeroy < 1.3 recommended to limit the risk of corrosion (according to Leroy) Legrand & Leroy (1995)  
Buffer intensity (β)  CT, CO3 = total concentration of carbonate species, all in moles/L No known threshold AwwaRF (1996)  
Index nameFormulaInterpretation for the evaluation of risk of corrosionReference
Langelier Index (LSI) LSI = pH – pHs
pHs is the pH at CaCO3 equilibrium 
LSI<0 – Aggressive water for CaCO3 – Risk of corrosion
LSI=0 – Water at equilibrium
LSI>0 – Scale-forming water – can limit corrosion 
Langelier (1936)  
Calcium Carbonate Precipitation Potential (CCPP) 
a = 2, b = −2[Ca2+]–Alk c = Alk[Ca2+] − 0.5(K′S)–[Ca2+
Quantity of CaCO3 that can precipitate (CCPP > 0) or be dissolved (CCPP < 0).
To limit corrosion 4 < CCPP < 10 mg/L 
Rossum & Merrill (1983)  
Driving Force Index (DFI)  Concentrations in mol/L DFI<1 – Aggressive water– Risk of corrosion
DFI=1 – Water at equilibrium
DFI>1 – Scale-forming water 
Rossum & Merrill (1983)  
Ryznar Index (RI)  RI < 5 – Highly scale-forming
5 > RI > 6 – Mildly scale-forming
6 > RI > 7 – Equilibrium
7 > RI > 7.5 – Mildly corrosive
RI > 7.5 – Very corrosive 
Ryznar (1944)  
Larson Index (La)  Concentrations in meq/L La > 0.5 – Risk of corrosion (according to Larson)
La < 2/3 – Recommendation in Finland
La < 1 – Recommendation in Germany 
Larson & Skold (1958)  
Modified Larson Index (MLI)  Concentration in mg/L, T in °C, Hydraulic Residence Time in days MLI > 0.5 – Risk of corrosion Imran et al. (2005)  
Feigenbaum Index (Y)  A = 3.5 × 10−4, B = 0.34, C = 19, H = ([Ca2+] × /[CO2])
concentrations expressed in mg/L
expressed in mg/L of CaCO3 
Y > 500 – Mildly corrosive water
500 < Y < 200 – Moderately corrosive water
Y < 200 – Corrosive water 
Feigenbaum et al. (1978)  
Riddick Index (RCI)  Alk and TH expressed in mg/L of CaCO3
The other concentrations expressed in mg/L 
RCI < 5 – Non-corrosive
6 > RCI > 25 – Very mildly corrosive
25 > RCI > 50 – Moderately corrosive
50 > RCI > 100 – Very corrosive
RCI > 100 – Extremely corrosive 
Singley (1981)  
Pisigan Index (IPisigan
β is the buffer intensity Cond is the conductivity 
No known threshold
An increase in the index value leads to a decrease in water corrosiveness

 
Pisigan & Singley (1987)  
Leroy Index (ILeroy Alk is alkalinity and TH is hardness in °f 0.7 < ILeroy < 1.3 recommended to limit the risk of corrosion (according to Leroy) Legrand & Leroy (1995)  
Buffer intensity (β)  CT, CO3 = total concentration of carbonate species, all in moles/L No known threshold AwwaRF (1996)  

RESULTS AND DISCUSSION

Corrosion rates of cast iron and steel coupons

Figure 2 presents the instantaneous corrosion rate measured by LPR (equipped with a cast iron probe) at day 45 as a function of the cumulative corrosion rate measured by mass loss after 45 days for cast iron and steel coupons. The data shows that the instantaneous corrosion rate is well correlated to the coupon mass loss analysis of each type of material. The correlation is different for the cast iron and the steel coupons which highlights the difference in corrosion behaviour of the two materials. This is particularly true for waters with low alkalinity (<50 mg/L CaCO3) and low conductivity (<135 μS/cm) that yield lower corrosion rates for steel and waters with low alkalinity (<50 mg/L CaCO3) and high conductivity (>350 μS/cm) that yield higher corrosion rates for steel. Overall, steel appears to be more sensitive to water quality as emphasized by its wider range of corrosion rates. These first results stress the practical difficulty of developing universal corrosion indices for iron pipes.

Figure 2

Instantaneous corrosion rate measured by LPR at day 45 as a function of (a) cast iron corrosion rate by coupon mass loss at day 45; (b) steel corrosion rate measured by coupon mass loss at day 45.

Figure 2

Instantaneous corrosion rate measured by LPR at day 45 as a function of (a) cast iron corrosion rate by coupon mass loss at day 45; (b) steel corrosion rate measured by coupon mass loss at day 45.

The instantaneous corrosion rates measured at day 45 were always lower than the corrosion rates measured by mass loss after 45 days, both for steel and cast iron. This was expected due to the fact that higher corrosion rates occur at the beginning of the trial runs, when coupons are new, and stabilize at lower levels over the 45 days of operation (Reiber et al. 1996). The initially higher rates are only captured via the cumulative measurements of corrosion by coupon mass loss. For the cast iron runs, the instantaneous corrosion rates ranged from 24 to 403 μm/yr for cumulative corrosion rates ranging from 73 to 513 μm/yr after 45 days. Corrosion rates greater than 200 μm/yr can be considered very high in regard to the expected loss of pipe material and reflect a highly corrosive character of the conveyed water. If pipe failure is assumed to occur when pipe wall thickness is reduced to 50% by internal corrosion, a 6 mm thick pipe can be expected to last 120 years if corrosion is maintained at 25 μm/yr. The expected lifetime is reduced to 15 years for a corrosion rate of 200 μm/yr. Differences in instantaneous corrosion rates observed for different waters correspond to a wide range of expected pipe lifetime levels covering a ratio of 1 to 16.

Particle release

The iron released by corrosion can remain attached to the pipe surface as a fixed deposit (obstruction) or behave as a loose particle that produces turbidity. Water turbidity was monitored online throughout the first 35 days of each trial run. The turbidity of day 35 was chosen as representative of the cumulative effect of corrosion on water quality for each pilot test. The day 35 turbidity ranged from 0.2 to 12 NTU, highlighting a difference in iron particle release behaviour depending on water quality. Due to the recirculation of the water in the pilot loops, particle accumulation cannot be neglected and the absolute turbidity values cannot be considered as representative of the levels occurring in actual water distribution networks. However, a comparison of the measured levels produced a classification of particle release potential on a relative scale based on the following thresholds:

  • Turbidity < 1 NTU – Low iron particle emissions

  • 1 NTU < Turbidity < 4 NTU – Moderate iron particle emissions

  • 4 NTU < Turbidity < 8 NTU – High iron particle emissions

  • Turbidity >8 NTU – Very high iron particle emissions

Figure 3 illustrates the instantaneous corrosion rate measured by LPR (equipped with a cast iron probe) after 45 days, as a function of the water turbidity measured at day 35. No significant correlation was found between these two parameters. These results highlight the role of water chemistry on corrosion itself and on the behaviour of the deposits formed on the surface of the pipe. Under certain conditions, iron particles released by corrosion tend to remain fixed onto a deposit layer on the pipe surface instead of migrating to the water column (Leroy 2012), thus having no effect on water turbidity.

Figure 3

Instantaneous corrosion rate of cast iron (day 45) as a function of water turbidity (day 35).

Figure 3

Instantaneous corrosion rate of cast iron (day 45) as a function of water turbidity (day 35).

The data showed that water which generates a low corrosion rate on cast iron material can nonetheless lead to discoloured water following significant release of iron oxide particles. This was the case for waters 5, 6 and 13 which all share an acidic pH (pH < 7). The correlation between corrosion rate and turbidity became significant and positive when those three waters were not taken into account. Overall, the apparent independent nature of internal corrosion effects on material degradation and on water quality degradation triggered the need for separate indices reflecting each of these two phenomena.

Evaluation of the relevance of conventional corrosion indices

The value of each corrosion index (CI) was calculated for each given water quality and compared to the experimental instantaneous corrosion rate data obtained from the LPR probe at day 45. The LPR probe data correlated well with the mass loss analysis of both cast iron and steel coupons (Figure 2) and was chosen as the reference value. To compute the modified Larson index (LMR), the hydraulic residence time was taken to be equal to 3. For the Riddick index, due to the absence of silica in the raw waters, the silica concentration parameter was chosen equal to 1.

Results were grouped into three categories, depending on the observed LPR corrosion rates: low for waters with corrosion rates under 50 μm/yr (9 waters), medium for corrosion rates between 50 and 100 μm/yr (4 waters) and high for waters with corrosion rates greater than 100 μm/yr. Table 4 presents the average, maximum and minimum values of each index for each water category, as well as the corresponding standard error of the mean.

Table 4

Comparison of instantaneous corrosion rates to values for conventional corrosion indices

Corrosion rate levelParameterLPR (day 45)LangelierCCPP (mg/L)TSRyznarLarsonModified Larson (HRT = 3)FeigenbaumRiddickPisiganLeroyβ
Low (<50 μm/yr) Average 36 −0.2 −0.6 1.2 7.6 0.8 0.2 164 49 0.2 0.8 2.0 
Max 48 0.4 38.8 2.5 10.3 2.0 0.2 409 95 0.5 0.9 4.5 
Min 24 −1.8 −44.7 0.0 6.6 0.3 0.1 21 19 0.0 0.6 0.8 
SEM 0.3 12.2 0.4 0.5 0.2 0.0 47.5 9.2 0.1 0.0 0.5 
Medium (50–100 μm/yr) Average 83 −0.3 −15.0 0.8 8.2 0.8 0.2 113 43 0.1 0.9 1.3 
Max 98 0.2 0.9 1.6 9.4 2.0 0.2 220 114 0.3 1.0 4.3 
Min 72 −1.0 −54.1 0.1 7.5 0.2 0.1 27 12 0.0 0.6 0.1 
SEM 0.3 13.2 0.4 0.4 0.4 0.0 39.9 23.9 0.1 0.1 1.0 
High (>100 μm/yr) Average 204 −0.1 −0.8 1.3 8.4 2.4 0.4 124 118 0.0 0.9 0.2 
Max 403 10 227 459 0.6 
Min 104 −1.0 −7.5 0.1 7.9 0.2 0.2 43 14 0.0 0.6 0.1 
SEM 33 0.2 1.2 0.3 0.2 1.2 0.1 23 54 0.0 0.1 0.1 
Corrosion rate levelParameterLPR (day 45)LangelierCCPP (mg/L)TSRyznarLarsonModified Larson (HRT = 3)FeigenbaumRiddickPisiganLeroyβ
Low (<50 μm/yr) Average 36 −0.2 −0.6 1.2 7.6 0.8 0.2 164 49 0.2 0.8 2.0 
Max 48 0.4 38.8 2.5 10.3 2.0 0.2 409 95 0.5 0.9 4.5 
Min 24 −1.8 −44.7 0.0 6.6 0.3 0.1 21 19 0.0 0.6 0.8 
SEM 0.3 12.2 0.4 0.5 0.2 0.0 47.5 9.2 0.1 0.0 0.5 
Medium (50–100 μm/yr) Average 83 −0.3 −15.0 0.8 8.2 0.8 0.2 113 43 0.1 0.9 1.3 
Max 98 0.2 0.9 1.6 9.4 2.0 0.2 220 114 0.3 1.0 4.3 
Min 72 −1.0 −54.1 0.1 7.5 0.2 0.1 27 12 0.0 0.6 0.1 
SEM 0.3 13.2 0.4 0.4 0.4 0.0 39.9 23.9 0.1 0.1 1.0 
High (>100 μm/yr) Average 204 −0.1 −0.8 1.3 8.4 2.4 0.4 124 118 0.0 0.9 0.2 
Max 403 10 227 459 0.6 
Min 104 −1.0 −7.5 0.1 7.9 0.2 0.2 43 14 0.0 0.6 0.1 
SEM 33 0.2 1.2 0.3 0.2 1.2 0.1 23 54 0.0 0.1 0.1 

The data presented highlights that none of the indices could be used as a stand-alone parameter for the evaluation of the risks of corrosion due to high variability. The same result was obtained in regard to the relevance of the indices to anticipate the risk of particle release into the water as illustrated in Table 5. Some indices, e.g. the Langelier saturation index, showed very little differentiation among the different water categories as defined by LPR values while others showed very high variability within each water category (e.g. CCCP or the TS index). The Ryznar, Larson, Modified Larson and Pisigan indices showed a relevant trend in predicting corrosion but did not systematically reflect the corrosion risk of a water sample reliably. The target threshold values commonly recommended for the CCPP (minimum 5 mg/L CaCO3) and Ryznar index (maximum 7.5) appear relevant to maintain very low corrosion rates but values did not correlate well outside of these limits. Overall, these results triggered the need to develop more robust indices to anticipate corrosion in network pipes.

Table 5

Comparison of turbidity levels to values for conventional corrosion indices

Particle release levelParameterTurbidity (day 35) NTULangelierCCPP (mg/L)TSRyznarLarsonModified Larson (HRT = 3)FeigenbaumRiddickPisiganLeroyβ
Low (<1 NTU) Average 0.2 9.6 1.9 7.6 0.6 0.2 170 31 0.1 0.9 0.9 
Max 0.4 38.8 2.5 8.1 2.0 0.3 409 95 0.1 1.0 2.4 
Min 0.0 0.0 1.0 6.6 0.2 0.1 77 12 0.0 0.6 0.1 
SEM 0.1 5.1 0.2 0.2 0.2 0.0 33.4 9.2 0.0 0.1 0.3 
Moderate (1–4 NTU) Average −0.5 −8.6 0.9 8.7 1.8 0.3 111 95 0.2 0.9 0.5 
Max 0.4 2.4 2.5 10.3 6.4 0.7 227 329 0.5 1.0 1.3 
Min −1.8 −39.6 0.0 7.5 0.3 0.2 21 19 0.0 0.6 0.1 
SEM 0.3 6.4 0.4 0.4 1.0 0.1 38.0 47.9 0.1 0.1 0.2 
High to Very High (>4 NTU) Average −0.5 −21.4 0.3 8.4 2.9 0.4 103 145 0.1 0.8 1.9 
Max 12 −1 10 220 459 4.5 
Min −1.0 −54.1 0.1 7.4 0.3 0.1 45 15 0.0 0.6 0.1 
SEM 0.1 11.6 0.1 0.4 1.8 0.2 31 80 0.0 0.1 1.0 
Particle release levelParameterTurbidity (day 35) NTULangelierCCPP (mg/L)TSRyznarLarsonModified Larson (HRT = 3)FeigenbaumRiddickPisiganLeroyβ
Low (<1 NTU) Average 0.2 9.6 1.9 7.6 0.6 0.2 170 31 0.1 0.9 0.9 
Max 0.4 38.8 2.5 8.1 2.0 0.3 409 95 0.1 1.0 2.4 
Min 0.0 0.0 1.0 6.6 0.2 0.1 77 12 0.0 0.6 0.1 
SEM 0.1 5.1 0.2 0.2 0.2 0.0 33.4 9.2 0.0 0.1 0.3 
Moderate (1–4 NTU) Average −0.5 −8.6 0.9 8.7 1.8 0.3 111 95 0.2 0.9 0.5 
Max 0.4 2.4 2.5 10.3 6.4 0.7 227 329 0.5 1.0 1.3 
Min −1.8 −39.6 0.0 7.5 0.3 0.2 21 19 0.0 0.6 0.1 
SEM 0.3 6.4 0.4 0.4 1.0 0.1 38.0 47.9 0.1 0.1 0.2 
High to Very High (>4 NTU) Average −0.5 −21.4 0.3 8.4 2.9 0.4 103 145 0.1 0.8 1.9 
Max 12 −1 10 220 459 4.5 
Min −1.0 −54.1 0.1 7.4 0.3 0.1 45 15 0.0 0.6 0.1 
SEM 0.1 11.6 0.1 0.4 1.8 0.2 31 80 0.0 0.1 1.0 

Impact of water quality parameters on corrosion rate

In addition to the evaluation of the various corrosion indices, this study gave interesting insights into the role of the various water quality parameters that were controlled over the course of the project.

The buffer intensity (β) of the water was shown to be negatively and non-linearly correlated to the corrosion rate of cast iron and, to a lesser extent, steel. This confirms previous observations that showed that the rate of iron corrosion decreased in a uniform manner as β increases (Leroy 2012). The role played by the buffer intensity is to slow down the oxidation rate of the divalent iron. This can take place in one of two ways. First, as corrosion occurs, OH ions are produced, inducing a local increase in pH (Leroy 2012). This change in pH increases the rate of oxidation of divalent to trivalent iron and thus the corrosion rate. The buffer intensity limits the variations of pH and hence slows the corrosion rate. A second mechanism is related to the increased levels in the free radical scavenger in the presence of calcium at higher buffer intensity which limits the oxidation of divalent iron by dissolved oxygen (Leroy 2012).

The concentration of chloride and sulphate ions showed a positive correlation with the corrosion rate. It is suspected that these ions combine with the divalent iron to form complexes around the anodic area of the pipe material. This results in an increase in the release of divalent iron and an increase in the corrosion rate. The results of this study showed that, for waters with similar Larson index values, a higher buffer intensity limited the effect of the chloride and sulphate ions on the corrosion rate. It thus appeared that the corrosion inhibiting role of the buffer intensity was greater than the corrosion promoting role of the chloride and sulphate ions.

The scale-forming or aggressive nature of the water, as defined by the Langelier saturation index, did not correlate well with the corrosion rate of pipes. This is particularly true for waters with low levels of alkalinity, which can be defined as scale-forming but produce high corrosion rates and for aggressive waters with high buffer intensity, which can yield slow corrosion rates. The results of this study did show, however, that if the alkalinity of the water is high enough, its scale-forming character can limit the effects of corrosion both on material degradation and iron particle release.

When evaluating the iron particle release phenomenon, it appeared that a pH lower than 7 caused increased turbidity even at low corrosion rates. This is hypothesized to be due to a slower oxidation reaction of divalent iron by oxygen at low pH. Thus, the ferrous ions produced during the corrosion process dissolve directly into the water without forming a deposit on the pipe. The dissolved iron is then oxidized in the water column, forming suspended particles that increase turbidity.

Development of novel corrosion indices

Two novel indices were developed during this project, one for each aspect of corrosion in a distribution system: pipe material degradation and water quality degradation. They build on the three existing indices that showed the most relevant correlations with the observed effects of corrosion during the pilot trials: the Buffer intensity (β), the Langelier saturation index (LSI) and the Larson index (La).

The first is referred to as the CI (Equation (1)) and was developed empirically to assess the risk of internal corrosion on the degradation of iron material in pipes.  
formula
(1)
where:
  • β = Buffer intensity in mmol/L as determined by chemical simulation of the injection of HCl

  • La = Larson index

  • p = 1 if LSI < 0 and p = −1 if LSI ≥ 0, where LSI is the Langelier saturation index

  • K = 1 if pH > 7 and K = La/(1 + La) if pH ≤ 7

Figure 4 presents the data of the instantaneous corrosion rate of cast iron measured at day 45 as a function of the CI. Data from trial 19, which produced excessively high corrosion rates due to very high chloride levels, was excluded from the correlation analysis. These results show that there was a very strong correlation between the CI and the corrosion of cast iron. The correlation of this index with the corrosion of steel was not as strong but remained significant if values for the cumulative level of corrosion are considered (F-test, p < 0.05).

Figure 4

CI as a function of (a) cast iron corrosion rate measured by LPR at day 45; (b) steel corrosion rate measured by coupon mass loss at day 45.

Figure 4

CI as a function of (a) cast iron corrosion rate measured by LPR at day 45; (b) steel corrosion rate measured by coupon mass loss at day 45.

In particular, some trial runs showed steel corrosion results that could not be properly predicted by the CI. For instance, trial runs 5 and 6 each showed high corrosion rates despite low values for the CI. This appears to be linked to the high buffer intensity and aggressive nature of these two waters. This underlines that the buffer intensity does not mitigate corrosion of steel as it does for cast iron and that chloride and sulphate ions have a higher impact on this material. This phenomenon was further emphasized by trial 11, which had a low corrosion rate despite a weak buffer intensity but very low levels of chloride and sulphate ions.

Overall, the data produced during this study was used to define thresholds corresponding to different levels of corrosion risk associated with specific ranges of the CI primarily for cast iron. Table 6 summarizes the proposed interpretation of the CI in regard to the anticipated effect of corrosion on the degradation of cast iron pipes, as well as the corresponding life expectancy associated to each level. The application to steel is less significant. The service life estimate was based on the equivalent loss of 50% wall thickness of a pipe with an internal diameter of 125 mm.

Table 6

Corrosion index interpretation scale to evaluate risk of internal pipe wall degradation by corrosion in cast iron pipes

CILevel of risk of internal corrosionCorrosion rate (μm/year)Estimated service life of cast iron pipes
CI ≤ 2 Low CR ≤ 50 >80 years 
2 < CI ≤ 9 Moderate 50 < CR ≤ 100 80–40 years 
9 < CI ≤ 16 High 100 < CR ≤ 150 40–25 years 
CI > 16 Very high CR > 150 <25 years 
CILevel of risk of internal corrosionCorrosion rate (μm/year)Estimated service life of cast iron pipes
CI ≤ 2 Low CR ≤ 50 >80 years 
2 < CI ≤ 9 Moderate 50 < CR ≤ 100 80–40 years 
9 < CI ≤ 16 High 100 < CR ≤ 150 40–25 years 
CI > 16 Very high CR > 150 <25 years 

Figure 5 presents the water turbidity in the pilot circuit after 35 days as a function of the CI. It is apparent that the CI does not correlate with the potential for iron particle release.

Figure 5

Water turbidity as a function of the CI.

Figure 5

Water turbidity as a function of the CI.

A second index was therefore developed to anticipate the impact of internal corrosion on the degradation of water quality conveyed by cast iron pipes. Similar to the CI, the particle release index (PRI) was empirically derived and based on the same three existing indices and is calculated according to Equation (2).  
formula
(2)
where:
  • β = Buffer intensity in mmol/L as determined by chemical simulation of the injection of HCl

  • La is the Larson index

  • p = 1 if LSI < 0 and p = −1 if LSI ≥ 0, where LSI is the Langelier saturation index

Figure 6 illustrates the water turbidity data as a function of the PRI for cast iron pipes. The observed correlation is less strong than for the LPR corrosion rates but remains significant (R2 = 0.82). Four different levels of the index were defined using this data. Waters with values of the PRI below 1 are considered as presenting a very low risk of particle release. PRI values ranging from 1 to 4 are deemed as presenting a moderate risk of particle release. Finally, the risk increased for higher PRI values classified as high (4 < PRI < 10) to very high (PRI > 10).

Figure 6

Water turbidity as a function of the PRI.

Figure 6

Water turbidity as a function of the PRI.

CONCLUSIONS

The objective of this study was the evaluation of the relevance of existing corrosion indices to anticipate the risk of internal corrosion in regard to two different consequences for a distribution system: pipe material degradation and water quality degradation. For this purpose, pilot-scale studies were set-up to assess the impact of 20 different water qualities on cast iron and steel coupons through the monitoring of corrosion rates and iron particle release. The calculated values for eleven conventional corrosion indices were correlated with the experimental data for turbidity as well as for cast iron and steel corrosion rates measured by coupon mass loss and LPR. The specified thresholds for the interpretation of each index were also compared to measured corrosion rates. The study showed that none of the indices found in the literature performed reliably as a stand-alone predictor of the risk of corrosion on material degradation or particle release.

Certain indices (Ryznar, Larson, Modified Larson, Pisigan) showed relevant trends in forecasting corrosion rates of cast iron but did not fully reflect observed phenomena. Recommended thresholds to interpret CCPP and Ryznar indices were found to be relevant to detect low corrosion risk but values outside of this condition did not correlate well with the results. The other indices evaluated during this study proved to be inadequate in forecasting any trend in terms of corrosion risk (Langelier saturation index, DFI, Feigenbaum). Overall, none of the conventional indices were systematically in accordance with the level of corrosion measured for the 20 different water qualities tested on the pilot loops.

In light of these results, two new indices were developed, each looking at a specific outcome of corrosion. The first was the CI, designed to assess the risk of internal degradation of the cast iron and steel material in pipes. The second was the PRI that aims to predict the risk of water quality degradation by iron particle release.

The CI showed a very good linear correlation with the corrosion rate of cast iron. In particular, this index systematically determined whether a particular water would cause severe or mild corrosion. The CI was less reliable in predicting the risk of corrosion for steel pipes but the correlation for this material remained relevant.

The PRI showed a good linear correlation with the turbidity of water in contact with cast iron. An increase in the index value corresponded to an increase in turbidity and, therefore, in the risk of discoloured water.

The two indices proposed showed a strong potential to anticipate the risk of internal corrosion due to the chemistry of the distributed water and can be promising tools to establish suitable treatment quality targets defined to mitigate water discolouration and to extend the service lifetimes of iron-based pipes. Water utilities looking to implement these tools would only need knowledge of their water pH, temperature, mineral content and a simple software to estimate their buffer capacity and equilibrium pH. The indices can then be computed and analysed at a chosen frequency, taking into account resource variability.

ACKNOWLEDGEMENTS

The authors would like to acknowledge the help and feedback of Anthony Audo and Pierre Leroy over the course of this study.

REFERENCES

REFERENCES
ASTM Standard
1999
G4-95
.
Guide for Conducting Corrosion Coupon Tests in Field Application
.
ASTM International
,
West Conshohocken, PA, USA
.
ASTM Standard
2005
D2688
.
Standard Test Method for Corrosivity of Water in the Absence of Heat Transfer (Weight Loss Method)
.
ASTM International
,
West Conshohocken, PA, USA
.
ASTM Standard
2013
G96-90
.
Standard Guide for Online Monitoring of Corrosion in Plant Equipment (Electrical and Electrochemical Methods)
.
ASTM International
,
West Conshohocken, PA, USA
.
AWWA
2001
Reinvesting in Drinking Water Infrastructure
.
American Water Works Association Research Foundation
,
Denver, CO, USA
.
AWWA
2002
WaterStats: Distribution System Survey
.
American Water Works Association Research Foundation
,
Denver, CO, USA
.
AWWA
2011
Buried No Longer: Confronting America's Water Infrastructure Challenge
.
American Water Works Association Research Foundation
,
Denver, CO, USA
.
AWWA
2015
2015 AWWA State of the Water Industry Report
.
American Water Works Association Research Foundation
,
Denver, CO
.
AwwaRF
1996
DVGW Technologiezentrum Wasser, Internal Corrosion of Water Distribution Systems
, 2nd edn.
American Water Works Association Research Foundation
,
Denver, CO
.
Feigenbaum
,
C.
,
Gal-Or
,
L.
&
Yahalom
,
J.
1978
Scale protection criteria in natural waters
.
Corrosion
34
(
4
),
133
137
.
French Drinking Water Standards
2007
Code de la santé publique, aux articles R
.
1321-1 à R. 1321–66 - arrêté du 11 janvier 2007
.
Grigg
,
N. S.
2010
Secondary Impacts of Corrosion Control on Distribution System and Treatment Plant Equipment
.
Water Research Foundation
,
Denver, CO
.
Imran
,
S. A.
,
Dietz
,
J. D.
,
Mutoti
,
G.
,
Taylor
,
J. S.
&
Randall
,
A. A.
2005
Modified Larson ratio incorporating temperature, water age, and electroneutrality effects on red water release
.
Journal of Environmental Engineering ASCE
131
(
11
),
1514
1520
.
Langelier
,
W. F.
1936
The analytical control of anti-corrosion water treatment
.
Journal – American Water Works Association
28
(
10
),
1500
.
Legrand
,
L.
&
Leroy
,
P.
1995
Prévention de la corrosion et de l'entartrage dans les réseaux de distribution d'eau. CIFEC 1995
.
McNeill
,
L. S.
&
Edwards
,
M.
2001
Iron pipe corrosion in distribution systems
.
Journal – American Water Works Association
93
(
7
),
88
100
.
Merrill
,
D. T.
&
Sanks
,
R. L.
1977
Corrosion control by deposition of CaCO₃ films: part 1, A practical approach for plant operators
.
Journal – American Water Works Association
69
(
11
),
592
599
.
NHMRC
2004
National Water Quality Management Strategy – Australian Drinking Water Guidelines 6
.
Pisigan
,
R. A.
&
Singley
,
J. E.
1987
Influence of buffer capacity, chlorine residual, and flow rate on corrosion of mild steel and copper
.
Journal – American Water Works Association
79
(
2
),
62
70
.
Rajani
,
B.
&
Kleiner
,
Y.
2012
External and internal corrosion of large diameter cast iron mains
.
Journal of Infrastructure Systems
19
,
486
495
.
Rajeev
,
P.
,
Kodikara
,
J.
,
Robert
,
D.
,
Zeman
,
P.
&
Rajani
,
B.
2014
Factors contributing to large diameter water pipe failure
.
Water Asset Management International
10
,
9
14
.
Reiber
,
S.
,
Ryder
,
R. A.
&
Wagner
,
I.
1996
Corrosion assessment technologies
. In:
Internal Corrosion of Water Distribution Systems
.
Chapter 9
.
AWWA Research Foundation
,
Denver, CO, USA
, pp.
445
486
.
Rezaei
,
H.
,
Ryan
,
B.
&
Stoianov
,
I.
2015
Pipe failure analysis and impact of dynamic hydraulic conditions in water supply networks
.
Procedia Engineering
119
,
253
262
.
Rossum
,
J. R.
&
Merrill
,
D. T.
1983
An evaluation of the calcium carbonate saturation indexes
.
Journal – American Water Works Association
75
(
2
),
95
100
.
Ryznar
,
J. W.
1944
A new index for determining amount of calcium carbonate scale formed by water
.
Journal – American Water Works Association
36
(
4
),
472
486
.
Singley
,
J. E.
1981
The search for a corrosion index
.
Journal – American Water Works Association
73
(
11
),
578
582
.
Stumm
,
W.
1960
Investigation on the corrosive behaviour of waters
.
Journal of the Sanitary Engineering Division
86
(
SA6
),
27
45
.
USEPA
2009
2006 Community Water System Survey
.
EPA 815-R-09-001
.
US Environmental Protection Agency
,
Washington, DC, USA
. .
WHO
2011
Guidelines for Drinking-water Quality
, 4th edn.
World Health Organization
,
Geneva, Switzerland
.