⚡ electrical engineering

Comparative Evaluation of Observed and Predicted Erosion in Hydraulic Turbines Using IEC 62364:2019 Model

This study validates the IEC 62364:2019 model for predicting hydro-abrasive erosion in Kaplan, Francis, and Pelton turbines by demonstrating its high accuracy and reliability through strong statistical correlation with observed field and experimental data.

Original authors: Sandeep Sagar, Manish Pandey, Naman Arora

Published 2026-08-24

📖 3 min read☕ Coffee break read

CC BY 4.0

Original authors: Sandeep Sagar, Manish Pandey, Naman Arora

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Rivers that carve through young, rugged mountains are powerful engines of energy, but they carry a hidden cost. As water rushes down steep valleys, it picks up a heavy load of sand, silt, and tiny stones, including hard minerals like quartz and feldspar. When this gritty water is channeled into a hydroelectric power plant to spin a turbine, the solid particles act like sandpaper against the metal blades. Over time, this constant grinding wears away the machinery, thinning the metal, roughening the surfaces, and eventually causing the plant to lose efficiency or shut down for expensive repairs. Engineers have long needed a reliable way to predict exactly how fast this wear happens so they can design stronger turbines and plan maintenance before a breakdown occurs. For years, they have relied on various formulas to guess the damage, but these guesses often varied wildly depending on the specific type of turbine or the unique conditions of the river.

A team of researchers set out to test the most recent and comprehensive standard for predicting this damage, known as the IEC 62364:2019 model. This standard is a unified set of rules designed to estimate how deep the wear will be on different parts of a turbine based on the speed of the water, the hardness and size of the sand, and the material of the metal itself. The researchers gathered data from a wide range of real-world power plants and laboratory experiments involving the three main types of turbines used in the industry: the Kaplan, which handles low water pressure but high flow; the Francis, which works in medium-pressure conditions; and the Pelton, which is built for high-pressure jets. They compared the actual wear measured on these machines against the wear predicted by the standard's calculations.

The results showed that the standard is remarkably accurate for most situations. When the researchers plotted the predicted wear against the actual damage found on the blades, guide vanes, and other components, the numbers lined up closely. For the vast majority of the cases they examined, the model's prediction fell within a margin of error of plus or minus 25 percent of the observed reality. In statistical terms, the agreement between what the model said would happen and what actually happened was nearly perfect, with a correlation so strong that it explained almost all the variation in the data. This suggests that the standard provides a trustworthy tool for engineers to estimate how much material will be lost over time, allowing them to make better decisions about which metals to use and when to schedule repairs.

However, the study also highlighted where the model's confidence begins to waver. The researchers found that the standard works best when the wear happens in a steady, linear fashion over time. In some cases, particularly on the outer edges of turbine blades after they have been running for a long time, the wear did not follow a straight line. Instead, the damage accelerated and then slowed down as the shape of the metal changed, altering how the water and sand hit the surface. Because the standard assumes a steady rate of wear, it sometimes overestimated the damage in these long-term scenarios. Despite this limitation, the study concluded that the model is robust enough to be used as a primary guide for managing hydroelectric plants in sediment-rich rivers. By confirming that the standard can reliably predict wear across different turbine types and operating conditions, the research offers a clear path toward more durable and efficient hydropower generation in some of the world's most challenging environments.

Technical Summary: Comparative Evaluation of Observed and Predicted Erosion in Hydraulic Turbines Using IEC 62364:2019 Model

Problem Statement
Hydro-abrasive erosion remains a critical degradation mechanism for hydraulic turbines operating in sediment-rich rivers, particularly in mountainous regions where flows carry high concentrations of hard minerals like quartz and feldspar. This erosion compromises turbine durability, reliability, and efficiency, leading to increased maintenance costs and shortened operational lifespans. While various empirical and semi-empirical models have been proposed over the decades, they often lack standardization or are limited to specific turbine types and site conditions. The International Electrotechnical Commission (IEC) introduced the IEC 62364:2019 standard to provide a unified framework for estimating erosion depth based on sediment characteristics, material properties, and flow parameters. However, a systematic, cross-comparative evaluation of this standard's predictive capability against observed field and experimental data across different turbine types (Kaplan, Francis, and Pelton) has been lacking. Furthermore, the standard's assumption of linear temporal erosion progression has not been rigorously addressed in the context of long-term field validation.

Methodology
This study conducted a comprehensive comparative evaluation by compiling a dataset of 78 observed-versus-predicted erosion pairs from six published field and experimental studies. The dataset encompassed three primary turbine types:

Kaplan Turbines: Data from low-head, high-discharge conditions (e.g., Mohammadpur HPP), focusing on blade tips, outlets, and chambers.
Francis Turbines: Data from medium-head schemes (e.g., Maneri Bhali-II, Bhilangana-III HPP), covering guide vanes, runner inlets/outlets, facing plates, and labyrinth seals.
Pelton Turbines: Data from high-head schemes (e.g., Toss HPP), focusing on nozzles, spear needles, and bucket splitters.

The authors applied the IEC 62364:2019 semi-empirical model to these cases. Key input parameters included relative velocity ( $u$ ), material resistance coefficients ( $K_m$ ), flow coefficients ( $K_f$ ), geometric scaling exponents ( $p, R_s$ ), and particle load ( $P_l$ ). The particle load was calculated by integrating suspended sediment concentration (SSC), particle size distribution (PSD), shape factors, hardness ratios, and exposure duration.

The model's performance was validated using two complementary approaches:

Graphical Analysis: Plotting computed erosion depth against observed depth on 1:1 axes with $\pm25\%$ error envelopes for each study.
Statistical Analysis: Calculating six performance indices (Correlation Coefficient, Coefficient of Determination, Mean Absolute Error, Mean Squared Error, Root Mean Squared Error, and Mean Absolute Percentage Error) across the aggregate dataset. Additional diagnostics included Bland-Altman agreement plots, residual histograms, boxplots, and Q-Q plots.

Key Results
The evaluation demonstrates that the IEC 62364:2019 model provides a robust prediction of hydro-abrasive erosion across diverse turbine geometries and operating conditions:

Statistical Performance: The model exhibited a strong correlation between predicted and observed values, with a Correlation Coefficient (CC) of 0.997 and a Coefficient of Determination ( $E^2$ ) of approximately 0.995. This indicates the model accounts for over 99.5% of the variance in the observed data.
Error Bands: Approximately 83% of the compiled observations fell within the $\pm25\%$ error band, which the authors define as an acceptable level of engineering accuracy for this class of hydraulic systems. Only 17% of predictions deviated substantially beyond this threshold.
Bias and Residuals: The Bland-Altman analysis revealed a mean bias close to zero, indicating no systematic tendency to over- or under-predict. Residual diagnostics confirmed that errors were randomly distributed around zero with approximate normality, suggesting the model is unbiased.
Component-Specific Trends:
- Kaplan: The model successfully identified blade tips and outlets as high-wear zones, though some overprediction was noted at turbine outlets under extended operational durations.
- Francis: Predictions aligned closely with field measurements for guide vanes (severe wear) and labyrinth seals (moderate wear), with data points tightly clustering around the 1:1 line.
- Pelton: The model correctly identified the nozzle and spear as critical wear zones compared to the runner buckets, consistent with field observations of jet impingement damage.

Limitations Identified
The study highlights a specific limitation regarding the model's assumption of linear temporal erosion progression. The IEC 62364 formulation calculates particle load as a linear accumulation of exposure time. However, field data suggests that erosion often follows a nonlinear trajectory: initial surface roughening may accelerate wear, while subsequent geometric changes (e.g., blade thinning, gap enlargement) can alter flow patterns and reduce impact severity over time. Consequently, the model may overpredict erosion in components undergoing significant geometric evolution over extended periods, particularly in outlet regions.

Significance and Claims
The paper asserts that the IEC 62364:2019 framework serves as a reliable, standardized tool for evaluating and mitigating hydro-abrasive wear in sediment-rich hydropower environments. By validating the model against a consolidated dataset of 78 data points across three turbine types, the study confirms its applicability for:

Design and Maintenance: Supporting informed decisions regarding turbine design, material selection, and maintenance planning.
Risk Assessment: Providing a trustworthy first-order predictor for erosion depth when calibrated with site-specific sediment characteristics (SSC, PSD, hardness).
Standardization: Offering a unified methodology that overcomes the fragmentation of previous site-specific empirical models.

The authors conclude that while the model is not perfect—particularly for long-term predictions involving complex geometric evolution—the achievement of 83% of predictions falling within the $\pm25\%$ error band represents a satisfactory level of engineering accuracy. The study emphasizes that continued coupling of field validation with geometry-evolving simulations and site-specific material optimization is essential for further enhancing the reliability of erosion prediction in sediment-rich river environments.

More like this