Statistical method for evaluating hot gas path parts of large industrial gas tubines

Hisato Arimura, Koichi Akagi, Yoichi Sato, Toshio Kojo2004 · DOI

Abstract: This paper describes the approach for managing hot gas path parts for large industrial gas turbines, derived from plant operating experience and maintenance/field inspection records. Statistical analysis of this information provides a better understanding of the gaps in design assumptions and aids in establishing corrective actions. A proactive approach that continually monitors and compares field experience with design capability is essential for achieving high reliability and durability.

Introduction

Engineers from Mitsubishi Heavy Industries (MHI) have developed a robust statistical framework to manage the "hot gas path" components of large-scale industrial gas turbines. Modern turbines operate at extreme temperatures, reaching up to 1500°C, to maximize efficiency. These conditions impose significant thermal and mechanical stress on critical parts such as stationary vanes and rotor blades.

Traditionally, maintenance schedules relied on conservative design estimates. However, a "gap" often exists between theoretical performance and actual field degradation. Because power plants operate in diverse modes—ranging from continuous base-load operation to frequent start-stop cycles—a uniform maintenance approach is often inefficient. The challenge lies in accurately quantifying degradation to prevent failures while avoiding the premature disposal of viable components.

Methodology: A Data-Driven Feedback Loop

MHI’s approach bridges the gap between design and reality through a closed-loop system based on three pillars:

  1. Remote Monitoring System (RMS): This global network tracks real-time operating data, including hours, starts, and load patterns. It enables early anomaly detection and provides a comprehensive stress history for each machine.
  2. Inspection Capture Device (ICD): During maintenance, inspectors use this device to digitize field records. Data on crack lengths, oxidation, and wear are entered directly into a database, replacing manual notes with structured digital information.
  3. Statistical Analysis (Weibull Distribution): This is the core analytical engine. Weibull plots correlate crack occurrence and propagation with both operating hours and start-up cycles. This allows engineers to calculate the probability (e.g., 50% or 90%) of a part reaching a specific degradation threshold over time.
Operating history of F-Series and G-Series turbines, illustrating the diversity in plant operation modes (starts vs. hours).

Fleet Analysis and Datasets

The study draws on field data from MHI’s extensive fleet, specifically the D-series, F-series, and G-series turbines. As shown in the progression of turbine technology, increasing the Turbine Inlet Temperature (TIT) is essential for higher efficiency and capacity, but it necessitates more sophisticated monitoring.

Evolution of gas turbine capacity and inlet temperature across MHI's D, F, and G series.

The datasets include detailed operating histories and inspection records of crack lengths on stationary vanes. Performance metrics are further verified at MHI’s in-house combined cycle power plant.

Key Insights and Design Feedback

A critical finding is the non-linearity of damage. Degradation is not a simple function of time; the interplay between operating hours and start-up cycles varies significantly by plant mode. Managing parts based on a single metric is insufficient for modern high-temperature machines.

Statistical findings are directly fed back into the design process. For instance, field data indicated that 2nd-stage stationary vanes in F-series turbines were susceptible to thermal fatigue cracks. In response, MHI redesigned these components from integrated segments to a "single-vane bolted structure," which successfully suppressed crack generation in subsequent field applications.

Furthermore, MHI utilizes probabilistic maintenance. By using the 50% probability line of crack propagation to set intervals, the company shifts from reactive to predictive repairs. This is supported by advanced repair technologies like "Low Flow Brazing" and automated welding, which allow for the restoration of components previously deemed unrepairable.

Conclusion and Future Directions

The Weibull analysis has proven successful in establishing correlations between crack propagation and operating parameters, enabling accurate predictions of component life. While the framework is robust, its effectiveness depends on data volume. For newer models with fewer operating hours, the "gap" between design and reality remains a focus for ongoing data collection.

As power grids become more dynamic, future work will focus on incorporating granular load-following data into statistical models. Additionally, integrating statistical outcomes with deeper metallurgical analysis of material aging will further refine the predictive capabilities of the system.

Summary updated

Abstract

This paper describes the approach for managing the hot gas path parts for large industrial gas turbine witch is lead from the plant operating experience and maintenance/field inspection records. The statistic analysis result of these information provide the better understanding for the gaps in the design assumptions and for establishing corrective actions. A proactive approach that continually monitors and compares field experience, and design capability is an important aspect of achieving the goals of high reliability and durability.
BibTeX
@inproceedings{arimura2004statistical,
  author = {Hisato Arimura and Koichi Akagi and Yoichi Sato and Toshio Kojo},
  title = {{Statistical method for evaluating hot gas path parts of large industrial gas tubines}},
  booktitle = {Doryoku,},
  year = {2004},
  doi = {10.1299/jsmepes.2004.9.149},
}
Figure