Content area
Dynamic stress data are essential for evaluating structural fatigue life. To address the challenges of complex test data formats, low data reading efficiency, and insufficient visualization, this study systematically analyzes the .raw and .sie file formats from IMC and HBM data acquisition systems and proposes a unified parsing approach. A lightweight .dac format is designed, featuring a “single-channel–single-file” storage strategy that enables rapid, independent retrieval of specific channels and seamless cross-platform sharing, effectively eliminating the inefficiency of the .sie format caused by multi-channel coupling. Based on Python v3.11, an automated format conversion tool and a PyQt5-based visualization platform are developed, integrating graphical plotting, interactive operations, and fatigue strength evaluation functions. The platform supports stress feature extraction, rainflow counting, Goodman correction, and full life-cycle fatigue damage assessment based on the Palmgren–Miner rule. Experimental results demonstrate that the proposed system accurately reproduces both time- and frequency-domain features, with equivalent stress deviations within 2% of nCode results, and achieves a 7–8× improvement in file loading speed compared with the original format. Furthermore, multi-channel scalability tests confirm a linear increase in conversion time (R2 > 0.98) and stable throughput across datasets up to 10.20 GB, demonstrating strong performance consistency for large-scale engineering data. The proposed approach establishes a reliable data foundation and efficient analytical tool for fatigue life assessment of structures under complex operating conditions.
Full text
1. Introduction
Since the inauguration of the world’s first high-speed railway, the Shinkansen, in Japan in 1964, global high-speed rail technology has evolved for more than six decades. The opening of the Beijing–Tianjin Intercity Railway in 2008 marked China’s official entry into the high-speed rail era. However, as operational mileage and speeds continue to increase, reliability issues with high-speed trains have emerged. Fatigue cracks have been frequently observed in critical structural components of the bogie, such as connecting beams, hangers, traction seats, and anti-yaw damper brackets [1], which can eventually lead to structural fatigue failure and pose serious safety risks. Consequently, structural health assessment and fatigue life prediction have gained increasing attention. Line operation tests are typically conducted to measure dynamic stress responses of vehicle structures and key components under actual service conditions. These test results are then used as the basis for service life prediction and fatigue evaluation, ultimately leading to reasonable structural improvement schemes that extend the operational lifespan of the vehicle [2]. Within the overall process of structural fatigue life assessment, dynamic stress data serve as the essential source linking physical testing, signal analysis, and fatigue life prediction. In engineering practice, structural stress responses under operational loads are usually collected using dedicated data acquisition systems (e.g., imc Test & Measurement (IMC, Berlin, Germany), and Hottinger Baldwin Messtechnik (HBM, Darmstadt, Germany) devices, including the eDAQ Data Acquisition System (eDAQ) and the SOMAT XR Data Acquisition System (SOMAT XR)) and stored in specific formats such as .raw and .sie [3,4,5]. These raw datasets not only reflect the stress characteristics of the structure but also serve as the fundamental basis for fatigue life estimation and cumulative damage analysis. However, most current research and fatigue analysis systems focus primarily on algorithmic aspects such as signal filtering, peak extraction, and damage evaluation [6], while relatively little attention has been paid to fundamental aspects including raw data parsing, standardized storage structures, cross-platform sharing, and visualization mechanisms. In fact, the accuracy of dynamic stress data parsing directly determines the reliability of fatigue analysis inputs, and achieving structured, lightweight, and standardized data storage is a prerequisite for realizing full-process automation and traceability in structural fatigue life assessment.
Currently, valuable progress has been made in the field of dynamic stress data analysis and software development. Chen Yu [7] developed a dedicated data processing software, CodeBlade, for line dynamic stress testing and completed the parsing of data formats generated by IMC equipment. Jun Wang et al. [8] created a frequency-domain-based vibration fatigue analysis module within the SABRE software platform, integrating the Power Spectral Density (PSD) method with the Dirlik model. Similarly, Xin Jin [9] presented an integrated software system for asphalt pavement service performance, combining material aging modeling and fatigue life prediction to achieve a complete workflow from physical model construction to fatigue life computation. Jinkang Xu [10] conducted secondary development based on the Nastran kernel and the Qt platform to establish a software system for dynamic load and fatigue life analysis of vehicle-mounted radar structures. Greaves [11] proposed a fatigue analysis method based on the theory of multiple continua and fracture mechanics, accompanied by the simulation software Hellus: Fatigue, which enables fatigue life prediction from the material level to full-scale blades. Nazar Adamchuk [12] investigated an integrated software chain for the casting industry to predict the fatigue life of spheroidal graphite iron (SGI) components induced by manufacturing processes.
Beyond these algorithmic and modeling advances, several studies have addressed specific implementation aspects but revealed persistent limitations in data management and interoperability. Xian Xiaoyu [13] implemented the algorithmic modules of the next-generation fatigue evaluation software Vulcon, yet did not address the unified parsing and conversion of multi-source data formats. Sun Dianju [14] and Zhu Wenting [15] stored data and results in text files, which occupy large storage space, suffer from slow loading speeds, and cannot be shared with international fatigue analysis software such as nCode. Chris Wallbrink et al. [16] created the Computerized General Analysis Platform (CGAP), which integrates both strain-life and crack-propagation fatigue analyses. However, the dynamic stress data used in CGAP are derived from public experimental databases, and the platform does not provide end-to-end integration with dedicated data acquisition systems.
On the other hand, while some systems have demonstrated practical engineering applications, they still lack comprehensive solutions for multi-source data parsing and format conversion. Chen Li et al. [17] established a fatigue life prediction software for lifting machinery that enables damage calculation and life prediction under random loading, but it does not involve automated processing of multi-source data parsing and format conversion. These limitations underscore the need for a unified framework that addresses both algorithmic sophistication and fundamental data infrastructure challenges.
In structural dynamic stress testing, the data scale is enormous, often reaching tens or even hundreds of terabytes. Such large-scale dynamic stress data present two critical challenges in the fatigue life assessment process. First, the data file formats generated by existing acquisition systems are not mutually compatible, with limited public documentation and a lack of open-source parsing tools. This leads to low parsing efficiency and poor cross-platform interoperability, thereby affecting the accuracy and efficiency of subsequent fatigue feature extraction and life calculation. Second, most existing engineering software neglects the optimization of lightweight storage structures, rapid access, and efficient retrieval mechanisms, resulting in low performance in data loading, conversion, retrieval, and visualization under massive data conditions, which has become a bottleneck in the fatigue life assessment workflow. Although commercial software such as nCode and FAMOS offers powerful capabilities [18,19], the practicality of open-source solutions has been steadily increasing due to their cost advantages. This trend provides a clear direction for developing customized and cost-effective fatigue life analysis software, particularly in the contexts of academic research and specialized industrial applications [20].
In engineering applications that demand automation and high-efficiency response, it is imperative to establish an open-source, standardized, and rapidly interpretable multi-source dynamic stress data processing framework to enable seamless integration from raw stress data to structural fatigue life assessment. To address this need, this study focuses on the development of a multi-source dynamic stress data parsing and visualization software oriented toward structural life evaluation, and conducts a systematic investigation into representative dynamic stress data formats. The main contributions and innovations of this study are as follows: Perform an in-depth analysis of the structural characteristics and encoding schemes of mainstream data-acquisition file formats, and design a unified low-level parsing algorithm to achieve open and automated format interpretation; Propose a standardized and lightweight “single-test-channel–single-file” storage format (.dac) that supports rapid retrieval and cross-platform sharing, suitable for large-scale batch processing and fast access; Develop a Python-based format conversion and data processing toolkit integrating key fatigue evaluation algorithms, including signal preprocessing, rainflow counting, and stress spectrum analysis, enabling efficient conversion from raw dynamic stress data to fatigue characteristic parameters; Build an interactive visualization platform based on PyQt5, integrating time-domain and frequency-domain displays, interactive operations, and equivalent stress calculation modules. This platform establishes a complete analytical loop from raw data to fatigue strength assessment, significantly enhancing the intuitiveness and reliability of the evaluation process.
The structure of the manuscript is as follows. Section 2 analyzes the internal characteristics of the raw data format (RAW) and the Somat Insight Environment (SIE) data format and presents the corresponding parsing methods. Section 3 introduces the design of the proposed lightweight Dynamic Analysis Container (DAC) format. Section 4 describes the dynamic stress preprocessing procedures and fatigue damage evaluation methods. Section 5 outlines the architecture and implementation of the software platform. Section 6 provides the experimental validation. Section 7 concludes the paper and discusses future work.
2. Data File Characteristics and Parsing Methods
This section investigates the internal characteristics of the RAW and SIE data formats, which constitute the primary sources of dynamic stress measurements. A precise understanding of their binary organization, metadata structures, and encoding mechanisms is essential for achieving reliable and unified multi-source data interpretation. To this end, a byte-level examination of both formats is conducted using the UltraEdit binary viewer, based on which Python-based parsing algorithms are developed to identify structural elements, extract metadata, decode measurement payloads, and reconstruct physically meaningful time-domain signals. These analyses establish the technical basis for the parsing and preprocessing methods presented in the subsequent subsections.
2.1. IMC Equipment and RAW File Structure
The .raw file is a proprietary storage format used by the IMC testing system to record raw structural dynamic stress data. Its display in the UltraEdit (version 2023.1) software is shown in Figure 1: the left column represents the byte storage addresses, the middle column shows the corresponding hexadecimal values, and the right column displays the decoded results under a specific character encoding. This interface intuitively reveals the underlying storage structure and characteristics of the .raw file at the binary data level.
Byte-wise parsing of the .raw file using UltraEdit revealed that its internal structure is highly modular and compact, consisting mainly of a sequence of structured segments identified by key tags. Each key tag begins with the character | and comprises two identifier characters followed by associated parameters, following the pattern |Cx, …; or |Nx, …; where C denotes a critical key whose correct interpretation is essential for preserving the data semantics, and N denotes a non-critical key that supplies supplementary information. Metadata fields are encoded in ANSI, while measured data are stored in little-endian order and encapsulated as int16 values within the corresponding key-tag segments. Individual key tags further divide into multiple parameter items separated by commas and terminated with a semicolon. The observed storage rules are summarized as follows:
|KeyName, Parameter1, Parameter2, Parameter3, …; |KeyName, Parameter1, Parameter2, Parameter3, …;
It can be seen that the entire .raw file constructs a tightly coupled storage model between metadata and time-domain data through these key tags. The key structural information within the .raw file is summarized in Figure 2.
2.2. Design of the RAW File Parsing Algorithm
The core of parsing a .raw file lies in the accurate identification and interpretation of key tags, including the extraction of essential information such as channel definitions, sampling rules, unit conversion factors, and offsets. This enables the transformation of raw integer values stored in binary form into physically meaningful time-domain signals. The core parsing algorithm for the .raw file was designed and implemented in Python.
To facilitate understanding of the algorithm’s logical structure, Figure 3 illustrates the key steps and data flow of the .raw file parsing algorithm. The entire parsing process establishes a mapping from binary data to physical quantities, ensuring consistency and reliability across data storage, conversion, and visualization.
2.3. HBM Device and SIE File Structure
The EDAQ and Somat XR sensors both belong to HBM Test and Measurement Technology. Their output files, with the extension .sie, are a proprietary binary data format widely used for the acquisition and analysis of structural dynamic signals. A byte-level analysis of these files using UltraEdit software reveals that the .sie file adopts a block-structured stream stacking storage method. Each block contains fields such as block size, group ID, synchronization word, payload (measurement data), and CRC checksum, which together ensure data integrity.
The internal block structure is shown in Table 1. Several key group stream blocks are defined in the file: group = 0 corresponds to the metadata block; group = 1 to the data index block; group = 2 to the device log block; group = 3 to the device configuration block.
while group > 3 stores the actual measured signal data.
In addition, the .sie file embeds XML-formatted metadata on top of its binary block structure, which provides a detailed description of channel definitions, dimensional labels, linear transformation formulas, and decoding rules.
As illustrated in Figure 4, within the channel definition, the
It can be seen that the .sie file, through the coupling of binary block structure and XML metadata, achieves self-descriptiveness of measurement data. This design eliminates the dependence on external description files required by traditional formats and enables dynamic parsing and scalable processing of raw data based on the embedded decoder rules.
2.4. Design of the SIE File Parsing Algorithm
The parsing of the .sie file begins with identifying valid data block structures and concatenating blocks that share the same group identifiers to reconstruct the complete data stream. Each block is verified using its synchronization word, and both the payload data and checksum values are extracted to ensure data integrity through CRC validation. Following this, the embedded XML metadata is retrieved and analyzed to extract channel definitions, decoding rules, and calibration parameters. It is observed that the XML section at the end of the file is not properly closed; however, by manually appending the missing end tag, the metadata can be successfully parsed using standard XML processing methods. The binary payload data are then decoded based on the rules defined in the
2.5. Robust Parsing and Error Handling Mechanisms
To ensure reliability when processing heterogeneous and potentially corrupted acquisition files, the parsing algorithms adopt a combination of format-specific error-handling strategies and a unified top-level exception framework.
For .raw files, the parser performs strict header validation, checking whether all mandatory metadata fields (sampling interval, scaling coefficients, channel name, data length) are present and type-correct. Any missing or malformed field triggers an immediate ValueError, preventing further processing and avoiding inconsistent results caused by invalid input. In contrast, the .sie parser employs a more tolerant, recovery-oriented strategy suited to its block-structured format. Synchronization-word checking, block skipping, and CRC verification are used to detect and isolate corrupted segments. Upon encountering errors, the parser attempts to resynchronize with the next valid data block. Embedded XML metadata undergo well-formedness checks, and minor structural issues (e.g., missing end tags) can be automatically repaired when feasible, enabling extraction of valid data even from partially damaged files.
Both parsers are encapsulated within a unified top-level exception handler that logs unrecoverable errors and reports them to the Graphical User Interface (GUI), ensuring that faults in individual files do not interrupt the batch-parsing workflow while providing users with clear and actionable diagnostic information.
3. Design and Implementation of the Lightweight DAC Format
The future development trend of data file formats is inevitably toward smaller storage size, higher precision, and improved data security. Meanwhile, establishing a unified standard for data file formats is essential to achieve seamless data sharing and minimize data loss or errors during format conversion [21]. In time-domain approaches for structural fatigue strength evaluation, a major limitation lies in the requirement for long-duration recordings and high sampling rates to accurately capture dynamic stress signals, including peaks and valleys. As a result, the generated numerical time-series data consume substantial storage resources and impose significant computational and storage burdens [22]. Therefore, it is essential to design a lightweight and efficient file storage format that optimizes data compression, reduces redundancy, and improves the overall efficiency of fatigue data processing and analysis.
In this study, a lightweight .dac file format is designed based on a “single test channel–single file” storage paradigm. The .dac file consists of three primary components: the file header, the data block, and the file footer, as summarized in Table 2. The header section stores key metadata and statistical parameters of the test channel—such as the number of sampling points, sampling frequency, sampling interval, maximum and minimum values, mean, standard deviation, and root mean square (RMS)—so that these values can be directly retrieved without recalculation, greatly improving data loading and analysis efficiency. The data block section stores the time-domain waveform data in float32 format with 4-byte alignment, ensuring consistency and efficiency across platforms. Finally, the footer section records the channel name and acquisition timestamp, ensuring the integrity and traceability of the test data.
This design not only achieves a clear structure, convenient retrieval, and rapid data access, but also provides advantages in storage optimization and cross-platform portability. It is fully compatible with internationally recognized fatigue analysis software such as nCode, enabling seamless data exchange across different platforms and software environments. Consequently, it is well-suited for the efficient storage, management, and analysis of large-scale experimental datasets.
In addition, the .dac format incorporates robust mechanisms for data integrity and security to ensure both reliability and traceability. During format conversion, integrity is enforced through CRC-32 verification of source .sie data blocks, statistical consistency checks between the source and converted data (mean, standard deviation, and value range), and dimensional validation to guarantee that the number of sampling points remains fully consistent. Regarding security, all conversion operations are executed entirely within the local environment without any external data transmission. Optional AES-256 encryption is provided for sensitive datasets, and comprehensive audit logs are maintained for all file operations to enhance data protection and traceability.
4. Signal Preprocessing and Fatigue Strength Assessment
4.1. Signal Preprocessing
Raw dynamic stress signals often contain baseline drift, high-frequency noise, and power-frequency interference that must be addressed before fatigue analysis.
To correct baseline drift caused by thermal effects, sensor drift, or ambient temperature variations, this study adopts the Trend Removal Method (TRM) based on a segmented linear-drift assumption [23]. The signal is first partitioned into fixed-length segments, and the mean value of each segment is computed. Using these segment means, the endpoint values of the estimated trend line are determined as follows:
(1)
To quantitatively assess baseline drift, the Mann–Kendall trend test is introduced. The test statistic is defined as:(2)
where the sign function is given by:(3)
The standardized Mann–Kendall statistic is then computed by:(4)
where denotes the variance of .In practice, TRM iteratively removes the estimated trend component and performs the Mann–Kendall test after each iteration. The procedure terminates once |Z|< 1.96 (95% confidence level), indicating that the drift has been sufficiently removed without introducing oscillatory artifacts. Figure 6a presents the overall workflow—including signal input, trend estimation, iterative correction, and termination criteria—while Figure 6b illustrates the principle of the TRM.
Based on the structural dynamic characteristics of railway bogie components, a Butterworth low-pass filter is employed to suppress high-frequency random noise. To eliminate phase distortion, forward–backward filtering (via scipy.signal.filtfilt()) is applied, ensuring a zero-phase response—an essential property for accurate cycle counting in fatigue analysis. A second-order IIR notch filter is applied to remove the 50 Hz mains interference and its harmonics (e.g., 100 Hz, 150 Hz). For each interference frequency , a notch filter with a specified quality factor is designed using scipy.signal.iirnotch(), and implemented with zero-phase filtering. Multiple notch filters can be cascaded to suppress several interference components simultaneously. Detailed mathematical formulations of these standard signal processing techniques can be found in references [13].
4.2. Fatigue Damage and Strength Calculation
The time-domain fatigue strength evaluation process begins with acquiring the stress response time-history signals of the structure under actual loading conditions. These stress histories, typically representing complex and random load sequences, are then decomposed into a series of complete stress cycles using the rainflow counting method, from which a stress range histogram is statistically generated. Based on this statistical representation, the Palmgren–Miner linear cumulative damage rule is applied in conjunction with the material’s S–N (stress–life) curve to calculate the fatigue damage contribution corresponding to each stress level [24]. The total damage is obtained by summing the contributions from all stress cycles, thereby enabling a quantitative assessment of fatigue damage and prediction of the overall fatigue life of the structure. The detailed computational procedure for time-domain fatigue damage estimation based on the rainflow counting algorithm is illustrated in Figure 7.
The stress experienced by a structure during its service life exhibits significant randomness; therefore, constructing a fatigue stress spectrum based on measured data is a fundamental prerequisite for evaluating the structural fatigue reliability. Since the duration of line operation tests is typically much shorter than the full life cycle of the vehicle structure, a common approach is to use the measured data obtained from such tests to compile a representative fatigue stress spectrum corresponding to typical operating conditions, and then extrapolate it to the full service life to achieve lifetime fatigue strength assessment. The procedure for constructing the fatigue stress spectrum can be found in [25]. In practical applications, the cycle counting method is used to extract discrete stress cycles from the random stress time history, thereby transforming the complex random stress process into a quantifiable variable-amplitude stress–time history, i.e., a variable-amplitude stress spectrum. Among various cycle counting techniques, the Rainflow Counting Method is one of the most classical and widely adopted algorithms in the field of fatigue analysis [26]. The stress spectrum obtained through the rainflow method is a two-dimensional spectrum with stress mean and stress amplitude as its coordinates. For subsequent fatigue life calculation, it is necessary to transform the two-dimensional spectrum into an equivalent one-dimensional stress spectrum under symmetric cyclic loading, where the mean stress is zero. In this study, the Goodman method is employed to perform the transformation from the two-dimensional to the one-dimensional stress spectrum [27]. The specific conversion equation is given as follows:
(5)
In the above equation, denotes the amplitude of the original stress cycle, represents the mean stress of the original cycle, is the equivalent stress amplitude obtained through the equal-life conversion, and is stress limit. The physical significance of this equation lies in transforming the fatigue damage effect of a stress cycle with a nonzero mean stress into an equivalent damage effect caused by a series of fully symmetric stress cycles.For variable-amplitude stress, if a structure experiences cycles at each of the k stress levels , the total accumulated fatigue damage can be expressed according to the Palmgren–Miner linear damage accumulation theory [28] as follows:
(6)
In the above equation, represents the fatigue life, i.e., the number of cycles to failure under the corresponding stress level . When the total accumulated damage D reaches 1, the structure is considered to have failed due to fatigue. To determine at each stress level, this study adopts the Basquin power-law form of the S–N curve [29], which can be expressed mathematically as follows:
(7)
In this equation, and are empirical material-dependent parameters. According to the International Institute of Welding (IIW) standard IIW-1823–2008 [30], this study adopts a fatigue damage evaluation method based on a 95% confidence level, ensuring the conservativeness and reliability of the selected parameters. By substituting the Basquin equation into the Palmgren–Miner criterion, the total fatigue damage under variable-amplitude stress spectra can be expressed as:(8)
where denotes the equivalent stress amplitude corresponding to the -th stress level.In the railway industry, the safe service mileage of a vehicle structure is typically regarded as its fatigue life. For China’s high-speed electric multiple units (EMUs), the designed service life is 30 years, corresponding to an approximate total mileage of 12 million kilometers, assuming an annual mileage of 400,000 km [31,32]. In this study, the fatigue strength evaluation value is determined based on the principle of equivalent damage, under the assumption that the relationship between running mileage and fatigue damage is linear. The effects of nonlinear factors such as corrosion environments and extreme loading conditions are neglected, implying that the damage rate remains constant [33]. Accordingly, a linear relationship model between accumulated running mileage and fatigue damage can be expressed mathematically as:
(9)
In this equation, denotes the total fatigue damage accumulated over the designed full-life operational mileage , represents the total fatigue damage corresponding to the measured variable-amplitude stress spectrum obtained from the line operation test, and is the actual test mileage during the line operation experiment.
Assuming that the entire service life of the structure is equivalent to a constant-amplitude, fully reversed stress cycle, the corresponding number of loading cycles at the material’s fatigue strength limit produces a total damage of = 1 [34]. Thus, the following relationship can be expressed:
(10)
In the above equation, represents the number of cycles corresponding to the material’s fatigue strength, and denotes the equivalent stress amplitude over the entire service life. By combining the experimental damage equation with the full-lifetime damage model and substituting the above relationship, the mathematical expression for the equivalent stress amplitude over the full service life can be derived as follows:(11)
This formula integrates measured data, the cumulative damage theory, and the lifetime extrapolation model, thereby providing a rigorous quantitative framework for evaluating the fatigue strength of vehicle structures over their full service life. Based on the proposed fatigue damage calculation model, the developed software system realizes the automated computation of equivalent stress, which has been subsequently validated through experimental testing.5. Design and Implementation of Data Visualization Software
A comprehensive software system was developed in Python that integrates data file parsing and conversion, time- and frequency-domain data visualization, and fatigue life evaluation. The system is designed to provide a unified platform for waveform display, interactive analysis, and equivalent stress computation of multi-source experimental data formats such as .raw and .sie. Its goal is to enable intuitive visualization, rapid validation, and efficient analysis of test data across different measurement systems.
5.1. Overall System Architecture
The developed software system adopts a multi-layered modular architecture that systematically organizes functional components from low-level data acquisition to high-level user services. The complete architecture consists of seven distinct layers, as illustrated in Figure 8, with each layer encapsulating specific responsibilities while maintaining clear interfaces with adjacent layers.
The proposed software suite adopts a modular, layered architecture that ensures interoperability, scalability, and maintainability. The Software Platform Layer provides the execution foundation, built on Python 3.11 and core scientific libraries such as NumPy v2.3.1, SciPy v1.8.3, PyQt5 v5.15.11, pyqtgraph v0.13.7, and Matplotlib v3.10.3, with fatigue-analysis toolkits supporting rainflow counting and related computations. The Data Acquisition Layer ingests heterogeneous test data from multiple engineering systems, including IMC (.raw), HBM (.sie), and standardized formats (.dac, .csv), and exposes a unified input interface regardless of the original instrumentation. The Access and Parsing Layer implements the essential decoding logic for proprietary binary formats. Major functions include .sie XML metadata extraction, CRC-based integrity verification, .raw/.sie binary parsers, and timestamp validation, enabling accurate reconstruction of physical stress signals. The Data Storage and Indexing Layer manages structured organization of measurement data, incorporating binary data handlers, metadata registries, memory caching for high-frequency access, and indexing structures to support fast retrieval and batch operations. The Computing and Analysis Layer provides the complete fatigue-evaluation processing pipeline, supporting batch computation for large-scale datasets and ensuring consistent signal preprocessing, spectrum estimation, stress-cycle analysis, and equivalent-stress evaluation. The Presentation Layer delivers interactive visualization through a PyQt5-based GUI, integrating time-domain and frequency-domain views with zooming, region selection, highlighting, and rapid rendering of million-sample waveforms using the pyqtgraph engine. Finally, the Management and Service Layer provides system-level services, including logging, performance monitoring, structured error reporting, and optional user-permission control. This layered architecture with clear interface boundaries facilitates extensibility toward advanced features such as machine-learning-based life prediction and cloud-based collaborative analysis.
The software development environment and library versions used in this study are listed in Table 3, ensuring reproducibility of the implementation.
5.2. User Interface Design and Implementation
As shown in Figure 9, the developed software system features a clean, intuitive, and user-friendly graphical interface, designed to enhance the efficiency and comfort of human–computer interaction. The layout adheres to ergonomic design principles, ensuring that key functions are easily accessible and logically organized. The main window is structured with a title bar, menu bar, and toolbar positioned at the top, providing convenient access to essential operations such as file management, parameter configuration, view switching, and execution control. This modular design facilitates smooth workflow navigation and minimizes user learning cost. On the left side of the interface lies the system file directory tree, which intuitively presents the hierarchical structure of data files within the current working directory, along with their associated metadata and attributes. This design enables users to efficiently browse, locate, and select target files for further processing. The upper-right section serves as the data file parsing and conversion area, supporting multi-format conversions from .sie and .raw files into .dac or .csv formats. Real-time feedback is provided through a path display box and a progress bar, allowing users to monitor file locations and conversion status throughout the execution process.
The lower-right section of the interface is dedicated to the data visualization and plotting area, where users can select converted target files and generate corresponding time-domain waveforms. This area supports a range of interactive operations, including mouse-wheel zooming, drag-based panning, region selection (as shown in Figure 10), and highlighting of specific data segments. Additionally, it provides waveform playback and statistical analysis functionalities, enabling dynamic and intuitive exploration of measurement data. At the bottom of the interface lies the log output area, which records file processing and plotting results in a structured list format. Beyond reporting conversion status and target filenames, it also displays key data metrics—such as maximum, minimum, and mean values, sampling interval, as well as X/Y-axis titles and channel names—allowing users to efficiently verify results and trace processing history.
In addition to the conventional visualization and interaction of time-domain signals, the software integrates a Power Spectral Density (PSD) analysis module to enhance the depth of data interpretation. By right-clicking within the waveform display area and selecting the “Power Spectrum (FFT)” option, users can transform the original time-domain waveform into its corresponding frequency-domain representation, thereby visualizing how signal energy is distributed across different frequency components. This functionality provides an intuitive means to identify the dominant frequency characteristics of the measured signals, enabling users to rapidly detect noise sources, power-line interference, resonance phenomena, and other periodic patterns that may influence structural performance. Through this analysis, the system facilitates a deeper understanding of the dynamic behavior of structures and provides valuable diagnostic insights. Moreover, the PSD visualization serves as a critical analytical tool for fatigue life evaluation, structural integrity assessment, and signal preprocessing, thereby bridging the gap between experimental data and engineering interpretation. The detailed implementation interface of this module is presented in Figure 11.
5.3. Full-Process Fatigue Assessment Using the Developed Software
As a case study, the developed software system was tested and validated using measured dynamic stress data collected from a high-speed electric multiple unit (EMU) operating on the Hu–Yu–Rong Railway, specifically along the Yichang North–Hanchuan North section. The selected measurement point was located on the outer side of the fillet weld beneath the motor hanger and crossbeam, where dynamic stresses were recorded under actual service conditions. As illustrated in Figure 12, the corresponding time-domain waveform indicates that the raw signal is measured in strain units, which must be converted into equivalent stress by multiplying it with the material’s elastic modulus. Since no zero-level calibration was performed during data acquisition, a slight baseline drift can be observed over time. Moreover, the gradual variation in amplitude is likely attributed to thermal strain effects induced by temperature fluctuations during prolonged operation of the structure. To enhance signal stability and remove low-frequency trends, low-pass filtering or detrending techniques should be applied during the preprocessing stage.
The frequency-domain analysis, presented in Figure 13, reveals the presence of strong power-frequency interference, with distinct spectral peaks appearing at 50 Hz, 100 Hz, and their higher harmonics. Such interference components are characteristic of electromagnetic coupling or power system noise in onboard electrical environments. To improve signal quality and ensure the accuracy of subsequent fatigue analysis, it is necessary to apply notch filtering at these specific frequencies during the preprocessing stage, thereby effectively suppressing interference and enhancing the overall reliability of the dynamic stress data.
The construction of a fatigue stress spectrum serves as the fundamental prerequisite for fatigue damage evaluation, while the equivalent stress acts as a key criterion for assessing the fatigue reliability of structures. The software developed in this study enables a fully automated workflow that integrates the entire process—from time-domain dynamic stress signal acquisition, to fatigue stress spectrum generation, and finally to equivalent stress computation.
In the stage of fatigue stress spectrum compilation, a systematic data processing procedure is employed to transform the raw strain time-history signal into a standardized stress spectrum suitable for quantitative fatigue damage analysis. First, the strain time-series data are converted into stress time-series based on the mechanical properties of the material. Subsequently, the rainflow counting algorithm is applied to extract cyclic features from the random stress waveform, yielding a complete set of stress cycles characterized by stress amplitude and mean stress. To manage large volumes of cyclic data, a two-dimensional stress spectrum statistical matrix is constructed for data normalization [35]. The ranges of stress amplitude and mean stress are divided into discrete bins—typically with 16 or 32 levels—transforming continuous stress data into discretized stress intervals. The number of occurrences within each amplitude–mean stress bin is then counted, producing a visualized stress distribution map (as shown in Figure 14a). To satisfy the computational requirements of the linear cumulative damage theory, the two-dimensional stress spectrum is further transformed into a one-dimensional equivalent spectrum using the Goodman life-equivalent model (Equation (5)). By introducing the ultimate strength of the material, this model establishes an equivalence relation that accounts for the influence of mean stress, converting actual stress cycles with non-zero mean stress into equivalent zero-mean stress cycles with the same damage effect. The resulting one-dimensional stress spectrum (Figure 14b) contains the equivalent stress amplitudes and their corresponding cycle counts. Using the full life-cycle equivalent stress calculation formula (Equation (11)) introduced in Chapter 4, the equivalent stress at the dynamic stress measurement points located on the motor hanger and the outer side of the fillet weld on the lower surface of the crossbeam was computed. The obtained equivalent stress value was 14.9 MPa, which is significantly lower than the fatigue limit of steel welds (70 MPa). Therefore, the structure can be considered safe and reliable under the evaluated operational conditions.
In summary, the developed software system achieves an integrated platform that combines file management, data format conversion, time-domain waveform visualization, frequency-domain analysis, and equivalent stress computation within a unified framework. This comprehensive integration not only significantly enhances the efficiency of large-scale experimental data processing and analysis but also improves the intuitiveness and convenience of user interaction. The system provides a robust computational foundation for subsequent structural fatigue life assessment and engineering applications. Moreover, the proposed framework is highly scalable and transferable, allowing potential applications in a wide range of engineering domains, including automotive systems, marine structures, construction machinery, power equipment, and civil infrastructures such as bridges [36,37].
6. Experimental Validation
To meet the engineering requirements of structural fatigue life assessment, this chapter presents a series of experimental validations focusing on four key aspects: waveform consistency, equivalent stress computation accuracy, file loading performance, and multi-channel conversion scalability. By comparing waveform visualization results, evaluating equivalent stress calculation errors, analyzing data loading efficiency, and assessing conversion performance across datasets of varying scales, a comprehensive evaluation is conducted to verify the reliability, computational accuracy, scalability, and engineering applicability of the proposed methodology and the developed software system.
6.1. Waveform Consistency Verification
Figure 15 and Figure 16 illustrate a direct comparison of the time-domain dynamic stress waveforms obtained from the same measurement point, processed, respectively, using the software platform developed in this study and the commercial fatigue analysis software nCode v9.0. The experimental data were collected from fatigue tests conducted on a high-speed EMU operating along the Chongqing–Guiyang railway line, with the measurement point located at the weld seam of the bogie frame. The comparison results demonstrate that the waveforms from both platforms show excellent agreement in amplitude, phase, and local variation details, thereby validating the accuracy and robustness of the parsing algorithm proposed in this study.
Furthermore, the DAC-format files generated by the proposed software can be directly imported and recognized in nCode, indicating that the format is highly compatible and portable across internationally used fatigue analysis platforms. In addition, the same time-domain signal was subjected to Power Spectral Density (PSD) analysis using both software systems, as shown in Figure 17 and Figure 18. The two spectra display identical dominant frequency components and spectral energy distributions, which further confirms the correctness, precision, and reliability of the frequency-domain analysis algorithm implemented in the developed system.
6.2. Verification of Equivalent Stress Calculation
To further validate the accuracy and engineering applicability of the proposed dynamic stress data parsing and format conversion algorithms in the context of fatigue life assessment, the same measured dynamic stress signal from a single test channel was selected as a comparative sample. Equivalent stress calculations were independently performed using both the self-developed software platform presented in this study and the commercial fatigue analysis software nCode. In the developed software system, the time-domain dynamic stress data obtained through parsing were first analyzed using the Rainflow Counting Method to identify and quantify load cycle characteristics, thereby extracting the stress amplitude and mean stress information for each cycle. Subsequently, the Nominal Stress Method was employed to compute the stress amplitude distribution, which was then combined with the full life-cycle linear extrapolation model and the equivalent stress calculation formula established in Chapter 4 (Equation (11)) to obtain the equivalent stress values under full-life conditions. Figure 19a presents a comparison of the equivalent stress results obtained from both software systems, while Figure 19b illustrates the correlation between their results and the corresponding relative error distribution. Comparative analysis shows a high degree of consistency between the two approaches, with the relative deviation of equivalent stress results remaining within 2%.
These results strongly validate the correctness and stability of the proposed dynamic stress data parsing and format conversion algorithm in terms of amplitude preservation, cycle feature reconstruction, and stress parameter transmission. They demonstrate that the algorithm effectively ensures the accuracy and consistency of key stress parameters throughout the fatigue life evaluation process.
6.3. Evaluation of File Loading Efficiency
To systematically evaluate the performance advantages of the proposed .dac file format in engineering fatigue data processing, this study utilized the same set of raw experimental data, which were, respectively, converted into the .dac format and retained in the original .sie format. The file loading performances of the two formats were then compared under identical hardware configurations and testing environments. As shown in Table 4, the .dac files exhibited a significantly shorter average loading time and a much higher data reading rate, achieving an improvement of approximately 7–8 times compared with the .sie format. This notable enhancement in performance is primarily attributed to the binary compressed storage structure and the optimized data indexing mechanism employed by the .dac format, which effectively reduce disk I/O operations and the computational complexity of data parsing.
Further performance analysis indicates that the .dac format substantially reduces system resource consumption through data preprocessing and structured storage, while fully preserving key characteristics of the original stress signals—including the integrity of time-domain waveforms, the spectral energy distribution in the frequency domain, and load-cycle information. This efficiency advantage becomes particularly prominent in large-scale fatigue testing scenarios involving multi-channel and long-duration datasets, enabling real-time waveform visualization, rapid frequency-domain analysis, and efficient fatigue damage computation.
6.4. Multi-Channel Conversion Performance Evaluation
To assess scalability across different data scales, systematic conversion performance tests were conducted on six datasets (file sizes from 647 MB to 10.20 GB, a 15.8× range); dataset details are given in Table 5. Performance metrics were recorded at 10-channel intervals during each conversion.
As shown in Figure 20a, the conversion time increases linearly with channel count across all datasets (R2 > 0.98), and the slope scales with the per-channel data volume, corroborating the algorithm’s O(n) time complexity. Figure 20b further shows that the initial 10–20 channels experience a short warm-up phase attributable to one-time initialization procedures, including module loading, buffer allocation, I/O cache establishment, and stabilization of the operating system’s file-access pipeline. After reaching the steady state, the throughput stabilizes at 153 ± 14 MB/s and remains consistent for 30–60 channels, regardless of file size. The uniform performance across a 15.8-fold variation in data volume confirms the strong scalability of the proposed “single-channel–single-file” architecture. This linear behavior also enables reliable extrapolation; for example, a 120-channel dataset of approximately 20 GB is estimated to require only about 7 min for complete conversion.
The comprehensive experimental results demonstrate that the proposed multi-source dynamic stress data parsing and visualization system exhibits excellent accuracy and engineering applicability in the key stages of structural fatigue life assessment. The waveform consistency validation confirms the correctness and reliability of both the proposed parsing algorithm and the DAC file format; the equivalent stress calculation verification further proves the precision and stability of the developed fatigue life evaluation model; the multi-channel scalability evaluation validates linear performance characteristics (R2 > 0.98) across a 15.8× file size range with consistent throughput independent of channel count; and the file loading efficiency assessment highlights the system’s high performance and scalability when handling large-scale datasets.
In summary, the developed software system not only enables efficient parsing, visualization, and equivalent stress computation of dynamic stress data, but also provides a unified, efficient, and reliable technical framework for structural fatigue analysis and life assessment. This work lays a solid foundation for the future development of data-driven intelligent structural life assessment methods and their practical implementation in complex engineering applications.
7. Conclusions
To address the critical challenges in structural life assessment—including the complexity of multi-source dynamic stress data parsing, the inconsistency of file formats, low storage efficiency, and insufficient visualization capability—this study systematically investigated the key technologies of data parsing, unified storage, equivalent stress computation, and visual analysis of dynamic stress signals. The major contributions and findings are summarized as follows: Unified parsing and conversion method for dynamic stress data files
Comprehensive analyses were conducted on the .raw and .sie file formats of mainstream data acquisition systems such as IMC and HBM, revealing their binary storage mechanisms and XML metadata structures. Based on Python, a generalized data parsing and automated format conversion algorithm was developed, achieving transparent access to file structures and unified processing of heterogeneous data sources. This method significantly improves parsing and conversion efficiency, providing a robust foundation for multi-platform data compatibility and engineering data sharing. 2.. Design and implementation of a lightweight unified storage format
A lightweight “single-test-channel–single-file” storage structure (.dac format) was proposed. The file header preserves precomputed statistical quantities to avoid redundant calculations, while the data section adopts a binary compressed storage scheme. This design substantially enhances file loading and retrieval efficiency, achieving an average data reading rate approximately 7–8 times faster than that of the original formats. Moreover, it supports cross-platform data sharing and efficient access, offering a high-performance solution for managing and analyzing large-scale dynamic stress datasets. 3.. Development and validation of an integrated fatigue life evaluation system
An integrated fatigue analysis software platform was developed using PyQt5, incorporating modules for file parsing and conversion, time-domain visualization, frequency-domain analysis, and equivalent stress computation. The system integrates core algorithms such as signal preprocessing, rainflow counting, stress spectrum analysis, and equivalent stress evaluation, thereby enabling automated computation and visualization of fatigue-related parameters from raw dynamic stress data. Experimental validation shows that the equivalent stress results obtained from the proposed software are in excellent agreement with those from the commercial fatigue analysis tool nCode, with a maximum relative error below 2%, demonstrating the accuracy and reliability of the proposed methods in fatigue life assessment.
In summary, this study establishes a comprehensive technological framework for multi-source dynamic stress data parsing and lightweight storage tailored for structural life assessment. The proposed methods form a complete technical chain—from data acquisition, parsing, and analysis to fatigue life computation—providing an efficient, scalable, and unified data support platform for fatigue analysis and life prediction of engineering structures. Furthermore, this research lays a solid foundation for the digitalization and intelligentization of structural life evaluation. Future work will explore the integration of artificial intelligence (AI) techniques, drawing inspiration from neural network-based fatigue analysis approaches (as discussed in Ref. [38]), to develop data-driven life prediction models. This advancement aims to realize an intelligent, closed-loop, and highly efficient structural life assessment process, promoting its broader application in practical engineering scenarios.
Conceptualization, Q.L. and Z.L.; methodology, Q.L.; software, Q.L.; validation, Y.C., Q.L. and Z.L.; formal analysis, Y.C.; investigation, Q.L.; resources, Z.L.; data curation, Y.C.; writing—original draft preparation, Q.L.; writing—review and editing, Q.L.; visualization, Q.L.; supervision, Z.L.; project administration, Y.C.; funding acquisition, Y.C. All authors have read and agreed to the published version of the manuscript.
The author declares that this study is not aimed at humans or animals.
Not applicable.
Unable to obtain data due to privacy restrictions.
The authors declare no conflicts of interest.
Footnotes
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Figure 1 Display of the .raw file in UltraEdit. Blue, green, and red frames indicate the byte addresses, hexadecimal values, and decoded results, respectively.
Figure 2 Internal structure of the .raw files.
Figure 3 Flowchart of the .raw file parsing algorithm.
Figure 4 (a) XML metadata channel definition. (b) XML metadata decoder definition.
Figure 5 Flowchart of the .sie file parsing algorithm.
Figure 6 (a) TRM algorithm flowchart. (b) Principle of the trend estimation and removal process.
Figure 7 Procedure of fatigue damage estimation in time domain using rainflow counting algorithm.
Figure 8 Complete seven-layer architecture of the software system.
Figure 9 Data visualization software interface design.
Figure 10 Waveform after regional selection reconstruction.
Figure 11 Power spectrum density analysis in data visualization software.
Figure 12 Dynamic stress time-domain signal measured at the outer side of the fillet weld located beneath the motor hanger and crossbeam.
Figure 13 Power spectrum density analysis of dynamic stress time-domain signals.
Figure 14 (a) Amplitude–mean stress histogram obtained by rainflow counting. (b) Equivalent stress amplitude histogram derived using the Goodman model.
Figure 15 The waveform displayed in the software developed in this study.
Figure 16 The waveform displayed in the nCode software.
Figure 17 The Power Spectral Density analysis performed in the software developed in this study.
Figure 18 The Power Spectral Density (PSD) analysis performed in the nCode software.
Figure 19 (a) Comparison of equivalent stress results obtained from nCode and the proposed software. (b) Correlation and relative error distribution of equivalent stress results.
Figure 20 (a) Conversion time versus channel count for different datasets, demonstrating linear scalability of the SIE-to-DAC conversion pipeline. (b) Conversion throughput as a function of channel count, illustrating the initial warm-up phase and subsequent steady-state performance.
Internal block structure of the .sie files.
| Name | Offset (Bytes) | Size (Bytes) | Description |
|---|---|---|---|
| Size | 0 | 4 | Block size (including header and footer) |
| Group | 4 | 4 | Group identifier (Group ID) |
| Syncword | 8 | 4 | Synchronization word (0x51EDA7A0) |
| Payload | 12 | n | Measurement data (payload) |
| Checksum | 12 + n | 4 | CRC32 checksum |
| Size2 | 16 + n | 4 | Block size (repeated) |
Internal structure of the .dac files.
| File Region | Address | Description | Data Type |
|---|---|---|---|
| Header Section (512 bytes) | 0x00–0x03 | Total number of sampling points | float32 |
| 0x04–0x07 | Sampling frequency | float32 | |
| 0x08–0x0B | Time of the first sampling point | float32 | |
| 0x0C–0x0F | Sampling interval | float32 | |
| 0x10–0x13 | Mean value of Y-axis | float32 | |
| 0x14–0x17 | Standard deviation of Y-axis | float32 | |
| 0x20–0x23 | Maximum value of Y-axis | float32 | |
| 0x24–0x27 | Minimum value of Y-axis | float32 | |
| 0x44–0x47 | Root mean square (RMS) of Y-axis | float32 | |
| 0x70–0x73 | Index of maximum Y value | unsigned int | |
| 0x74–0x77 | Index of minimum Y value | unsigned int | |
| 0x78–0x7B | Index of maximum absolute Y value | unsigned int | |
| 0x7C–0x7F | Total number of sampling points | unsigned int | |
| 0x164–0x17B | Y-axis label | windows-1252 | |
| 0x17C–0x193 | X-axis label | windows-1252 | |
| 0x1B0–0x1BF | Y-axis unit | windows-1252 | |
| 0x1C0–0x1CF | X-axis unit | windows-1252 | |
| 0x1F4–0x1F7 | Same as 0x70–0x73 | float32 | |
| 0x1F8–0x1FB | Same as 0x74–0x77 | float32 | |
| 0x1FC–0x1FF | Same as 0x78–0x7B | float32 | |
| Data Block Section | 0x200–m | Experimental sampling data | float32 |
| File Tail Section | L-512–L-385 | Channel name | windows-1252 |
| L-384–L-257 | Acquisition timestamp | windows-1252 |
Software development environment and library versions.
| Component | Version | Purpose |
|---|---|---|
| Python | 3.11.2 (64-bit) | Core programming language |
| PyQt5 | 5.15.11 | GUI framework |
| Pyqtgraph | 0.13.7 | High-performance data visualization |
| Numpy | 2.3.1 | Numerical array operations |
| Scipy | 1.8.3 | Signal processing (FFT, filtering) |
| Rainflow | 3.2.0 | Rainflow counting algorithm |
| Matplotlib | 3.10.3 | Publication-quality plotting |
| IDE | PyCharm 2024.3.2 | Python IDE for coding and debugging |
| Operating System | Windows 10 (64-bit) | Development and runtime platform |
Comparison of file loading performance between .sie and .dac formats.
| No. | .sie File Size/Loading Time | .dac File Size/Loading Time |
|---|---|---|
| 1 | 891 MB/7.73 s | 1.61 GB/2.31 s |
| 2 | 719 MB/6.20 s | 1.25 GB/1.66 s |
| 3 | 682 MB/6.33 s | 1.17 GB/1.55 s |
| 4 | 998 MB/11.20 s | 1.79 GB/2.18 s |
| 5 | 887 MB/9.76 s | 1.57 GB/1.93 s |
| Average loading rate | 101.31 MB/s | 785.81 MB/s |
Multi-channel conversion test datasets.
| Dataset | File Size | Channels | Scenario |
|---|---|---|---|
| Test1 | 10.20 GB | 80 | Extended multi-day monitoring |
| Test2 | 8.00 GB | 66 | Full-day comprehensive test |
| Test3 | 6.62 GB | 66 | Long-duration operation |
| Test4 | 2.58 GB | 64 | Standard line section test |
| Test5 | 1.64 GB | 64 | Medium-duration assessment |
| Test6 | 647 MB | 64 | Rapid diagnostic test |
1. Yang, G.; Wei, Y.; Zhao, G.; Liu, Y.; Zeng, X.; Xing, Y.; Lai, J.; Zhang, Y.; Wu, H.; Chen, Q.
2. Zhao, G.; Liu, Z.; Chen, Y. Research on Subway Dynamic Stress Test Data Management and Data Visualization System. Comput. Appl. Softw.; 2022; 39, pp. 101-107. (In Chinese) [DOI: https://dx.doi.org/10.3969/j.issn.1000-386x.2022.09.015]
3. imc Test & Measurement GmbH. imc DEVICES User Manual; imc Test & Measurement GmbH: Berlin, Germany, 2023; Available online: https://www.imc-tm.cn/download-center/product-downloads/imc-c-serie/manuals (accessed on 19 December 2025).
4. HBK (Hottinger Brüel & Kjær). HBM Data Acquisition Systems: Product Overview and Technical Specifications; HBK: Darmstadt, Germany, 2024; Available online: https://www.hbkworld.com/en/products (accessed on 19 December 2025).
5. HBK. eDAQ Data Acquisition System: User Guide and Technical Documentation; HBK: Darmstadt, Germany, 2023; Available online: https://www.hbkworld.com/en/products/instruments/mechanical-structural-daq/edaq (accessed on 19 December 2025).
6. Yue, J.; Yang, K.; Peng, L.; Guo, Y. A Frequency–Time Domain Method for Ship Fatigue Damage Assessment. Ocean. Eng.; 2021; 220, 108154. [DOI: https://dx.doi.org/10.1016/j.oceaneng.2020.108154]
7. Chen, Y. Fatigue Life Study of High-Speed Bogies under Different Working Conditions. Master’s Thesis; Beijing Jiaotong University: Beijing, China, 2012; (In Chinese) [DOI: https://dx.doi.org/10.7666/d.Y2428039]
8. Wang, J.; Tian, L.L.; Guo, Y.C.; Chang, L. Development of a Frequency-Domain Vibration Fatigue Analysis Module on SABRE Software. J. Phys. Conf. Ser.; 2025; 3127, 012003. [DOI: https://dx.doi.org/10.1088/1742-6596/3127/1/012003]
9. Jin, X. Contributions to an Improved Oxygen and Thermal Transport Model and Development of Fatigue Analysis Software for Asphalt Pavements. Master’s Thesis; Texas A&M University: College Station, TX, USA, 2009.
10. Xu, J.; Zhou, X. Dynamic Load Analysis Software Design of Vehicle Radar Structure. Advances in Artificial Intelligence, Big Data and Algorithms; IOS Press: Amsterdam, The Netherlands, 2023; pp. 656-662.
11. Greaves, P. Fatigue Analysis and Testing of Wind Turbine Blades. Ph.D. Thesis; Durham University: Durham, UK, 2013.
12. Adamchuk, M.S.N. On Establishment of a Software Chain in the Foundry Industry for the Estimation of Process-Induced Fatigue Life. Ph.D. Thesis; Clausthal University of Technology: Clausthal-Zellerfeld, Germany, 2025.
13. Xian, X. Development of Dynamic Stress Data Processing and Fatigue Evaluation Software System. Master’s Thesis; Beijing Jiaotong University: Beijing, China, 2017; (In Chinese)
14. Sun, D.J. Design and Application of Dynamic Stress Test Data Analysis Software for Bogies. Master’s Thesis; Southwest Jiaotong University: Sichuan, China, 2010; (In Chinese) [DOI: https://dx.doi.org/10.7666/d.y1687816]
15. Zhu, W.T. Design and Application of Fatigue Strength Analysis Software Based on Dynamic Stress Testing. Master’s Thesis; Southwest Jiaotong University: Sichuan, China, 2014; (In Chinese)
16. Wallbrink, C.; Hu, W.P. Development of CGAP for Fatigue Damage and Crack Growth Analysis: Verification, Validation and Examples of Application. Adv. Mater. Res.; 2014; 891–892, pp. 702-707. [DOI: https://dx.doi.org/10.4028/www.scientific.net/AMR.891-892.702]
17. Chen, L.; Liu, G.; Ding, K. Development and Engineering Application of Fatigue Life Analysis Software for Hoisting Machinery. China Saf. Sci. Technol.; 2016; 12, pp. 138-145. (In Chinese) [DOI: https://dx.doi.org/10.11731/j.issn.1673-193x.2016.09.025]
18. Silva, J.A.; Caetano, D.A.; Ribeiro Filho, S.L.M.; Teixeira, F.N.; Guimarães, L.G.M. Calculation and Enhancement of Fatigue Life by ε–N Approach and Corrosion Fatigue in Steam Turbine Rotor. Int. J. Damage Mech.; 2022; 31, pp. 845-863. [DOI: https://dx.doi.org/10.1177/10567895221082209]
19. Shi, X.; Guo, W.; Wang, J.; Li, G.; Lu, H. Compilation of Load Spectrum of Loader Working Device and Application in Fatigue Life Prediction. Sensors; 2025; 25, 5585. [DOI: https://dx.doi.org/10.3390/s25175585]
20. Gandhi, R.; Maccioni, L.; Concli, F. Significant Advancements in Numerical Simulation of Fatigue Behavior in Metal Additive Manufacturing—Review. Appl. Sci.; 2022; 12, 11132. [DOI: https://dx.doi.org/10.3390/app122111132]
21. Li, Y.; Shang, Y.; Yuan, Y.; Chen, J.; Li, D.; Wang, Y.; Liu, C.; Dou, Y. Data File Formats in 3D Printing Technology. J. Beijing Univ. Technol.; 2016; 42, pp. 1009-1016. (In Chinese) [DOI: https://dx.doi.org/10.11936/bjutxb2015070019]
22. Chaaban, R. Frequency-Domain Fatigue Analysis of Wind Turbine Structures and Fatigue Damage Detection: Performance Evaluation of Spectral-Based Methods Against the Rainflow Counting Algorithm. Ph.D. Thesis; University of Siegen: Siegen, Germany, 2021.
23. Zhao, G.W.; Li, N.; Sun, Y.X. A Novel Spike Detection Model for Dynamic Stress Monitoring of Bogie Frame. Adv. Mech. Eng.; 2024; 16, 16878132241277638. [DOI: https://dx.doi.org/10.1177/16878132241277638]
24. Liu, C.; Sun, S.; Li, Q. The Impact of Gear Meshing in High-Speed EMU Gearboxes on Fatigue Strength of the Gearbox Housing. Technologies; 2025; 13, 311. [DOI: https://dx.doi.org/10.3390/technologies13080311]
25. Li, L.; Wang, B.; Wang, W.; Wang, H. A Method for Constructing the Load Spectrum of Metro Bogie Frame for Fatigue Damage Prediction. Measurement; 2025; 247, 116790. [DOI: https://dx.doi.org/10.1016/j.measurement.2025.116790]
26. Leng, J.; Zhao, L.; Mao, H. Online Fatigue Life Prediction Method of Jacket Structures Based on Proper Orthogonal Decomposition and Rainflow Counting Algorithm. Structures; 2025; 74, 108517. [DOI: https://dx.doi.org/10.1016/j.istruc.2025.108517]
27. Kumar, R.; Sharma, T.; Shekhar, A.; Vyas, N.S. Primary Suspension Failure Analysis in FIAT Type LHB Bogies and Life Estimation. Eng. Fail. Anal.; 2022; 138, 106320. [DOI: https://dx.doi.org/10.1016/j.engfailanal.2022.106320]
28. Li, G.; Guo, D.; Deng, Y.; Zhang, X. Fatigue Life Prediction of the Aluminum Box for Aircraft Parts Turnover Based on the Rainflow Counting Method. J. Mech. Sci. Technol.; 2025; 39, pp. 2651-2662. [DOI: https://dx.doi.org/10.1007/s12206-025-0424-x]
29. D’Antuono, P. An Analytical Relation between the Weibull and Basquin Laws for Smooth and Notched Specimens and Application to Constant Amplitude Fatigue. Fatigue Fract. Eng. Mater. Struct.; 2020; 43, pp. 991-1004. [DOI: https://dx.doi.org/10.1111/ffe.13175]
30. Zong, L.; Shi, G.; Wang, Y.Q.; Li, Z.X.; Ding, Y. Experimental and Numerical Investigation on Fatigue Performance of Non-Load-Carrying Fillet Welded Joints. J. Constr. Steel Res.; 2017; 130, pp. 193-201. [DOI: https://dx.doi.org/10.1016/j.jcsr.2016.12.010]
31. Li, F.; Wu, P.; Zeng, J. Method for Compiling Fatigue Test Load Spectrum of Underframe Equipment Supporting Structure. J. Mech. Eng.; 2016; 52, pp. 99-106. (In Chinese) [DOI: https://dx.doi.org/10.3901/JME.2016.24.099]
32. Xiao, Q.; Huang, W.; Chen, D.; Sun, S.; Li, Q.; Shi, X. Fatigue Reliability Assessment of Bogie Frame Considering Nonlinear Cumulative Damage Model Correction. J. Mech. Eng.; 2025; pp. 1-13. Available online: https://link.cnki.net/urlid/11.2187.th.20250624.1555.022 (accessed on 3 November 2025). (In Chinese)
33. Chen, D.; Xiao, Q.; Mou, M.; Yang, W.; Liu, X.; Zeng, Y. Fatigue Reliability Evaluation of Heavy-Haul Locomotive Car Body Underframe Based on Measured Strain and Virtual Strain. Int. J. Fatigue; 2023; 172, 107661. [DOI: https://dx.doi.org/10.1016/j.ijfatigue.2023.107661]
34. Chen, D.; Xiao, Q.; Mou, M.; Sun, S.; Li, Q. Study on Establishment of Standardized Load Spectrum on Bogie Frames of High-Speed Trains. Acta Mech. Sin.; 2019; 35, pp. 812-827. [DOI: https://dx.doi.org/10.1007/s10409-019-00841-6]
35. Peng, B.; Wu, X.; Zhang, Z.; Yang, N.; Hu, F.; Wang, P.; Wu, S. Determination of Wheel Polygonal Wear Limit and Fatigue Life of Railway Bogie Frames Considering Wheel/Rail Excitation. Eng. Fail. Anal.; 2025; 169, 109220. [DOI: https://dx.doi.org/10.1016/j.engfailanal.2024.109220]
36. Cui, E. Development of Engineering Structure Data Processing and Fatigue Strength Evaluation System. Master’s Thesis; Beijing Jiaotong University: Beijing, China, 2005; (In Chinese) [DOI: https://dx.doi.org/10.7666/d.y741198]
37. Ju, Y.; Zhou, R.; Guo, Y.; Li, J.; Fang, B. Development and Key Technologies of Fatigue Analysis Software for Ship Shafting in Ice Regions. J. Dalian Marit. Univ.; 2023; 49, pp. 48-57. (In Chinese) [DOI: https://dx.doi.org/10.16411/j.cnki.issn1006-7736.2023.02.006]
38. Jia, X.L.; Dai, J.T.; Luo, Z.M.; He, J.; Liu, H.; Wang, Y.; Yu, T. Development of Fatigue Analysis Software for Coiled Tubing Based on Artificial Neural Networks. Oil Field Equip.; 2025; 54, pp. 1-6. (In Chinese) [DOI: https://dx.doi.org/10.3969/j.issn.1001-3482.2025.05.001]
© 2026 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license. Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.