The materials for this experiment were U71Mn rail steels with varying levels of manganese. Two commercially availible materials were studied: a regular U71Mn rail steel, which is referred to as RM, and a modified version with lower manganese concentration, termed HM (Zhao et al., 2022). Table 1 illustrates the results of the chemical analysis in which it was determined that the manganese content of HM steel was 0.94%, which is approximately 0.1% less than RM steel’s manganese content of 1.03%. The chemical compositions are measured by heat analysis, following GB/T 222-2006 and TB/T 3276-2011 (Zhao et al., 2022). Both steels fulfilled the criteria of TB/T 3276-2011 standard specification for U71MnG rail steel (Ruijie et al., 2023). The analysis of the chemical composition uniformity at different positions of the rail head, web and foot showed little difference, thus proving the rail section is homogenous compositionally.

The experimental study precisely controlled the flash welding parameters, based on optimized methodologies for high-carbon rail steels (Pang et al., 2022). Key parameters were determined through systematic welding trials (Saita et al., 2013), including flashing speed (0.85–1.15 mm/s), upset amount (13–16 mm), and forging force (950–1150 KN). Three distinct cooling rates were implemented: rapid cooling (2.4°C/s), medium cooling (1.6°C/s), and slow cooling (0.8°C/s), following established rail welding protocols (Kuchuk-Yatsenko et al., 2016). The temperature-time cooling curves for these three cooling rates are shown in Figure 1. The curves clearly demonstrate the distinct cooling behaviors, with the rapid cooling rate (2.4°C/s) showing the steepest temperature decline, while the slow cooling rate (0.8°C/s) exhibits a more gradual descent. This controlled cooling process is critical for the development of desired microstructural features in the heat-affected zone.

Post-weld heat treatment (PWHT) optimization was applied based on recent advances in heat treatment technology for rail steels (Porcaro et al., 2019). Normalization was performed at 610°C ± 10°C for 2 h, followed by controlled cooling to 300°C with rate of 140°C/h and then air cooling. The HM steel, with its optimized manganese content of 0.94% compared to 1.03% in RM steel, demonstrated improved properties without post-weld heat treatment, aligning with contemporary research on composition-property relationships (Ghazanfari and Tehrani, 2021).

Rail specimens were welded using thermally controlled flash butt welding techniques. Thermal simulations were carried out with three distinct cooling rates which included fast cooling at 2.4°C/s, medium cooling at 1.6°C/s, and slow cooling at 0.8°C/s (Mortazavian et al., 2022). After normalizing, PWHT was performed on RM specimens, whereas HM specimens were evaluated in the as welded state. Weldability of U71Mn, U75V, and four experimental compositions with different manganese and chromium quantities were modelled with JMATpro (Zhang B. et al., 2022). The values of carbon equivalent, shown in Table 2, were calculated using International Institute of Welding (IIW) and Japanese Industrial Standard (JIS) standards. The U71Mn standard was surpassed by the experimental composition 202109–1 which had a slightly lower carbon equivalent, while 202109–4 had higher values, and all new compositions exhibited a microstructure of 100% pearlite.

Multiple rail sections were analyzed in terms of microstructure and rail performance independently. An analysis specimens were collected from the head, web, and foot of the rail for microstructural evaluation (Wróblewsk et al., 2023). The network ferrite was measured using optical microscope and was quantified at a ×100 magnification. Scanning electron microscopy (SEM) was used to study martensitic microstructure in the heat-affected zone by sampling 30 × 30 mm² areas (shown in Figure 2a), while assessing the heat treatment response, the area fraction of martensite was determined at varying cooling rates.

All mechanical properties were evaluated through the standard tensile, impact, and hardness tests. The machining of the tensile specimens was done as shown in sampling diagram Figure 2b. Testing was done for yield strength (Rp0.2), ultimate tensile strength (Rm), and elongation (A%) (Gu and Wang, 2022). Systematic comparison between domestic and foreign rail steel was also carried out, as shown in Table 3. Impact testing was performed with vertical specimens taken as per Figure 2c with primary dependence on manganese content and impact toughness correlation. Examination of fracture surfaces was conducted using SEM to study the microcrack formation and propagation mechanisms.

Statistical methods were applied to study the connection between the chemical make-up and the microstructural features of the rail steels. The martensite formation was correlated to the carbon equivalent by fitting analysis which rendered a value of 66% correlation coefficient (Du et al., 2023). Manganese segregation analysis was conducted using energy dispersive spectroscope (EDS) on the SEM samples where segregation rates for both martensite and matrix regions were determined. The dependence of martensite content on manganese content was rather weakly correlated with 27%, while the correlation of base material with martensite manganese content was 66%, which is presented in Table 4 (Zhu et al., 2024). The analysis of segregation provided the result of the average, over ten rail steel grades, manganese segregation rate of 77%, which was much greater than the chromium segregation rate of 55%, thereby illustrating the considerable degree of manganese segregation during welding.

This section presents a comprehensive evaluation of rail steel performance, including microstructure, mechanical properties, and comparisons between domestic and international rail grades. The investigation focuses particularly on the effects of the reduced manganese content then conventional RM steel.

The martensitic transformation in the heat-affected zone has a decisive influence on the performance of welded joints. Quantitative analysis of the martensite content in different rail steels under various cooling conditions reveals the intrinsic relationship between martensite transformation behavior, chemical composition, and process parameters.

Software simulation using JMATpro was conducted to analyze the weldability of U71Mn, U75V, and four experimental compositions. Analysis showed that experimental composition 202109–1 achieved a lower carbon equivalent compared to standard U71Mn, while maintaining 100% pearlitic microstructure. The CCT curves obtained through simulation (Figure 5) demonstrated systematic shifts with varying chromium content, showing a rightward displacement of approximately 2 s with increasing Cr levels.

As shown in Table 11 and Figures 6, 7, the martensite content increases significantly with increasing cooling rate and exhibits distinct differences among rail steel grades. Under the same cooling conditions, the martensite content from highest to lowest is as follows: Pangang U71Mn, Angang U71Mn, Voestalpine DO350, French HY260, French HY350, and Japanese R350.

Further analysis reveals that the tendency for martensite formation has a strong correlation with the carbon equivalent. The higher the carbon equivalent, the greater the driving force for martensitic transformation and the higher the volume fraction of martensite. Statistical analysis indicates that the correlation coefficient between carbon equivalent and martensite content is as high as 66%, which is significantly stronger than the correlation with the single alloying element manganese content (27%). This suggests that when assessing the martensite tendency of rail welded joints, the combined effect of various alloying elements should be comprehensively considered, and using the carbon equivalent as a criterion is more reasonable.

The synergistic effects between alloying elements demonstrate significant influence on microstructural evolution and phase transformation characteristics. Statistical analysis of elemental interactions reveals complex relationships governing martensite formation and segregation behavior. The manganese segregation rate reaches 77% compared to chromium at 55%, indicating a strong tendency for manganese to form MnS inclusions. More significantly, the carbon equivalent exhibits a robust correlation with martensite content (R = 0.66), while manganese content alone shows a weaker correlation (R = 0.27), suggesting that manganese primarily influences phase transformation through modifying carbon activity rather than direct effects.

Systematic investigation of transformation behavior under controlled cooling conditions provides crucial insights into the martensite formation mechanisms. Under rapid cooling conditions (2.4°C/s), specimens with higher carbon equivalent (Ceq = 2.28) developed 3.69% martensite, while those with lower carbon equivalent (Ceq = 1.86) formed only 0.05% martensite. CCT curve analysis reveals a systematic rightward shift of approximately 2 s with increasing chromium content, directly impacting transformation kinetics. Local composition analysis demonstrates significant manganese enrichment in martensitic regions (2.14%) compared to the matrix (1.17%), confirming the critical role of compositional segregation in determining local hardenability.

The optimization of silicon and aluminum contents (Si = 0.33%, Al = 0.003%) resulted in superior microstructural homogeneity, as evidenced by quantitative metallography. These relationships are clearly illustrated in Supplementary Figure S1. It is worth noting that the martensite content in welded joints exhibits distinct spatial distribution characteristics. The martensite content in the rail web region is generally higher than that in the rail head and foot regions, mainly due to the relatively faster cooling rate and more severe composition segregation in the web region. EDS analysis shows that the average segregation rate of manganese is as high as 77%, significantly higher than that of chromium (55%), indicating that manganese preferentially segregates and has a greater influence on local hardenability.

The modified chemical compositions, particularly through manganese reduction and chromium addition, produced several significant effects on weldability. The lower manganese content reduced segregation tendency during welding, while the carefully controlled chromium additions helped maintain adequate hardenability without promoting excessive martensite formation.

In summary, the martensitic transformation behavior is closely related to the carbon equivalent of the rail steel and is significantly influenced by welding process parameters (such as cooling rate) and composition segregation. In optimizing the performance of welded joints, the martensite structure should be regulated through rational composition design and process control to avoid brittle fracture caused by excessively high local hardness.

Gray spots remain a critical quality concern in flash-welded joints. Macroscopic examination reveals that these defects appear as elliptical smooth regions with a slightly darker surface compared to the surrounding fracture surface, as shown in Figure 8. These defects exhibit a characteristic distribution pattern, predominantly concentrated within 30 mm from the rail foot edges. Statistical analysis of 50 welded joints reveals that approximately 65% of gray spots occur in the foot region, 28% in the web region, and only 7% in the head region, with an average gray spot size of 6.8 ± 0.5 mm in length and 1.0 ± 0.1 mm in width.

The formation of gray spots involves complex chemical reactions during the welding process. High-speed photography analysis (Figure 9) reveals that during continuous flash welding of the rail web, void formation occurs through three distinct stages: electromagnetic explosion, metal droplet ejection, and void formation. In the initial stage, electromagnetic forces expel metal from the contact interface at temperatures exceeding 1,500°C. The chemical reactions proceed according to the following sequence: 2 Mn + O 2   →   2 MnO  T   >   1500 ° C Si + O 2   →  SiO 2   T   >   1500 ° C 2 MnO + SiO 2   →  Mn 2 SiO 4   manganese silicate formation

Chemical analysis of gray spot regions shows elevated concentrations of manganese (2.14% ± 0.15%), silicon (1.58% ± 0.12%), and oxygen (0.86% ± 0.08%) compared to the surrounding matrix. The spatial distribution of these elements varies systematically across different regions of the weld, as detailed in Table 12. The edge zones consistently show higher manganese concentrations (2.23% ± 0.16%) compared to the center zones (1.89% ± 0.13%), indicating preferential segregation during the welding process.

The spatial distribution of key elements involved in gray spot formation was further investigated using EDS mapping. As shown in Figure 10, manganese, silicon, and oxygen exhibit clear concentration gradients across different regions of the weld. The edge zones consistently show higher manganese concentrations compared to the center zones, corroborating the quantitative EDS results presented in Table 12. This visual evidence strengthens the understanding of preferential manganese segregation as a driving force behind gray spot formation in flash-welded rail joints.

Manganese plays a dual role in gray spot formation. First, dissolved manganese in the steel matrix readily oxidizes at elevated temperatures, forming MnO which subsequently reacts with SiO2 to form complex silicates. Second, pre-existing manganese sulfide inclusions decompose during welding, releasing additional manganese that contributes to oxide formation. This dual-source mechanism explains why reducing the bulk manganese content by 0.1% (from 1.03% to 0.94%) resulted in a significant 35% reduction in gray spot frequency.

The investigation revealed that void formation typically occurs within approximately 1.5 s of initial flash formation. These voids, measuring 6–8 mm in length and approximately 1 mm in width, provide channels for atmospheric interaction with the hot metal surface. The subsequent rapid closure of these voids by the movement of the dynamic frame effectively seals the silicate inclusions within the weld interface. The process occurs most frequently in regions experiencing higher thermal gradients and mechanical constraints, explaining the preferential formation near the rail foot edges.

Experimental trials with reduced manganese content demonstrate a clear correlation between bulk manganese content and gray spot formation. Statistical analysis of weld cross-sections shows that reducing manganese content from 1.03% to 0.94% decreases the average number of gray spots per weld from 8.3 to 5.4, with a corresponding reduction in average spot size from 7.2 ± 0.6 mm to 6.5 ± 0.5 mm. This improvement is attributed to both reduced manganese availability for oxide formation and decreased manganese sulfide inclusion content in the base material.

The study concludes that effective control of gray spot defects requires a multi-faceted approach combining composition control and process optimization. By rationally designing the composition (particularly reducing manganese and silicon content), improving surface preparation processes, optimizing flashing parameters (such as shortening the flashing time and increasing current density), and maintaining adequate forging pressure, gray spot formation can be significantly suppressed. These findings provide practical guidelines for improving the quality and reliability of flash-welded rail joints in high-speed railway applications.

Rail foot micro-cracks are another typical defect in flash-welded joints. Metallographic analysis reveals that micro-cracks are primarily distributed on both sides of the rail foot, exhibiting various morphologies along the weld interface (Figures 11, 12a,b. These defects are most concentrated in the overheated zone, 0.2–0.8 mm away from the weld center, and frequently occur within 0–5 mm below the rail foot surface. Statistical analysis of 50 welded joints reveals that approximately 65% of micro-cracks occur in the foot region, 28% in the web region, and only 7% in the head region, with an average micro-crack size of 6.8 ± 0.5 mm in length and 1.0 ± 0.1 mm in width.

The formation mechanism of these micro-cracks is intimately linked to the deformation behavior of MnS inclusions during the welding process. In the base material, MnS inclusions exhibit a horizontally elongated banded morphology, as shown in Figure 12c where the left image shows the schematic distribution of inclusions, and the right image presents the actual metallographic observation of the banded structure. During flash welding, these inclusions undergo a two-stage deformation process. The first stage involves bending deformation during the upsetting stage, causing joint bulging as illustrated in Figure 12d. The continuous arrow lines in the schematic diagram demonstrate how the originally horizontal banded structure gradually deforms following the plastic deformation of the joint region during the upset process. The actual microstructural observations confirm this systematic reorientation of MnS inclusions in response to the applied thermomechanical forces during welding.

Following the upsetting stage, the trimming tool moves from the right end to the left end of the joint to remove the flash upset. This operation induces a second deformation in the surface-adjacent banded structure, as depicted in Figure 13a, where the near-surface banded structure bends in the trimming direction, creating conditions conducive to micro-crack formation. Figure 13b illustrates how MnS inclusions at the ends of the high-temperature overheated zone experience stress state changes during trimming, leading to crack formation along the inclusion direction due to the release of stress perpendicular to the surface. The greater deformation angle on the left side of the weld results in a higher concentration of micro-cracks in this region.

Fatigue testing provides important insights into the practical implications of these micro-cracks. Analysis of four flash-welded joints that passed standard fatigue testing (2 × 10⁶ load cycles according to TB/T 1632.2-2014) revealed the presence of micro-cracks in all specimens. This suggests that while micro-cracks are prevalent, they can be tolerated to a certain extent by the joint’s service performance. Statistical analysis of these fatigue-tested joints showed varying degrees of micro-crack severity, with some specimens exhibiting up to 49 cracks with maximum lengths reaching 571 μm, yet still maintaining acceptable fatigue resistance.

The analysis indicates that micro-cracks are inherent in flash-welded joints due to the presence of MnS inclusions in rail steel. These defects typically distribute in the overheated zone 0.2–0.8 mm from the weld center and within 0–5 mm of the trimmed surface. While complete elimination of micro-cracks may be difficult, their impact can be minimized through several approaches: reducing MnS inclusions in the rail steel, modifying the trimming strategy by increasing the trimming allowance, optimizing the geometric dimensions of the rail foot (such as transition fillets), and improving the precision machining quality and surface roughness of the rail foot. The presence of micro-cracks, while persistent, does not necessarily compromise joint integrity when maintained within acceptable limits through proper material and process control.