A method for detecting surface cracks in steel rails

Abstract

This invention relates to a method for detecting surface cracks in steel rails. By controlling changes in the ambient temperature of the rail to induce a hysteresis in the thermal response, slight asynchronous thermal fluctuations with discontinuous and asymmetric thermal diffusion occur in localized areas of the material surface, exposing potential crack boundaries. A non-contact, low-intensity laser thermal disturbance multi-angle scanning is used to perform a disturbance penetration test, eliminating non-structural thermal anomalies and identifying the actual crack area. The thermal response decay curves are recorded after multiple rounds of disturbance excitation, and the integrity of the microstructure beneath the crack is deduced based on the disturbance settling time to determine the structural fatigue level. A steady-state environmental comparison is performed on the same detection area. By comparing changes in temperature, humidity, wind speed, and radiation environmental parameters before and after the disturbance, if the crack area continues to exhibit abnormal thermal response under constant environmental conditions, the thermal characteristic is confirmed to be caused by a structural crack; otherwise, misjudgments caused by occasional external interference are ruled out.

Description

Technical Field

[0001] This invention relates to rail crack detection, specifically a method for detecting cracks on the surface of rails. Background Technology

[0002] Current technologies for detecting surface cracks in rails, such as Chinese patent CN111678948A (High-speed non-destructive testing method and implementation device for rail surface defects), while offering significant advantages in detection speed, operational efficiency, and hardware integration, still suffer from several technical bottlenecks and systemic deficiencies in handling complex operating environments and identifying fine cracks. Firstly, regarding its detection mechanism, this patent uses a jet flame as the excitation source, acquiring thermal response images and analyzing thermal flow disturbances in the crack area using an infrared thermal imager. While flame excitation offers advantages in high heating efficiency and low cost at high speeds, this method is essentially a macroscopic heating approach. Its sensitivity to minute differences in thermal resistance and subsurface cracks is low, especially in the presence of surface interference factors such as impurities, rust, and oil. The temperature field disturbance signal in the infrared image is easily masked, leading to misjudgments or missed detections of cracks. In addition, this method emphasizes controlling the angle between the flame angle and the direction of movement between 100° and 170°, and uses a linear flame array to improve heating uniformity. Although this can improve the quality of the visible image to a certain extent, the spatial distribution of flame control and the heat diffusion path are highly dependent on physical symmetry and are easily affected by the bending, deformation and structural non-uniformity of the rail surface, which causes spatial distortion of the thermal disturbance response and reduces the accuracy of the detection results.

[0003] Secondly, this method does not establish an effective classification and discrimination mechanism for the multi-type structural characteristics of cracks (such as closed, open, and bifurcated types), nor does it have the ability to identify the direction of crack boundaries. This makes it difficult to identify the crack direction and propagation trend in complex structures with multiple intersecting cracks, and it also lacks dynamic monitoring methods for the evolution paths of different structural cracks. More importantly, the detection method proposed in this method is a single-excitation mode of flame thermal disturbance, and its temperature change response mainly depends on the heat conduction process. However, in the actual high-speed rail operating environment, the rail structure is often affected by multiple environmental factors such as wind speed disturbance, lightning-induced electric field, train vibration, and diurnal temperature difference. These factors produce a significant background interference on the thermal imaging results after flame excitation. This patented method does not perform baseline calibration for the steady state of the environment and lacks the ability to track and discriminate the nonlinear change trend of the thermal response, making it difficult to effectively distinguish between structural cracks and occasional disturbances, and posing a certain degree of risk of misidentification.

[0004] Furthermore, this method does not incorporate a thermal response attenuation analysis mechanism under multi-cycle perturbation, nor does it design a quantitative analysis model to assess the material integrity and fatigue level beneath the crack. It relies solely on surface thermal images for apparent identification, failing to predict crack evolution trends or provide risk warnings. This is particularly evident in the early stages of dynamic microcrack propagation, where the method's identification capability is relatively slow. In addition, this flame excitation method still suffers from issues such as gas supply stability, complexity in electronically controlled flame adjustment, and safety hazards related to wind and explosion protection in outdoor environments. The uneven heating problem is particularly pronounced in cold, humid, or high-wind areas, severely limiting its real-time adaptability and large-scale deployment. Summary of the Invention

[0005] The purpose of this invention is to provide a method for detecting cracks on the surface of rails, thereby addressing some of the drawbacks and shortcomings pointed out in the background art.

[0006] The present invention addresses the aforementioned technical problems by employing the following technical solution: a method for detecting surface cracks in rails, comprising: controlling the temperature difference change in the rail environment to induce a hysteresis phenomenon in the thermal response of the rail metal in the stress concentration area, thereby inducing slight asynchronous thermal fluctuations with discontinuous and asymmetric thermal diffusion in a local area of ​​the material surface to expose potential crack boundaries; and applying a directional short-period external disturbance signal, including laser thermal pulses or ambient acoustic interference, based on the asynchronous thermal fluctuation area, to cause thermal phase disturbances at the crack edge due to discontinuous heat conduction and to form a phase difference with misaligned response time.

[0007] For the region where the phase difference is generated, a non-contact low-intensity laser thermal perturbation multi-angle scanning is used to perform a perturbation penetration test to identify the difference between the laser thermal perturbation transparent area and the weak response deposition area, eliminate non-structural thermal anomalies and determine the real crack area; the thermal response of the real crack area after multiple rounds of perturbation excitation is recorded as a decay curve, and the integrity of the microstructure below the crack is deduced based on the perturbation settling time to determine the structural fatigue level;

[0008] After completing the fatigue level determination, a steady-state environmental comparison is performed on the same test area. By comparing the changes in temperature, humidity, wind speed, and radiation environmental parameters before and after the disturbance, if the crack area continues to show abnormal thermal response under constant environmental conditions, the thermal characteristic is confirmed to be caused by structural cracks; otherwise, misjudgment caused by occasional external interference is ruled out.

[0009] Furthermore, inducing the asynchronous thermal fluctuations includes pre-calibrating the material regions on the rail surface before thermal excitation to identify low heat capacity regions as crack-sensitive candidate points; controlling the ambient temperature difference includes setting alternating heating and slow cooling processes to construct an excitation sequence for thermal reversal inertia.

[0010] Furthermore, during the application of the perturbation signal, the laser pulse frequency is dynamically adjusted according to the amplitude of the thermal response delay at the crack boundary, so as to amplify the phase response differences between crack types including closed / open / forked cracks; the laser thermal perturbation adopts an alternating method of vertical excitation and tangential excitation to detect the directional thermal conduction barrier at the crack edge, so as to determine the crack direction.

[0011] Furthermore, the phase difference is extracted by vectorized reconstruction of the relative time delay of the temperature peak; the perturbation multi-angle scanning adopts clockwise and counterclockwise dual-sequence cross-excitation to detect the asymmetry of the thermal perturbation path in the crack region and to determine whether the crack penetrates the cross section; the perturbation penetration test results are mapped by the time difference between the transparent area and the weak response area to form a real crack region distribution map based on the delay topology.

[0012] Furthermore, the confirmation result of the actual crack area is used to trigger a secondary low-frequency thermal disturbance to re-verify the thermal path stability, so as to eliminate the misjudgment response caused by surface impurities or oil stains; the temperature, humidity and wind data collected during the environmental steady-state comparison process are synchronously recorded in the crack response log, and anomaly backtracking analysis is performed by comparing the change trend of the excitation response before and after.

[0013] Furthermore, when a structural crack is detected in the detection area by comparing the environmental steady-state conditions, a supplementary detection task is triggered in the adjacent area. The supplementary detection range is not statically set, but rather an adaptive risk diffusion radius is constructed based on the current crack response behavior and thermal diffusion characteristics, thereby activating an active scanning mechanism with this risk radius as the boundary. To quantify the evolution trend of the crack, a thermal response nonlinear enhancement trend function is constructed, defined as follows:

[0014] ;

[0015] in:

[0016] To indicate in time The crack thermal response risk integral value at any given time is used to determine whether the crack has entered a high-risk zone for crack propagation; For the core region of the crack in time The thermal response temperature value at that time; The first derivative of the thermal response represents the instantaneous temperature rise / thermal disturbance rate; The second derivative of the thermal response represents the acceleration trend of the response, i.e., a sharp change in temperature response. The square of the temperature difference between the current moment and the reference state before excitation is used to highlight the change in response intensity; This is an exponential suppression term that controls the weight of the influence of historical responses. It is the attenuation factor; The crack region in spatial coordinates The response morphology function at the location reflects its geometric deformation, orientation, and boundary perturbation characteristics; The empirical adjustment coefficient controls the weights of different response features on the risk function. To monitor the start time, the integral is calculated from the initial identification to the current time;

[0017] This function is a high-order thermal dynamic behavior feature aggregation formula: it integrates first-order and second-order thermal behavior, nonlinear amplitude, decay memory, and spatial features; it can be calculated in real time, and the results... With preset threshold After comparison, it is used to determine whether the crack has entered the critical zone for propagation; once... If this occurs, the system will mark the area as a high-risk zone for crack propagation and enter dynamic monitoring mode; simultaneously, it will... The mapping range is extended to the neighboring region as input, realizing an active scanning closed loop driven by the risk diffusion radius;

[0018] The above The derivation of the function:

[0019] Observations show that the temperature response behavior of the crack region is not stable and constant over continuous excitation cycles, but rather exhibits typical nonlinear fluctuations, including a significant increase in the heating rate (i.e., the first derivative) in some cycles. (Increase), or the acceleration of temperature rise also increases significantly in certain cycles (i.e., the second derivative). (Increase), and at the same time, the temperature difference between the crack region and the reference thermal state also increases quadratically; hence, a first derivative term is introduced. As a core indicator of the rate of change of thermal disturbance, it reflects the instantaneous response capability of the crack to external excitation; then, the second derivative term is introduced. This is used to measure the drasticness and abrupt evolution of temperature response trends; in addition, the square of the temperature difference between the current time and the initial time also needs to be introduced. To capture the nonlinear enhancement behavior of thermal response intensity, this square structure has a mathematical amplification effect that enhances violent fluctuation behavior;

[0020] In real physical systems, historical responses have a certain inertial memory effect on the current state, but this effect cannot be infinitely accumulated. Therefore, an exponentially decaying weighting function is introduced. This is used to limit the contribution of historical responses far from the current time in the time integral, thereby making the responses of the most recent time period more prominent;

[0021] To combine all the aforementioned thermal behavior characteristics into a continuous integral function, they are uniformly constructed into an integral expression. In this context, the time variable starts from the initial excitation time. Points accumulated to the current time It covers responses across multiple excitation cycles; simultaneously, to incorporate spatial factors into the risk calculation relationship, a spatial thermal response function is introduced. , indicating the crack in spatial coordinates The thermal morphological coupling characteristics, including edge perturbations, orientation, and symmetry factors, are used as coefficient multipliers for the entire integral structure, thereby linking temporal and spatial thermal behaviors in a unified thermal risk expression. Finally, empirical control coefficients are assigned to each behavioral term. The relative weights of thermal change rate, first-order abrupt change, and response intensity to total risk are controlled respectively.

[0022] In summary, the following original integral risk function is derived. In practical systems, this function exhibits the following behavior: when there is a sustained thermal strengthening trend in the crack region, both its first and second derivatives increase positively, while the squared temperature difference term amplifies the overall risk integral value; if the response speed tends to stabilize, or the fluctuations tend to weaken, the integral curve will tend to saturate, eventually... Can be used with preset thresholds The system performs real-time comparisons, and when the value exceeds the threshold, it determines that the current area has entered a high-risk state for crack propagation. It then automatically activates the dynamic monitoring status and the active scanning mechanism for neighboring areas, achieving a complete closed loop from physical process modeling to control strategy linkage.

[0023] Furthermore, the risk diffusion radius is adjusted based on the local thermal conductivity gradient change of the rail material to identify potential abnormal heat conduction paths; when the active scanning mechanism identifies multiple abnormal thermal response regions in the vicinity, it clusters each response region according to the similarity of its thermal behavior to the original crack to determine whether they belong to the same extended chain.

[0024] Furthermore, if the same nonlinear response pattern appears in the detection of the neighboring region, the current region and the neighboring region are marked as structural co-decay regions and enter a synchronous dynamic monitoring state; the nonlinear response enhancement trend is captured by the change of the second derivative of the thermal response to identify the dynamic growth rate of the crack and predict the expansion time window.

[0025] Furthermore, after entering the dynamic monitoring state, a dynamic thermal excitation power adjustment mechanism is adopted to make the excitation energy linearly matched with the crack response intensity; wherein the duration of the dynamic monitoring state is determined by the stability change amplitude of the crack response in multiple consecutive cycles, and if the fluctuation value is less than the set threshold, it is downgraded to the periodic scanning mode.

[0026] Furthermore, under the dynamic monitoring state, the system synchronously records environmental disturbance indicators including wind speed, vibration, and lightning electric field, which are used to dynamically adjust the nonlinear trend judgment threshold; the continuous thermal response data collected under the dynamic monitoring state is processed by differential fitting and used to determine whether the crack has entered the critical state of thermal hysteresis and thermal destability, as a basis for high-level risk identification.

[0027] The beneficial effects of this invention are as follows: This invention provides a method for detecting surface cracks in rails, combining a thermophysical response induction mechanism with a dynamic perturbation scanning strategy. This allows for the precise exposure of crack boundaries by inducing asynchronous thermal fluctuations due to stress concentration through environmental temperature differences, before the cracks become visible. Simultaneously, by applying directional short-period perturbations using laser thermal pulses and acoustic interference, and through thermal phase difference reconstruction and non-contact penetration testing, the crack direction, type, and depth are effectively identified, improving the adaptability for detecting various types of cracks. Furthermore, this method calculates the structural fatigue level through thermal response decay behavior and perturbation settlement time analysis, and introduces environmental steady-state comparison to eliminate misjudgments, significantly improving the accuracy and engineering robustness of the detection.

[0028] Furthermore, this invention constructs an active scanning mechanism based on the risk diffusion radius and a nonlinear response clustering model, which can identify crack propagation chains and dynamic evolution trends in real time. Through dynamic excitation power adjustment, environmental disturbance perception, and second-order derivative differential fitting, it accurately assesses whether cracks have entered the critical states of thermal hysteresis and thermal destability, thereby achieving a closed-loop control of "identification-prediction-response." The overall method has advantages such as high sensitivity, high anti-interference, and high intelligence, and is suitable for early crack detection and risk level management in complex operating environments such as high-speed railways and heavy-haul tracks. Attached Figure Description

[0029] Figure 1 This invention simplifies the main process of thermal detection of rail cracks.

[0030] Figure 2 This is the main process for asynchronous thermal response detection of cracks in this invention;

[0031] Figure 3 This is the main process for dynamic diffusion and adaptive monitoring of crack risk in this invention;

[0032] Figure 4 This is the main process for crack detection in mountainous railways according to Embodiment 1 of the present invention;

[0033] Figure 5 This is a flowchart of the intelligent monitoring process for crack propagation in mountainous railways, as described in Embodiment 2 of the present invention. Detailed Implementation

[0034] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0035] Combined with appendix Figure 1 This invention discloses a method for detecting surface cracks in rails. During static inspection or non-operational periods, a controlled alternating heat-cold gradient is applied to the rail surface using an external environmental temperature difference control device. This temperature difference is achieved by alternating operation of an infrared heater and a forced air convection cooling module. The purpose of this step is to utilize the difference in thermal response inertia exhibited by metallic materials under abrupt temperature changes, especially in stress concentration areas—a thermal inertial hysteresis effect. This causes a measurable difference in the thermal diffusion capacity of potential crack locations within the microstructure compared to surrounding crack-free areas, thereby inducing a slight asynchronous thermal fluctuation in the thermal image response—a "discontinuous and asymmetric" fluctuation. Asynchronous thermal fluctuations refer to the microscopic perturbation changes in thermal conductivity or heat capacity near the crack, causing the temperature rise rate of this area to be slower or faster than that of the surrounding uniform area after heating. This results in a phase delay or advance in the temperature curve on the time axis, manifesting as irregular hot spot boundaries in thermal imaging equipment. These abnormal boundaries are precisely the potential path indication signals of the crack. Secondly, after identifying the region with asynchronous thermal response characteristics, the system continues to apply a highly directional, short-period external perturbation excitation signal to this region. The perturbation signal can be a laser thermal pulse, directional microwave excitation, or fixed-frequency acoustic wave excitation as the carrier. Laser thermal pulses are preferentially suitable for enhancing the response of narrow cracks, while acoustic wave excitation is suitable for the response diffusion excitation of through-cracks or bifurcated cracks. The excitation signal is based on low power consumption, high repetition rate, and small pulse width as its parameter control basis, and the excitation direction must have a certain angle with the normal direction of the crack's asynchronous thermal boundary, thereby achieving effective energy coupling in the crack edge region. This excitation will induce an instantaneous thermal response shift at the crack boundary due to the discontinuity of the heat conduction path. Specifically, this manifests as an observable misalignment in the thermal perturbation phase at the crack edge with the uncracked regions on either side, i.e., the so-called thermal phase perturbation. The thermal phase misalignment will expand or intensify to a certain extent as the excitation intensity or frequency changes, thus forming a clear and identifiable phase difference mapping. This mapping can be captured in real time by a highly sensitive infrared detection device, and its thermal phase vector can be analyzed by a calculation module to ultimately achieve accurate identification of the crack edge and extraction of the crack initiation path.

[0036] After identifying the thermal phase disturbance region at the crack edge, a non-contact, low-intensity laser thermal disturbance scan is performed on the target area generating the thermal phase difference. Using a laser pulse source below a preset power threshold, coupled with a multi-angle rotating scanning mirror assembly, the region is repeatedly disturbed at different incident angles and scanning paths to perform a so-called disturbance penetration test. During this test, the system continuously records the instantaneous temperature response changes after laser pulse excitation at various angles to determine whether there are differences in the thermal disturbance penetration capability of local areas. If a local area exhibits extremely low thermal response or a significantly delayed response time under multi-angle laser excitation, the system marks it as a weak response deposition area. Such areas are often caused by structural cracks, thermal cavities, or micro-damage. Conversely, if an area consistently recovers rapidly or has a linear response under excitation, it is marked as a laser disturbance transparent area. This type of response is usually caused by surface contaminants, uneven thermal conductivity, or measurement errors. Through the above-mentioned identification of differences in regional thermal response penetration, the system can effectively eliminate non-structural thermal anomaly response areas and screen out real crack areas with continuously decaying hysteresis characteristics in thermal behavior without relying on image edge algorithms. Furthermore, multiple rounds of perturbation excitation experiments were conducted on the selected real crack regions. That is, without changing the excitation parameters, the same laser thermal perturbation was periodically and repeatedly applied. After each round of excitation, the system recorded the thermal response decay curve of the crack region in real time, including thermal behavior parameters such as the maximum temperature value, response rise and fall time, and steady-state recovery time, and constructed a time-series response model. During this process, if the crack region exhibited a slower temperature rise rate, a longer heat deposition time, or a continuously slower cooling rate after excitation, it indicated that its internal thermal diffusion path was obstructed. Using this phenomenon, the perturbation settling time was defined as a criterion, i.e., the time required for the crack region to decay from the peak of the excitation response to a steady state. This time can quantify the degree of thermal resistance of the internal structure, thereby inferring the integrity of the microstructure beneath the crack. That is, if the settling time is prolonged, it often corresponds to significant cavities, delamination, or grain boundary deterioration within the crack; conversely, if the settling time is short and stable, it indicates that the crack is more likely to be a surface fine crack or an early micro-crack.

[0037] Further, a steady-state environmental comparison step is implemented in the same detection area to ensure that the abnormality of the crack thermal response behavior originates from the inherent defects in the material structure, rather than being misjudged due to occasional interference from environmental factors. After multiple rounds of laser thermal disturbance and response recording are completed, the system temporarily stops the excitation input and enters the steady-state environmental acquisition stage. Through the integrated multi-parameter environmental sensing components, high-precision readings of multiple physical parameters, including temperature (T), relative humidity (RH), wind speed (V), and radiation intensity (IR), are acquired in real time to construct a feature set of the environmental state after the disturbance. Subsequently, the system backtracks to the environmental parameters of the same type recorded within a time window before the start of the disturbance excitation to construct a feature set of the environmental state before the disturbance. The environmental parameters of the two stages are compared item by item, including indicators such as the rate of change of temperature ΔT, the rate of change of humidity ΔRH, the wind speed disturbance ΔV, and the radiation difference ΔIR. If the change amplitude of the above parameters is less than the set environmental stability threshold (i.e., within the natural fluctuation tolerance), the system will be considered stable. Within the specified range, the system can determine that the current detection window is in a constant environmental state. Under this premise, if the crack area continues to exhibit obvious abnormal thermal response characteristics, including temperature response peak shift, abnormally slow response speed, or stable deviation of the response curve from the normal trajectory, then the abnormal thermal response is determined to have a stable structural source, and can be confirmed as thermal behavior caused by structural cracks. Conversely, if any parameter such as temperature, humidity, wind, or radiation fluctuates beyond the threshold in the environmental comparison, or if the abnormal thermal response behavior highly coincides with the time point of environmental change, the system will mark the result as an anomaly related to occasional environmental interference and will not input it as a structural crack signal into the next stage of the evaluation model, thereby avoiding false detection problems caused by external interference, climate fluctuations, equipment errors, and other factors.

[0038] Combined with appendix Figure 2To improve the significance of asynchronous thermal fluctuation response and enhance the early identifiability of crack regions, the system first performs a pre-calibration operation of the material region before implementing thermal excitation. This includes using a high-sensitivity infrared scanning device to map the basic thermal characteristics of the rail surface under inspection, collecting parameters such as the initial heating rate, heat capacity response curve, and surface thermal diffusion efficiency at different locations under standard excitation conditions. Combined with a known crack generation probability distribution model, regions with heat capacity significantly lower than the regional average are marked as thermally sensitive candidate points. These regions often exhibit low heat capacity due to microscopic voids, uneven structure, or fatigue accumulation, and are therefore more prone to thermal response anomalies during subsequent thermal excitation, thus possessing the potential for preferential crack manifestation. This step constructs a thermal response potential map in space, providing high-confidence region guidance for the subsequent induction of asynchronous thermal fluctuations. Next, to achieve effective asynchronous thermal response excitation, a thermal excitation sequence of alternating heating and slow cooling was designed. This thermal process is executed in concert by a precisely controllable heat source and cooling system. In the heating stage, a multi-point heating method is used to rapidly raise the surface temperature of the rail to the target range while maintaining a constant excitation rate to ensure that a rate difference is formed between the normal heat conduction area and the crack-sensitive area in the heat diffusion process. Subsequently, the slow cooling stage is entered, and the temperature is slowly reduced by switching to a low-speed cooling module or natural convection. During the cooling process, the crack area exhibits more obvious response lag and spatial thermal distortion due to thermal inertia asymmetry, thereby inducing a thermal reversal inertia effect. That is, the thermal behavior of the crack area in the heating and cooling directions shows opposite response directions and response speeds, producing response asymmetry on the time axis. This asymmetry is the physical basis of asynchronous thermal fluctuation.

[0039] To enhance the response sensitivity and type discrimination capability of crack boundary identification, the system dynamically adjusts the excitation frequency of the laser pulse in real time based on the thermal response delay amplitude exhibited at different locations along the crack boundary during the application of laser thermal perturbation signals to the crack region. In the initial stage of laser excitation, the system first scans the crack boundary region at a standard frequency, acquiring the temperature rise response delay time at different scanning points in real time and comparing it with the laser pulse triggering sequence to form a thermal response time delay spectrum. If the delay amplitude in a certain region exceeds the system's set asynchronous response threshold, the system identifies that region as a thermal phase-sensitive segment. Subsequently, the laser pulse frequency is increased or decreased according to the magnitude of the response amplitude in that segment. High-frequency pulses are applied to high-delay regions to enhance thermal inertial coupling and amplify their phase misalignment characteristics, while low-frequency pulses are used to avoid excessive energy perturbation causing thermal response masking, thereby achieving segmented enhancement and adjustment of the crack's thermal phase response characteristics. This frequency dynamic control strategy effectively amplifies the thermal shielding phase shift caused by adhesion at the two side boundaries of closed cracks, the synchronous collapse-type thermal response abrupt change caused by the thermal channel rupture of open cracks, and the multi-directional thermal interference characteristics exhibited by bifurcated cracks at the splitting point. Ultimately, it forms a set of response differences with crack type distinction in the thermal phase space, facilitating subsequent category identification and grade assessment. Simultaneously, to further obtain the thermal conductivity directionality information of the crack edge, an alternating excitation method of vertical and tangential excitation is adopted during laser thermal perturbation. That is, the laser beam is alternately applied with the laser beam incident perpendicularly to the crack boundary normal direction and tangentially along the crack edge direction. This allows for comparison of the thermal response changes in the same region under different excitation incident angles. If the response is strong under vertical excitation and weak under tangential excitation, it indicates that the crack region has a directional thermal conductivity barrier, and heat flow diffusion along the edge direction is limited. Conversely, if the response is enhanced in the tangential direction, it indicates that the crack structure has a deep open form.

[0040] To achieve refined capture and structural recognition of the thermal disturbance behavior of cracks, a thermal phase difference extraction operation is performed on the crack region after the excitation disturbance response. The phase difference is not determined based on the image edge contour or intensity threshold, but is vectorized and reconstructed using the relative time delay of the temperature peak. The system continuously records the temperature change curves of each coordinate point within the crack region under multiple rounds of laser excitation, extracts the time of occurrence of the first main peak temperature for each point, constructs a time-domain vector group of the main peak time coordinates of all recorded points, and then normalizes it to calculate the temperature peak time delay difference between any adjacent points, thereby forming a phase delay vector field with the thermal response delay gradient as a feature. This vector field can clearly restore the local response asynchrony behavior caused by the obstruction of the heat conduction path within the crack region, solving the problem of misidentification of crack boundaries caused by edge blurring and unstable thermal maps in traditional methods. Subsequently, to further improve the resolution of crack geometric features under thermal perturbation, a clockwise and counterclockwise dual-sequence cross-excitation method is adopted in the laser scanning path design. That is, the laser scanning mirror system is controlled to repeatedly execute multi-angle low-power perturbations in opposite rotation directions, so that the crack region receives thermal excitation energy input from different directions. The system records the temperature response trajectory under the two sets of excitation sequences and performs difference analysis. If the same region shows significant thermal response shift or response in both directions, the system will detect the difference. The system identifies the asymmetry in the delay curve as an asymmetry in the thermal disturbance path. This asymmetry is typically caused by a crack penetrating the cross-section of the rail. This is because penetrating cracks impede the transverse and longitudinal thermal diffusion paths to varying degrees, and the energy distribution on both sides of the crack reverses when the rotation direction changes. Based on the above detection results, the system further performs response time mapping on the abnormal areas identified in the disturbance penetration test. Specifically, the time delay from the application of the laser disturbance to the manifestation of the temperature response is used as the mapping parameter. Areas with shorter delay times are marked as thermal disturbance transparent areas, while areas with longer delay times and a slow increase in response amplitude are marked as weak response deposition areas. The system establishes a topological mapping relationship between these two types of areas through time difference, forming a crack thermal response topology map based on delay behavior. This map not only intuitively reflects the spatial distribution of the crack area but also reveals the dynamic depth structure of its thermal response, providing a highly reliable physical basis and spatial evidence for subsequent real crack location, expansion trend judgment, and risk level classification.

[0041] To improve the accuracy and anti-interference capability of crack identification results, after identifying the actual crack area, the system triggers a secondary low-frequency thermal perturbation based on the preliminary response results for response verification. This perturbation uses low-frequency, low-power excitation parameters to ensure that the thermal response behavior of the target area is re-excited without causing thermal saturation or thermal diffusion overlap. This secondary perturbation has a frequency much lower than the standard excitation frequency in the initial identification process, making the thermal response curve smoother and more structurally dependent. If the actual crack area still exhibits stable delay, phase lag, or asymmetric boundary response characteristics in the thermal response path under the secondary low-frequency excitation, the system can further confirm that the thermal behavior originates from a structural thermal conduction barrier and has a physical basis for thermal path blocking. Conversely, if the response of the area suddenly returns to normal or exhibits irregular response perturbation under the low-frequency perturbation, the system identifies it as an occasional heating anomaly caused by impurities, oil, rust, or contamination layers on the rail surface and excludes it. This strategy significantly reduces the impact of short-term surface thermal... To mitigate the risk of misjudging cracks due to abnormal properties, a physical verification closed loop for crack thermal response results was established. During this verification process, an environmental steady-state comparison analysis module was further integrated. Environmental meteorological parameters, including ambient temperature, relative humidity, and ground wind speed, collected during and before / after the disturbance were simultaneously acquired and recorded in real-time in the crack response log system. This log file provides a unified timestamp for the crack detection task, ensuring a precise mapping between each excitation-response-analysis process and the corresponding environmental variables. After completing the secondary disturbance response data acquisition, the system performs a correlational backtracking analysis between the crack thermal response curve's changing trend and the corresponding environmental log data. By comparing the changes in temperature, humidity, and wind before and after the excitation with the slope, phase difference, and peak amplitude of the thermal response curve, if a significant environmental change is found and the crack response curve also shifts synchronously, the response result is marked as an environment-dominant anomaly. If the environment is stable but the abnormal response behavior persists, it is classified as a confirmed structural anomaly.

[0042] Combined with appendix Figure 3 When a structural crack is determined to exist in a detection area through multiple rounds of perturbation and environmental steady-state comparison, the system uses this determination as a trigger to activate a supplementary detection task in the vicinity of that area. This supplementary detection does not employ fixed geometric boundaries or manually set distance ranges. Instead, the system constructs a set of response extrapolation boundaries based on the nonlinear behavior characteristics of the current crack in the thermal response dimension, its own thermal diffusion behavior, and directional delay performance. The area formed by this boundary is called the risk diffusion radius. Based on this, the system establishes an active scanning closed-loop mechanism to achieve rapid extrapolation identification and local expansion monitoring of high-risk areas. Furthermore, to further quantify the evolution trend of the current crack area, the system constructs a nonlinear enhancement trend function for the crack's thermal response to continuously track its dynamic behavior. This function is defined as follows:

[0043] ;

[0044] in:

[0045] Indicates time The thermal response risk integral value of the crack area at any given time is used to determine whether it has entered a high-risk zone for crack propagation;

[0046] For the core region of the crack in time The thermal response temperature value;

[0047] It represents the first derivative, i.e. the rate of temperature change, which reflects the speed at which the crack responds to excitation;

[0048] It represents the second derivative, which describes the abrupt change in the trend of thermal response, i.e., the acceleration of temperature change;

[0049] This is the squared term of the nonlinear response amplitude relative to the initial excitation temperature, used to amplify the signal of thermal fluctuation intensity change;

[0050] It is an exponential decay factor. This is used to assign weight decay to historical responses in order to improve the sensitivity of the latest response data;

[0051] The crack region in spatial coordinates The response morphology function reflects its geometric orientation, structural boundary characteristics, and non-uniform thermal behavior;

[0052] These are empirical control coefficients, which control the contribution weights of the three response terms to the risk integral, respectively.

[0053] The integration interval is the start time for incentivizing response monitoring. .

[0054] The aforementioned function integrates first-order thermal response, second-order thermal trend, nonlinear response amplitude, historical decay behavior, and spatial location features to construct a continuously computable high-order integral model. Its initial purpose is to replace traditional image thresholding with the evolution rate of response behavior and static boundary extraction with dynamic trend prediction, thereby achieving a continuous quantitative description of crack occurrence and evolution. Each term in the function has a clear physical meaning: when a sharp increase in thermal response occurs in the crack region, the first and second derivative terms rise rapidly, and the overall value of the integral function also increases accordingly. If the temperature difference amplitude also increases simultaneously, the squared term further strengthens this result, ultimately leading to a higher value in the function. The curve exhibits nonlinear inflection point behavior; however, if the function value continuously approaches or exceeds a preset risk assessment threshold during system operation... ,Right now:

[0055] ;

[0056] The system immediately marks the area as a high-risk zone for crack propagation and enters enhanced dynamic monitoring mode, while simultaneously updating the corresponding spatial response function in the function. The mapped boundary range is incorporated into the risk diffusion radius of the next round, driving the active scanning closed loop to continue expanding the detection, thereby realizing a closed-loop linkage mechanism from the acquisition of original physical behavior signals to nonlinear trend function modeling, then to critical state identification, and finally to diffusion scanning activation. The derivation of this integral function is based on physical heat conduction behavior. Through multi-cycle excitation experiments, it was found that the temperature response in the crack region is not stable, but exhibits a significant acceleration in response speed (i.e., Increase), trend change (i.e.) (surge) and the temperature difference continues to widen (i.e. To capture such complex thermal dynamics, three response dimensions are introduced into the function simultaneously, and an exponential memory inhibition term is constructed to ensure the weight of new data. Spatial function coupling is used to form the ability to identify surface risks, and finally a crack risk quantification model that can be embedded in edge devices and has high-speed real-time judgment capabilities is generated.

[0057] To enhance the ability to identify structural crack propagation trends and improve risk warning response in adjacent areas, the risk diffusion radius activated based on environmental steady-state comparison and dynamic thermal response function calculation is not defined by a fixed geometric size. Instead, it is constructed based on the material thermal properties of the crack area to create an adaptive risk propagation boundary. The core mechanism is to dynamically adjust the shape and radius of the risk boundary based on the local thermal conductivity gradient change exhibited by the rail material under the condition of microstructural continuity failure. The system calculates the difference in thermal conductivity per unit length between different points using spatial thermal response data collected by laser thermal perturbation, which is equivalent to the spatial gradient distribution of local thermal conductivity. In areas with cracks or micro-damage, the thermal conductivity often exhibits a nonlinear, abrupt decrease in spatial distribution. Therefore, the system uses this gradient change region as an extension reference to construct a dynamic risk diffusion radius, which preferentially extends in the direction of drastic thermal conductivity change and contracts in the thermally stable direction, thus forming a thermal response extension layer that conforms to the abnormal thermal conduction path of the material. This layer provides the optimal energy excitation boundary for active scanning. Within this dynamic boundary constructed based on the thermal conductivity gradient, the active scanning mechanism begins to perform low-intensity perturbation excitation detection on the adjacent area. If the system identifies multiple clusters of points with obvious abnormal thermal response characteristics, i.e., multiple regions with response time lag, abrupt response changes, or abnormal peak amplitudes, to avoid misjudging unrelated local anomalies as the same crack system extension structure, the system associates the above-mentioned abnormal response regions with the original crack. The thermal behavior of the crack region is compared with multiple parameters to extract its core feature vector, including five types of dynamic features: heating rate after excitation, first-order thermal phase shift amplitude, thermal decay time, maximum response temperature difference, and response symmetry factor. A multi-dimensional distance metric algorithm (such as weighted cosine similarity) is used to calculate the response similarity between each abnormal point and the original crack region, forming a thermal behavior feature clustering space. If multiple abnormal points belong to the same cluster as the original crack in this space, the system determines that these points are crack chain extension structures caused by the same micro-damage mechanism or stress coupling chain, and expands the crack risk path model based on this to achieve holistic identification and topology reconstruction of complex crack network structures. Conversely, if the cluster distance between the abnormal point and the original crack in the response feature space increases significantly, the system marks it as an isolated thermal anomaly, excluding its structural correlation within the extension chain, thereby effectively avoiding non-correlated false alarms and improving the judgment accuracy and topology modeling capability of the crack detection system.

[0058] When the system actively scans the area adjacent to the crack to detect its expansion, if the detection results show a nonlinear thermal response mode that is the same as or highly similar to the original crack area, the system immediately identifies this adjacent area and the current crack area as a structural co-deterioration zone and simultaneously activates dynamic monitoring to achieve continuous tracking and unified evaluation of the multi-point coupled crack system. The so-called "same nonlinear response mode" refers to a highly consistent dynamic enhancement trend in the temperature response curves of different regions during the thermal disturbance cycle, especially in the acceleration dimension of the thermal response, where the second derivative of temperature with respect to time is consistently positive and shows an increasing trend. This reflects a sharp increase in the response of the crack area to external stimuli, indicating that the internal heat conduction path of the crack is further restricted or that the micro-cracks are rapidly expanding. Therefore, the system calculates the second derivative term of the thermal response function in real time. The system captures the curvature changes of the thermal response curve in each excitation cycle. If the derivative value is found to rise continuously in multiple regions simultaneously with a consistent growth rate trend, it can be determined that these regions not only have similar thermal behavior but also exhibit synchronicity in structural coupling degradation. This indicates that they are affected by a common stress field or thermal fatigue mechanism and belong to a synergistic crack system under the same potential propagation link. Based on this, the system labels the thermal response behavior as a nonlinear response enhancement trend. After confirming the synergistic degradation relationship, the system enters a synchronous dynamic monitoring state for the entire synergistic region and continuously fits the trend of its second derivative change to calculate the crack growth rate in the thermal response dimension. Furthermore, by combining the change in the response enhancement slope and time interval from the previous cycle to the current cycle, a prediction model is constructed to estimate the time window before the future crack enters critical instability. This time window reflects the warning period boundary of the crack propagation to irreversible structural failure. The system outputs this result to the maintenance platform for early intervention and risk classification linkage.

[0059] Once the crack region enters dynamic monitoring mode after initial assessment, the system activates a dynamic thermal excitation power adjustment mechanism to achieve continuous and highly sensitive tracking of the crack evolution process. This mechanism adjusts the excitation energy output in real time based on the thermal response intensity exhibited by the crack in each excitation cycle, ensuring a linear match between the excitation power and the crack response amplitude. This prevents excessive excitation from causing thermal saturation or insufficient excitation from causing response masking. This matching relationship is determined by the system's response to the maximum temperature rise in the crack region during the previous cycle. The peak temperature rise in the cracked area during a single cycle, ranging from 0.5 to 12.0℃, is recorded. Combined with the response amplitude variation trend curve, a preset linear response function is applied. Calculate the thermal excitation power required for the next excitation cycle, where The current cycle excitation power, The response gain coefficient (ranging from 0.5 to 5.0 W / ℃, used to control the amplification ratio of the excitation energy with the response; generally, a value of 1 to 2 can be used to avoid excessive disturbance). A base energy bias term (range 3–8W) ensures that the energy excitation is both sufficient and not excessive. Under this dynamic monitoring state, the system continuously evaluates the thermal response stability of the crack region. Specifically, it acquires the temperature change curves over multiple consecutive periods and extracts the maximum, average, and slope change amplitude of the response for each period, calculating the fluctuation index of the response during the period. (Value range: 0.1–3.0℃) If this fluctuation range is continuously lower than the system's set threshold... (Value range 0.3–0.5℃) indicates that the crack thermal response has entered a steady state or tends to passivate. At this time, the system judges that the current structure is in a non-proliferation period or the response evolution rate has decreased significantly. Therefore, it triggers the monitoring mode degradation mechanism, switching the dynamic monitoring state to the periodic scanning mode, that is, using a longer time interval to perform low-frequency disturbance scanning to save energy consumption and computing resources. Conversely, if the response fluctuation is found to be increasing or the fluctuation amplitude exceeds the threshold during dynamic monitoring, the system continues to maintain the dynamic high-frequency tracking state and can choose to increase the sampling density to improve the response resolution, thereby realizing the graded response and adaptive control of resource scheduling for crack risk state.

[0060] To improve the accuracy of identifying the actual evolution of cracks and suppress false alarms under dynamic monitoring conditions, the system simultaneously activates a multi-source environmental disturbance acquisition module upon entering dynamic monitoring mode. This module records key external disturbance indicators in real time, such as wind speed, structural vibration intensity, and lightning electric field changes, within the crack detection area. Wind speed information can be used to correct the surface convective heat transfer rate, vibration parameters reflect the frequency and amplitude of local mechanical disturbances in the track body, and the lightning electric field is used to identify thermal noise or induced heating anomalies caused by sudden strong electromagnetic disturbances. These environmental disturbance indicators are continuously injected into the thermal response analysis module and used as dynamic weight adjustment factors in the threshold update process of the nonlinear trend function. This ensures that the system maintains effective accuracy in identifying the thermal response behavior of cracks even when facing different external disturbance intensities. Simultaneously, to further evaluate the response evolution characteristics of cracks during dynamic thermal excitation cycles, the system performs differential fitting on the thermal response time series collected from multiple consecutive excitation cycles. The system takes the response temperature curve of each cycle as input, performs first- and second-order difference calculations on the change in temperature rise amplitude and the trend of response curve slope between adjacent cycles, and extracts the response fluctuation pattern through regression fitting. If the fitted curve shows a continuous increase in response delay and the peak response no longer rises synchronously with the excitation, the system identifies it as a crack entering a thermal hysteresis state, reflecting local blockage of the heat conduction path in the crack area or non-uniform phase transformation behavior of the material. If the response curve is found to exhibit reverse fluctuation, rapid oscillation, or sudden instability under constant excitation power, the system determines that the region has entered a critical state of thermal destability, indicating that the crack structure has a violent discontinuous evolution trend and a high risk of unstable expansion. Based on this, the system uses this state as a high-level risk identification marker, triggering alarm linkage, data reporting, and extended scanning strategies to ensure sufficient response capability and protection redundancy for the violent thermal dynamic behavior at the end of crack evolution.

[0061] Example 1:

[0062] Combined with appendix Figure 4In this embodiment, the K327+300 meter section of a mountainous trunk railway is used as a detection example. This section is affected by the diurnal temperature range throughout the year, and the material used for the track is typical high-manganese steel, which has a certain degree of temperature change sensitivity. The maintenance unit selected this section for early crack identification. The detection process first involves the detection vehicle carrying a high-precision infrared thermal imaging system to perform the first material area pre-calibration at 4:00 AM. Using low-power short-pulse thermal excitation, a linear temperature-controlled scan is performed on the surface of the rail section at 2-meter intervals. The excitation power is set to 4W, and the thermal imaging sampling frequency is 10Hz. The track section is identified as being in the interval 327+314. Between 6 and 327+315.3 meters, there exists a region approximately 0.7 meters long with abnormally low lateral heat capacity. Its maximum temperature rise is only 1.4℃, significantly lower than the average level of 2.1–2.5℃ in the surrounding area. The system initially identified this region as a potential crack-sensitive candidate point. Subsequently, during the environmental temperature difference control phase, maintenance personnel used a heating-ventilation integrated unit to alternately heat and slowly cool this region. The heating cycle was set to increase by 1.5℃ per minute, and after reaching 45℃, the cooling rate was set to 1.2℃ / minute. After three consecutive excitation cycles, the thermal reversal inertial response was recorded, revealing that this candidate region was in a critical cooling phase. The segment still exhibited a delayed temperature rise tail of 0.9℃, significantly higher than the background fluctuation zone of 0.2–0.3℃. Based on this, the system confirmed the presence of a potential abnormal heat conduction path in this region and proceeded to the next step of laser perturbation response testing. During laser perturbation, the detection system first applied a low-frequency pulse excitation of 6Hz in the vertical direction. The thermal response time delay in the crack boundary region was 0.37 seconds. Subsequently, the system increased the excitation frequency to 12Hz and re-excited the region, finding that the delay intensified to 0.61 seconds. This indicates that the crack boundary exhibits a nonlinear response amplification trend to the excitation frequency, consistent with the heat conduction hysteresis characteristics of a closed crack. Next, to further... To determine the crack morphology, the system switched the laser path to the tangential direction excitation mode and alternately acquired response thermograms under the two excitation directions. Through difference analysis, it was found that the heat conduction velocity of the crack decreased by 35% under the tangential excitation direction, while it only decreased by 12% in the vertical direction. This indicates that the thermal conductivity barrier of the crack is mainly distributed along the track direction. The system determined that it is an open longitudinal crack with a bifurcation trend. Finally, the detection system defined the crack area as a structural thermal conductivity anomaly point and marked it to enter a dynamic monitoring state. The system re-examined it with low-power excitation every 30 minutes until its response parameters stabilized or destabilized.

[0063] In the crack identification process within the K327+314.6 to 327+315.3 meter segment, after completing laser excitation directionality testing and preliminary crack contour confirmation, the system further conducts phase difference vectorization reconstruction and delay topological distribution analysis to accurately reconstruct the crack structure evolution behavior. In this stage, the system uses infrared imaging equipment to continuously record the temperature response curve induced by laser excitation. Twenty-one thermal sampling points are selected at the crack edge and its adjacent area. The peak time of the thermal wave is extracted from the temperature-time curve collected at each point, serving as a reference point. (Located 10cm in front of the crack center) is the phase reference point; the time of its peak appearance is recorded as follows. And the crack boundary point and The peaks are respectively delayed by , The relative delay is calculated as , The system vectorizes the relative time difference to establish a thermal response matrix containing a 21-dimensional time delay vector. Combined with the spatial coordinates, a high-dimensional mapping is performed to reconstruct the temporal thermal phase delay map of the crack region, providing a basis for reconstructing its geometric contour and internal perturbation response structure. Next, to determine whether the crack penetrates the track cross-section, the system employs a dual-sequence perturbation excitation strategy. Under the same excitation power, multi-angle laser scanning excitation is performed along the track surface in clockwise and counterclockwise directions, recording the rise delay and maximum response amplitude of the corresponding thermal response channels. In the forward excitation sequence, the maximum thermal response delay at the left edge is 0.53s, and at the right edge, it is 0.31s. In the reverse sequence, the delays on both sides are reversed (the right side becomes 0.54s, and the left side becomes 0.33s). This asymmetric path response phenomenon indicates directional thermal conduction hindrance within the crack and demonstrates structural inertial asymmetry in response to the scanning direction. Based on this, the system determines that the crack has a tendency to penetrate deep into the track and is no longer a surface micro-crack. To accurately depict the crack distribution morphology, the system further extracted the spatial distribution boundary between the transparent response region (i.e., thermal response delay < 0.25s) and the weak response deposition region (delay > 0.6s) based on the perturbation penetration test data. The average response time difference between the transparent region and the deposition region was calculated to be 0.38s. Combining this with the two-dimensional coordinate mapping relationship, a crack distribution map based on the delay topology was constructed. The image shows a main crack with a double-branched structure, with a total length of approximately 73cm. This region was marked as the true crack area. To eliminate thermal anomaly interference caused by non-structural factors such as surface oil or metal oxide layers, the system applied a secondary low-frequency thermal perturbation excitation (excitation period 5s, power reduced to 3W) to this true crack area and re-detected the stability of the thermal conduction response path. If the peak thermal fluctuation change was less than 5% and the path remained unchanged, it was considered a true thermally conductive structural response, confirming it as a structural crack signal. Simultaneously, within the entire monitoring window, the system synchronously collected temperature, humidity, and wind speed data around the track, which were 28.4℃, 64%, and 3.2 m / s, respectively, and recorded them in the crack response log as environmental steady-state samples. Subsequently, the trend of the maximum response temperature rise during the three rounds of scanning before and after excitation was compared. The temperature rose from 2.91℃ and 2.89℃ to 2.87℃, with fluctuations less than 0.05℃, indicating that the system was in a stable thermal response period under constant environmental conditions, thus enhancing support for high-confidence crack identification. Finally, combining the crack area delay topology distribution map, thermal response trend, asymmetric disturbance verification results, and environmental steady-state data backtesting analysis, the system determined the crack to be a structural bifurcation crack with cross-sectional tendency and stable heat conduction path, and output a risk assessment level of Level 3 (medium-high), recommending that track segment intensified scanning and crack extension tracking maintenance be performed within the next 15 days.

[0064] Example 2:

[0065] Combined with appendix Figure 5Based on Example 1, in the section from K327+314.6 to 327+315.3 meters where structural cracks have been identified, the system confirms the existence of cracks in this area through environmental steady-state comparison and triggers a supplementary detection task in the adjacent area. The supplementary detection range no longer uses a preset radius, but instead collects thermal response behavior and thermal diffusion velocity data at every 0.1 meter point within 0.5 meters of the crack point and its surrounding area. The local thermal conductivity is then inverted based on the thermal response amplitude and diffusion time data. And calculate its gradient. If the gradient exceeds a threshold of 0.8 W / (m·K²) in a certain direction, the risk diffusion radius extends along that direction to the gradient fallback position, thereby dynamically determining the scan boundary profile. Taking the measured data as an example, the maximum gradient at the crack center reaches 1.2 W / (m·K²), and the gradient recedes to 0.5 at ±0.4 m along the track direction, thus determining the risk diffusion radius to be 0.4 m. Subsequently, the system dynamically establishes the following high-order thermal response nonlinear strengthening trend function within this range and calculates it in real time:

[0066] ;

[0067] Using parameter settings , , , Spatial response function A value of 1 is applied at the crack core, and 0.6 is applied at the edge. Monitoring is initiated at... Data is collected continuously for 10 seconds. The data at a specific moment is set as follows: exist s is 5.2°C. °C / s, °C / s², °C². Substituting this into the formula, the instantaneous integral contribution at that moment is approximately:

[0068] ;

[0069] Points accumulated to s time overall If the preset threshold is If the threshold is exceeded, the system determines that the area has entered a high-risk zone for crack propagation and initiates dynamic monitoring. Simultaneously, The mapping range is incorporated into the next round of diffusion boundary closure.

[0070] The system uses the integral risk function from the previous step. The region has been determined to be in a high-risk state for crack propagation, and this has been determined based on the spatial response morphology function mapped by the function. The risk diffusion radius was adaptively defined, and the system then performed an active scan within a proximity of ±0.5 meters based on this radius, and measured the local thermal conductivity of the rail in this area in real time. The high-frequency transient thermal response method was employed, heating each 5cm block with laser pulses and recording the temperature response rate, using the formula:

[0071] ;

[0072] in For heat flux density, For density, Given the specific heat capacity, the thermal conductivity is calculated and its gradient distribution is determined. At K327+314.9 m, the local thermal conductivity is measured to decrease from a central value of 43 W / m·K to 34 W / m·K within a 50 cm range, with a gradient value reaching [value missing]. The thermal response (W / (m²·K)) was the highest in the surrounding area, and the system included this area in the active scanning path accordingly, continuously collecting thermal disturbance data. As a result, three local response hotspots were found in this area, located at K327+314.85, 314.93, and 315.00 meters, respectively. The system extracted the thermal response feature vectors of these three areas. and the original core area of ​​the crack (denoted as) Calculate the Euclidean distance between the response vectors of (e.g.) Features are And point K327+314.93 is The distance between the two regions was approximately 0.32, lower than the set similarity clustering threshold of 0.5. The system clustered these regions into the same "crack propagation chain," labeling it as chain number CH-14. Based on this clustering result, the system further analyzed whether there was a nonlinear response enhancement trend in this region, specifically whether the change in the second derivative of the thermal response was continuously increasing. During the excitation cycle at 314.93m, the recorded second derivatives of the thermal response were 0.26, 0.31, 0.35, and 0.39°C / s², showing a clear stepwise upward trend. Based on this, the system determined that the region was in a nonlinear enhancement state, highly matching the original crack response curve, confirming it as a structural decay propagation node within the same evolutionary chain. Therefore, the system marked the original crack region and its cluster neighbors as a structural co-decline region and initiated a synchronous dynamic monitoring mode. To further predict the crack propagation time window, the system used a second derivative slope fitting method to regress the enhancement trend: fitting the above derivative values ​​to a straight line slope. Set the critical response threshold as ,according to This indicates that the system anticipates entering a high-heat instability risk zone approximately 7 seconds later. Based on this, the system triggers an early warning for track inspection and slows down the track, and locks onto the high-frequency monitoring window.

[0073] After the section from K327+314.6 to K327+315.3 meters was officially marked as the structural co-deterioration zone by the system, the crack identification system entered a dynamic monitoring state and, for the first time, activated a dynamic thermal excitation power adjustment mechanism to achieve linear adaptive matching between the excitation energy and the crack response intensity. During this process, the system monitored the maximum temperature rise in the crack region during the previous cycle. To dynamically adjust the laser power for the next cycle According to the matching function: Where the adjustment coefficient This is based on empirical settings. For example, when the system detects that the maximum temperature rise in the thermal response of the crack region during the previous cycle is... The excitation power is then adjusted as follows: If the temperature rise in the next cycle is Then the power is updated to This ensures that the excitation energy and the actual crack response intensity remain linearly synchronized. Under dynamic monitoring conditions, the system uses a 5-second detection cycle, continuously monitoring for three thermal response cycles. Afterward, the fluctuation range of the crack response is statistically analyzed, and a fluctuation index is defined.

[0074] ;

[0075] Taking the maximum temperature rise of the three cycles as an example, which is 2.9°C, 2.8°C, and 2.86°C respectively, then... If the value falls below the system's set threshold by 5%, the system determines that the response has stabilized, and the dynamic monitoring state is downgraded to a periodic scanning mode after the fourth cycle, switching to periodic detection every 30 minutes. To enhance control over environments prone to misjudgment, the system simultaneously records three environmental disturbance indicators—wind speed, vibration, and lightning electric field—during dynamic monitoring. The current wind speed is 3.8 m / s, the vertical vibration intensity of the track is 0.12 g, and the lightning electric field value is 15 V / m. The system maps these parameters to the thermal response trend judgment function through a disturbance sensitivity coefficient. If the wind speed exceeds 5 m / s, the vibration exceeds 0.2 g, or the electric field intensity exceeds 30 V / m, the system will adjust the threshold upwards. A threshold is set to prevent false triggering of thermal disturbances; however, since the current values ​​are all below the disturbance impact threshold, the threshold remains unchanged. Meanwhile, continuous thermal response data of the crack region is processed through differential fitting, and the system collects the temperature rise sequence of the crack center point within 10 seconds. C, the system calculates its first-order difference. Furthermore, the system fitted the data to show that the peak response time lags behind the excitation center by more than 3 seconds, indicating a thermal hysteresis state. If the differential results show alternating positive and negative fluctuations and the fitted curve trend is reversed, it is determined that the system has entered a critical state of thermal destability. The current response data shows a response plateau and negative fluctuations after 6 seconds. Based on the steady state of the environment and the absence of obvious external disturbances, the system confirmed that the thermal behavior was caused by the non-uniform evolution of the internal heat conduction path of the material structure. This phenomenon was recorded as a "thermal hysteresis + thermal destability" composite event, rated as a high-risk level Level-4, and reported to the dispatch center. At the same time, the area and 50cm on both sides were designated as high-priority monitoring areas, and subsequent track inspection vehicles were forced to enter silent mode and activate the infrared and laser joint scanning strategy.

[0076] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting surface cracks in steel rails, characterized in that, include: By controlling the temperature difference changes in the rail environment, a hysteresis phenomenon is generated in the thermal response of the rail metal in the stress concentration area, inducing slight asynchronous thermal fluctuations with discontinuous and asymmetric thermal diffusion in local areas of the material surface, thereby exposing potential crack boundaries. The asynchronous thermal fluctuations refer to the microscopic perturbation changes in thermal conductivity or heat capacity near the crack, causing the temperature rise rate of the crack area after heating to be slower or faster than that of the surrounding uniform area. This results in a phase delay or advance in the temperature curve of the crack area on the time axis, which appears as an irregular hot spot outline boundary in the thermal imaging equipment. Based on the asynchronous thermal fluctuation area, a directional short-period external perturbation signal, including laser thermal pulse interference, is applied to cause thermal phase perturbation at the crack edge due to discontinuous heat conduction, forming a phase difference with misaligned response time. For the region generating the phase difference, a non-contact, low-intensity laser thermal perturbation multi-angle scanning is used to perform a perturbation penetration test to identify the difference between the laser thermal perturbation transparent region and the weak response deposition region, eliminate non-structural thermal anomalies, and determine the true crack region. If a local area exhibits extremely low thermal response or significantly delayed response time under multi-angle laser excitation, the system marks it as a weak response deposition region. Such regions are often caused by structural cracks, thermal cavities, or micro-damage. If a region always recovers rapidly or responds linearly under excitation, it is marked as a laser perturbation transparent region. This region is usually caused by surface contaminants, uneven thermal conductivity, or measurement errors. The thermal response of the actual crack region after multiple rounds of disturbance was recorded by attenuation curves, and the integrity of the microstructure below the crack was deduced based on the disturbance settlement time to determine the fatigue level of the structure. The disturbance settlement time is the time required for the crack region to decay from the peak of the excitation response to a steady state. After completing the fatigue level determination, a steady-state environmental comparison is performed on the same test area. By comparing the changes in temperature, humidity, wind speed and radiation environmental parameters before and after the disturbance, if the crack area continues to show abnormal thermal response under constant environmental conditions, it is confirmed to be caused by structural cracks; otherwise, misjudgment caused by occasional external interference is excluded.

2. The method for detecting surface cracks in rails according to claim 1, characterized in that, Inducing the asynchronous thermal fluctuations includes pre-calibrating the material regions on the rail surface before thermal excitation to identify low heat capacity regions as crack-sensitive candidate points; controlling the ambient temperature difference includes setting alternating heating and slow cooling processes to construct an excitation sequence for thermal reversal inertia.

3. The method for detecting surface cracks in rails according to claim 1, characterized in that, During the application of the perturbation signal, the laser pulse frequency is dynamically adjusted according to the amplitude of the thermal response delay at the crack boundary, so as to amplify the phase response difference between crack types, including closed, open, or bifurcated cracks. The laser thermal perturbation adopts an alternating method of vertical excitation and tangential excitation to detect the directional thermal conduction barrier at the crack edge and to determine the crack direction.

4. The method for detecting surface cracks in rails according to claim 1, characterized in that, The phase difference is extracted by vectorization reconstruction of the relative time delay of the temperature peak; the perturbation multi-angle scanning adopts clockwise and counterclockwise dual-sequence cross-excitation to detect the asymmetry of the thermal perturbation path in the crack region and to determine whether the crack penetrates the cross section; the perturbation penetration test results are mapped by the time difference between the transparent area and the weak response area to form a real crack region distribution map based on the delay topology.

5. The method for detecting surface cracks in rails according to claim 1, characterized in that, The confirmation result of the actual crack area is used to trigger a secondary low-frequency thermal disturbance to re-verify the thermal path stability and eliminate misjudgment response caused by surface impurities or oil stains; the temperature, humidity and wind data collected during the environmental steady-state comparison process are synchronously recorded in the crack response log, and anomaly backtracking analysis is performed by comparing the change trend of the excitation response before and after.

6. The method for detecting surface cracks in rails according to claim 5, characterized in that, If the environmental steady-state comparison result determines that it is a structural crack, a supplementary detection task in the adjacent area is triggered to form an active scanning mechanism based on the risk diffusion radius; if the thermal response of the crack area shows a trend of nonlinear response enhancement in more than two excitation cycles, the system marks it as a high-risk area for crack propagation and enters a dynamic monitoring state.

7. The method for detecting surface cracks in rails according to claim 6, characterized in that, The risk diffusion radius is adjusted based on the local thermal conductivity gradient change of the rail material to identify potential abnormal heat conduction paths. When the active scanning mechanism identifies multiple abnormal thermal response areas in the vicinity, it clusters each response area according to the similarity of its thermal behavior to the original crack to determine whether they belong to the same extended chain.

8. The method for detecting surface cracks in rails according to claim 7, characterized in that, If the same nonlinear response pattern appears in the detection of the neighboring region, the current region and the neighboring region are marked as structural collaborative decay regions and enter the synchronous dynamic monitoring state. The enhanced trend of the nonlinear response is captured by the change in the second derivative of the thermal response to identify the dynamic growth rate of the crack and predict the expansion time window.

9. The method for detecting surface cracks in rails according to claim 8, characterized in that, After entering the dynamic monitoring state, a dynamic thermal excitation power adjustment mechanism is adopted to make the excitation energy linearly matched with the crack response intensity. The duration of the dynamic monitoring state is determined by the stability change of the crack response over multiple consecutive cycles. If the fluctuation value is less than the set threshold, it is downgraded to the periodic scanning mode.

10. The method for detecting surface cracks in rails according to claim 9, characterized in that, Under the dynamic monitoring state, the system synchronously records environmental disturbance indicators including wind speed, vibration, and lightning electric field, which are used to dynamically adjust the nonlinear trend judgment threshold. The continuous thermal response data collected under the dynamic monitoring state is processed by differential fitting to determine whether the crack has entered the critical state of thermal hysteresis and thermal destability, serving as the basis for high-level risk identification.

Images