Fiber bragg grating pretension maintaining and temperature decoupling method based on invar steel composite fixation

Abstract

The invention relates to the technical field of data processing and analysis, in particular to a fiber bragg grating pretension maintaining and temperature decoupling method based on invar steel composite fixing, which comprises the following steps of: 1, assembling a fiber bragg grating by adopting an invar steel composite fixing component to form a pretension maintaining body, the mirror image template comprises virtual mark peaks which are arranged according to a fixed sequence and a fixed number, and the virtual mark peaks are used as anchor points for subsequent registration and cancellation; step 2, on the basis of the mirror image template and input reflection data, executing a compensation virtual steady state fusion algorithm, and writing a corresponding compensation action increment into a temperature decoupling mapping table; and 3, compensating the input reflection data in real time according to the temperature decoupling mapping table, and returning to execute the step 2 to re-estimate and update the temperature decoupling mapping table until the in-service maintenance requirement is met. According to the invention, long-term stable work of the fiber bragg grating can be maintained in a complex environment, and the accuracy of a pretension index and the controllability of a temperature drift index are ensured.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of data processing and analysis, and particularly relates to a fiber grating pre-tension maintaining and temperature decoupling method based on invar composite fixation. BACKGROUND

[0002] As a key device widely used in optical fiber communication, optical fiber sensing and structural health monitoring, fiber grating can reflect specific wavelength light signals through the Bragg reflection effect and produce sensitive response to external strain and temperature changes. Due to the advantages of small size, high sensitivity, anti-electromagnetic interference and long-distance transmission, fiber grating has been rapidly developed in the fields of civil engineering monitoring, aerospace component measurement, power equipment operation monitoring and state perception in precision manufacturing. However, one of the core challenges faced by fiber grating in practical application is how to ensure that the grating can maintain a stable pre-tension state under long-term service conditions, while effectively decoupling temperature effect and strain effect.

[0003] In the prior art, the fixing methods of fiber grating mainly include mechanical clamping, adhesive curing and metal packaging. Mechanical clamping usually relies on components such as springs, micro-teeth or pressure plates to maintain the fiber under a certain tension, which has the characteristics of simple operation and strong repeatability. However, mechanical clamping has the problem of long-term tension decay, especially under the action of temperature cycle change and material aging, micro-slippage will occur at the clamping position, which will cause the shift of fiber strain reference, and thus the reliability of grating output signal will decrease. The adhesive curing method fixes the fiber on the substrate through adhesives such as epoxy resin and ultraviolet curing glue, and the initial tension can be applied and maintained before curing. However, with the passage of time, the adhesive interface will gradually weaken the mechanical transmission due to creep, delamination and micro-crack expansion, etc., and thus it cannot guarantee long-term stability. The metal packaging method usually embeds the fiber grating in a metal tube or a metal block, and then fixes it through welding or pressure bonding. The advantage of this method is strong environmental resistance, but the disadvantage is that there is a difference in the coefficient of thermal expansion between the metal and the fiber, which will easily introduce additional stress when the environmental temperature fluctuates, thus increasing the complexity of temperature strain coupling. SUMMARY

[0004] The main purpose of the present invention is to provide a fiber Bragg grating pre-tension maintenance and temperature decoupling method based on Invar composite fixation, a stable pre-tension maintenance body is formed through the synergistic effect of the Invar base layer, the shape reset layer and the micro-tooth clamping seat, and a dual-channel alignment and virtual steady-state reference mechanism is established in combination with the mirror template and the virtual marker peak, thereby realizing effective distinction and cancellation of temperature effect and strain effect; on this basis, through the dynamic update of the temperature decoupling mapping table and the sequential driving of the compensation action on the shape reset layer, the present invention can maintain the long-term stable operation of the fiber Bragg grating in a complex environment, ensure the accuracy of the pre-tension index and the controllability of the temperature drift index, thereby significantly improving the reliability, precision and service life of the fiber Bragg grating sensing system, and has an application value superior to the existing technology.

[0005] In order to solve the above problems, the technical solution of the present invention is achieved as follows: A method for maintaining pre-tension and temperature decoupling of fiber Bragg gratings based on Invar composite fixation, the method comprising: Step 1: Assemble the fiber Bragg grating (FBG) using an Invar composite fixture assembly to form a pre-tensioned retainer. The Invar composite fixture assembly consists of an Invar substrate layer, a shape-reset layer, and a micro-toothed clamping seat. A mirror template is created based on a primary reference reflection spectrum and reference entries are set in the temperature decoupling mapping table. The mirror template contains virtual marker peaks set in a fixed order and number, which serve as anchor points for subsequent registration and cancellation. Step 2: Based on the mirror template and input reflection data, the virtual steady-state fusion algorithm is executed to establish the real channel and the mirror channel. The virtual steady-state reference is used to complete the cancellation and perform robust processing, and the pretension index and temperature drift index are output. When the index passes the verification, the corresponding compensation action increment is written into the temperature decoupling mapping table. Step 3: Perform real-time compensation on the input reflection data based on the temperature decoupling mapping table, and apply the compensation result to the shape reset layer to drive the Invar composite fixing component to perform micro-redistribution, thereby maintaining the stable state of the pre-tension retainer. When the pre-tension index or temperature drift index triggers an out-of-bounds, return to step 2 to re-estimate and update the temperature decoupling mapping table until the in-service retention requirements are met.

[0006] Furthermore, in step 1, the reference reflection spectrum is used as the only input, the peak group is identified according to the spectral position sequence and the relative position of each peak in the spectral position sequence is recorded; the relative position sequence is completely reversed to form a mirror sequence; the virtual marker peak is embedded in the mirror sequence; the setting rule of the virtual marker peak is: a boundary virtual marker peak is set at each end of the mirror sequence, and an intermediate virtual marker peak is set at the midpoint of the spectral position of the adjacent real peak in the mirror sequence, so that the virtual marker peak and the real peak form an alternatingly distributed anchor point sequence. After the embedding is completed, a mirror template is generated.

[0007] Further, in step 1, the process of creating the reference entry of the temperature decoupling mapping table includes: writing a reference to the mirror template into the reference entry, writing a fixed number and a fixed order of virtual marker peaks into the reference entry, writing the anchor point sequence into the reference entry; in the reference entry, setting the pre-tension indicator field to an uncalculated state, setting the temperature drift indicator field to an uncalculated state, setting the compensation action field to 0, and setting the entry state field to available; after the writing of the reference entry is completed, the reference entry is set to a read-only state.

[0008] Further, step 2 specifically includes: Step 2.1: Establish a real channel and a mirror channel, the real channel carries input reflection data, and the mirror channel loads a mirror template and retains an anchor point sequence; Step 2.2: Identify a plurality of peaks in the real channel, divide the spectrum into a plurality of segments according to the valley values between adjacent peaks to obtain a peak group sequence; perform segment-by-segment registration between the virtual marker peaks in the mirror channel, select a registration path using the joint criterion of the least sequence violation, the smallest shape difference within a segment, and the uniqueness of the peak top correspondence, and obtain a one-to-one correspondence between the peak groups in the real channel and the anchor points in the mirror channel; Step 2.3: At each pair of registered peak tops, generate a virtual steady-state reference point at the midpoint of the peak top position in the real channel and the corresponding anchor point position in the mirror channel, and connect adjacent virtual steady-state reference points into a virtual steady-state reference sequence in a linear interpolation manner within a segment; perform point-by-point cancellation between the real channel and the virtual steady-state reference sequence to obtain a cancellation residual curve; perform two-level robust processing based on the cancellation residual curve, including: performing intra-segment smoothing to suppress isolated spikes, and performing peak top consistency checking to remove abnormal points that do not meet the peak width continuity and peak shape symmetry, to obtain a stable residual curve; Step 2.4: In the stable residual curve, read the offset direction and offset distance of each peak according to the peak top correspondence obtained by registration, sum the offset information of all peaks within a segment and normalize it according to the number of peaks to obtain a pre-tension indicator, and read the average offset distance of the overall spectrum phase relative to the anchor points of the mirror channel in the virtual steady-state reference sequence to obtain a temperature drift indicator; Step 2.5: Check the obtained pre-tension indicator and temperature drift indicator, and the pass condition of the check is that the indicator value is within a preset range and the registration process is stable; when the check passes, write the compensation action increment corresponding to the temperature drift indicator as a new entry into the temperature decoupling mapping table in real time; when the check fails, automatically rollback to step 2.2 to re-register and repeat steps 2.2 to 2.5 until the check passes.

[0009] Further, step 2.2 specifically comprises: traversing the real channel from left to right point by point, identifying a plurality of peaks in the mode of continuous rise, peak top, and continuous fall, and recording each peak top and its left and right nearest valley values; dividing the spectrum line into a plurality of continuous segments with adjacent valley values to form a peak group sequence arranged in order of spectrum position, and recording the start valley value, end valley value, and the number of peaks in each segment; for each segment, selecting the interval between the two adjacent virtual marker peaks in the mirror channel as the candidate registration interval, which contains one or more anchor points; calculating the joint criterion for each candidate registration pair, which consists of the number of violations of the order, the shape difference metric in the segment, and the peak top corresponding to the unique marker; first, eliminate the registration pairs with more than 0 violations of the order; then sort them from small to large according to the shape difference metric in the segment; if multiple candidate registration pairs have the same peak top corresponding to the unique marker, these registration pairs are excluded; use forward cumulative cost search to simultaneously optimize in the full segment range, retain the connection pointer of the current best registration pair in each segment and the last best registration pair, and get the registration path from the first segment to the last segment after backtracking, and output the one-to-one correspondence between the peak group of the real channel and the anchor points of the mirror channel.

[0010] Further, in step 2.3, for each pair of registered peak top and anchor point obtained in step 2.2, a virtual steady-state reference point is generated at the midpoint of the spectrum position of the two; at the start valley value and the end valley value of each segment, one boundary virtual steady-state reference point is generated according to the trend of the connecting line of the two adjacent registered peak top and anchor point, which is used to constrain the transition at the end of the segment; when the number of the nearest registered pairs to the ends of the mirror template is less than 2, a supplementary virtual steady-state reference point is generated along the direction of the anchor point sequence to prevent the breakage caused by sparse boundaries; the spectrum position distance between any adjacent virtual steady-state reference points is checked, and when the distance is greater than the preset upper limit, a transition virtual steady-state reference point is inserted between them until the maximum distance does not exceed the limit; when a continuous missing interval is detected, virtual steady-state reference points are evenly arranged in the missing interval according to the order of the original anchor point sequence, and are monotonically connected with the adjacent generated virtual steady-state reference points on both sides.

[0011] Further, in step 2.3, the virtual steady-state reference sequence is formed into an equal-length reference sequence by linear interpolation according to the sampling positions of the real channel; the cancellation residual curve is obtained by subtracting the real channel from the equal-length reference sequence point by point; within the guard band of each virtual marker peak, the sign consistency check is performed first, and when the cancellation results of the left and right adjacent guard bands are consistent, the cancellation is performed, otherwise the median value of the neighborhood of the guard band is used to replace the cancellation result in the interval to suppress the cross-segment crosstalk.

[0012] Further, in step 2.3, the process of intra-segment smoothing to suppress isolated spikes includes: setting the smoothing window to contain 5 sample points, step size of 1 sample point; when the number of peaks in a segment is less than 3, the window is reduced to 3 sample points, when the number of peaks in a segment is greater than 7, the window is increased to 7 sample points; for isolated spikes lasting only 1 sample point and higher than the average of both sides, replace them with the average of the neighborhood; for spikes lasting 2 sample points, replace them with linear interpolation of adjacent endpoints; intra-segment smoothing also includes protection smoothing, i.e. setting a protection zone of 1 sample point at the beginning and end of each segment, and using half-window smoothing within the protection zone to avoid excessive blurring at the boundaries.

[0013] Further, in step 2.3, the peak consistency check of the cancellation residual curve after intra-segment smoothing includes: calculating the peak width of each peak as the sum of the distance from the peak top to the nearest valley on the left and the distance from the peak top to the nearest valley on the right; when the peak width difference between adjacent peaks is greater than 2 sample points, use the outward expansion priority rule to symmetrically expand the one with narrower peak width until the peak width difference does not exceed 2 sample points; calculate the absolute value of the left and right distance difference as the peak shape symmetry deviation, when the deviation is greater than 3 sample points, first stretch both sides equally; if the symmetry deviation after stretching still does not meet the requirements, then perform peak crown back paving, paving with equal length on both sides of the peak top until the shapes on both sides tend to be symmetrical; if the peak top falls within the protection zone of the virtual marker peak, clamp the peak top position to the boundary of the protection zone and make the adjacent two sides transition according to linear interpolation.

[0014] The method for maintaining pre-tension and decoupling temperature of the FBG fixed by the invar composite has the following beneficial effects: the stability and reliability of the FBG under long-term service conditions can be significantly improved. Unlike the existing methods that rely on single mechanical clamping or bonding fixation, the invar base layer and the controllable deformation characteristics of the shape resetting layer are combined in the present application, so that the FBG can maintain a stable pre-tension state when subjected to periodic changes in environmental temperature and external stress interference, avoiding the reference offset problem caused by slippage, curing failure or thermal expansion difference in traditional methods. At the same time, by establishing a mirror image template and embedding a virtual marker peak, the present application forms a dual-channel registration mechanism of real channel and mirror channel in the reflection data processing link, which can provide anchor point reference in spectral analysis and effectively distinguish temperature factors and strain factors. The introduction of the virtual steady-state reference sequence enables the residual error curve to accurately reflect the influence of real strain on pre-tension after robust processing, thereby obtaining pre-tension indicators and temperature drift indicators independent of temperature effects, avoiding the uncertainty caused by the dependence of the reference FBG or external materials in the existing compensation methods. Further, the present application directly acts on the shape resetting layer and sequentially completes the micro-redistribution in the longitudinal direction, which not only ensures the progressiveness and coordination of the compensation process, but also effectively avoids the cross coupling caused by simultaneous adjustment of multiple points, so that the pre-tension maintaining body can maintain high stability under dynamic working conditions. Through continuous updating and checking of the temperature decoupling mapping table, the present application realizes closed-loop control from spectral recognition, index extraction to compensation execution, with real-time and adaptive advantages. Therefore, the present application is superior to the prior art in terms of fixation reliability and compensation accuracy, and can significantly prolong the service life of the FBG sensing system in complex environments, reduce maintenance costs, improve the accuracy and consistency of overall measurement, and is particularly suitable for long-period operation and engineering application scenarios with extremely high measurement accuracy requirements. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 The method flowchart of the method for maintaining pre-tension and decoupling temperature of the FBG fixed by the invar composite provided by the embodiments of the present application is shown in the figure. Figure 2 The real channel and mirror channel registration experimental curve schematic diagram provided by the embodiments of the present application is shown in the figure. Figure 3 The virtual steady-state reference sequence generation experimental curve schematic diagram provided by the embodiments of the present application is shown in the figure. DETAILED DESCRIPTION

[0016] In order to make the person skilled in the art better understand the technical scheme of the present application, the technical scheme in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by the person skilled in the art without creative labor should belong to the protection scope of the present application.

[0017] Reference Figure 1 : the method for decoupling the pre-tension retention and temperature based on the invar composite fixation, the method comprising: Step 1: assembling the fiber Bragg grating by using the invar composite fixation assembly to form a pre-tension retention body, the invar composite fixation assembly being composed of an invar base layer, a shape reset layer and a micro-tooth clamping seat; establishing a mirror template based on a primary reference reflection spectrum and setting a reference entry in a temperature decoupling mapping table, the mirror template containing virtual marker peaks arranged in a fixed order and a fixed number, serving as anchor points for subsequent registration and cancellation; In the specific implementation process, first, the invar composite fixation assembly is provided, which is composed of an invar base layer, a shape reset layer and a micro-tooth clamping seat. The invar base layer is used to provide high rigidity and stable support, the shape reset layer is used to establish and maintain the reset ability of the pre-tension channel during assembly, and the micro-tooth clamping seat is used to form a stable non-slip clamping interface on the outer surface of the fiber Bragg grating. The fiber Bragg grating is laid along the assembly reference straight line of the invar base layer, so that the sensitive area is located in the clamping window of the micro-tooth clamping seat; the fiber Bragg grating is slowly straightened by means of the reset action of the shape reset layer, so that the fiber Bragg grating remains straight and untwisted in the unclamped state, and the free lead lengths on both sides of the clamping window are as symmetrical as possible to reduce the lateral deviation during subsequent engagement. Then, the micro-tooth clamping seats are closed in order from one end to the other, so that the micro-tooth and the fiber Bragg grating form a surface contact engagement. After the engagement is completed, it is detected whether there is slip and rebound by sliding along the axial direction in a light touch manner, and it is confirmed that the shape reset layer is in balanced stress and well adhered to the invar base layer, that is, a pre-tension retention body is formed. The pre-tension retention body is used to stably transmit the pre-tension and suppress the local displacement caused by the assembly stress in the subsequent process, so as to ensure the repeatability of the fiber Bragg grating reflection characteristics.

[0018] After the pre-tension maintaining body is formed, it is connected to the demodulation device, and only one reference reflection spectrum is collected. Before collection, the demodulation device is in a stable working state, and no repeated scanning or history superposition is performed during the collection process, so as to ensure that the embodiment does not rely on historical data. The obtained reference reflection spectrum is used to establish a mirror template and set a reference entry in a temperature decoupling mapping table. The establishment process of the mirror template is as follows: taking the reference reflection spectrum as the only input, identifying the multi-peak structure in the spectral order and recording the relative position relationship, completely reversing the relative position relationship to form a mirror order, embedding a virtual marker peak in the mirror order, the number of the virtual marker peak is determined at one time in the embodiment and remains fixed in the subsequent process, and the order of the virtual marker peak is determined at one time in the embodiment and remains fixed in the subsequent process, wherein a boundary virtual marker peak is arranged at each end of the mirror order, and a middle virtual marker peak is arranged at the spectral midpoint of adjacent real peaks in the mirror order, so that the virtual marker peak and the real peak form an alternating distribution of anchor points, which are used for subsequent registration and cancellation. After the embedding is completed, the integrity of the mirror template is checked, the checking content includes whether the fixed number of the virtual marker peak is consistent with the setting, whether the fixed order of the virtual marker peak is strictly monotonic, and whether the virtual marker peak and the real peak are positionally coincident, and after the checking passes, the mirror template is solidified and saved.

[0019] Step 2: Based on the mirror template and the input reflection data, a cancellation virtual steady-state fusion algorithm is executed to establish a real channel and a mirror channel, a virtual steady-state reference is used to complete cancellation and robust processing, and a pre-tension index and a temperature drift index are output; when the index passes the check, the corresponding compensation action increment is written into the temperature decoupling mapping table; Specifically, based on the mirror template and the input reflection data, the cancellation virtual steady-state fusion algorithm is executed, first, in principle, the temperature effect and the pre-tension effect are divided into two different spectral changes: under the constraint of the invar steel composite fixed assembly, the temperature effect mainly shows as a slow and same direction translation and a slow scale expansion of the whole spectrum, which belongs to the uniform drift across sections; the pre-tension effect affects the peak shape and the peak distance in the pre-tension maintaining body in the form of local stress redistribution, which belongs to the differential deformation at the section level and the peak level. In order to distinguish the two effects on the same processing chain, the embodiment simultaneously establishes a real channel and a mirror channel, the real channel only carries the input reflection data, and the mirror channel only carries the mirror template and retains the anchor point sequence composed of the virtual marker peaks. The mirror template is generated in the spectral order in step 1, and its reverse structure and the anchor point sequence together constitute a symmetric constraint, so that the subsequent registration must keep the peak sequence from crossing and the section boundary from exceeding, thereby providing a structured reference for separating the whole spectrum drift and the local deformation.

[0020] In the implementation process, the real channel forms a peak group sequence according to the valley between adjacent peaks, and the mirror channel limits the candidate matching interval of each segment by the built-in anchor point sequence. The peak top sequence is used as a strong constraint to search for a continuous anchor point sub-sequence in the mirror channel segment by segment, and the one-to-one correspondence is determined by the joint criterion of the least sequence violation, the minimum shape difference in the segment, and the unique correspondence of the peak top. The registration path from the first segment to the last segment is obtained by combining the forward cumulative cost and backtracking. This path guarantees two points in principle: first, the overall spectrum shift caused by temperature will be consistently mapped to the same direction offset of all segments, without destroying the peak sequence; second, the local peak shape and peak spacing change caused by pre-tension can only be absorbed in the corresponding segment and cannot spread across the anchor point sequence. Based on this correspondence, the spectral midpoint between each pair of registered peak top and anchor point is taken as a virtual steady-state reference point, and the virtual steady-state reference is formed by linear interpolation within the segment. Since one end of the virtual steady-state reference is connected to the anchor point sequence of the mirror channel, and the other end is connected to the peak top position of the real channel, when the input reflection data changes belong to the overall spectrum temperature drift, the virtual steady-state reference can produce a same direction displacement with the peak top position while maintaining the topological consistency with the anchor point sequence; when the change belongs to the local pre-tension effect, the virtual steady-state reference is restrained by the segment boundary of the anchor point sequence and suppresses the non-uniform distortion, thereby being insensitive to local deformation. Therefore, the cancellation result obtained by subtracting the virtual steady-state reference from the real channel is in principle equivalent to the difference quantity after removing the overall common component, becoming a cancellation residual curve representing only the local stress redistribution.

[0021] To ensure the stability of the cancellation residual curve, the embodiment performs robust processing under two types of noise. For sampling noise and transient disturbance, segment smoothing suppresses isolated spikes without changing the peak sequence and segment boundary, and the protection zone at the segment start and end points avoids excessive blurring of the boundary; for cross-segment crosstalk and uncertainty in the anchor point neighborhood, a protection zone is set around each virtual marker peak, the cancellation results on the left and right of the protection zone are checked first, and only when the signs are consistent, the cancellation is allowed, otherwise the median value in the neighborhood is used to replace the interval result, thereby preventing local anomalies from spreading to adjacent segments. After obtaining the stable residual curve through robust processing, the peak top correspondence relationship determined in the registration path is used to read the displacement direction and displacement distance of each peak in the stable residual curve, and the pre-tension index is obtained by summing and normalizing by the number of peaks within the segment; at the same time, the average displacement distance of the overall spectral phase relative to the anchor point sequence of the mirror channel is calculated on the virtual steady-state reference to obtain the temperature drift index. The pre-tension index is derived from the difference quantity after cancellation, reflecting the local stress redistribution within the pre-tension maintaining body; the temperature drift index is derived from the overall displacement of the virtual steady-state reference relative to the anchor point sequence, reflecting the overall spectrum drift. The two are not mixed in the algorithm structure, meeting the decoupling goal of the cancellation virtual steady-state fusion algorithm.

[0022] When the pre-tension index and the temperature drift index are verified, the temperature decoupling mapping table is incrementally written according to the corresponding relationship between the temperature drift index and the compensation action. The writing is constrained by the reference of the mirror template, the consistency of the anchor point sequence, and the current segment boundary state, ensuring that the compensation action is completely consistent with the alignment reference at the time of generation; since the temperature drift index represents the common component of the entire spectrum in principle, the relationship between it and the compensation action can be accumulated in the temperature decoupling mapping table as a stable and monotonous entry set, which is used for real-time compensation and closed-loop maintenance of the stable state of the pre-tension maintaining body in step 3. Through the above principles and implementations, the embodiment absorbs the entire spectrum temperature drift as a virtual steady-state reference follower in the same processing chain, and visualizes the local pre-tension effect as a cancellation residual, achieving real-time decoupling of the input reflection data and establishment of a traceable compensation basis.

[0023] Step 3: Real-time compensation is performed on the input reflection data according to the temperature decoupling mapping table, and the compensation result is applied to the shape reset layer to drive the invar composite fixed assembly to perform micro-redistribution, thereby maintaining the stable state of the pre-tension maintaining body; when the pre-tension index or the temperature drift index triggers an out-of-boundary, step 2 is executed to re-estimate and update the temperature decoupling mapping table until the in-service maintenance requirement is met.

[0024] In the process of implementation, the system first reads the input reflection data and calls step 2 to obtain the pre-tension index and temperature drift index, then accesses the temperature decoupling mapping table to retrieve the target entry matching the temperature drift index. When there is a complete matching entry, the compensation action is obtained directly; when there is no complete matching entry, the temperature drift index is selected adjacent entries above and below, and the compensation action is synthesized according to the relative position between the two entries to maintain the monotonic continuity of the compensation in the temperature dimension. The compensation action is converted into a driving description acting on the shape reset layer, and the driving description is implemented in sequence from one end to the other end along the pre-tension retention body in the longitudinal direction, so that the micro-redistribution of each position is completed in turn to avoid cross coupling caused by simultaneous action; keep the symmetry push in the transverse direction to prevent the fiber grating from producing transverse displacement in the micro-tooth clamping seat. During the compensation execution, the invar base layer provides a constant stiffness channel, and the shape reset layer generates a small reversible displacement according to the driving description, and the stress path inside the invar composite fixing assembly is adjusted to the state consistent with the target entry due to the micro-redistribution. After the completion of the compensation, the new input reflection data is collected immediately and the pre-tension index and temperature drift index are recalculated by calling step 2, and the new temperature drift index is compared with the corresponding value of the target entry; when the difference is within the allowable range set by the temperature decoupling mapping table, and the new pre-tension index is within the allowable range, it is determined that the compensation action is effective, and the compensation is ended; when the difference is not within the allowable range and the pre-tension index is within the allowable range, the compensation action with amplitude scaling or amplitude expansion is generated according to the same target entry without modifying the temperature decoupling mapping table, and the compensation is executed again until the new temperature drift index enters the allowable range, so as to ensure the closed-loop convergence of the whole spectrum drift.

[0025] When either the pre-tension index or the temperature drift index triggers an out-of-range, it is considered that the temperature decoupling mapping table is insufficient to describe the current environment and assembly state, and step 2 is immediately executed to re-estimate, and the new temperature drift index and the corresponding compensation action increment are written into the temperature decoupling mapping table or overwrite the related entry. The principle of this fallback and update mechanism is to use the separation ability of step 2 for the whole spectrum common component and the local difference component to ensure that the temperature decoupling mapping table always aligns with the latest anchor sequence state as a reference; after completing the update, real-time compensation is performed again according to the temperature decoupling mapping table, so that the correspondence between the compensation action and the mirror channel, the virtual steady-state reference and the real channel is kept consistent. Through the above closed-loop process, the compensation continuously offsets the uniform area drift caused by temperature in the form of micro-redistribution inside the invar composite fixing assembly, while keeping the local stress distribution of the pre-tension retention body stable, so that the pre-tension index and the temperature drift index are maintained within the allowable range for a long time under the in-service retention requirement.

[0026] Furthermore, the system first selects a target entry based on the temperature decoupling mapping table and converts the action partition, action direction, and action rhythm in the target entry into an executable drive description. The drive description activates each action partition in sequence from one end to the other in the vertical direction, maintaining left-right symmetry in the horizontal direction to prevent lateral offset of the fiber Bragg grating within the micro-tooth clamping seat. A protection zone is established at the anchor point sequence positions on both sides of each activated action partition to isolate the influence of adjacent partitions. A rhythmic process consisting of an ascending stage, a holding stage, and a descending stage is then executed within the action partition: in the ascending stage, the shape reset layer generates a small, gradual displacement in the action direction to introduce an equivalent reverse action; in the holding stage, the displacement is maintained and the input reflection data is immediately acquired. A new stable residual curve is generated by canceling the virtual steady-state fusion algorithm and the virtual steady-state reference; in the descending stage, when the criteria are met, the displacement is returned to the holding value to solidify the current fine-tuning. If it is detected in the sustained segment verification that the temperature drift indicator converges toward the target item and the pre-tension indicator is still within the allowable range, the current displacement is recorded as a valid step and the compensation action corresponding to the temperature drift indicator is incrementally accumulated in the temperature decoupling mapping table; if the temperature drift indicator converges and the pre-tension indicator deviates from the allowable range, reverse microstepping is performed within the action partition to retreat until the pre-tension indicator is restored; if both indicators do not reach the allowable range, additional microsteps are added in the same action direction while keeping the protection band open and the beat process is repeated until the allowable range is reached or the protection mode is triggered.

[0027] To suppress cross-coupling, when performing micro-redistribution in any active partition, the system freezes the drive descriptions of adjacent partitions, allowing only fine-tuning within the current active partition and its protection band to participate in the calculation. Anchor point sequence consistency checks are performed between beats. If a jump in the anchor point correspondence is detected, the most recent valid step is immediately canceled and retried with a smaller amplitude. To avoid energy accumulation at the boundaries, a sequence of multiple small steps with low amplitude is executed in the active partitions near the ends of the pretensioning body. After each small step, the median value replaces the cancellation result within the protection band to ensure the continuity of the stable residual curve at the segment ends. After completing micro-redistribution in each active partition, a rapid retest is immediately performed, comparing the new pretension index and the new temperature drift index with the previous record. When the changes in the direction of change in two consecutive retests are consistent and the amplitude of the change decreases, the active partition is judged to have reached local convergence, and the drive description is switched to the next active partition for further execution.

[0028] When any action partition appears out of position, out of limit or response delay during the beat process, the system enters protection mode, the partition weight of the action partition and the subsequent action partition is reduced by the original proportion, and the holding time in the beat is correspondingly prolonged; if the same action partition triggers the protection mode for 2 times in succession, the temperature drift index is switched to the low-speed channel entry in the temperature decoupling mapping table, and only the compensation action of the low-speed channel entry is continued to execute, until the new stable residual curve shows that the segment boundary difference between the partition and the adjacent partition is reduced to the allowed range. After completing all the action partitions, the system performs a longitudinal equalization sweep back, and each action partition is symmetrically reset with a small back-laying amount opposite to the last effective step, the purpose is to eliminate the residual bias introduced by the beat switching; after the equalization sweep back, the input reflection data is acquired again and the pre-tension index and the temperature drift index are reviewed through the cancellation virtual steady fusion algorithm, and whether the anchor point sequence in the mirror channel and the peak group sequence in the real channel remain one-to-one correspondence is checked. If both indexes are within the allowed range and the anchor point sequence is consistent, the execution track of the trace redistribution, the number of effective steps and the final driving description are archived together to the execution record of the corresponding entry in the temperature decoupling mapping table, for subsequent quick call and trace in service.

[0029] Further, step 2 specifically comprises: Step 2.1: Establish a real channel and a mirror channel, the real channel is used to carry the input reflection data, and the mirror channel is used to load the mirror template and reserve the anchor point sequence. The mirror template is composed of the mirror sequence and the virtual marker peak given in step 1, and the reverse ordering and the fixed number and fixed sequence of the anchor point sequence provide monotonic and non-crossing matching constraints in the spectral position dimension, so that the subsequent segments only match within the allowed anchor point neighborhood. Through the double-channel structure, any overall translation of the input reflection data is manifested as consistent offset relative to the anchor point sequence, and local changes within the segment are limited within the corresponding segment boundary and cannot spread across segments.

[0030] Step 2.2: Identify a plurality of peaks in the real channel, and divide the spectral line into a plurality of segments with the valley value between adjacent peaks to obtain a peak group sequence. With the peak top sequence of the peak group sequence as a constraint, the segments are registered between the virtual marker peaks in the mirror channel. The joint criterion adopts a combination of the least sequence violation, the least shape difference within the segment, and the only one-to-one correspondence of the peak top, the first two ensure that the registration path remains monotonic and shape consistent in the full spectrum range, and the last one ensures that the same peak top cannot coincide with multiple anchors. The forward cumulative cost and backtracking connection are used to form a registration path from the first segment to the last segment, the continuity of the path makes the overall spectrum change caused by the environmental temperature be uniformly mapped as the same direction offset of each segment, and the local stress redistribution caused by the pre-tension retention body only changes the peak shape and peak spacing within the corresponding segment, thereby separating the two types of actions in structure.

[0031] Step 2.3: A virtual steady reference is generated between each pair of registered peak top and anchor point with the midpoint of the spectrum position, and a virtual steady reference sequence is formed by linear interpolation within the segment. The virtual steady reference is driven by the real channel peak top position at one end and constrained by the mirror channel anchor point sequence at the other end. When the input reflection data has a global shift, the virtual steady reference moves with the peak top together, keeping the segment boundary and order unchanged. When there is a local distortion, the virtual steady reference is insensitive to non-uniform changes due to the segment boundary constraint. The real channel is point-by-point subtracted from the virtual steady reference sequence to obtain a subtraction residual curve, which mainly retains the difference within the segment. Two levels of robust processing are then performed: segment-wise smoothing to suppress isolated spikes with short duration and large amplitude, avoiding amplification of sampling disturbance; peak top consistency check to eliminate abnormal points that do not conform to the evolution rule within the segment based on peak width continuity and peak shape symmetry, obtaining a stable residual curve. After this processing chain, the common offset of the whole spectrum is absorbed in the virtual steady reference, and the difference component is concentrated in the stable residual curve.

[0032] Step 2.4: Using the peak top correspondence obtained by registration, the offset direction and offset distance of each peak are read from the stable residual curve, summed within the segment and normalized by the number of peaks to obtain the pre-tension index; at the same time, the average offset distance of the whole spectrum position relative to the mirror channel anchor point sequence is calculated in the virtual steady reference sequence to obtain the temperature drift index. The pre-tension index is derived from the difference after subtraction, reflecting the local stress redistribution within the pre-tension maintenance body; the temperature drift index is derived from the whole displacement of the virtual steady reference relative to the anchor point sequence, reflecting the whole spectrum drift caused by environmental temperature. Since the registration path remains monotonic, the anchor point sequence is fixed, and the virtual steady reference is linearly interpolated at the segment boundary, the two indexes are not aliasing in the generation chain and can be directly used for subsequent control.

[0033] Step 2.5: The pre-tension index and temperature drift index are checked. The check focuses on two aspects: first, whether the index value is within the allowed range set by the temperature decoupling mapping table, to ensure that the output can be stably executed; second, whether the registration process is stable, including whether the registration path is continuous, whether the peak top correspondence is unique, and whether the segment boundary is crossed. When the check passes, the compensation action increment corresponding to the temperature drift index is written as a new entry in the temperature decoupling mapping table in real time, and the current mirror template reference and anchor point sequence are associated to ensure that the compensation execution is consistent with the decoupling reference; when the check fails, automatically fall back to step 2.2 to re-register and repeat steps 2.2 to 2.5, to eliminate the influence of transient instability or local abnormalities on the index by updating the registration path and virtual steady reference. Through the above mechanism, the fiber grating pre-tension maintenance and temperature decoupling method based on invar composite fixation realizes the separation and quantification of whole spectrum drift and local distortion within a single data channel, providing reliable input for real-time compensation in step 3.

[0034] Further, in step 2.2, the mirror channel loads the mirror template and retains the anchor point sequence, and for each segment, only the candidate registration interval between the adjacent two virtual marker peaks is optimized, which is equivalent to setting a non-border-crossing "window" for each segment, limiting the uniform translation caused by the whole spectrum temperature to the approximately equivalent interval offset for all segments, and limiting the local peak shape and peak spacing changes caused by the pre-tension holding body to be absorbed within the corresponding segment. The evaluation of the candidate registration pair uses a joint criterion, in which the number of violations of the order constraint ensures that the mapping of the peak top and the anchor point remains monotonic, ensuring that the peak order does not cross; the shape difference metric within the segment focuses on the matching degree of the relative peak spacing, the rising and falling slope, and the consistency of the relative position of the peak top, so that the selected mapping not only has a similar position, but also has a consistent shape evolution; the unique marker corresponding to the peak top eliminates one-to-many and many-to-one situations, ensuring that the output relationship can be called by the subsequent virtual steady-state reference and cancellation processing. The registration pair with more than 0 violations of the order constraint is removed in advance to avoid introducing irreversible sequence errors before global optimization; then, the remaining candidate registration pairs are sorted in ascending order of the shape difference metric within the segment, so that the most similar shape is concentrated in the remaining candidate registration pairs; when multiple candidate registration pairs have the same unique marker corresponding to the peak top, they are directly removed to prevent path branching and drifting caused by the same peak occupying the anchor point.

[0035] The forward cumulative cost search is used in the whole segment range, and the connection pointer is used to record the association between the current best registration pair of each segment and the best registration pair of the previous segment, so that the decision depends not only on the local optimum of a single segment, but also on the continuity and compatibility with the previous and subsequent segments. After backtracking, the registration path from the first segment to the last segment is obtained, which naturally satisfies three properties: first, monotonicity, the mapping does not back off along the spectrum position direction, ensuring that the peak order of the real channel to the mirror channel is consistent; second, boundary consistency, the mapping of any segment falls within its candidate registration interval, and the segment boundary is not crossed, thereby providing a reliable boundary for subsequent generation of virtual steady-state references within the segment and linear splicing; third, separability, the whole spectrum translation is reflected as approximately uniform offset in the path, and the inhomogeneous deformation within the segment does not propagate to the adjacent segment, facilitating the absorption of common components and the visualization of difference components through cancellation operation in step 2.3. The final output is the one-to-one correspondence between the peak group of the real channel and the anchor point of the mirror channel, which not only satisfies the local consistency under the joint criterion, but also satisfies the global coherence under the cumulative cost, providing a structured basis for the formation of stable residual curves and the accurate extraction of pre-tension indicators and temperature drift indicators in the subsequent steps.

[0036] Further, in step 2.3, for each pair of registered peak top and anchor point obtained in step 2.2, a virtual steady reference point is generated at the midpoint of the two spectral positions, which serves to symmetrically distribute the local drift of the real channel and the structural constraint of the mirror channel, so that the common displacement of the whole spectrum caused by ambient temperature is equivalently absorbed, and the intra-segment difference caused by the pre-tension holding body is only reflected in the subsequent cancellation residual curve in the residual amount. To avoid discontinuity at the transition of the segment end, at the valley position of the start and end of each segment, a boundary virtual steady reference point is generated according to the trend of the connecting line of the adjacent two pairs of registered peak top and anchor point, so that the segment interpolation maintains the continuity of slope and trend when entering and leaving the segment boundary, preventing the introduction of false distortion by boundary truncation. When close to the ends of the mirror template, if the number of the nearest registered pairs to the ends is less than 2, a supplementary virtual steady reference point is generated in the direction of the anchor point sequence, which is used to provide a minimum support density in the sparse anchor point area, avoid the virtual steady reference sequence from breaking or retreating at the boundary, and ensure the topological consistency with the anchor point sequence.

[0037] In terms of density control, the spectral position interval of any adjacent virtual steady reference point is checked, and a transition virtual steady reference point is inserted in it until the maximum interval does not exceed the upper limit when the interval is greater than the preset upper limit. This limits the local curvature and change rate of the segment interpolation, so that the subsequent linear interpolation does not amplify high-frequency disturbances, while maintaining the sensitivity of the virtual steady reference sequence to the slow drift of the whole spectrum. When a continuous missing interval is detected, the virtual steady reference points are arranged uniformly in the missing interval in the order of the original anchor point sequence, and are monotonically connected with the generated virtual steady reference points on the adjacent two sides, so that the reference skeleton is reconstructed without destroying the peak sequence and segment boundary, ensuring that the registration relationship between the real channel and the mirror channel can still be linearly mapped and robustly evaluated in the interval.

[0038] After the virtual steady reference sequence is completed, an equal-length reference sequence is formed by linear interpolation according to the sampling positions of the real channel. This process maps the reference skeleton to the sampling grid consistent with the input reflection data, avoiding pseudo-differences caused by sampling misalignment, so that point-by-point cancellation only reflects physical mechanisms rather than sampling errors. The real channel is subtracted from the equal-length reference sequence point by point to obtain the cancellation residual curve, the common displacement of the whole spectrum is absorbed by the virtual steady reference sequence in the construction, and the non-uniform change in the segment is retained as the difference signal, which is convenient for extracting the pre-tension index and temperature drift index in the subsequent link.

[0039] To suppress cross-segment crosstalk and ensure segment boundary stability, a guard band is set at each virtual marker peak position and a sign consistency check is performed first. When the results of the cancellation on the left and right sides of the guard band are consistent, the cancellation is performed, indicating that the change trends on both sides of the segment boundary are in the same direction and there is no risk of phase reversal; when the signs are inconsistent, the cancellation results in the interval are replaced by the median value in the neighborhood of the guard band, which suppresses the cross-segment misjudgment caused by noise, local mismatch or transient disturbance through the anti-exceptional characteristics of the median value, and maintains the continuity and monotonicity of the cancellation residual curve at the segment boundary. Through the above construction, the midpoint generation ensures the symmetric absorption of the common component, the boundary virtual steady-state reference point and the supplementary virtual steady-state reference point ensure the continuity of the segment boundary and the template edge, the transition virtual steady-state reference point and the monotonicity constraint ensure the smoothness and topological invariance of the reference trajectory, the equal-length reference sequence eliminates the influence of sampling misalignment, and the guard band and the sign consistency check provide the necessary barrier for segment isolation and robust noise suppression, thereby comprehensively realizing the stable input basis required by the cancellation virtual steady-state fusion algorithm.

[0040] Further, in step 2.3, the technical mechanism of the intra-segment smoothing and peak consistency check is described. The intra-segment smoothing aims to stabilize the residual curve, and through local processing within the segment, it removes sampling disturbances and transient noise from the cancellation residual curve while maintaining the segment boundary structure consistent with the anchor sequence. The smoothing window is set to contain 5 sampling points with a step size of 1 sampling point, so that each update only depends on the nearest information, avoiding cross-segment information leakage. When the number of peaks within the segment is less than 3, the window is reduced to 3 sampling points to prevent excessive averaging from weakening the peak top in a sparse peak environment; when the number of peaks within the segment is greater than 7, the window is increased to 7 sampling points to enhance local statistical stability in a dense peak environment. For isolated spikes that last only 1 sampling point and are higher than the average on both sides, the neighborhood average is used to replace them, so that the energy of isolated pulses does not enter the peak shape metric; for spikes that last 2 sampling points, linear interpolation of adjacent endpoints is used to replace them, so that short-term mutations are absorbed by smooth transitions without changing the trend within the segment. Intra-segment smoothing also introduces a protection smoothing, which sets a guard band at the beginning and end of each segment, and uses half-window smoothing within the guard band to limit the sliding window from crossing the segment boundary, thereby avoiding the blunting effect on the steepness and phase of the true segment boundary, and ensuring that the subsequent virtual steady-state reference can maintain monotonicity and continuity at the segment end.

[0041] After the in-segment smoothing, the peak consistency check with geometric constraints ensures that the peak shape and peak width in the residual meet the in-segment gradual change rule, so that the temperature-induced overall spectrum shift continues to be left in the virtual steady-state reference channel, and the local difference related to the pre-tension is stabilized and visualized. The peak width of each peak is calculated as the sum of the distance from the peak top to the nearest valley value on the left and the distance from the peak top to the nearest valley value on the right. When the peak width difference between adjacent peaks is greater than 2 sampling points, the outward expansion priority rule is used to symmetrically expand the one with smaller peak width until the peak width difference does not exceed 2 sampling points, ensuring the continuity and comparability of the in-segment energy distribution. The absolute value of the left and right distance difference is taken as the peak shape symmetry deviation. When the deviation is greater than 3 sampling points, the two sides are first stretched by the same amplitude, so that the asymmetry is converted from a translation type error to a controllable shape type error; if the symmetry deviation after stretching still does not meet the requirements, peak crown back paving is performed, and the peak crown area is paved with equal length on both sides, so that the curvature of the peak crown area is balanced on both sides, avoiding the bias caused by unilateral sharpness into the pre-tension index. If the peak top falls within the protection zone of the virtual marker peak, the peak top position is clamped at the boundary of the protection zone and the adjacent two sides are transitioned by linear interpolation, so that the protection zone constrains the phase of the segment boundary, preventing the small mismatch in the anchor point neighborhood from spreading to the cross-segment deformation. Through the above constraint chain, the in-segment smoothing provides noise suppression and boundary conservation, and the peak consistency check provides shape geometry consistency check. Both of them together ensure that the stable residual curve only carries information related to the local stress redistribution of the pre-tension maintaining body, and the uniform drift caused by temperature is still borne by the virtual steady-state reference, so as to provide stable input for the reliable extraction of the subsequent pre-tension index and temperature drift index.

[0042] Let the sampling grid be (the index of the th sampling position, in units of sampling points), In this example, we take . Let the reference reflectance spectrum be , and the input reflectance data be . Let the set of reference peak positions be (the position of the th peak top on the sampling axis), and in this example, we obtain main peaks in step 1, with positions .

[0043] Construct an image template based on the reference reflectance spectrum. Let the set of virtual marker peaks in the image template be (the position of the th virtual marker peak on the sampling axis, with a fixed number and order), and in this example, we obtain according to the embedding rule of the two end boundaries and the midpoint of the adjacent real peak, so as to obtain the anchor point sequence . Let the temperature decoupling mapping table be , whose reference entries are written into the reference to the image template and fixed number and fixed order.

[0044] In step 2, the real channel and the mirror channel are established. After the input reflection data in the real channel is segmented and registered by peak group, a one-to-one correspondence between each peak top and the anchor point of the mirror channel is obtained. For the sake of brevity, the anchor point corresponding to the i-th peak is denoted as (the value is taken from the specific element in ), and the registration result of the present example is .

[0045] Suppose the peak top position of the input reflection data is (the peak top position detected on ). The present example measures under the combined action of the increase of the ambient temperature and the partial stress fine adjustment of the pre-tension maintaining body. In step 2.3, the virtual steady-state reference point is generated according to the midpoint of the registered pair. Suppose the position of the i-th virtual steady-state reference point is , which is defined as . Accordingly, we obtain .

[0046] Suppose the equal-length reference sequence formed on the sampling grid by linear interpolation within the segment is . The subtraction of the real channel and the equal-length reference sequence point by point obtains the cancellation residual curve . In the smoothing link within the segment, suppose the smoothing window length is (an odd number of sampling points), the step is , and the moving average is adopted .

[0047] When the number of peaks within the segment is less than , take ; when the number of peaks within the segment is between and , take ; when the number of peaks within the segment is greater than , take . Only the isolated sharp peak lasting for sampling points and higher than the average value of the two sides is replaced by the neighborhood average value; the sharp peak lasting for sampling points is replaced by the linear interpolation of the adjacent endpoints; a protection band is set at the start and end positions of each segment and the half-window smoothing is adopted. Suppose the curve after smoothing and peak top consistency check is the stable residual curve .

[0048] In the index extraction link, suppose the local offset of the i-th peak in the stable residual sense is , and its physical meaning is the local offset of the peak relative to the virtual steady-state reference. The present example calculates .​​ .

[0049] Let the pre-tension index be defined as the root mean square measure after de-meaning .

[0050] Substitute the data in this example, .

[0051] Let the temperature drift index be defined as the average displacement of the virtual steady state reference from the anchor point .

[0052] Substitute the data .

[0053] Thus, describes the magnitude and direction of the overall spectrum uniform drift, describes the strength of the intra-segment difference after removing the uniform drift. The two are decoupled: dominated by the consistent displacement of each peak from the anchor point, dominated by the dispersion of each peak from the consistent displacement.

[0054] In step 2.5, let the allowed threshold be and . In this example, .

[0055] When and , the check passes. The compensation action increment corresponding to is then written to the temperature decoupling mapping table . Let the compensation action vector of the shape reset layer be ( is the number of partitions), let the partition weight be and , let the proportionality coefficient of temperature drift to displacement be . In this example, , , , then , and is obtained. The entry is incremented in .

[0056] In real-time compensation in step 3, according to act on the shape reset layer in order from one end to the other to form a small amount of redistribution. Let the new peak position obtained by re-measuring immediately after one compensation be , and calculate the new and If where and is the allowed deviation of in-service retention (in this example, the , ), then end compensation; otherwise generate a magnitude scaling or expansion entry and repeat the beat once until the threshold is met, if out of bounds, then back up to step 2 to re-estimate and with new and override corresponding entry.

[0057] In summary, the present example gives the numerical process of fixed sampling size, clear peak position, anchor point and virtual steady-state reference; through , the pre-tension index and temperature drift index are output simultaneously, and the shape reset layer is driven to complete closed-loop compensation and entry backwrite, meeting the engineering landing requirements of the fiber grating pre-tension retention and temperature decoupling method based on invar composite fixation.

[0058] The experiment curve shows the core registration process of the cancellation virtual steady-state fusion algorithm in the application, the wavelength is taken as the horizontal coordinate, and the reflectivity is taken as the vertical coordinate, the upper half part shows the real channel data, and the lower half part shows the mirror channel data. It can be observed from the experimental data that the real channel carries the actually measured input reflection data, and the typical fiber grating multi-peak reflection characteristics are presented in the wavelength range of 1548.0 nm to 1552.0 nm. The registration process is strictly performed according to the joint criterion: first, the system identifies four main peaks in the real channel in the mode of continuous rise, peak top and continuous decline, which are located at 1549.0 nm, 1549.5 nm, 1550.5 nm and 1551.5 nm respectively. Then the spectrum is divided into three continuous segments by the valley value between adjacent peaks, and each segment contains the complete structure of 1-2 peaks. In the mirror channel, the system loads the pre-established mirror template and retains the anchor point sequence, and these anchor points are distributed in the corresponding spectral position interval in a fixed order and a fixed number. The key technology in the registration process lies in the application of the joint criterion: the violation of the order minimization principle ensures the physical rationality of the registration path, the shape difference minimization principle selects the best match by calculating the shape similarity between the real peak and the mirror anchor point, and the peak top corresponds to the unique marker to prevent one-to-many or many-to-one false registration. The experimental results show that, through the forward cumulative cost search algorithm, the system successfully establishes a one-to-one correspondence between the peak group of the real channel and the anchor point of the mirror channel, and the registration accuracy reaches ±0.02 nm, which provides reliable basic data for subsequent virtual steady-state reference sequence generation. Figure 2

[0059] Figure 3 ​​The generation process of the virtual steady-state reference sequence and its technical implementation details are described in detail. The experimental curve is generated by specific geometric rules based on the registration results of Figure 2 The virtual steady-state reference points are generated and connected into a continuous sequence. It can be clearly observed from the curve that the system first generates virtual steady-state reference points at the spectral midpoint of each pair of registered peak top and anchor point, and these midpoints accurately reflect the geometric center position of the real channel and the mirror channel. The generation of boundary virtual steady-state reference points uses extrapolation technology: at the starting valley and the ending valley of each segment, the system performs linear extrapolation according to the trend of the connecting line of the adjacent two pairs of registered peak top and anchor point, to ensure the smoothness of the sequence at the segment end transition. When the number of the nearest registered pairs to the ends of the mirror template is less than 2, the system generates supplementary virtual steady-state reference points in the direction of the anchor point sequence, effectively preventing the sequence from being broken due to boundary sparseness. The insertion of transition virtual steady-state reference points follows the interval control rule: the system automatically detects the spectral interval of adjacent virtual steady-state reference points, and when the interval exceeds the preset upper limit of 0.5 nm, transition reference points are uniformly inserted in the interval until the maximum interval meets the requirements. For the detected continuous missing interval, the system uniformly arranges virtual steady-state reference points in the missing interval according to the order of the original anchor point sequence, and maintains monotonous connection with the generated reference points on the adjacent two sides through linear interpolation. The finally generated virtual steady-state reference sequence has good continuity and stability, providing a high-quality reference benchmark for the subsequent point-by-point cancellation operation, and the cancellation accuracy can reach ±0.005 nm.

[0060] The above-described and above-embodied examples are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for maintaining pre-tension and temperature decoupling of fiber Bragg gratings based on Invar composite fixation, characterized in that: The method comprises: Step 1: Assemble the fiber Bragg grating (FBG) using an Invar composite fixture assembly to form a pre-tensioned retainer. The Invar composite fixture assembly consists of an Invar substrate layer, a shape-reset layer, and a micro-toothed clamping seat. A mirror template is created based on a primary reference reflection spectrum and reference entries are set in the temperature decoupling mapping table. The mirror template contains virtual marker peaks set in a fixed order and number, which serve as anchor points for subsequent registration and cancellation. Step 2: Based on the mirror template and input reflection data, the virtual steady-state fusion algorithm is executed to establish the real channel and the mirror channel. The virtual steady-state reference is used to complete the cancellation and perform robust processing, and the pretension index and temperature drift index are output. When the index passes the verification, the corresponding compensation action increment is written into the temperature decoupling mapping table. Step 3: Perform real-time compensation on the input reflection data based on the temperature decoupling mapping table, and apply the compensation result to the shape reset layer to drive the Invar composite fixing component to perform micro-redistribution, thereby maintaining the stable state of the pre-tension retainer. When the pre-tension index or temperature drift index triggers an out-of-bounds, return to step 2 to re-estimate and update the temperature decoupling mapping table until the in-service retention requirements are met.

2. The fiber Bragg grating pre-tensioning and temperature decoupling method based on Invar composite fixation according to claim 1, characterized in that: In step 1, the reference reflectance spectrum is used as the only input, the peak group is identified according to the spectral position sequence and the relative position of each peak in the spectral position sequence is recorded; the relative position sequence is completely reversed to form a mirror sequence; the virtual marker peak is embedded in the mirror sequence; the setting rule of the virtual marker peak is: a boundary virtual marker peak is set at each end of the mirror sequence, and an intermediate virtual marker peak is set at the midpoint of the spectral position of the adjacent real peak in the mirror sequence, so that the virtual marker peak and the real peak form an alternating anchor point sequence. After the embedding is completed, a mirror template is generated.

3. The fiber Bragg grating pre-tensioning and temperature decoupling method based on Invar composite fixation according to claim 2, characterized in that: In step 1, the process of creating a baseline entry of the temperature decoupling mapping table includes: writing a reference to the mirror template into the baseline entry, writing a fixed number and a fixed order of virtual marker peaks into the baseline entry, and writing an anchor point sequence into the baseline entry; in the baseline entry, setting the pretension indicator field to an uncalculated state, setting the temperature drift indicator field to an uncalculated state, setting the compensation action field to 0, and setting the entry status field to available; the baseline entry is set to a read-only state after the writing is completed.

4. The fiber Bragg grating pre-tensioning and temperature decoupling method based on Invar composite fixation according to claim 3, characterized in that: Step 2 specifically includes: Step 2.1: Establish a real channel and a mirror channel. The real channel carries the input reflection data, and the mirror channel loads the mirror template and retains the anchor point sequence. Step 2.2: Identify multiple peaks in the real channel and divide the spectrum into segments based on the valley values ​​between adjacent peaks to obtain a peak cluster sequence. Using the peak top order of the peak cluster sequence as a constraint, perform segment-by-segment registration between the virtual marker peaks in the mirror channel. Select the registration path using the joint criteria of minimal order violations, minimal intra-segment shape differences, and unique peak top correspondence to obtain a one-to-one correspondence between the peak clusters in the real channel and the anchor points in the mirror channel. Step 2.3: At each pair of registered peaks, generate a virtual steady-state reference point at the midpoint between the peak position of the real channel and the anchor point position corresponding to the mirror channel. Connect adjacent virtual steady-state reference points to form a virtual steady-state reference sequence using intra-segment linear interpolation. Perform point-by-point cancellation of the real channel and the virtual steady-state reference sequence to obtain a cancellation residual curve. Based on the cancellation residual curve, perform two-level robust processing, including intra-segment smoothing to suppress isolated spikes and peak consistency checking to eliminate outliers that do not meet peak width continuity and peak shape symmetry, to obtain a stable residual curve. Step 2.4: In the stable residual curve, read the offset direction and offset distance of each peak according to the peak top correspondence obtained by registration. Sum the offset information of all peaks within the segment and normalize it by the number of peaks to obtain the pretension index. At the same time, read the average offset distance of the entire spectrum relative to the anchor point of the mirror channel in the virtual steady-state reference sequence to obtain the temperature drift index. Step 2.5: Calibrate the obtained pretension index and temperature drift index. The pass condition for the calibration is that the index value is within the preset range and the alignment process is stable. If the calibration passes, the compensation action increment corresponding to the temperature drift index is immediately written as a new entry into the temperature decoupling mapping table. If the calibration fails, it automatically returns to step 2.2 to re-align and repeats steps 2.2 to 2.5 until the calibration passes.

5. The fiber Bragg grating pre-tensioning and temperature decoupling method based on Invar composite fixation according to claim 4, characterized in that: Step 2.2 specifically includes: traversing the real channel from left to right point by point, identifying multiple peaks in the pattern of continuous rise, peak top, and continuous decline, and recording each peak top and its nearest valley value on the left and right; dividing the spectrum into several continuous segments based on adjacent valley values, forming a sequence of peak groups arranged in spectral position order, and recording the valley value at the starting point of the segment, the valley value at the end of the segment, and the number of peaks in the segment for each segment; for each segment, selecting the interval between two adjacent virtual marker peaks in the mirror channel as the candidate registration interval, which contains one or more anchor points; calculating the joint criterion for each set of candidate registration pairs, and the joint criterion is determined by violating The algorithm is composed of the number of inverse order, the shape difference measure within the segment and the unique mark corresponding to the peak; first, the registration pairs with the number of order violations greater than 0 are screened out; then the registration pairs are sorted from small to large according to the shape difference measure within the segment; if multiple candidate registration pairs have the same unique mark corresponding to the peak, these registration pairs are eliminated; the forward cumulative cost search is used to jointly optimize the entire segment, retaining the connection pointer between the current best registration pair and the best registration pair of the previous segment in each segment, and after completing the backtracking, the registration path from the first segment to the last segment is obtained, and the one-to-one correspondence between the peak group of the real channel and the anchor point of the mirror channel is output.

6. The fiber Bragg grating pre-tensioning and temperature decoupling method based on Invar composite fixation according to claim 5, characterized in that: In step 2.3, for each pair of registered peaks and anchor points obtained in step 2.2, a virtual stable reference point is generated at the midpoint of their spectral positions. At the valley values ​​at the start and end of each segment, one boundary virtual stable reference point is generated based on the trend of the connection between the two adjacent pairs of registered peaks and anchor points to constrain the segment end transition. When there are fewer than two pairs of nearest registered pairs at both ends of the mirror template, one supplementary virtual stable reference point is generated by extending along the anchor point sequence to prevent breakage caused by sparse boundaries. The spectral position spacing of any adjacent virtual stable reference points is checked. When the spacing is greater than the preset upper limit, a transition virtual stable reference point is inserted until the maximum spacing does not exceed the limit. When a continuous missing interval is detected, the virtual stable reference points are evenly arranged in the missing interval in the order of the original anchor point sequence, and are kept monotonically connected with the generated virtual stable reference points on both sides.

7. The fiber Bragg grating pre-tensioning and temperature decoupling method based on Invar composite fixation according to claim 4, characterized in that: In step 2.3, the virtual steady-state reference sequence is linearly interpolated according to the sampling position of the real channel to form an equal-length reference sequence; the real channel and the equal-length reference sequence are subtracted point by point to obtain the cancellation residual curve; A sign consistency check is first performed within the guard band of each virtual marker peak. Cancellation is performed when the signs of the cancellation results immediately to the left and right of the guard band are consistent. Otherwise, the cancellation result in this interval is replaced by the median value of the guard band neighborhood to suppress cross-segment crosstalk.

8. The fiber Bragg grating pre-tensioning and temperature decoupling method based on Invar composite fixation according to claim 7, characterized in that: In step 2.3, the process of performing intra-segment smoothing to suppress isolated spikes includes: setting the smoothing window to include 5 sampling points and a step size of 1 sampling point; when the number of peaks in the segment is less than 3, the window is reduced to 3 sampling points, and when the number of peaks in the segment is greater than 7, the window is increased to 7 sampling points; for isolated spikes that only last for 1 sampling point and are higher than the mean on both sides, they are replaced with the neighborhood mean; for spikes that last for 2 sampling points, they are replaced by straight lines connecting the adjacent endpoints; intra-segment smoothing also includes protection smoothing, that is, setting a protection band at the starting point and end point of each segment, and using half-window smoothing within the protection band to avoid excessive boundary blunting.

9. The fiber Bragg grating pre-tensioning and temperature decoupling method based on Invar composite fixation according to claim 8, characterized in that: In step 2.3, the peak consistency check is performed on the cancellation residual curve after intra-segment smoothing, including: calculating the peak width of each peak as the sum of the distance from the peak top to the nearest valley on the left and the distance from the peak top to the nearest valley on the right; when the peak width difference between adjacent peaks is greater than 2 sampling points, the peak with narrower width is symmetrically expanded using the outward expansion priority rule until the peak width difference does not exceed 2 sampling points; calculating the absolute value of the left and right distance difference as the peak shape symmetry deviation, when the deviation is greater than 3 sampling points, first stretch both sides with equal amplitude; if the symmetry deviation still does not meet the requirements after stretching, perform peak crown back-spreading, and back-spread with equal length on both sides of the peak top until the shapes of both sides tend to be symmetrical; if the peak top falls within the protection band of the virtual marker peak, the peak top position is clamped to the protection band boundary and the adjacent two sides are linearly interleaved.

Images