A rail grinding strategy optimization method based on a grinding template

Abstract

The application discloses a rail grinding strategy optimization method based on a grinding template, and comprises the following steps: collecting a measured profile, a target profile, grinding template parameters and engineering constraints; determining a grinding pass search range through template capacity evaluation; screening a candidate template combination sequence by using a beam search algorithm; classifying and evaluating the candidate sequence and performing directional optimization according to the engineering constraints to obtain a grinding scheme meeting the constraints; generating a single-side grinding scheme after full-dimension comprehensive verification; and finally performing bilateral synchronous processing to generate a bilateral synchronous grinding scheme. The application realizes automatic and refined design of the grinding strategy, effectively improves grinding precision and efficiency, and ensures bilateral grinding consistency.

Description

Technical Field

[0001] This invention relates to the field of rail grinding technology, and more specifically, to a method for optimizing rail grinding strategies based on grinding templates. Background Technology

[0002] As the core load-bearing component of railway lines, rails directly bear the wheel-rail interaction forces during train operation. Their profile accuracy and surface condition are closely related to the safety, stability, and operational efficiency of railway transportation. During long-term operation, rails inevitably experience wear and deformation due to factors such as wheel-rail contact friction and dynamic impacts, requiring regular grinding operations to restore their designed profile and repair surface defects.

[0003] The formulation and execution of traditional rail grinding operation plans have long relied heavily on the practical experience of on-site operators, lacking unified and scientific execution standards. Specific problems manifest in the following aspects:

[0004] (1) The design of the grinding scheme is highly subjective: there is a pre-set grinding template library, but there is no standardized template combination, grinding pass number, power adjustment design logic. The grinding method is selected based on experience, and it is difficult to formulate a targeted grinding strategy according to the actual profile deviation of the rail. The problem of insufficient grinding or over-grinding is likely to occur.

[0005] (2) Insufficient work accuracy and consistency: A high-precision matching system between the measured profile and the target profile has not been established. The error between the rail profile after grinding and the design target is difficult to control within the optimal range of the project. The grinding effect varies greatly among different operators and at different work times.

[0006] (3) Lack of scientific optimization of grinding process: The combination sequence and usage order of grinding templates were not systematically planned, and the process design was not combined with engineering constraints, resulting in redundant grinding passes and low work efficiency.

[0007] (4) Asynchronous grinding of the two sides of the track: In traditional operation, the grinding coordination of the left and right sides of the track is not considered. The number of grinding passes and parameter settings of the two sides are independent of each other, which easily leads to inconsistent contour repair effects on the two sides.

[0008] Therefore, there is an urgent need for an optimization method for rail grinding strategy based on grinding templates to solve the above problems. Summary of the Invention

[0009] This invention aims to solve the technical problems of existing rail grinding operations, such as reliance on manual experience, lack of scientific standards for template selection and combination, low grinding accuracy, poor efficiency, and asynchronous grinding on both sides. It provides an optimized rail grinding strategy based on grinding templates.

[0010] To achieve the above objectives, this invention provides a method for optimizing rail grinding strategies based on a grinding template, comprising:

[0011] Collect the measured profile of the rail, the target profile, the predefined grinding template parameters, and the engineering constraint parameters;

[0012] Based on the polishing template parameters, the template capability of each predefined polishing template is evaluated, and the search range for the number of polishing passes is obtained based on the capability evaluation results.

[0013] Using a beam search algorithm, starting from the measured profile and aiming to approximate the target profile, candidate template combination sequences are selected from the polishing template within the search range;

[0014] The candidate template combination sequence is constrained and evaluated according to the engineering constraint parameters. Based on the constraint evaluation results, the candidate template combination sequence is divided into different categories of schemes. Targeted optimization is performed on the candidate sequences of different categories of schemes to obtain a polishing scheme that meets the engineering constraints.

[0015] The grinding scheme was comprehensively verified in all dimensions, and after the verification was passed, it was determined to be a single-sided rail grinding scheme.

[0016] Preferably, the grinding template parameters include information on N predefined grinding templates, each template containing M grinding heads, and specifying the grinding angle and grinding power of each grinding head. The engineering constraint parameters include the maximum allowable grinding surface length, profile error threshold, and smoothness threshold, which are set according to the differences in grinding angles.

[0017] Preferably, the template capability assessment specifically includes: simulating the grinding effect of each predefined template on a standard rail specimen, calculating the actual effective removal capacity of a single grinding pass; statistically analyzing the actual effective removal capacity of all predefined templates and calculating their average value; multiplying the average value by a preset safety factor to obtain a grinding capability benchmark value; calculating the estimated number of grinding passes based on the area of ​​the profile deviation between the measured profile and the target profile and the grinding capability benchmark value; and setting a minimum search pass and a maximum search pass, centered on the estimated number of grinding passes, to form the search range of the bundle search algorithm.

[0018] Preferably, the beam search algorithm uses the combination sequence of the polishing templates as the search variable to construct a comprehensive objective function. Wherein, error is the root mean square error of the normal direction, smoothness is the average curvature of the profile, and step is the number of polishing passes. , , These are the preset weighting coefficients.

[0019] Preferably, the beam search algorithm includes:

[0020] An initial search state is created, wherein the template sequence of the initial search state is empty, the profile is the measured profile, and the error is the root mean square error of the initial normal of the measured profile and the target profile. The cost of the initial search state is calculated according to the comprehensive objective function. The initial search state is added to the bundle set and the bundle width parameter K is preset. The bundle width parameter K represents the number of search states retained in each iteration.

[0021] For each search state in the bundle set, all predefined polishing templates are used sequentially. The new candidate search states generated by each polishing template are counted, and the profile, error, cost, and corresponding template sequence of each new candidate state are calculated. All newly generated candidate states are collected, and inferior candidates whose error increases by more than a preset threshold compared to the parent state are removed. The preset threshold is 0.05mm, which can be adjusted according to the polishing accuracy requirements. The remaining candidate states are sorted from low to high cost, and the top K are selected as the bundle set for the next round. The search state with the lowest cost in the current bundle set is recorded and compared with the historical global optimal solution. If the cost is lower, the global optimal solution is updated.

[0022] Determine whether the preset termination condition is met; if so, stop the iteration. The termination condition includes at least one of the following: reaching the maximum number of search iterations, the error of the global optimal solution being lower than the profile error threshold, no update of the global optimal solution for multiple consecutive rounds, or the calculation time exceeding the preset duration.

[0023] Output the template combination sequence corresponding to the global optimal solution, and also output the top M candidate template sequences with the lowest cost, along with their corresponding error, smoothness, and number of polishing passes, where M is the preset number of output candidates, and M < K.

[0024] Preferably, in creating an initial search state, search efficiency can be improved through heuristic sorting, parallel computing, and early pruning.

[0025] Preferably, the candidate template combination sequence is divided into different categories of schemes based on the constraint evaluation results, including: category A schemes, category B schemes, category C schemes and category D schemes.

[0026] Option A satisfies all engineering constraints: error ≤ profile error threshold, polished surface length ≤ maximum allowable polished surface length, and smoothness ≤ smoothness threshold.

[0027] Scheme B is defined as follows: error > profile error threshold, polished surface length ≤ maximum allowable polished surface length, and smoothness ≤ smoothness threshold.

[0028] Category C schemes are those where the error is less than or equal to the profile error threshold, and the length of the polished surface is greater than the maximum allowable length of the polished surface or the smoothness is greater than the smoothness threshold.

[0029] Option D is one of the following three cases:

[0030] Error > profile error threshold and polished surface length > maximum allowable polished surface length;

[0031] Error > Profile error threshold and smoothness > Smoothness threshold;

[0032] Error > profile error threshold, polished surface length > maximum allowable polished surface length and smoothness > smoothness threshold.

[0033] Preferably, the targeted optimization of the B-type scheme includes: taking the B-type scheme as the scheme to be optimized, and setting a maximum allowable increase in search depth. Starting from the increase in search depth, the following search process is executed: each predefined polishing template is sequentially inserted at the end of the current scheme to be optimized, and different adjustable power levels are sequentially selected for the inserted templates to generate multiple candidate schemes. The post-polishing profile error of each candidate scheme is calculated. If there is a candidate scheme whose error meets the profile error threshold, it is output as the optimized scheme. If the error of all candidate schemes does not meet the profile error threshold, candidate schemes are filtered according to preset pruning rules: schemes exhibiting over-polishing are removed, schemes with error improvement amounts lower than the profile error threshold are removed, and the remaining schemes are sorted from high to low according to error improvement amounts, retaining a preset number of schemes as the schemes to be optimized in the next round of search. The search depth is increased by 1, and the above search process is repeated until a scheme meeting the profile error threshold is found or the search depth reaches the maximum allowable increase in search depth, where the search depth is the number of polishing steps currently attempted in the iterative deepening search.

[0034] The targeted optimization of the C-type scheme includes: when the C-type scheme has a grinding surface length greater than the maximum allowable grinding surface length, adding a low-power, wide-angle template grinding pass to the C-type scheme, updating the C-type scheme, and recalculating the error and grinding surface length. If the grinding surface length is less than or equal to the maximum allowable grinding surface length and the error is less than or equal to the profile error threshold, then the optimization is complete; if the grinding surface length is greater than the maximum allowable grinding surface length, then continue to increase the number of grinding passes with the same parameters until the grinding surface length is less than or equal to the maximum allowable grinding surface length or the error is greater than the profile error threshold.

[0035] When the smoothness of the C-type scheme is greater than the smoothness threshold, add a low-power, uniform angle template grinding pass to the C-type scheme, update the C-type scheme and recalculate the error and smoothness. If the smoothness is less than or equal to the smoothness threshold and the profile error is less than or equal to the profile error threshold, then the optimization is complete. If the smoothness is greater than the smoothness threshold, continue to increase the number of grinding passes with the same parameters until the smoothness is less than or equal to the smoothness threshold or the error is greater than the profile error threshold.

[0036] The targeted optimization of scheme D includes: using the grinding surface length constraint and smoothness constraint as optimization objectives, and constructing a second comprehensive objective function together with the error constraint. The second comprehensive objective function is... Different template and power combinations are sequentially inserted into the current scheme to generate candidate schemes. The error, polishing surface length, and smoothness of each candidate scheme are calculated, and the candidate scheme that simultaneously satisfies all three constraints is selected as the optimized scheme output. If no scheme satisfying all three constraints is found at the current search depth, the search depth is increased to continue searching until a scheme satisfying the constraints is found or the maximum allowable increase in search depth is reached.

[0037] Preferably, the comprehensive verification includes: verifying whether the root mean square error of the normal direction between the polished profile and the target profile is less than or equal to the profile error threshold, and whether there is any over-polishing; verifying whether the polished surface length corresponding to each angle of the rail is less than or equal to the maximum allowable polished surface length corresponding to that angle; and verifying whether the average curvature of the polished profile is less than or equal to a preset smoothness threshold, and whether the profile has no obvious fluctuations or abrupt changes. If any of the above verifications fails, the polishing scheme is returned to the constraint evaluation and directional optimization stage for re-optimization.

[0038] Preferably, the generation of the dual-sided synchronous rail grinding scheme includes: independently generating a left-side rail grinding scheme and a right-side rail grinding scheme, and recording their template combination sequence and the number of grinding passes on the left and right sides respectively; taking the larger value between the left-side and right-side grinding passes as the target number of grinding passes for dual-sided synchronous grinding; for the side with fewer grinding passes, calculating the number of additional grinding passes P = target number of passes - original number of grinding passes on that side; adding P passes of low-power, uniform angle template grinding to the original grinding scheme on that side to obtain the synchronous grinding scheme for that side; merging the synchronous grinding schemes on the left and right sides to generate a dual-sided synchronous grinding scheme table, which specifies the template combination, grinding power, and grinding angle corresponding to each grinding pass.

[0039] The beneficial effects of this invention are as follows:

[0040] This invention uses a combined bundle search algorithm to make the measured profile continuously approach the target profile in order to minimize the profile error. Furthermore, through comprehensive verification of error, polished surface, and smoothness, the error between the polished profile and the target profile is strictly controlled within the engineering allowable threshold.

[0041] Secondly, this invention scientifically defines the search range for the number of polishing passes through template capability assessment to avoid invalid searches, and then uses a bundle search algorithm to quickly find the optimal template combination sequence. At the same time, it takes the minimum number of polishing passes as one of the optimization objectives to reduce the number of polishing passes while meeting quality requirements.

[0042] Furthermore, this invention evaluates and classifies the polishing scheme based on triple engineering constraints of profile error, polishing surface length, and smoothness, and implements targeted algorithm optimization and process adjustment for different quality problems to ensure that the final polishing scheme meets all engineering constraints.

[0043] Finally, this invention restores the standard profile of the rails through fine grinding, improves the wheel-rail contact relationship, and the synchronous grinding design of the rails on both sides ensures the consistency of the rail profiles on both sides, effectively reducing vibration and noise during train operation.

[0044] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0045] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a flowchart of the rail grinding strategy optimization method of the present invention;

[0047] Figure 2 This is a schematic diagram of the normal error of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0049] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be defined and explained again in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0050] Example 1:

[0051] This embodiment details the specific implementation process of an optimized rail grinding strategy based on a grinding template.

[0052] like Figure 1 As shown, the method includes the following steps:

[0053] Step 1: Collect the measured profile of the rail, the target profile, the predefined grinding template parameters, and the engineering constraint parameters.

[0054] The grinding template parameters include information on N predefined grinding templates, each template containing M grinding heads, and specifying the grinding angle and grinding power of each grinding head. The engineering constraint parameters include the maximum allowable grinding surface length, set according to the differences in grinding angles. Profile error threshold and smoothness threshold .

[0055] In some specific embodiments, the template is polished. It includes: the first grinding head has an angle of -15° and a grinding power of 30%; the second grinding head has an angle of -10° and a grinding power of 30%; and the third grinding head has an angle of -5° and a grinding power of 30%.

[0056] Step 2: Evaluate the template capability of each predefined grinding template based on the grinding template parameters, and obtain the search range for the number of grinding passes based on the capability evaluation results.

[0057] The template capability assessment specifically includes: simulating the grinding effect of each predefined template on a standard rail specimen, calculating the actual effective removal capacity of a single grinding pass; statistically analyzing the actual effective removal capacity of all predefined templates and calculating their average value; multiplying the average value by a preset safety factor to obtain a grinding capability benchmark value; calculating the estimated number of grinding passes based on the area of ​​the profile deviation between the measured profile and the target profile and the grinding capability benchmark value; and setting a minimum search pass and a maximum search pass, centered on the estimated number of grinding passes, to form the search range of the bundle search algorithm.

[0058] This embodiment first obtains the search range for the number of polishing passes through template capability assessment to avoid invalid searches, and then uses a bundle search algorithm to quickly find the optimal candidate template combination sequence. At the same time, with "minimum number of polishing passes" as one of the optimization objectives, the number of polishing passes is significantly reduced while meeting quality requirements. Compared with the traditional method of manually designing polishing schemes, this significantly shortens the scheme design cycle and improves the overall efficiency of on-site polishing construction.

[0059] Step 3: Using the beam search algorithm, starting from the measured profile and aiming to approximate the target profile, candidate template combination sequences are selected from the polishing template within the search range.

[0060] The beam search algorithm uses the combination sequence of the polishing templates as search variables to construct a comprehensive objective function. Wherein, error is the root mean square error of the normal direction, smoothness is the average curvature of the profile, and step is the number of polishing passes. , , These are the preset weighting coefficients.

[0061] The beam search algorithm specifically includes:

[0062] An initial search state is created, wherein the template sequence of the initial search state is empty, the profile is the measured profile, and the error is the initial root mean square error of the normal between the measured profile and the target profile (see [reference]). Figure 2 The cost of the initial search state is calculated based on the comprehensive objective function; the initial search state is added to the bundle set and a bundle width parameter K is preset, wherein the bundle width parameter K represents the number of search states retained in each iteration;

[0063] For each search state in the bundle set, all predefined polishing templates are used sequentially. The new candidate search states generated by each polishing template are counted, and the profile, error, cost, and corresponding template sequence of each new candidate state are calculated. All newly generated candidate states are collected, and inferior candidates whose error increases by more than a preset threshold compared to the parent state are removed. The remaining candidate states are sorted from low to high cost, and the top K are selected as the bundle set for the next round. The search state with the lowest cost in the current bundle set is recorded and compared with the historical global optimal solution. If the cost is lower, the global optimal solution is updated.

[0064] Determine whether the preset termination condition is met; if so, stop the iteration. The termination condition includes at least one of the following: reaching the maximum number of search iterations, the error of the global optimal solution being lower than the profile error threshold, no update of the global optimal solution for multiple consecutive rounds, or the calculation time exceeding the preset duration.

[0065] Output the template combination sequence corresponding to the global optimal solution, and output the top M candidate template sequences with the lowest cost, along with their corresponding errors, smoothness, and number of polishing passes. M is the preset number of output candidates, and M < K, where 10 ≤ M ≤ 15.

[0066] In algorithm initialization, search efficiency can be improved through heuristic sorting, parallel computing, and early pruning.

[0067] This embodiment uses a beam search algorithm to continuously approximate the measured profile with the target profile in order to minimize the error. Furthermore, through comprehensive verification of error, grinding surface length, and smoothness, the error between the ground profile and the target profile is strictly controlled within the engineering allowable threshold. This effectively avoids the profile matching deviation problem caused by experience judgment in traditional grinding methods, and significantly improves the accuracy of rail profile repair.

[0068] Step 4: Perform constraint evaluation on the candidate template combination sequence according to the engineering constraint parameters, and divide the candidate template combination sequence into different categories based on the constraint evaluation results. Perform targeted optimization on the candidate sequences of different categories to obtain a polishing scheme that meets the engineering constraints.

[0069] Based on the constraint evaluation results, the candidate template combination sequence is divided into different categories of schemes, including: Class A schemes, Class B schemes, Class C schemes and Class D schemes.

[0070] Option A is where the error is less than or equal to the profile error threshold. The length of the polished surface is less than or equal to the maximum allowable length of the polished surface. Smoothness ≤ Smoothness threshold It satisfies all engineering constraints.

[0071] Option B is when the error exceeds the profile error threshold. The length of the polished surface is less than or equal to the maximum allowable length of the polished surface. And smoothness ≤ smoothness threshold .

[0072] Category C schemes are those where the error is less than or equal to the profile error threshold. And the length of the polished surface is greater than the maximum allowable length of the polished surface. Or smoothness > smoothness threshold .

[0073] Type D solutions are those where the error simultaneously exceeds the profile error threshold. And the length of the polished surface is greater than the maximum allowable length of the polished surface. Or there may be an error greater than the profile error threshold. And smoothness > smoothness threshold Or there may be an error greater than the profile error threshold. The length of the polished surface is greater than the maximum allowable length of the polished surface. And smoothness > smoothness threshold .

[0074] The targeted optimizations for different types of solutions are as follows:

[0075] Targeted optimization of scheme B includes: using scheme B as the scheme to be optimized, and setting a maximum allowable increase in search depth. Starting from a search depth of 1, the following search process is executed: Each predefined polishing template is sequentially inserted at the end of the current scheme to be optimized, and different adjustable power levels are sequentially selected for the inserted templates to generate multiple candidate schemes. The post-polishing profile error of each candidate scheme is calculated; if an error exists that meets the profile error threshold... If the candidate solutions are not found, then the optimized solution is output. If the error of any of the candidate solutions does not meet the profile error threshold, then the optimized solution is output. Then, candidate solutions are selected according to preset pruning rules: solutions exhibiting excessive wear are removed, solutions with error improvement below the profile error threshold are removed, and the remaining solutions are sorted from highest to lowest error improvement, retaining a preset number of solutions as optimization solutions for the next round of search. The search depth is increased by 1, and the above search process is repeated until a solution that meets the profile error threshold is found. The scheme or search depth reaches the maximum allowed increase in search depth.

[0076] Targeted optimization of solution type C includes:

[0077] When the C-type scheme has a grinding surface length greater than the maximum allowable grinding surface length In this case, add a low-power, wide-angle template grinding step to the C-type scheme, update the C-type scheme, and recalculate the profile error and grinding surface length. If the grinding surface length is less than or equal to the maximum allowable grinding surface length... And the error is less than or equal to the profile error threshold. If the length of the polished surface is greater than the maximum allowable length of the polished surface, then the optimization is complete; Then continue increasing the number of polishing passes with the same parameters until the length of the polished surface is less than or equal to the maximum allowable length of the polished surface. Or error > profile error threshold .

[0078] When the C-type scheme has a smoothness greater than the smoothness threshold In the case of a low-power, uniform angle template grinding process, a new C-type solution is added. After updating the C-type solution, the error and smoothness are recalculated. If the smoothness is less than or equal to the smoothness threshold... And the error is less than or equal to the profile error threshold. If the smoothness is greater than the smoothness threshold, then the optimization is complete; Then continue increasing the number of polishing passes with the same parameters until the smoothness is less than or equal to the smoothness threshold. Or error > profile error threshold .

[0079] The low power refers to a low-level polishing power, and in some specific embodiments, the low power is a polishing power of 10% to 30%. The principle for selecting this power level is that it can make fine adjustments or smooth the profile without significantly changing the error of the already achieved profile.

[0080] In some specific embodiments, the grinding angle range of the wide-angle template is 20°~55°, which can cover the main working area of ​​the rail head. A large area of ​​the profile can be ground in a single grinding, thereby effectively shortening the excessively long grinding surface.

[0081] In some specific embodiments, the uniform angle template also adopts an angle range of 20° to 55°, but its power level is lower, such as 10% power for polishing, in order to slightly smooth the profile, eliminate local abrupt changes, and not change the overall geometric features of the profile.

[0082] Targeted optimization of scheme D includes:

[0083] The length and smoothness constraints of the polished surface are used as optimization objectives, which, together with the error constraint, constitute the second comprehensive objective function. Different templates and grinding power combinations are sequentially inserted into the current scheme to generate candidate schemes. The error, grinding surface length, and smoothness of each candidate scheme are calculated, and the candidate scheme that simultaneously satisfies all three constraints is selected as the optimized scheme output. If no scheme satisfying all three constraints is found in the current search iterations, the search depth is increased to continue searching until a scheme satisfying the constraints is found or the maximum allowed number of search iterations is reached.

[0084] This embodiment evaluates and classifies different types of solutions by considering triple engineering constraints—profile error, grinding surface length, and smoothness—and implements targeted optimization and process adjustments for different quality issues. This ensures that the final grinding solution meets all engineering constraints, solves the problem of traditional grinding quality being affected by human experience and fluctuating greatly, and achieves standardized and stable control of rail grinding quality.

[0085] Step 5: Conduct a comprehensive verification of the grinding scheme from all dimensions. If the verification is successful, it is determined to be a single-sided rail grinding scheme.

[0086] The comprehensive verification includes: verifying whether the root mean square error of the normal direction between the polished profile and the target profile is less than or equal to the profile error threshold. And there is no over-grinding; verify that the length of the grinding surface corresponding to each angle of the rail is less than or equal to the maximum allowable grinding surface length corresponding to that angle. Verify whether the average curvature of the profile after polishing is less than or equal to the preset smoothness threshold. Furthermore, the profile shows no significant fluctuations or abrupt changes. If any of the above verifications fails, the refinement plan will be returned to the constraint assessment and targeted optimization stage for re-optimization.

[0087] Step 6: Based on the single-sided rail grinding scheme, perform double-sided synchronous processing to generate a double-sided synchronous rail grinding scheme.

[0088] The process of generating a dual-sided synchronous rail grinding scheme includes: independently generating a left-side rail grinding scheme and a right-side rail grinding scheme, and recording their template combination sequence and the number of grinding passes on the left and right sides respectively; taking the larger of the left-side and right-side grinding passes as the target number of grinding passes for dual-sided synchronous grinding; for the side with fewer grinding passes, calculating the number of additional grinding passes P = target number of passes - original number of grinding passes on that side; adding P passes of low-power, uniform angle template grinding to the original grinding scheme on that side to obtain the synchronous grinding scheme for that side; merging the synchronous grinding schemes on the left and right sides to generate a dual-sided synchronous grinding scheme table, which specifies the template combination, grinding power, and grinding angle corresponding to each grinding pass.

[0089] In some specific embodiments, the final dual-sided synchronous polishing scheme is shown in Table 1.

[0090] Table 1. Simultaneous Grinding Scheme for Both Sides

[0091]

[0092] The polishing template is , where i is the total number of templates, here There are a total of 8 polishing templates.

[0093] This embodiment ensures the consistency of the rail profile on both sides by synchronously processing the rails on both sides, avoiding the wheel-rail contact imbalance caused by grinding on one side, effectively reducing vibration and noise during train operation; it also avoids secondary adjustments after grinding on one side, further reducing the overall cost of on-site construction.

[0094] Example 2:

[0095] This embodiment uses the grinding and optimization of 60-gauge railway rails as the application scenario.

[0096] Step S1: Collect the measured profile of the rail, the target profile, the predefined grinding template parameters, and the engineering constraint parameters.

[0097] The measured profile of the 60-track component to be ground and the target profile of the 60-track standard component are collected. This embodiment predefines 8 grinding templates. Each template contains 10 grinding heads, with each grinding head having a grinding angle range of -15° to 70° and a grinding power range of 10% to 70%.

[0098] Simultaneously preset engineering constraint parameters: profile error threshold =0.02mm; Smoothness threshold =0.05;

[0099] The maximum allowable grinding surface length is set according to the differences in grinding angle. When the grinding angle is -1.91° to 1.91°, the maximum allowable grinding surface length is set. =10mm; when the grinding angle is -13.01° to -1.91° or 1.91° to 13.01°, set the maximum allowable grinding surface length. =7mm; when the grinding angle is -15° to -13.01° or 13.01° to 70°, set the maximum allowable grinding surface length. =5m.

[0100] right Eight predefined grinding templates were used to simulate the grinding effect on 60-gauge steel rails. The actual effective removal capacity of each template in a single grinding pass was calculated to be 0.03-0.06 mm² / standard unit. The average removal capacity of all templates was calculated to be 0.045 mm² / standard unit. Multiplying this by a safety factor of 1.2, the baseline grinding capacity value was obtained as 0.054 mm² / standard unit.

[0101] Based on the area of ​​the profile deviation between the measured profile and the target profile, the total removal requirement is calculated to be 0.324 mm². Dividing the total removal amount by the sanding capacity benchmark value yields an estimated sanding pass of 6 passes. With 6 passes as the center value, the minimum search pass is set to 3 passes, and the maximum search pass is set to 8 passes.

[0102] Constructing a comprehensive objective function Wherein, error is the root mean square error of the normal direction, smoothness is the average curvature of the profile, and step is the number of polishing passes. , , The preset weighting coefficients are here. = 0.6、 =0.2、 =0.2.

[0103] Algorithm initialization: Create an initial search state with an empty template sequence. The current profile is the measured profile, and the current error is the initial root mean square error of the normal between the measured profile and the target profile, which is 0.15 mm. Calculate the initial cost based on the objective function. Add the initial state to the bundle set and set the bundle width parameter K=20.

[0104] Iterative search: Apply the following to each state in the bundle set in turn. There are 8 polishing templates, each generating a new candidate state, for a total of 160 candidate states. The post-polishing profile, error, cost, and corresponding template sequence are calculated for each new state. All candidate states are collected, and inferior candidates whose error exceeds a preset threshold of 0.05mm compared to their parent state are removed. The remaining candidates are sorted by cost from low to high, and the top 20 are selected as a new bundle set. The state with the lowest cost in the current bundle is recorded, compared with the historical global optimal solution, and updated.

[0105] Termination condition: When the iteration reaches step 6, the error of the global optimal solution drops to 0.018 mm, which is lower than the profile error threshold of 0.02 mm, thus satisfying the termination condition and stopping the iteration.

[0106] Output: The globally optimal template sequence is output as follows. , , , , , It also outputs the top 12 candidate template sequences with the lowest cost, along with the corresponding error, smoothness, and number of polishing passes for each sequence.

[0107] All candidate solutions were evaluated for engineering constraints, with the evaluation dimensions being profile error, polished surface length, and smoothness. The results were then categorized and processed accordingly.

[0108] (1) Class A solution processing: candidate sequence , , , After polishing, the error is 0.017mm ≤ 0.02mm, and the length of the polished surface is 4-8mm ≤ , smoothness 0.045≤ It was determined to be a Class A solution and proceeded directly to the comprehensive verification stage.

[0109] (2) Class B scheme processing: candidate sequence , , , After grinding, the error is 0.028mm > 0.02mm, and the length of the ground surface is 4-8mm ≤ the corresponding angle. The smoothness is 0.04 ≤ 0.05, classifying it as a Class B solution. An iterative deepening search algorithm is used for optimization: starting with an increase of 1 iteration, polishing templates are sequentially inserted at the end of the current solution. Try power levels from 10% to 70%. Search for templates. When using 50% power, the profile error after polishing is reduced to 0.019mm ≤ 0.02mm, meeting the error constraint. The optimized and supplemented […]. The 50% power sequence is connected to the original sequence to form the final polishing sequence. , , , , .

[0110] (3) Class C scheme processing: candidate sequence , , , , After grinding, the error is 0.019mm≤0.02mm, and the smoothness is 0.06>0.05, which is classified as a Class C solution (insufficient smoothness). To improve this solution, a second grinding pass with a low-power (10%), uniform angle (20°-55°) T4 template is added. After updating and recalculating, the smoothness is 0.042≤0.05, and the profile error remains 0.019mm≤0.02mm, thus completing the optimization.

[0111] (4) D-type scheme processing: candidate sequence , , , , , After polishing, the error is 0.025mm > 0.02mm, the smoothness is 0.06 > 0.05, and the polished surface length is 4-8mm ≤ The solution was classified as Class D. The grinding surface length and smoothness constraints were incorporated into the comprehensive objective function, and an iterative deepening search was used for optimization. After two searches, the template combination [ 30% power], [ The polishing effect at 20% power meets three constraints: profile error 0.018mm ≤ ε_error, smoothness 0.042 ≤ The length of the polished surface should be kept 4-8mm or less. The two optimized and supplemented steps are then combined with the original sequence to form the final polishing sequence.

[0112] The optimized schemes were verified: the error between the measured profile after grinding and the target profile was ≤0.02mm and there was no over-grinding; the length of the grinding surface corresponding to each angle of the rail was ≤ the maximum allowable grinding surface length for the corresponding angle. The average curvature of the profile after polishing is ≤0.05 and shows no significant fluctuation. After successful verification, the optimized sequence of type D is determined. , , , , , , , The final grinding plan for the left rail is 8 passes.

[0113] Independently generate the right-side rail grinding scheme and obtain the optimal template sequence as follows: , , , , , The grinding process is to be repeated 7 times. The larger of 8 times on the left and 7 times on the right is taken as the target number of grinding passes for simultaneous grinding on both sides. For the right rail, an additional grinding pass of P=8-7=1 is required. One additional low-power (10%), uniform angle (20°-55°) pass is added to the end of the original grinding plan on the right side. Template polishing. Merge the left and right side schemes to generate a dual-side synchronous polishing scheme table, specifying the template combination, polishing power, and polishing angle corresponding to each polishing pass.

[0114] It should be noted that the specific manner in which each module performs its operation in the apparatus described in the above embodiments has been described in detail in the embodiments of the method, and will not be elaborated here.

[0115] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0116] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for optimizing rail grinding strategies based on grinding templates, characterized in that, include: Collect the measured profile of the rail, the target profile, the predefined grinding template parameters, and the engineering constraint parameters; The grinding template parameters include the predefined grinding angle and grinding power; The engineering constraint parameters include the maximum allowable grinding surface length, the profile error threshold, and the smoothness threshold. Based on the polishing template parameters, the template capability of each predefined polishing template is evaluated, and the search range for the number of polishing passes is obtained based on the capability evaluation results. The template capability assessment includes: simulating the grinding effect of each predefined template on a standard rail specimen, and calculating the actual effective removal capacity of a single grinding operation; The actual effective removal capacity of all predefined templates is statistically analyzed and their average value is calculated; the average value is multiplied by a preset safety factor to obtain the grinding capacity benchmark value; The estimated number of polishing passes is calculated based on the area of ​​the profile deviation between the measured profile and the target profile, and the polishing capability benchmark value. Using the estimated number of polishing passes as the center value, the minimum number of search passes and the maximum number of search passes are set respectively to form the search range of the bundle search algorithm; Using a beam search algorithm, starting from the measured profile and aiming to approximate the target profile, a sequence of candidate template groups is selected from the polishing template within the search range; The beam search algorithm uses the combination sequence of the polishing templates as search variables to construct a comprehensive objective function. ; Wherein, error is the root mean square error in the normal direction, smoothness is the average curvature of the profile, and step is the number of polishing passes. , λ are preset weighting coefficients; The beam search algorithm includes: Create an initial search state and calculate the cost of the initial search state according to the comprehensive objective function; add the initial search state to the bundle set and preset the bundle width parameter K; For each search state in the bundle set, all predefined polishing templates are used sequentially to calculate the profile, error, cost, and corresponding template sequence for each new candidate state; Candidate states are sorted by cost, and K front-beam width parameters are selected as the bundle set for the next round. The search state with the lowest cost in the current bundle set is recorded and compared with the historical global optimal solution. If the cost is lower, the global optimal solution is updated. The candidate template combination sequence is constrained and evaluated according to the engineering constraint parameters. Based on the constraint evaluation results, the candidate template combination sequence is divided into different categories of schemes. Targeted optimization is performed on the candidate sequences of different categories of schemes to obtain a polishing scheme that meets the engineering constraints. Based on the constraint evaluation results, the candidate template combination sequence is divided into different categories of schemes, including: Class A schemes, Class B schemes, Class C schemes and Class D schemes; The targeted optimization includes: for the B-type scheme, increasing the number of polishing passes at the end of the current scheme to be optimized and trying different predefined combinations of polishing templates and polishing power until the error meets the engineering constraints or reaches the maximum allowable increase in search depth; For the C-type solution, add low-power, wide-angle or uniform-angle template grinding to the C-type solution until the length or smoothness of the grinding surface meets the engineering constraints or the error exceeds the corresponding threshold. For scheme D, the grinding surface length constraint, smoothness constraint and error constraint are combined as optimization objectives for a comprehensive search. The grinding scheme was comprehensively verified in all dimensions, and after the verification was passed, it was determined to be a single-sided rail grinding scheme. The comprehensive verification includes: verifying whether the root mean square error of the normal direction between the polished profile and the target profile is less than or equal to the profile error threshold, and whether there is any over-polishing. Verify that the length of the grinding surface corresponding to each angle of the rail is less than or equal to the maximum allowable length of the grinding surface corresponding to that angle; Verify that the average curvature of the profile after polishing is less than or equal to the preset smoothness threshold, and that the profile has no obvious fluctuations or abrupt changes; If any of the above verifications fails, the polishing plan will be returned to the constraint assessment and targeted optimization stage for re-optimization; Based on the single-sided rail grinding scheme, a double-sided synchronous processing is performed to generate a double-sided synchronous rail grinding scheme. The dual-side synchronous processing includes: taking the larger value between the number of polishing passes on the left side and the number of polishing passes on the right side as the target number of polishing passes for dual-side synchronous processing; For the side with fewer grinding passes, supplement with low-power, uniform-angle template grinding to obtain a synchronous grinding solution for that side.

2. The method for optimizing rail grinding strategy based on grinding template according to claim 1, characterized in that, The beam search algorithm specifically includes: An initial search state is created, wherein the template sequence of the initial search state is empty, the profile is the measured profile, and the error is the root mean square error of the initial normal of the measured profile and the target profile. The cost of the initial search state is calculated according to the comprehensive objective function. The initial search state is added to the bundle set and the bundle width parameter K is preset. The bundle width parameter K represents the number of search states retained in each iteration. For each search state in the bundle set, all predefined polishing templates are used sequentially. The new candidate search states generated by each polishing template are counted, and the profile, error, cost, and corresponding template sequence of each new candidate state are calculated. All newly generated candidate states are collected, and inferior candidates whose error increases by more than a preset threshold compared to the parent state are removed. The remaining candidate states are sorted from low to high cost, and the top K are selected as the bundle set for the next round. The search state with the lowest cost in the current bundle set is recorded and compared with the historical global optimal solution. If the cost is lower, the global optimal solution is updated. Determine whether the preset termination condition is met; if so, stop the iteration. The termination condition includes at least one of the following: reaching the maximum number of search iterations, the error of the global optimal solution being lower than the profile error threshold, the global optimal solution not being updated for multiple consecutive rounds, or the computation time exceeding the preset value. Output the template combination sequence corresponding to the global optimal solution, and also output the top M candidate template sequences with the lowest cost, along with their corresponding error, smoothness, and number of polishing passes, where M is the preset number of output candidates, and M < K.

3. The method for optimizing rail grinding strategy based on grinding template according to claim 1, characterized in that, Based on the constraint evaluation results, the candidate template combination sequence is divided into different categories of schemes, including: Class A schemes, Class B schemes, Class C schemes and Class D schemes; The Class A solution meets all engineering constraints: error ≤ profile error threshold, polished surface length ≤ maximum allowable polished surface length, and smoothness ≤ smoothness threshold. The B-type scheme is defined as follows: error > profile error threshold, polished surface length ≤ maximum allowable polished surface length and smoothness ≤ smoothness threshold. The C-type scheme is defined as follows: error ≤ profile error threshold, and polished surface length > maximum allowable polished surface length or smoothness > smoothness threshold. The Class D scheme is defined as follows: error > profile error threshold and polished surface length > maximum allowable polished surface length; or error > profile error threshold and smoothness > smoothness threshold; or error > profile error threshold, polished surface length > maximum allowable polished surface length and smoothness > smoothness threshold.

4. The method for optimizing rail grinding strategy based on grinding template according to claim 3, characterized in that, The B-type solution is subjected to targeted optimization, which includes: Using the aforementioned scheme B as the scheme to be optimized, a maximum allowable increase in search depth is set; Starting with a search depth of 1, perform the following search process: sequentially insert each predefined polishing template at the end of the current solution to be optimized, and sequentially select different adjustable polishing power levels for the inserted templates to generate multiple candidate solutions. Calculate the profile error after polishing for each candidate scheme. If there is a candidate scheme whose error meets the profile error threshold, then output it as the optimized scheme. If the error of all candidate solutions does not meet the profile error threshold, then candidate solutions are selected according to the preset pruning rules: solutions that exhibit over-grinding are removed, solutions with error improvement amounts lower than the profile error threshold are removed, and the remaining solutions are sorted from high to low according to error improvement amounts, and a preset number of solutions are retained as optimization solutions for the next round of search. Increase the search depth by 1 and repeat the above search process until a solution that meets the profile error threshold is found or the search depth reaches the maximum allowable increase in search depth.

5. The method for optimizing rail grinding strategy based on grinding template according to claim 3, characterized in that, Targeted optimization of the C-type scheme includes: When the length of the polished surface in the C-type scheme is greater than the maximum allowable length of the polished surface, add a low-power, wide-angle template polishing pass to the C-type scheme, update the C-type scheme, and recalculate the error and the length of the polished surface. If the length of the polished surface is less than or equal to the maximum allowable length of the polished surface and the error is less than or equal to the profile error threshold, then the optimization is complete. If the length of the polished surface is greater than the maximum allowable length of the polished surface, continue to increase the number of polishing passes with the same parameters until the length of the polished surface is less than or equal to the maximum allowable length of the polished surface or the error is greater than the profile error threshold. When the smoothness of the C-type scheme is greater than the smoothness threshold, add a low-power, uniform angle template grinding pass to the C-type scheme, update the C-type scheme, and recalculate the profile error and smoothness. If the smoothness is less than or equal to the smoothness threshold and the error is less than or equal to the profile error threshold, then the optimization is complete. If the smoothness is greater than the smoothness threshold, continue to increase the number of grinding passes with the same parameters until the smoothness is less than or equal to the smoothness threshold or the error is greater than the profile error threshold.

6. The method for optimizing rail grinding strategy based on grinding template according to claim 3, characterized in that, Targeted optimization of the D-type scheme includes: The length and smoothness constraints of the polished surface are used as optimization objectives, and together with the error constraint, a second comprehensive objective function is constructed. Starting with a search depth of 1, different combinations of templates and polishing power are sequentially inserted into the current solution to generate candidate solutions; Calculate the error, polishing surface length, and smoothness of each candidate solution, and select the candidate solution that simultaneously satisfies all three constraints as the optimized solution output; If no solution satisfying the three constraints is found in the current search iterations, the search depth is increased to continue searching until a solution satisfying the constraints is found or the maximum allowed search depth is reached.

7. The method for optimizing rail grinding strategy based on grinding template according to claim 1, characterized in that, The method for generating synchronous rail grinding on both sides includes: Generate the left rail grinding scheme and the right rail grinding scheme independently, and record their template combination sequence and the number of grinding passes for the left and right sides respectively; Take the larger value between the number of polishing passes on the left and the number of polishing passes on the right as the target number of polishing passes for simultaneous polishing on both sides; For the side with fewer polishing passes, calculate the number of additional polishing passes required, P = target number of passes - original number of polishing passes on that side; By adding P passes of low-power, uniform angle template grinding to the original grinding scheme on this side, a synchronous grinding scheme on this side is obtained. The simultaneous polishing schemes on the left and right sides are merged to generate a dual-side simultaneous polishing scheme table, which clearly specifies the template combination, polishing power and polishing angle corresponding to each polishing pass.

Images