Blasting parameter selection model establishment method and system and parameter selection method

Abstract

The invention discloses a tunnel smooth blasting parameter selection method, which aims at the tunnel back-break and back-break problem and comprises the following steps of: constructing a blasting parameter optimization model by determining the back-break and back-break amount minimization as a target function and taking the tunnel section size, the lithology grade and the geological condition as constraint conditions; an improved genetic algorithm which introduces dynamic constraint, hierarchical coding and multi-stage fitness evaluation is utilized, and optimized blasting parameters under the minimum back break amount are screened out through continuous selection, intersection and mutation operations. Compared with a traditional genetic algorithm, global convergence and parameter practicability are improved, and meanwhile the problems of premature convergence and invalid solutions of the traditional genetic algorithm are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel blasting and computer intelligent algorithm, and in particular to a blasting parameter optimization model establishment method, a blasting parameter selection method and an application thereof. Background Art

[0002] As my country's modernization continues to advance, transportation infrastructure, including highways, railways, and urban rail transit, has become a crucial cornerstone supporting national prosperity and development and strengthening people's well-being. Guided by the "Strong Transportation Nation" strategy, the traditional transportation infrastructure sector is accelerating its deep integration with artificial intelligence. From BIM+AI intelligent construction systems to 5G+edge computing intelligent operations and maintenance platforms, AI technology is reshaping the entire construction lifecycle. In particular, the large-scale application of AI technology in key engineering areas, such as precise exploration under complex geological conditions, health monitoring of ultra-large structures, and early warning of construction safety risks, will not only improve project quality and efficiency but also hold milestone significance in ensuring the implementation of the "New Infrastructure" strategy and promoting the upgrading of the transportation industry.

[0003] According to the "2024 China Tunnel Engineering Yearbook," drill-and-blast still accounts for 68.3% of mountain tunnel construction. This method remains the mainstream construction method, owing to its strong adaptability to surrounding rock grades V-VI and a 30%-50% lower per-shift cost compared to TBM. However, its "instantaneous impact load" (burst-induced stress wave peaks reach 100-300 MPa, with a duration of less than 10 ms) makes controlling the excavation profile a common challenge in the industry. Factors contributing to tunnel over- and under-excavation primarily include geological control, process control, and management factors. Over-excavation increases tunnel slag discharge and lining concrete backfill, creating difficulties for subsequent operations. This leads to excessive consumption of manpower, machinery, and materials, while also impacting project quality and progress. Under-excavation requires additional manpower or machinery for clearing and removal, and supplementary blasting increases the risk of over-excavation, which impacts construction progress and increases tunnel project costs. Severe under- and over-excavation can also negatively impact the acceptance and stability of the tunnel's surrounding rock.

[0004] Therefore, over-excavation and under-excavation will increase construction costs, delay construction schedules, and even affect tunnel operation safety. Therefore, how to accurately determine tunnel smooth blasting parameters to improve tunnel structural safety and subsequent construction efficiency is an urgent problem that needs to be solved. Summary of the Invention

[0005] In response to the shortcomings of the above-mentioned prior art, the present invention proposes a method for selecting parameters for tunnel smooth blasting. The method takes minimizing over-excavation and under-excavation as the objective function and uses tunnel cross-sectional dimensions, lithology grade, and geological conditions as constraints. A dynamic constraint-guided crossover mutation strategy, a hierarchical coding mechanism, and a multi-stage fitness evaluation method are added to the traditional genetic algorithm. The blasting parameters are optimized through continuous selection, crossover, and mutation operations to ultimately obtain the optimal tunnel smooth blasting parameters.

[0006] The technical solution adopted by the present invention comprises the following steps:

[0007] The method includes determining optimization variables, an objective function, and constraints, wherein:

[0008] The optimization variables are the peripheral hole spacing and the thickness of the light explosion layer;

[0009] The constraints are that the error of the actual diameter of the tunnel after blasting is within the allowable error range, the spacing between the peripheral holes is less than the limit value, and the thickness of the smooth blasting layer is less than the maximum allowable thickness;

[0010] The objective function is to minimize the over-excavation and under-excavation after smooth blasting of the tunnel.

[0011] Therefore, the present invention adopts the above-mentioned method for selecting parameters for tunnel smooth blasting, which has the following beneficial effects:

[0012] First, this invention builds upon the traditional genetic algorithm framework by incorporating dynamic constraint embedding, dynamic repair, and multi-stage evaluation, optimizing blasting parameters through continuous selection, crossover, and mutation operations. Compared to traditional genetic algorithms, this optimized algorithm improves global convergence and parameter practicality while avoiding the problems of premature convergence and invalid solutions often encountered in traditional genetic algorithms.

[0013] Second, the present invention adopts a hierarchical hybrid coding mechanism to distinguish primary and secondary variables through coding. The peripheral hole spacing and the thickness of the light explosion layer are determined as core control parameters, and the charge density and the detonation sequence are determined as auxiliary parameters. The core parameters with a significant impact on the objective function are optimized first and then directly embedded into the feasible interval through a dynamic constraint embedding mechanism, thereby reducing invalid searches and saving time.

[0014] Third, the present invention uses a multi-stage "rough screening-precision" evaluation mechanism to perform targeted repairs on individuals that violate constraints during genetic operations such as crossover and mutation, ensuring that the population always evolves within the feasible domain. Furthermore, during fitness calculations, the constraint violation degree is weighted to the objective function through a penalty function, forcing the algorithm to prioritize meeting engineering requirements and only generate parameter combinations that meet the current lithology.

[0015] Fourth, the present invention performs neighborhood perturbations on the selected optimal individuals to generate local offspring to replace individuals with low fitness in the population. A certain degree of randomness is retained through random perturbations, which prevents the population from falling into a single solution due to forced correction, thereby improving the convergence accuracy and solving the problems of premature convergence and insufficient practicality of solution sets in traditional genetic algorithms.

[0016] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The flowchart of the method proposed in the present invention is shown in FIG.

[0018] Figure 2 This is the flowchart of the traditional genetic algorithm.

[0019] Figure 3 This is a flow chart of the improved genetic algorithm of the present invention.

[0020] Figure 4 ANSYS simulation results in the embodiment.

[0021] Figure 5 The figures are the comparison results of the present invention, SGA algorithm and PSO algorithm in the embodiments.

[0022] Figure 6 The relationship between the optimal blasting parameters and fitness in the embodiment Figure 3 dimensional distribution graph. DETAILED DESCRIPTION

[0023] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art will make various changes or modifications to the present invention, and these equivalent forms also fall within the scope defined by the appended claims of the application.

[0024] like Figure 1 As shown, the tunnel smooth blasting parameter selection method of the present invention includes the following steps:

[0025] 1. Construct blasting parameter optimization model

[0026] (1) Determine the objective function

[0027] The most direct evaluation index for judging whether the selected tunnel blasting parameters are reasonable is the smooth blasting effect. The evaluation index of the smooth blasting effect includes multiple factors such as the rock blockiness after blasting and the amount of over-excavation and under-excavation. Among them, the amount of over-excavation and under-excavation can directly reflect the degree of deviation between the blasting profile and the designed section. At the same time, it directly affects the safety of the tunnel structure, the support cost and the subsequent construction efficiency. It is the core index for measuring the excavation quality and can be used to quantify the blasting effect and guide the direction of parameter optimization. Therefore, the present invention selects the amount of over-excavation and under-excavation as the evaluation index. Based on this evaluation index, the optimization goal of the objective function is determined to minimize the amount of over-excavation and under-excavation after the smooth blasting of the tunnel. The amount of over-excavation and under-excavation is the overall deviation between the actual tunnel profile after blasting and the designed profile. Therefore, the objective function is defined as:

[0028]

[0029] Where E represents the sum of over-excavation and under-excavation, h i is the elevation of the ith measured contour point after blasting, H is the design elevation, and n represents the total number of measuring points in the tunnel section;

[0030] (2) Determine the constraints

[0031] Before determining the constraints, a detailed analysis of the various factors influencing the tunnel blasting process is required. First, the dimensional error after tunnel blasting cannot be too large or too small, otherwise it will affect structural safety or increase repair costs. For example, excessive dimensional error may prevent the subsequent installation of support structures or require additional concrete filling, which is both expensive and delays the construction period. Second, the hardness of different rocks directly affects the selection of drill hole spacing. For example, hard rock can withstand larger drill hole spacing, while soft rock requires more dense drilling, otherwise it will easily become uneven after blasting. Finally, when blasting in areas with poor geological conditions, due to the inherent instability of the rock in these areas, the thickness of each blast must be reduced to avoid excessive blasting thickness that may cause landslides. Therefore, the present invention selects the peripheral hole spacing and the thickness of the light blasting layer as the constraints of the objective function.

[0032] Before designing the blasting parameter plan, the tunnel section size is determined first, and then the surrounding hole spacing and the thickness of the smooth blasting layer are constrained based on the section size.

[0033] The allowable error range of the actual diameter of the tunnel after blasting is:

[0034] D 设计 ×(1-η)≤D 实际 ≤D 设计 ×(1+η)

[0035] Where D 实际 is the actual diameter of the tunnel after blasting, D 设计is the design diameter of the tunnel, η is the allowable error ratio, and in the embodiment of the present invention, η is set to 5%.

[0036] ① Peripheral hole spacing constraints

[0037] Since the spacing between peripheral holes is affected by the lithology grade, the surrounding rock hardness is first quantified by the Proctor coefficient. The maximum allowable value of the peripheral hole spacing increases with the increase of the Proctor coefficient. Therefore, the range of the peripheral hole spacing is dynamically limited according to the Proctor coefficient value as follows:

[0038] S min ≤S≤S max (f)

[0039] Where, f is the Proctor coefficient, which is obtained through geological exploration, and f < 3 indicates soft rock, 3 ≤ f < 5 indicates medium-hard rock, and f ≥ 5 indicates hard rock; S is the spacing between the surrounding holes; S min is the minimum value of the peripheral hole spacing; S max (f) is the maximum value of the peripheral hole spacing, which can be determined by the Proctor coefficient; S min Determined by the minimum drilling construction requirements, refer to engineering experience, S min Set to 0.3D 设计 ;

[0040] The above method of dynamically adjusting the peripheral hole spacing according to the Proctor coefficient can avoid collapse caused by excessive peripheral hole spacing in soft rock, or waste of resources caused by excessive peripheral hole spacing in hard rock.

[0041] in:

[0042] S min =0.3D 设计

[0043] S max =k·D 设计

[0044] Where k is the correction coefficient, k = 0.4 + 0.1 × (f-1);

[0045] Based on engineering specifications and historical data, a mapping relationship table of lithology, Proctor coefficient, and peripheral hole spacing is constructed, namely the lithology-parameter mapping rules shown in Table 1.

[0046] Table 1 Lithology-parameter mapping rules

[0047]

[0048] ② Constraints on the thickness of the light explosion layer under different geological conditions

[0049] Under conventional geological conditions, the maximum allowable thickness is determined based on the actual diameter D and lithology grade of the tunnel before blasting. Conventional geological conditions specifically refer to stable geological areas with intact rock mass and no obvious adverse geological structures (such as faults, fracture zones, karst caves, etc.);

[0050] The maximum permissible thickness is determined to be:

[0051] W max =(0.05~0.12)·D

[0052] Among them, W max is the maximum allowable thickness of the light explosion layer under conventional geology;

[0053] In fault or fracture zone areas, the thickness of the light explosion layer can be determined during the geological exploration stage before construction through geological radar detection, borehole scanning, CT scanning, etc.:

[0054] W≤β·W max

[0055] Where β is the safety factor, which is set according to the geological risk assessment report, and W is the thickness of the flash layer under the fault or fracture zone.

[0056] For soft rock or high stress areas, the maximum allowable thickness is adjusted according to the following formula in combination with the lithology grade:

[0057]

[0058] Among them, W′ max The maximum allowable thickness of the flash layer in soft rock or high stress areas.

[0059] (3) Establishing a blasting parameter optimization model

[0060] According to the above objective function and constraints, the following blasting parameter optimization model is established:

[0061]

[0062] Where D is the actual diameter of the tunnel before blasting; E is the sum of overbreak and underbreak;

[0063] By defining an objective function and constraints, the present invention transforms practical engineering problems into the mathematical optimization model described above. This model is able to find blasting parameters that minimize over- and under-break while satisfying three constraints: cross-sectional dimensions, peripheral hole spacing, and geological conditions. These blasting parameters include peripheral hole spacing S, smooth blasting layer thickness W, charge density ρ, and detonation sequence T. According to the "Technical Specifications for Tunnel Smooth Blasting," peripheral hole spacing S and smooth blasting layer thickness W are core parameters that significantly influence the objective function, while charge density ρ and detonation sequence T are auxiliary parameters. Because auxiliary parameters have a minimal impact on the blasting effect, they are often discrete options or empirical values, meaning they are typically fixed values, often based on empirical values.

[0064] 2. Using dynamic constraint-guided genetic algorithm to solve the blasting parameter optimization model

[0065] Genetic algorithm is a heuristic optimization algorithm used to solve optimization problems. It searches for the global optimal solution in the solution space by simulating mechanisms such as natural selection, crossover and mutation. Figure 2 The flowchart of the traditional genetic algorithm shows that it mainly includes chromosome encoding (generating the initial population), fitness function (assessing the quality of individuals), genetic operators (selection, crossover, mutation) and operating parameters (determining the termination condition). However, when the population diversity of the traditional genetic algorithm is insufficient, it may converge too early and the computational overhead is high. To address these problems, the present invention optimizes the traditional genetic algorithm based on the traditional genetic algorithm framework. Figure 3 As shown, the present invention solves the blasting parameter optimization model by introducing a dynamic constraint guidance mechanism, multi-stage fitness evaluation and local refined search (the red box is the new module added by the present invention). It can screen out the blasting parameter combination with the smallest over-excavation and under-excavation amount under the premise of satisfying dynamic constraints.

[0066] (1) Genetic algorithm based on dynamic constraint guidance

[0067] Compared with the traditional genetic algorithm, the improved genetic algorithm of the present invention (genetic algorithm based on dynamic constraint guidance) has the following improvements:

[0068] ① Dynamic constraint guidance mechanism

[0069] During tunnel blasting, the tunnel is typically divided into multiple construction sections to better control blasting effectiveness and safety. The goal of this invention is to obtain the optimal blasting parameters for each construction section. Within each construction section, the lithology and geological structure are relatively uniform, with negligible differences. Therefore, the blasting schemes designed for each construction section are identical. However, when local anomalies exist within a section, such as the sudden passage through a fault or an area of ​​lithologic variation during tunnel construction, the initial feasible domain may not cover all situations. Therefore, this invention introduces a dynamic constraint guidance mechanism. During the iterative process of the genetic algorithm, changes in geological conditions are monitored in real time. If changes in geological conditions are detected, the dynamic constraint mechanism dynamically increases or decreases the range of values ​​for the peripheral hole spacing S and the maximum allowable thickness W of the flash layer based on the new geological data.

[0070] Specifically, the present invention flexibly sets the allowable range of blasting parameters (core parameters) based on the physical properties of the surrounding rock (current lithology grade and geological conditions), rather than using fixed values. In the gene generation stage of the improved genetic algorithm of the present invention, the feasible interval of blasting parameters (core parameters) is dynamically set according to the current lithology grade and geological conditions, that is, the feasible interval of the peripheral hole spacing and the thickness of the light blasting layer is set. This process is constraint embedding. For example, if f≥3, it is judged to be hard rock, and the hole spacing is allowed to be increased. Otherwise, it is judged to be soft rock and the hole spacing is reduced. The improved genetic algorithm of the present invention automatically limits the search range of subsequent parameters and only generates parameter combinations that meet the current lithology. By directly integrating engineering constraints such as lithology and geology into the gene generation and mutation process, it is possible to avoid the traditional genetic algorithm from generating a large number of invalid solutions due to random search.

[0071] ②Multi-stage fitness evaluation

[0072] First, individuals with excessive over-excavation and under-excavation are quickly screened and eliminated using a simplified elastic wave model. The elastic wave model is a simplified mechanical model used to quickly estimate the disturbance effect of blasting energy transfer on the surrounding rock. Its core formula is:

[0073] E 预估 =∑(A·e -k·d )

[0074] Among them, E 预估 is the estimated over-excavation and under-excavation amount, A is the charge energy of a single hole, k is the rock mass attenuation coefficient, and d is the distance between the measuring point and the blast source;

[0075] The single-hole charge energy, rock mass attenuation coefficient and distance between the measuring point and the blast source are all set when the construction plan is determined;

[0076] Secondly, in the coarse screening stage, the estimated over-excavation and under-excavation amount E can be quickly calculated based on the simplified elastic wave model mentioned above. 预估 , E 预估 and the over-under excavation threshold E threshold For comparison, if E预估 >E threshold , then the individual is screened out, thereby screening out multiple individuals with values ​​greater than the over-under-excavation threshold, so that they do not need to enter the finite element simulation, which takes longer time. The over-under-excavation threshold E threshold Determined in accordance with the provisions of the Technical Specifications for Highway Tunnel Construction;

[0077] Thirdly, in the actuarial stage, the remaining individuals after the coarse screening stage are simulated by finite element blasting to obtain the accurate over-excavation and under-excavation amount E 精确 ,When conducting finite element blasting simulation, LS-DYNA software or ANSYS software can be used to perform high-precision numerical simulation of the blasting process;

[0078] Finally, based on the obtained accurate over-excavation and under-excavation E 精确 , establish a fitness function calculation formula, through which the individual quality can be evaluated;

[0079] The fitness function combines the precise over-excavation and under-excavation with the constraint violation degree and introduces a dynamic penalty factor. The specific fitness function calculation formula is:

[0080]

[0081] Among them, F is the fitness value, the larger the value, the better the solution; α is the dynamic penalty factor; C is the constraint violation degree;

[0082] In the formula, the dynamic penalty factor α can adjust the weight of the constraint violation C in the fitness calculation, and its expression is:

[0083] α=α0+γ·t

[0084] Wherein, α0 is the initial penalty value, which is set to 0.1 in the embodiment of the present invention; γ is the growth coefficient, which is set to 0.01 in the embodiment of the present invention; t is the current iteration number;

[0085] The calculation formula of constraint violation degree C is:

[0086]

[0087] Among them, P overlimit is the parameter exceeding the limit, S upper is the upper limit of the constraint;

[0088] The parameter over-limit value refers to the difference between the actual value of the parameter and the upper limit of the constraint, and the upper limit of the constraint is the maximum peripheral hole spacing and the maximum allowable thickness of the light explosion layer.

[0089] In summary, the present invention improves calculation efficiency and optimization accuracy through the "rough screening-precision calculation" multi-stage evaluation mechanism.

[0090] ③ Local refined search

[0091] After selecting high-quality individuals through a multi-stage evaluation mechanism, the neighborhood perturbation is performed on the best individual to generate local offspring to replace the low-fitness individuals in the population. After the neighborhood perturbation, crossover, mutation, evaluation, and perturbation are continued until the termination condition is met.

[0092] When the neighborhood is disturbed, based on the dynamic constraints of the previous step, that is, based on the upper limit of the maximum peripheral hole spacing and the upper limit of the maximum allowable thickness of the light explosion layer, the upper limit of the maximum peripheral hole spacing and the upper limit of the maximum allowable thickness of the light explosion layer are adjusted within the range of ±5%, and the new upper limit of the maximum peripheral hole spacing and the maximum allowable thickness of the light explosion layer are obtained. Within the new upper limit of the maximum peripheral hole spacing and the maximum allowable thickness of the light explosion layer, high-quality individuals are re-screened to generate local offspring;

[0093] The present invention reduces invalid searches through constraint embedding, lowers computational costs through multi-stage evaluation, and improves convergence accuracy through local disturbance. It solves the problems of premature convergence and insufficient practicality of solution sets in traditional genetic algorithms, and ultimately outputs a blasting parameter combination that meets the constraints of dynamic geological conditions and minimizes over-excavation and under-excavation.

[0094] (2) Solving the blasting parameter optimization model

[0095] In the initialization stage of the genetic algorithm, its core task is to generate an initial population that meets the engineering constraints, thereby converting actual engineering parameters (such as peripheral hole spacing, light explosion layer thickness, etc.) into genes that can be operated by the genetic algorithm.

[0096] The traditional genetic algorithm initializes the population through chromosome encoding, while the present invention distinguishes between primary and secondary variables during chromosome encoding, giving priority to optimizing the core parameters (primary variables) and then optimizing the auxiliary parameters (secondary variables), and directly embedding the optimized core parameters into the feasible solution interval (such as the peripheral hole spacing does not exceed 60% of the tunnel diameter) to reduce invalid searches and save time. During encoding, it is divided into main chain encoding and auxiliary chain encoding. The two encodings are carried out in parallel, and the core parameter encoding is on the main chain, which specifically adopts real number encoding. At the same time, the auxiliary parameter encoding is on the auxiliary chain, which specifically adopts binary encoding. The specific steps include:

[0097] ① Main chain processing flow

[0098] S1: Use Latin hypercube sampling (LHS) to generate the initial core parameter population;

[0099] First, the range of each core parameter (peripheral hole spacing, light explosion layer thickness) is divided into N subintervals (N is the population size). Then, a value is randomly sampled from each subinterval of each core parameter, ensuring that each subinterval is sampled only once. Finally, the different sampled values ​​are randomly combined to generate N individuals. Each parameter is evenly distributed within its range and no subintervals are repeated.

[0100] Next, during the gene generation phase (i.e., the initial population generation phase), the present invention dynamically sets the feasible interval of the main chain parameters according to the current lithology grade and geological conditions, so that the genes (individuals) screened out are high-quality genes, thereby avoiding the traditional genetic algorithm generating a large number of invalid solutions due to random search;

[0101] According to historical blasting data and similar engineering cases, N is usually set to 50-200. Due to limited computing resources, in order to balance accuracy and speed, N is set to 100 in the embodiment of the present invention.

[0102] S2: Perform crossover and selection operations on the generated initial parameter population to generate preliminary offspring;

[0103] First, simulated binary crossover is used as the crossover operation. By mixing the parent gene values, the offspring gene value is determined. For example, if the parent AS = 5.2m and the parent BS = 5.8m, the average of the two parent genes is taken, and the offspring S = 5.5m.

[0104] Then, through the selection operation, high-quality parents are selected from the current population, the positions of gene crossovers are randomly determined, and some genes are exchanged to generate preliminary offspring. The specific selection operation is to screen the core parameters (S, W) within the determined feasible interval and select the core parameters that are within the feasible interval and close to the middle value of the feasible interval as the high-quality parents.

[0105] S3: Perform mutation operation on the preliminary offspring to obtain the corrected parameters;

[0106] The hard truncation method is used to correct the parameters of the initial generation that exceed the dynamic constraints, and the parameters are directly corrected to the nearest boundary value. For example, taking the core parameter S as an example, if S max =6.0m, and the offspring S=6.3m, then the offspring parameter S=6.3m is directly corrected to S=6.0m;

[0107] This step ensures that all parameter combinations meet engineering safety requirements by forcing out-of-limit parameters back to the current dynamic constraint range, preventing invalid solutions from entering the next generation.

[0108] S4: Perform local refined search to screen out the optimal core parameter combination;

[0109] First, through the above-mentioned multi-stage fitness evaluation, the fitness value is calculated using the above-mentioned fitness function. In each generated generation, the core parameter combination with the smallest exact over-undermining (i.e., the highest fitness value) is selected as the candidate core parameters. Then, by introducing a small random perturbation to the selected candidate core parameters, local offspring are generated in their neighborhood to replace individuals with low fitness.

[0110] When performing neighborhood perturbation, taking the adjustment of the peripheral hole spacing S as an example, the parameters are adjusted using the following formula:

[0111] S new =S 修正后 +δ·N(0,1)

[0112] Among them, δ is the disturbance amplitude, N(0,1) is the standard normal distribution random number, S 修正后 is the peripheral hole spacing after correction using the hard cutoff method, S new is the new peripheral hole spacing after applying neighborhood perturbation;

[0113] For example, based on S=6.0m, press S new =6.0+0.1×random number, we may get S=6.03m;

[0114] Similarly, when adjusting the thickness of the light explosion layer W, a similar formula as above is used, such as W new =W 修正后 +δ·N(0,1), the thickness of the light burst layer W is also disturbed in the neighborhood to replace individuals with low fitness in the original population.

[0115] If the parameter exceeds the specified dynamic constraint interval again after the disturbance, the hard truncation method is used again for secondary correction to correct the parameter to the boundary value until the corrected parameter is within the dynamic constraint interval;

[0116] This dynamic repair method can make the final parameters strictly meet the dynamic constraints, and retain a certain degree of randomness through random perturbations, preventing the population from falling into a single solution due to forced corrections;

[0117] Repeat steps S1-S4 until the parameters converge and the optimal core parameter combination is obtained.

[0118] ②Auxiliary chain processing process

[0119] Since most auxiliary parameters are discrete options or empirical values, that is, they are generally fixed values, the traditional genetic algorithm is still used in the chromosome encoding stage in the auxiliary chain;

[0120] For auxiliary parameters (charge density, detonation sequence), the auxiliary chain uses binary coding to achieve discrete optimization of parameters.

[0121] S1: population initialization;

[0122] First, the number of binary digits is determined according to the number of discrete options of the auxiliary parameter. That is, for an auxiliary parameter with M discrete values, the number of binary digits n of the generated binary code string should satisfy 2 n ≥M;

[0123] Secondly, the population is initialized by randomly generating binary strings, where each segment of the binary string corresponds to a different parameter;

[0124] The binary coding mapping rules corresponding to the detonation sequence and charge density in the present invention are shown in Table 2:

[0125] Table 2 Discrete parameter binary coding settings and mapping rules

[0126]

[0127] S2: Perform a crossover operation and select the parent individuals with higher fitness;

[0128] The high-quality parents are selected through the crossover operation, that is, the parents with higher fitness (also refers to the gene value of the offspring), and some binary fragments are exchanged through single-point crossover. For example, if the charge density parent A = 01 and the parent B = 10, the offspring obtained after crossover is 00;

[0129] S3: Perform mutation operation on the screened individuals;

[0130] Flip a binary code in a binary string, flipping 0 to 1 and 1 to 0, thereby introducing randomness to maintain population diversity;

[0131] S4: binary encoding conversion;

[0132] The binary codes are converted into actual engineering parameters through the mapping rules between the binary codes of the auxiliary parameters and the engineering parameters (Table 2).

[0133] ③ Blasting parameter combination screening

[0134] By combining the actual engineering parameters obtained from the auxiliary chain with the continuous parameters obtained from the main chain, the optimal blasting parameter combination can be obtained;

[0135] After obtaining the optimal blasting parameter combination, the fitness value F is calculated using the fitness function calculation formula mentioned above. The quality of the blasting parameter combination is evaluated based on the size of the fitness value F to ensure that the algorithm can screen out the globally optimal blasting parameter combination while satisfying dynamic constraints.

[0136] ④Termination conditions

[0137] The algorithm is terminated when the optimal fitness change rate ΔF is less than 1% for 10 consecutive generations or when the maximum number of iterations is reached (in this scheme, the maximum number of iterations is set to 100). The final output is the parameter combination with the smallest over-excavation and under-excavation, which is the optimal solution or the approximately optimal solution set.

[0138] Example

[0139] Blasting construction data for the Guangming and Mingyueshan Tunnels were collected. Excess and underbreak measurements were obtained by cross-sectional scanning to measure the tunnel contours after blasting. A total of 443 data sets were collected during the blasting phase. Due to data incompleteness during the blasting process, missing data and data that clearly did not conform to the established data patterns were removed to ensure data integrity and accuracy. A total of 324 data sets were selected. The selected data sets shown in Table 3 were collected for each blasting process.

[0140] Table 3 Original dataset

[0141]

[0142] Before determining the peripheral hole spacing and the thickness of the flash layer, the tunnel section size is determined. The diameter D of this tunnel section is 12m, the surrounding rock grade is Class III, the corresponding Proctor coefficient is 4, and there is a local fault fracture zone. Therefore, the flash layer thickness is limited to W≤0.8W max .

[0143] Based on the tunnel smooth blasting parameter selection method proposed in the present invention, the specific steps of obtaining the optimal blasting parameters include:

[0144] ① Initial population generation: Using Latin Hypercube Sampling (LHS) guided by engineering experience and combined with the distribution of historical blasting data shown in Table 3, 100 initial individuals were generated. Statistics were performed on these 100 initial individuals, and 80% of them met the constraints S ≤ 0.6D = 7.2m and W ≤ 1.2m. The generated initial population is shown in Table 4:

[0145] Table 4 Examples of initial population parameters

[0146]

[0147] A dynamic constraint mechanism is used to further screen individuals in the initial population generated above (including 100 initial individuals), forcing W ≤ 0.8 × 1.2 = 0.96 m and limiting S ≤ 6.0 m. W and S are corrected to the nearest boundary value through hard truncation. The parameters after screening are shown in Table 5:

[0148] Table 5 Example of parameters after dynamic constraint correction

[0149]

[0150] The main chain adopts simulated binary crossover and Gaussian mutation, the auxiliary chain adopts single point crossover and bit flip, and the over-limit parameter is S new =S max -k(f)·δ correction, the individual parameter examples after the cross-mutation operation are shown in Table 6:

[0151] Table 6 Crossover mutation operation examples

[0152]

[0153] ②Multi-stage fitness evaluation: In the coarse screening stage, the elastic wave model is used to calculate E 预估 , eliminate individuals with E>20cm, eliminate 60% of individuals in this stage, and perform finite element blasting simulation on the remaining individuals in the actuarial stage to calculate the accurate E 精确 , the fitness evaluation results are shown in Table 7:

[0154] Table 7 Multi-stage fitness evaluation results (top 5 individuals)

[0155]

[0156] ③Finally, the best individuals of each generation are fine-tuned within the range of S±5% and W±5% to generate offspring. The adjusted individuals are shown in Table 8:

[0157] Table 8 Local refinement search results

[0158]

[0159] Finally, the optimal parameter combination obtained by the improved genetic algorithm in the engineering scenario of a tunnel with a diameter of 12m, surrounding rock grade III, and local fault fracture zone is as follows: peripheral hole spacing S = 5.83m, flash layer thickness W = 0.76m, charge density ρ = 1.0kg / m 3 , detonation time T = 50ms, under this blasting scheme, the over-excavation and under-excavation can be reduced to 7.6cm, and the dynamic constraint conditions are fully met. The optimized parameter combination is simulated by ANSYS, and the simulation results are as follows Figure 4 As shown in Figure 9, the total deviation between the actual blasting profile and the designed cross-section, E = 7.6 cm, is calculated based on profile comparison, significantly below the regulatory limit. Therefore, this can be considered the theoretically optimal blasting plan under current geological conditions. To verify the effectiveness of the improved model, this model was compared with the SGA (standard genetic algorithm) and PSO (particle swarm optimization) models. The experimental comparison indicators and results are shown in Table 9:

[0160] Table 9 Comparison of experimental results

[0161]

[0162]

[0163] From Table 9, we can see that the E=8.2cm of the present invention is 35% lower than that of SGA and 22% lower than that of PSO. This is mainly due to the fact that dynamic constraints avoid invalid solutions and local search improves the accuracy of the algorithm. SGA causes a surge in over-excavation and under-excavation due to parameter over-limit. At the same time, through multi-stage evaluation, the present invention reduces the number of finite element simulations to 40 times / generation, with a total time consumption of 45 minutes. PSO has no coarse screening mechanism and needs to simulate all 100 individuals, which takes longer. Figure 5 The following is a comparison of the convergence curves of the three algorithms. It can be seen from the figure that the blue solid line (DCGA) converges the fastest, with an F value of 0.15 after 35 generations and the lowest over-mining and under-mining; the red dotted line (SGA) converges slowly and the final F value is 0.08, showing premature convergence; the red dotted line (PSO) has mid-term oscillations and the final F value is 0.12. Figure 5 It can be concluded that the improved genetic algorithm (DCGA) is significantly superior to the traditional algorithm in terms of convergence speed and solution quality. Figure 6 This is a three-dimensional distribution diagram of parameters and fitness. The x-axis represents the peripheral hole spacing, the y-axis represents the thickness of the light explosion layer, and the z-axis represents the fitness. The fitness value gradually decreases from yellow to purple. It can be found from the figure that the red marked point is the global optimal solution, which is located in the high fitness area. The fitness of the parameter combination far away from the optimal area is significantly reduced.

[0164] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solutions of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for establishing a blasting parameter selection model, characterized in that: The method includes determining optimization variables, an objective function, and constraints, wherein: The optimization variables are the peripheral hole spacing and the thickness of the light explosion layer; The constraint conditions are that the peripheral hole spacing is less than the maximum peripheral hole spacing and greater than the minimum peripheral hole spacing, and the thickness of the light explosion layer is less than the maximum allowable thickness of the light explosion layer; The objective function is to minimize the amount of over-excavation and under-excavation after smooth surface blasting of the tunnel.

2. Blasting parameter selection system, characterized in that, The system includes a blasting parameter selection module, which includes a blasting parameter selection model. The blasting parameter selection model is constructed using a blasting parameter optimization model establishment method, and the expression of the blasting parameter selection model is: Where E is the sum of over-excavation and under-excavation, h i is the elevation of the measured contour point after blasting, H is the design elevation, f is the Proctor coefficient, S is the spacing between surrounding holes, S min is the minimum value of the peripheral hole spacing, S max (f) is the maximum value of the peripheral hole spacing, W max is the maximum allowable thickness under conventional geology, β is the safety factor, W is the thickness of the flash layer under the fault or fracture zone, and D is the tunnel diameter before tunnel blasting.

3. The blasting parameter selection method is characterized in that: The method comprises: The peripheral hole spacing and the thickness of the light explosion layer are determined as the core parameters, and the charge density and detonation sequence are determined as auxiliary parameters; Taking each core parameter and auxiliary parameter as an individual, the blasting parameter optimization model is solved using an improved genetic algorithm to obtain the core parameter combination and the auxiliary parameter combination. The core parameter combination and the auxiliary parameter combination are combined to obtain the global optimal blasting parameter combination.

4. The blasting parameter selection method according to claim 3, characterized in that: The improved genetic algorithm includes a parallel main chain processing branch and an auxiliary chain processing branch, wherein: On the main chain processing branch, according to the constraints of peripheral hole spacing and smooth blast layer thickness, determine the optimal peripheral hole spacing and smooth blast layer thickness that can minimize the over- and under-excavation under the constraints; In the auxiliary chain processing branch, binary coding is used to discretize the auxiliary parameters and generate the optimal charge density and detonation sequence individuals.

5. The blasting parameter selection method according to claim 4, characterized in that: In the main chain processing branch, based on the constraints of the peripheral hole spacing and the smooth-blast layer thickness, determining the optimal peripheral hole spacing and smooth-blast layer thickness that minimize the over- and under-break amount under the constraints includes: Obtain the surrounding rock grade and geological conditions of the current tunnel construction section, and dynamically determine the first feasible solution interval and the second feasible solution interval for the surrounding hole spacing and the maximum allowable thickness of the flash layer; Generate N initial individuals through Latin hypercube sampling to obtain an initial population of size N; Perform crossover and selection operations on the generated initial parameter population to generate preliminary offspring; In the mutation operation, the parameters of the offspring that are not in the first feasible solution interval or the second feasible solution interval in the preliminary offspring are modified respectively; The revised offspring parameters of each generation are coarsely screened and finely screened, and the optimal individuals are selected as candidate core parameters; Perform neighborhood perturbations on the candidate core parameters of each offspring, and correct the perturbed candidate core parameters; Repeat the above steps until the candidate core parameters converge, and take the candidate core parameters at this time as the optimal core parameter combination.

6. The blasting parameter selection method according to claim 5, characterized in that: In the mutation operation, the parameters of the preliminary offspring that are not in the first feasible solution interval or the second feasible solution interval are modified, including: The child generation parameters that exceed the first feasible solution interval or the second feasible solution interval are hard-truncated and corrected to the boundary value closest to the upper limit value of the first feasible solution interval or the second feasible solution interval to obtain the first corrected child generation parameters; wherein the child generation parameters include the peripheral hole spacing and the light explosion layer thickness.

7. The blasting parameter selection method according to claim 6, characterized in that: The coarse and fine screening of the revised offspring parameters of each generation and the selection of the best individuals as candidate core parameters include: Calculating the estimated over- and under-digging amount of each offspring individual, and comparing the estimated over- and under-digging amount with an over- and under-digging threshold to perform a rough screening, thereby screening out individuals with an over- and under-digging amount greater than the over- and under-digging threshold; The remaining individuals after coarse screening are subjected to finite element blasting simulation to obtain the precise over-excavation and under-excavation. A comprehensive fitness function is established based on the precise over-excavation and under-excavation. The comprehensive fitness values ​​of the remaining individuals are calculated through the comprehensive fitness function, and the p individuals with the highest fitness values ​​are screened out as candidate core parameters.

8. The blasting parameter selection method according to claim 7, characterized in that: The performing neighborhood perturbation on the candidate core parameters of each child generation and correcting the perturbed candidate core parameters includes: Add a small random perturbation to the candidate core parameters generated by each offspring to obtain the perturbed candidate core parameters; The individuals in the perturbation candidate core parameters that exceed the current first feasible solution interval or the second feasible solution interval are hard truncated and corrected again to the boundary value closest to the upper limit value of the first feasible solution interval or the second feasible solution interval.

9. A blasting parameter selection device, characterized in that: include: Parameter determination module, used to determine the peripheral hole spacing and light explosion layer thickness as core parameters, and charge density and detonation sequence as auxiliary parameters; The calculation module takes each core parameter and auxiliary parameter as an individual, uses the improved genetic algorithm to solve the blasting parameter optimization model, and obtains the core parameter combination and the auxiliary parameter combination; The selection module is used to merge the core parameter combination and the auxiliary parameter combination to obtain the global optimal blasting parameter combination.

10. A readable storage medium storing a program or instruction, wherein the program or instruction, when executed by a processor, implements any one of the above methods for establishing a blasting parameter selection model or a blasting parameter selection method.

Images