Modified coal gangue-based ecological restoration block and multi-objective optimization design method

Abstract

The application discloses a multi-objective optimization design method of modified coal gangue-based ecological restoration building blocks, relates to the field of mine ecological restoration and industrial solid waste resource utilization, and comprises the following steps: constructing a mixed formula-process parameter search space containing continuous variables and discrete variables; collecting multiple groups of initial samples in the search space, preparing the building blocks, and testing the performance to obtain an initial experimental data set; training a pre-constructed surrogate model based on the initial experimental data set, and setting a multi-objective optimization function containing at least the strength, pH value, water retention rate and cost of the building blocks; based on the trained surrogate model and a logarithmic expectation improvement acquisition function, iteratively performing candidate formula recommendation, experimental verification and surrogate model updating until a Pareto optimal formula set meeting the multi-objective requirements is obtained. Therefore, the problems of low optimization efficiency, easy falling into local optimum and optimization stagnation of traditional optimization algorithms in a high-dimensional mixed variable formula space are solved.

Description

Technical Field

[0001] The embodiments of this invention relate to the technical field of mine ecological restoration and industrial solid waste resource utilization, and particularly to modified coal gangue-based ecological restoration blocks and multi-objective optimization design methods. Background Technology

[0002] With the rapid development of my country's mining industry, extensive mining activities have created large areas of steep, exposed slopes, which not only damage the original ecological environment of the region but also easily trigger geological disasters such as landslides and debris flows. Ecological restoration of mine slopes has become an important issue in my country's ecological environmental protection field. Meanwhile, coal gangue generated during coal mining is one of the largest industrial solid wastes in my country in terms of annual emissions and cumulative stockpiles. Large stockpiles of coal gangue not only occupy valuable land resources but also pose multiple environmental risks, including dust pollution, leachate pollution, and spontaneous combustion. The high-value, harmless, and large-scale utilization of coal gangue is a common problem that the coal industry urgently needs to solve.

[0003] Currently, ecological restoration technologies for steep mine slopes mainly fall into three categories: First, traditional rigid concrete / masonry slope protection technology. While this technology can achieve stable mechanical support for the slope, the materials are highly alkaline and have a dense structure, making it difficult for plant roots to penetrate, easily leading to green desertification and failing to achieve true ecological revegetation. Second, flexible slope protection technology using ecological bags / vegetation mats. Although this technology can provide a basic environment for plant growth, the materials have low strength, are prone to aging and damage, and have a short service life. In steep slope scenarios, the long-term support effect is poor, and problems such as soil erosion and structural collapse are likely to occur. Third, conventional ecological block slope protection technology. This technology takes into account both structural support and ecological restoration needs and is currently the mainstream development direction for mine slope restoration. However, existing technologies still have many core defects: First, the blocks lack a systematic water retention and drainage structure, resulting in poor water retention capacity during dry seasons, low plant survival rates, and no comprehensive drainage and anti-loss design, which can easily lead to soil saturation and instability after rainfall; Second, existing coal gangue-based ecological blocks generally suffer from the difficulty of controlling the alkalinity of the geopolymer system, and the strongly alkaline environment after hydration cannot meet the needs of plant growth, and the bio-fermentation modification process lacks a standardized quality control system, resulting in poor batch performance stability of the products and failing to meet the requirements of large-scale engineering applications; Third, the construction of blocks relies heavily on large machinery and mortar, which cannot meet the rapid construction needs of steep slopes in mines with complex terrain and inconvenient transportation, resulting in low construction efficiency and high project costs.

[0004] Meanwhile, the formulation development of coal gangue-based eco-building materials is a typical high-dimensional mixed-variable, multi-objective, and multi-constraint optimization problem, requiring simultaneous consideration of multiple mutually restrictive optimization objectives such as material strength, pH neutrality, water retention, low cost, and environmental friendliness. Existing technologies mostly employ traditional trial-and-error methods such as single-factor rotation and orthogonal experimentation, which suffer from low optimization efficiency, high experimental costs, and a tendency to get trapped in local optima. In recent years, Bayesian optimization algorithms have been gradually applied to the field of material formulation development; however, traditional expectation improvement (EI) acquisition functions are prone to numerical underflow, gradient vanishing, and optimization stagnation in the later stages of iteration in high-dimensional spaces, and have poor adaptability to multi-objective, constrained, and parallel experimental scenarios, failing to meet the high-efficiency development requirements of coal gangue-based eco-block formulations. Summary of the Invention

[0005] The core objective of this invention is to overcome the shortcomings of existing technologies, provide modified coal gangue-based ecological restoration blocks and a multi-objective optimization design method, and innovatively construct a Bayesian optimization formula search system based on the LogEI family improvement. This solves the problems of low efficiency, easy getting trapped in local optima, and optimization stagnation in high-dimensional mixed variable formula spaces of traditional optimization algorithms, significantly reducing material research and development costs, shortening the research and development cycle, and achieving synergistic optimization of multiple performance indicators of blocks.

[0006] In a first aspect, embodiments of the present invention provide a multi-objective optimization design method for modified coal gangue-based ecological restoration blocks, comprising:

[0007] Construct a hybrid search space for formulation-process parameters that includes both continuous and discrete variables;

[0008] Multiple initial samples were collected in the search space, and block preparation and performance testing were performed on each initial sample to obtain an initial experimental dataset containing formula-process parameters and their corresponding block performance.

[0009] Based on the initial experimental dataset, a pre-built surrogate model was trained, and a multi-objective optimization function was set that included at least block strength, pH value, water retention rate, and cost.

[0010] Based on the trained surrogate model and the improved acquisition function using logarithmic expectation, the process iteratively executes candidate formulation recommendation, experimental verification, and surrogate model update until a Pareto optimal formulation set that satisfies multiple objectives is obtained.

[0011] Preferably, the logarithmic expectation improved acquisition function is constructed based on the logarithmic expectation improved framework, including:

[0012] The classic expectation-improved EI is reconstructed in the logarithmic field to obtain the basic LogEI acquisition function;

[0013] By integrating constraint processing capabilities into the basic LogEI acquisition function, a LogCEI acquisition function for single-objective constrained scenarios is obtained.

[0014] Log-domain reconstruction and operator smoothing approximation of the classical expected hypervolume improved EHVI are performed to obtain the qLogEHVI acquisition function for multi-objective unconstrained parallel scenarios.

[0015] By combining the multi-objective parallel optimization capability of the qLogEHVI acquisition function with the constraint handling capability of the LogCEI acquisition function, a qLogCEHVI acquisition function for multi-objective constrained parallel scenarios is obtained.

[0016] As a preferred approach, the acquisition function is improved based on the trained surrogate model and logarithmic expectation, and candidate formulation recommendation, experimental verification, and surrogate model update are iteratively performed until a Pareto-optimal formulation set that satisfies multiple objectives is obtained, including:

[0017] A hybrid optimization strategy is used to solve for the maximum value of the qLogCEHVI acquisition function in the multi-objective constrained parallel scenario, generating an experimental batch consisting of multiple candidate formulations;

[0018] The candidate formulation batches were prepared into blocks and their performance was tested. The experimental verification results were added to the initial dataset to complete the dataset update.

[0019] Determine whether the preset termination condition is met. If it is met, stop the iteration and output the Pareto optimal formula set. Otherwise, return to the surrogate model training step based on the updated dataset and continue the iteration.

[0020] The block material formulation and process parameters with the best overall performance were selected from the Pareto optimal formulation set.

[0021] Preferably, the method further includes: preparing modified coal gangue-based modular ecological restoration blocks according to the optimal block material formula and process parameters, wherein the block includes a block body, the block body adopts a shell-core double-layer composite structure, consisting of an outer shell layer and an inner core layer, wherein...

[0022] The shell layer is a high-strength modified coal gangue geopolymer material, and its leachate after hydration and hardening is weakly alkaline and meets the preset strength requirements.

[0023] The core layer is a bio-modified porous water-retaining material made of coal gangue, which is pre-filled with biodegradable plant seed slow-release capsules and has a biomimetic interconnected pore system that includes plant growth pores, capillary water conduction channels and water storage pores.

[0024] The side of the block body is provided with a tenon and mortise locking structure for realizing dry assembly and two-way interlocking between blocks;

[0025] The back of the block body that contacts the slope soil is provided with an inverted filter structure, which is connected to the drainage channel at the block joint to form a drainage and anti-loss system.

[0026] Preferably, the shell layer contains a composite acidic buffer regulator composed of organic acids, inorganic phosphates and humic acids, used to synergistically regulate the alkaline environment of the polymer system.

[0027] Preferably, the bio-modified porous water-retaining material of coal gangue is obtained by solid-state fermentation modification of coal gangue with composite microbial agents, and its leachate is neutral or weakly acidic.

[0028] Preferably, the biodegradable seed slow-release capsule comprises a biodegradable polymer shell, and plant seeds, slow-release fertilizer, and microbial agents encapsulated within the shell.

[0029] Preferably, modified coal gangue-based modular assembled ecological restoration blocks are prepared according to the block material formula and process parameters with optimal comprehensive performance, including the following steps:

[0030] Coal gangue was subjected to calcination modification and bio-fermentation modification pretreatment, and biodegradable plant seed sustained-release capsules were prepared.

[0031] The process employs a step-by-step pressing molding process. First, a shell layer blank is formed. Then, a core layer mixture containing bio-fermented modified coal gangue is filled into the mold, and the seed slow-release capsules are pre-embedded in a preset position. The mold is then closed and pressed to form the final product.

[0032] Preferably, the bio-fermentation modification includes:

[0033] Using coal gangue powder and organic matter as the base material, solid-state fermentation was carried out by inoculating compound microbial agents, and the pH value of the material, the degradation rate of organic matter and the seed germination index were used as quality control indicators to determine the fermentation endpoint.

[0034] Preferably, in the step-by-step pressing process, the seed-release capsule is pre-embedded in the area surrounding the plant growth pores of the core layer.

[0035] Compared with existing technologies, the present invention achieves the following beneficial effects:

[0036] (1) This invention achieves integrated synergy between slope mechanical reinforcement and ecological revegetation. The blocks of this invention adopt a shell-core double-layer composite structure. The outer shell layer meets the structural strength requirements of industry standards and can achieve long-term stable support for steep slopes with a design service life of ≥50 years. The inner core layer provides a neutral and highly water-retaining plant growth environment. Combined with built-in biodegradable plant seed slow-release capsules, it enables the natural germination and growth of endophytic plants, completely solving the industry pain point of "green desert" in traditional concrete slope protection. The natural survival rate of plants is ≥85%.

[0037] (2) This invention constructs a full-gradient water retention, water conduction, drainage and anti-loss system, which greatly improves the plant's stress resistance and slope stability. The three-level biomimetic interconnected pore system inside the block realizes rapid infiltration, uniform conduction and long-term storage of rainfall. During the dry season, it can slowly release water and the water holding period can reach more than 30 days. In areas with an annual rainfall of more than 400 mm, it can enable plants to survive naturally without artificial irrigation. The inverted filter groove structure on the back of the block and the drainage channel of the splicing joint form a drainage network that runs through the entire slope, which can quickly drain the water accumulated inside the slope and prevent soil loss, thereby improving the slope's anti-sliding stability by more than 40%.

[0038] (3) The modular self-locking structure of the present invention enables rapid construction in complex terrain and significantly reduces engineering costs. The cross-locking tenon and mortise structure of the present invention enables dry assembly of blocks without the need for mortar masonry and large construction machinery. Manual labor can quickly construct on steep mine slopes with complex terrain and inconvenient transportation, improving construction efficiency by more than 50% compared with traditional processes. At the same time, the blocks use coal gangue as the main raw material, with a total solid waste content of ≥70%, realizing on-site disposal of mine solid waste and reducing raw material costs by more than 30% compared with traditional concrete blocks.

[0039] (4) The present invention can stably control the pH value of the block leachate in the neutral range suitable for plant growth through a composite pH regulation system of rapid neutralization and long-term buffering, without affecting the hydration process and strength development of the geopolymer. Through standardized quality control of the bio-fermentation process and solidification of molding process parameters, the problem of large batch performance fluctuation of coal gangue-based materials is solved, and the batch performance variation coefficient of the product is ≤8%, which meets the quality requirements of large-scale engineering applications.

[0040] (5) The qLogCEHVI Bayesian optimization system innovatively constructed in this invention completely solves the problems of numerical underflow, gradient vanishing and optimization stagnation of traditional EI-type acquisition functions by reconstructing the logarithmic domain and smoothing the operator. It is suitable for multi-objective, multi-constraint and parallel experimental scenarios in mixed variable formulation space. It can quickly find the Pareto optimal formulation within 15 to 20 iterations and a total of no more than 100 sets of experiments, shortening the material development cycle by more than 60% and significantly reducing the experimental cost. It achieves the optimal balance of multiple objectives such as block strength, pH neutrality, water retention, environmental protection and low cost. Attached Figure Description

[0041] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings. The drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0042] Figure 1 This is a flowchart of the multi-objective optimization design method for modified coal gangue-based ecological restoration blocks provided in this embodiment of the invention;

[0043] Figure 2 This is a three-dimensional schematic diagram of the modified coal gangue-based modular assembly ecological restoration block provided in an embodiment of the present invention;

[0044] Figure 3 This is a cross-sectional structural schematic diagram of the modified coal gangue-based modular assembly ecological restoration block provided in an embodiment of the present invention;

[0045] Figure 4 This is a logical schematic diagram of the multi-objective optimization design method for modified coal gangue-based ecological restoration blocks provided in the embodiments of the present invention;

[0046] Figure 5 This is a flowchart of the preparation method of modified coal gangue-based modular assembly ecological restoration blocks provided in the embodiments of the present invention;

[0047] Among them, 1: shell layer; 2: core layer; 3: inverted filter structure; 4: cross-shaped tenon and mortise locking structure; 5: biomimetic interconnected pore system; 6: drainage channel; 7: plant growth hole. Detailed Implementation

[0048] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0049] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe operations (or steps) as being processed sequentially, many of these operations (or steps) may be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations may be rearranged. The process may be terminated when its operation is completed, but may also have additional steps not included in the figures. The process may correspond to a method, function, procedure, subroutine, subroutine, etc.

[0050] Example 1

[0051] Embodiment 1 of the present invention provides a modified coal gangue-based modular assembly ecological restoration block, with coal gangue solid waste as the main raw material and a total solid waste content of ≥70%.

[0052] like Figure 2 and Figure 3 As shown, the block includes a block body, which adopts a shell-core double-layer composite structure, consisting of an outer shell layer 1 and an inner core layer 2. The block body is a cuboid module with external dimensions of 500mm in length, 250mm in width, and 120mm in height.

[0053] The shell layer 1 is a high-strength modified coal gangue geopolymer material with a thickness of 20mm. Its leachate after hydration and hardening is weakly alkaline with a stable pH value of 7.5~8.5 and meets the preset strength requirements, such as an unconfined compressive strength of ≥35MPa after 28 days.

[0054] The core layer 2 is a bio-modified porous water-retaining material made of coal gangue, which occupies the remaining space inside the block body. The core layer is pre-filled with biodegradable plant seed slow-release capsules, which are evenly embedded around the plant growth holes. The core layer 2 also has a biomimetic interconnected pore system 5.

[0055] The four sides of the block are equipped with a cross-shaped tenon and mortise locking structure 4 to achieve dry assembly and horizontal-vertical bidirectional self-locking;

[0056] The back of the block body that contacts the slope soil is provided with an inverted filter structure 3, which is connected to the drainage channel at the block splice joint to form a drainage and anti-loss system.

[0057] Preferably, the raw materials for the shell layer are formulated as follows, based on parts by weight:

[0058] Pre-calcined modified coal gangue powder: 70-85 parts (coal gangue calcined at 600-800℃ for 2 hours, ground to a specific surface area of ​​400-600 m² / kg); Alkali activator: 10-20 parts (made from water glass with a modulus of 1.2-1.6 and NaOH, with a solid content of 35%-45%); Quartz sand: 10-20 parts (fine aggregate with a particle size of 0.15-0.6 mm); Composite acidic buffer regulator: 0.8-4 parts (made from citric acid, potassium dihydrogen phosphate, and humic acid in a mass ratio of 1:2:5); Polycarboxylate-based high-efficiency water-reducing agent: 0.3-1 parts (water reduction rate ≥30%); Mixing water: 8-15 parts.

[0059] This embodiment addresses the issue of strong alkalinity (initial pH > 12) in the coal gangue geopolymer system after hydration by employing a pH regulation mechanism of rapid neutralization + long-term buffering: citric acid rapidly neutralizes the free alkali in the initial stage of hydration; potassium dihydrogen phosphate reacts with calcium ions in the hydration products to generate stable calcium phosphate salts, continuously consuming hydroxyl groups and providing phosphorus to the plants; humic acid utilizes its carboxyl and phenolic hydroxyl groups to achieve long-term pH buffering. Testing showed that the shell layer had a 28-day compressive strength of 38.2 MPa, the leachate pH was 7.8, and the strength loss rate after 25 freeze-thaw cycles was 12%.

[0060] Preferably, the core layer raw materials are formulated as follows by weight:

[0061] Bio-fermented modified coal gangue powder: 40-60 parts (coal gangue powder fermented in solid state with compound microbial agents, pH 6.5-7.5); Coal gangue ceramsite: 20-30 parts (particle size 2-5mm, bulk density 800-1000kg / m3); Peat soil: 10-20 parts; Slow-release compound fertilizer (14-14-14): 2-5 parts (slow-release period 6-12 months); Modified coal gangue geopolymer binder: 5-10 parts (same as shell layer cementing system); Hydrogen peroxide foaming agent: 0.2-1.0 parts; Polyacrylamide water-retaining agent: 0.1-0.5 parts (water absorption ratio ≥300 times); Mixing water: 15-25 parts.

[0062] The core layer prepared by this formula has a volume porosity of ≥40%, a 24h water retention rate of ≥60%, a pH value of 6.5~7.5 for the leachate, and the heavy metal leaching concentration meets the requirements of GB5085.3-2007 standard, providing a long-term stable neutral, water-retaining, and fertile substrate environment for plant growth.

[0063] Preferably, the core layer contains a three-tiered biomimetic interconnected pore system consisting of "millimeter-level plant growth pores - micrometer-level capillary water-conducting channels - nanometer-level water-retaining pores".

[0064] Primary holes: 7 vertical plant growth holes with a diameter of 20~50mm are reserved in the core layer. 2~4 holes are set in each block, penetrating the core layer to provide space for plant root growth and to allow rainwater to infiltrate quickly.

[0065] Secondary channels: Inside the core layer, interconnected capillary water-conducting channels with a diameter of 10~100μm are formed through the stacking and foaming process of coal gangue ceramsite, so as to achieve uniform horizontal and vertical conduction of water and avoid local water accumulation and drought.

[0066] Tertiary pores: The nanoscale gel pores inside the core layer material and the water-retaining agent form water-storing pores with a diameter of 10~1000nm, which realize the adsorption and slow release of water, providing continuous water supply to plants during the dry season, and the water retention period can reach more than 30 days.

[0067] Preferably, plant seed slow-release capsules are pre-placed during the core layer molding process of the masonry block, with the specific design as follows:

[0068] Capsule shell: Made of biodegradable polylactic acid (PLA) / glycolic acid copolymer, with a thickness of 0.2~0.5mm. It completely degrades in 30~60 days under soil moisture ≥60%, avoiding seedling death caused by premature germination.

[0069] Capsule core components (by weight): Seeds of native plants for mining slopes: 30-50 parts (mixed with Amorpha fruticosa, Lespedeza bicolor, ryegrass, and tall fescue in a 3:3:2:2 ratio; the native varieties can be adjusted for northern / southern mining slopes); Slow-release compound fertilizer: 20-30 parts; Water-retaining agent: 10-20 parts; Rooting powder (naphthaleneacetic acid + indolebutyric acid compound): 1-5 parts; Compound microbial inoculant (nitrogen-fixing bacteria, phosphorus-solubilizing bacteria, potassium-solubilizing bacteria): 5-10 parts; Bentonite filler: balance.

[0070] Pre-installation method: During the block pressing process, capsules are evenly pre-embedded around the plant growth holes in the core layer at a depth of 10-20mm. Each block has 4-8 capsules pre-installed to ensure that the roots can quickly extend to the growth holes and slope soil after seed germination.

[0071] Preferably, the four sides of the block are equipped with a cross-shaped tenon and mortise locking structure 4 to achieve dry assembly and horizontal-vertical bidirectional self-locking:

[0072] The left and top sides of the block are provided with convex cross tenons. The tenon height is 15~20mm and the width is 1 / 3 of the block thickness. The end of the tenon is beveled at 1° to facilitate assembly and alignment.

[0073] The right and bottom sides of the block are provided with concave cross mortises. The mortise size is perfectly matched with the tenon, with a tolerance of ±0.5mm. An elastic sealing gasket is installed in the groove.

[0074] During assembly, the tenons and mortises of adjacent blocks fully interlock, forming a two-way self-locking mechanism in both horizontal and vertical directions. No mortar is required for masonry. The blocks resist slope soil pressure and horizontal thrust by relying on their own weight and locking structure. The anti-slip safety factor is ≥1.5, and the anti-overturning safety factor is ≥1.8.

[0075] Preferably, the back of the block in contact with the slope soil is equipped with an inverted filter groove structure, along with a complete drainage and anti-loss system:

[0076] The inverted filter tank is a horizontally continuous trapezoidal groove with a depth of 10-15mm and a width of 20-30mm. It is set with 2-3 grooves along the height of the block.

[0077] The groove is filled with graded crushed stone (particle size 2~5mm) and non-woven geotextile (200g / m2) to form a filter layer, which enables the rapid drainage of water inside the slope, while preventing the slope soil from being lost with the water flow, and avoiding slope hollowing and landslides.

[0078] A 10mm wide drainage channel 6 is reserved at the vertical splicing joint of the blocks and connected to the inverted filter groove to form a drainage network that runs through the entire slope. After rainfall, the water content of the slope soil can be reduced quickly, thereby improving the stability of the slope.

[0079] Preferably, the blocks are assembled in a stepped staggered joint manner, and are laid layer by layer from bottom to top along the slope surface. The staggered joint length between the upper and lower layers of blocks is ≥1 / 3 of the block length, forming an overall honeycomb / stepped slope protection structure. For steep slopes with a gradient greater than 1:0.75, full-length bonded anchor rods are installed between the block layers. The anchor rods have a diameter of 16~20mm and a spacing of 1.5~2.0m. The anchor rods pass through the reserved holes in the blocks and are anchored into the slope rock mass with an anchoring depth ≥1.0m, thereby improving the overall anti-sliding stability of the structure.

[0080] Based on the above embodiments, the core beneficial effects of the present invention are as follows:

[0081] (1) This invention achieves integrated synergy between slope mechanical reinforcement and ecological revegetation. The blocks of this invention adopt a shell-core double-layer composite structure. The outer shell layer meets the structural strength requirements of industry standards and can achieve long-term stable support for steep slopes with a design service life of ≥50 years. The inner core layer provides a neutral and highly water-retaining plant growth environment. Combined with built-in biodegradable plant seed slow-release capsules, it enables the natural germination and growth of endophytic plants, completely solving the industry pain point of "green desert" in traditional concrete slope protection. The natural survival rate of plants is ≥85%.

[0082] (2) This invention constructs a full-gradient water retention, water conduction, drainage and anti-loss system, which greatly improves the plant's stress resistance and slope stability. The three-level biomimetic interconnected pore system inside the block realizes rapid infiltration, uniform conduction and long-term storage of rainfall. During the dry season, it can slowly release water and the water holding period can reach more than 30 days. In areas with an annual rainfall of more than 400 mm, it can enable plants to survive naturally without artificial irrigation. The inverted filter groove structure on the back of the block and the drainage channel of the splicing joint form a drainage network that runs through the entire slope, which can quickly drain the water accumulated inside the slope and prevent soil loss, thereby improving the slope's anti-sliding stability by more than 40%.

[0083] (3) The modular self-locking structure of the present invention enables rapid construction in complex terrain and significantly reduces engineering costs. The cross-locking tenon and mortise structure of the present invention enables dry assembly of blocks without the need for mortar masonry and large construction machinery. Manual labor can quickly construct on steep mine slopes with complex terrain and inconvenient transportation, improving construction efficiency by more than 50% compared with traditional processes. At the same time, the blocks use coal gangue as the main raw material, with a total solid waste content of ≥70%, realizing on-site disposal of mine solid waste and reducing raw material costs by more than 30% compared with traditional concrete blocks.

[0084] (4) The present invention can stably control the pH value of the block leachate in the neutral range suitable for plant growth through a composite pH regulation system of rapid neutralization and long-term buffering, without affecting the hydration process and strength development of the geopolymer. Through standardized quality control of the bio-fermentation process and solidification of molding process parameters, the problem of large batch performance fluctuation of coal gangue-based materials is solved, and the batch performance variation coefficient of the product is ≤8%, which meets the quality requirements of large-scale engineering applications.

[0085] Example 2

[0086] like Figure 1 The image shows a multi-objective optimization design method 200 for modified coal gangue-based ecological restoration blocks provided in Embodiment 2 of the present invention, used to prepare blocks as described in Embodiment 1. The method includes:

[0087] Step S210: Construct a hybrid search space for formulation-process parameters that includes both continuous and discrete variables.

[0088] Preferably, a d-dimensional mixed variable search space is constructed using the block material formulation and molding process parameters as optimization variables. In this embodiment, the optimization variables include continuous variables such as coal gangue grinding fineness, alkali activator modulus and dosage, bio-fermented coal gangue dosage, foaming agent dosage, molding pressure, and standard curing time; categorical discrete variables such as acid regulator type; and ordered discrete variables such as coal gangue pre-calcination temperature. The value range and coding rules of each variable are shown in Table 1 below:

[0089] Table 1

[0090]

[0091] Step S220: Collect multiple sets of initial samples in the search space, and perform block preparation and performance testing on each set of initial samples to obtain an initial experimental dataset containing formula-process parameters and their corresponding block performance.

[0092] Preferably, in this embodiment, Latin hypercube sampling is used to collect initial samples of group N0=2d (d is the variable dimension) in the optimized variable space, and the performance indicators of the blocks, such as 28-day unconfined compressive strength, leachate pH value, 24-hour water retention rate, and unit raw material cost are tested. At the same time, the satisfaction of constraints such as volume porosity, freeze-thaw strength loss rate, and heavy metal leaching concentration are recorded to form an initial experimental dataset.

[0093] Step S230: Based on the initial experimental dataset, train the pre-built surrogate model and set a multi-objective optimization function that includes at least block strength, pH value, water retention rate and cost.

[0094] Preferably, this embodiment uses a hybrid Gaussian process (GP) with automatic relevance determination (ARD) as a surrogate model to fit the black-box mapping relationship between the formulation-process parameters and the block performance. The specific model design is as follows:

[0095] (1) Input to the surrogate model: a d-dimensional standardized vector of mixed variables ;

[0096] (2) Proxy model output: For each objective function and constraint function, construct a conditionally independent multi-output GP model for any input. Output posterior normal distribution ,in The posterior mean is... The posterior variance characterizes the uncertainty of the model regarding the prediction results;

[0097] (3) Hybrid kernel function design: For the mixed space of continuous and discrete variables, a hybrid kernel function in the form of tensor product is constructed to take into account both the correlation fitting of continuous variables and the class matching characteristics of discrete variables.

[0098] ;

[0099] : Hybrid kernel function, adapted to the continuous + discrete mixed variable search space of the ecological block formula optimization in this invention; : A continuous variable kernel function used to fit the nonlinear relationship between continuous formulation-process parameters such as coal gangue grinding fineness and alkali activator dosage and block performance in this invention. Maternity is employed. Kernel function implementation; : Kernel function for categorical variables; , : No. , Standardized mixed variable vector of block formulation and process parameters; , : No. , The continuous variable vector in the block formulation-process parameters corresponds to the seven continuous optimization variables in this invention. , : No. , The categorical discrete variable vector in the block formulation-process parameters corresponds to the unique thermal encoding vector of the acid regulator type in this invention.

[0100] Among them, the kernel function of continuous variables Using Matern The kernel function achieves accurate fitting of nonlinear relationships between continuous variables, and its expression is:

[0101] ;

[0102] in The Euclidean distance for a vector of continuous variables. For each dimension, there is a length scale parameter (ARD mechanism, where each continuous variable dimension corresponds to an independent length scale, enabling automatic correlation determination). For continuous variables, the signal variance; The Euclidean distance between two sets of continuous variable vectors of block formulations is used to measure the degree of difference between continuous parameters of different formulations. The length scale parameter of the kernel function for continuous variables is used in this invention. The ARD automatic correlation determination mechanism is adopted, and each continuous formula-process variable dimension corresponds to an independent length scale, which is used to automatically determine the influence weight of different optimization variables on the performance of blocks.

[0103] Categorical variable kernel function This invention employs a category matching kernel function to adapt to categorical discrete variables after one-hot encoding. The expression is as follows:

[0104] ;

[0105] in , The class encoding vectors for the two samples, This is an indicator function (value 1 when the category matches, value 0 when there is no match). This is the signal variance hyperparameter for categorical variables; this kernel function is suitable for a small number of categories ( In scenarios involving classes, efficient computation can be achieved; if more categorical variables (such as...) are introduced in the future... To avoid losing potential similarity information between different categories, a kernel function based on Hamming distance can be used instead. ,in The Hamming distance between two class encoding vectors. The length-scale hyperparameter of the Hamming distance can be optimized to automatically learn the similarity relationships between different categories, thereby improving the model's fitting accuracy.

[0106] (4) Hyperparameter optimization: The L-BFGS-B algorithm is used to optimize all hyperparameters of the hybrid kernel function through maximum likelihood estimation (MLE). ,in To observe the noise variance, where, : The set of length scale parameters corresponding to each dimension of the optimization variables in the hybrid kernel function, where d is the total dimension of the optimization variables in this invention. The degree of influence of each variable on the performance of the block is automatically learned through hyperparameter optimization.

[0107] (5) Setting of multi-objective optimization functions and constraints: Based on the core performance requirements of ecological building blocks, four mutually restraining optimization objectives are defined, and the corresponding values ​​are obtained through standardized experimental testing:

[0108] Objective 1 (Maximize): The 28-day unconfined compressive strength of the blocks, in MPa, must meet the requirements of JC / T standard Requirements;

[0109] Objective 2 (Minimization): The pH of the block leachate was reduced by 7 to create a neutral environment suitable for plant growth.

[0110] Objective 3 (Maximization): The water retention rate of the blocks over 24 hours, in units Characterizes the water retention performance of the material;

[0111] Objective 4 (Minimization): This represents the unit cost of raw materials for building blocks, expressed in yuan. .

[0112] Hard constraints: Boundary constraints on the range of values ​​for all optimization variables;

[0113] Black-box constraints: Volume porosity of masonry blocks ; Strength loss rate of blocks after 25 freeze-thaw cycles ; The leaching concentration of heavy metals in the building blocks complies with GB standards.

[0114] Step S240: Based on the trained surrogate model and logarithmic expectation, improve the acquisition function, iteratively perform candidate formulation recommendation, experimental verification and surrogate model update until a Pareto optimal formulation set that meets the multi-objective requirements is obtained.

[0115] Preferably, for multi-objective, constrained, and parallel experimental scenarios, this invention integrates the core idea of ​​logEI logarithmic domain reconstruction and innovatively constructs three types of acquisition functions to fundamentally solve the problems of numerical underflow and gradient vanishing in traditional EI / Expected Hypervolume Improvement (EHVI) acquisition functions:

[0116] (1) Basic analytical LogEI data acquisition function: The classic EI is reconstructed in the logarithmic domain, and the expression is:

[0117] ;

[0118] in, : An improved acquisition function based on the classic single-objective expectation; Mathematical expectation operator; Candidate formulation parameter vector in this invention The corresponding single-objective function value of the block performance is the output of the Gaussian process surrogate model; The currently observed single-objective optimal value of the block performance (incumbent) is the optimal objective value that has been experimentally verified during the formulation optimization process of this invention; Positive part operator, i.e. Only the positive value of the improvement amount of the candidate formulation is retained; if there is no improvement, it is taken as 0. Gaussian process surrogate model for candidate recipe parameter vectors The posterior variance corresponding to the objective function value; The normalized variable of the classic EI acquisition function is obtained by dividing the difference between the posterior mean of the candidate formulation and the current optimal value by the posterior variance, which is used to simplify the calculation of the EI function. The core calculation function of the classic EI acquisition function is obtained by combining the standard normal probability density function and the cumulative distribution function. It is the key intermediate function for calculating the EI value. : Standard normal probability density function, used for The core calculations of the function and the LogEI acquisition function; : Standard normal cumulative distribution function, and To achieve the present invention Calculation of functions; Gaussian process surrogate model for candidate recipe parameter vectors The posterior mean corresponding to the objective function value is the model's prediction of the block performance.

[0119] This invention reconstructs the logarithmic domain of classical EI and proposes a LogEI acquisition function, the expression of which is:

[0120] ;

[0121] in for The numerically stable logarithm calculation is implemented using a three-segment approach to ensure non-zero gradients and numerical stability across the entire input domain.

[0122] ;

[0123] In the formula, , For double-precision floating-point numbers with machine precision for Numerical stability was achieved. This is a scaled complementary error function; The improved acquisition function based on the fundamental analytical logarithmic expectation proposed in this invention reconstructs the logarithmic domain of the classical E1, solving the problems of numerical underflow and gradient vanishing, and is suitable for single-objective unconstrained optimization of block recipes; This invention The constant term of the acquisition function takes the value of , used for interval calculate; This invention The constant term of the acquisition function takes the value of , used for interval calculate; Double-precision floating-point numbers have machine precision and a value of approximately [missing information]. As the present invention The critical value parameters for interval division are calculated using a three-segment method for the data acquisition function; mexp: A numerically stable implementation method is provided to avoid the numerical underflow problem that occurs during direct calculation, and is used in this invention. interval calculate; : Scaling complementary error function, which is the basis of this invention interval Computation provides core mathematical support; :variable The absolute value of the value is used in this invention. interval calculate.

[0124] (2) Innovative Design of LogCEI with Constraints

[0125] For black-box constraint optimization scenarios, this invention extends LogEI and proposes the LogCEI acquisition function, with the expression:

[0126] ;

[0127] in, : Improved acquisition function with constrained logarithmic expectation, integrating constraint processing capabilities on top of LogEI. The total number of black-box constraints in block formulation optimization, in this invention This corresponds to three constraints: block volume porosity, freeze-thaw strength loss, and heavy metal leaching. The index variable of the constraint function, taking values ​​from 1 to... This is used to iterate through all black-box constraint functions; : No. The black-box constraint function takes a vector of candidate formulation parameters as input. Characterization of formulation For the first The satisfaction status of the project constraints; Satisfying constraints): The The posterior probability of satisfying the constraint.

[0128] (3) Innovative design of multi-objective parallel qLogEHVI and qLogCEI

[0129] For multi-objective optimization scenarios, the classic EHVI expression is:

[0130] ;

[0131] in, : Classic expectation hypervolume improved acquisition function; Mathematical expectation operator; : The supervolume index is used in this invention to evaluate the quality of the Pareto front in the multi-objective optimization of block formulations. The larger the index value, the wider the performance space of the high-quality formulations covered by the front. The currently observed Pareto front for multi-objective optimization of block formulations consists of a set of formulation performance indicators that have been experimentally verified. The reference point for multi-objective optimization is the worst feasible value of the four performance indicators of the block in this invention, and it is the benchmark parameter for calculating the hypervolume index. :Depend on Parallel experimental batches composed of candidate formulation parameter vectors, in this invention Pick It is suitable for high-throughput experimental research and development needs of block materials; Candidate formulation batch The corresponding set of multi-objective function values ​​is predicted by the Gaussian process (GP) surrogate model of this invention; : The positive part operator of ReLU, i.e. Only the overvolume positive improvement of the candidate formulation batch is retained, and the value is 0 when there is no improvement.

[0132] This invention improves upon the traditional EHVI by proposing the qLogEHVI acquisition function, with the following specific steps:

[0133] For the ReLU operator The problem of mathematical zero gradient is solved by using the fat-tailed soft-addition function (fatplus) for smooth approximation. The fatplus function is... ( (Take 0.1 to ensure monotonicity, convexity, and polynomial tail properties).

[0134] The fatmax function is used to smooth the approximation of the maximum value operator within a batch, which solves the problem that the gradient is only passed to the maximum value point;

[0135] For all smooth positive terms, a numerically stable Monte Carlo (MC) approximation of the logarithmic field is achieved using the logsumexp operator, and the final expression for qLogEHVI is:

[0136]

[0137] in, Parallel logarithmic hypervolume improved acquisition function; fatplus: fat-tailed soft addition function, a smooth approximation function for the positive part of ReLU operator, solving its mathematical zero gradient problem, and adapting to the gradient optimization requirements of the acquisition function; The mathematical expression for the fat-tailed soft-addition function; : The hyperparameter of the fatplus function, which is set to 0.1 in this invention; fatmax: fat-tailed maximum function; logsumexp: logarithmic summation exponential operator; Monte Carlo (MC) approximation: In this invention, a numerical approximation method is used to randomly sample the posterior distribution of the Gaussian process (GP) to achieve efficient calculation of multi-objective hypervolume improvement; Smoothing temperature parameter, the value taken in this invention This is used to adjust the smoothness of the logsumexp operator to adapt to the gradient optimization of the acquisition function; Candidate formulation batch The first in A vector of block recipe parameters The range of values ​​is ; : No. The first MC sample The hypervolume improvement value of each candidate formulation represents the degree of hypervolume improvement of the formulation to the current Pareto front under a single sampling. The numerically stable logarithm calculation method of the fatplus function is a core component of the qLogEHVI acquisition function of this invention, ensuring the numerical stability of the logarithm domain calculation.

[0138] This invention further integrates constraint processing capabilities and innovatively proposes the qLogCEHVI acquisition function, the expression of which is:

[0139] ;

[0140] This acquisition function can simultaneously support multi-objective collaborative optimization, black-box constraint handling, and parallel batch experiment recommendation, maintaining effective gradient signals throughout the entire search space and completely solving the optimization defects of traditional acquisition functions.

[0141] in, Constrained parallel log-expectation hypervolume improvement acquisition function. The number of candidate formulations in a parallel experimental batch, a value used in this invention. The throughput can be flexibly adjusted according to the actual material experiment. : Index variable for candidate formulation batches, with values ​​ranging from 1 to This is used to iterate through all candidate recipes within a batch; The index variable of the black-box constraint function, taking values ​​from 1 to... , used to traverse all block recipe optimizations of this invention using black-box constraints; The total number of black-box constraints for the optimization of the block formulation in this invention is 3, corresponding to the three constraints of block volume porosity, freeze-thaw cycle strength loss, and heavy metal leaching. The first part of the present invention The black-box constraint function takes the first term as input. Candidate formulation parameter vector Characterizing the effect of this formulation on the first The satisfaction status of the project constraints; : No. The candidate formulation satisfies the first... The posterior probability of a black-box constraint.

[0142] Preferably, the optimization and recommendation of candidate formulations includes the following steps:

[0143] Based on the trained agent model, and using the qLogCEHVI acquisition function constructed above, a hybrid optimization strategy is adopted to solve for the maximum value of the acquisition function;

[0144] The L-BFGS-B algorithm is used for multi-starting-point gradient ascent optimization, and the optimized ordered discrete variables are projected to preset discrete values.

[0145] Categorical discrete variables: For the current 4-class scenario, a local enumeration + Bayesian random search strategy is adopted. Near the optimal solution of the continuous variable in gradient optimization, all class values ​​are enumerated and the combination with the highest collection function value is selected. If more categorical variables or high cardinality discrete variables are introduced in the future, the following two alternative optimization strategies can be adopted: (1) Continuous relaxation method: Relax the categorical discrete variables into continuous probability vectors (such as relaxing the one-hot encoding into real value vectors in the interval [0,1], satisfying the constraint that the sum is 1). After gradient optimization in the continuous space, the discrete class is mapped back through argmax or random sampling. (2) Bayesian random mutation method: Combine the random search strategy of Markov Chain Monte Carlo (MCMC) or Particle Swarm Optimization (PSO) to perform random mutation and local search on the discrete variables near the solution of the continuous variable in gradient optimization, balancing exploration and utilization efficiency.

[0146] Final generation A batch consisting of one optimal candidate formulation and process parameters ( To represent the number of parallel experiments, take... (This is the recommended approach for the next round of experiments.)

[0147] Furthermore, experimental verification and dataset updates: for batches In The candidate schemes were used for block preparation and performance testing to obtain the corresponding performance results and constraint satisfaction. New experimental data were then added to the dataset to obtain the updated dataset. .

[0148] If any of the following termination conditions are met, the iteration stops and the Pareto optimal recipe set is output; otherwise, the iteration continues:

[0149] The number of iterations reaches the preset maximum value. ;

[0150] The cumulative number of experiments has reached the preset limit. ;

[0151] The hypervolume increment at the Pareto front is less than the threshold for three consecutive rounds. .

[0152] Finally, based on the actual engineering requirements, the block material formula and process parameters with the best overall performance were selected from the final Pareto optimal formula set.

[0153] Preferably, to verify the superiority of this method, this embodiment conducts a comparative experiment with the constrained parallel logarithmic expectation hypervolume improvement acquisition function (qLogCEHVI) and the classic expectation improvement acquisition function (EI) commonly used in the field of Bayesian optimization. This verifies the performance advantage of the algorithm of the present invention in the multi-objective optimization scenario of coal gangue-based ecological block formula, and proves that the present invention solves the technical problems of numerical underflow, gradient vanishing, optimization stagnation, and poor adaptability to multiple objectives and constraints of the traditional EI acquisition function.

[0154] Both algorithms use the same basic configuration to eliminate interference from irrelevant variables. The specific parameters are shown in Table 2 below:

[0155] Table 2

[0156]

[0157] Five representative formulations (high-strength group, neutral pH group, high water retention group, low-cost group, and comprehensive optimal group) were selected from the Pareto optimal solution sets after the iterations of the two algorithms. Real block synthesis and performance testing were conducted to verify the authenticity of the simulation results. The mean values ​​of the core indicators from 32 repeated experiments are compared in Table 3 below.

[0158] Table 3

[0159]

[0160] Explanation of key performance differences:

[0161] Convergence and optimization stagnation differences: Traditional EI's HV value growth basically stagnates after 10-12 iterations, resulting in severe optimization stagnation; the qLogCEHVI of this invention maintains continuous HV growth throughout all 20 iterations, reaching the HV level of traditional EI in the 20th iteration by the 10th iteration, more than doubling the convergence speed, and completely solving the optimization stagnation problem caused by gradient vanishing in traditional EI.

[0162] Differences in gradient effectiveness: As the number of iterations increases, traditional EI experiences numerical underflow in the collection function values ​​in most search regions. After 10 iterations, the gradient magnitude of more than 80% of the recommended points is <1e-10, and gradient optimization completely degenerates into random search. The present invention qLogCEHVI, through logarithmic domain reconstruction, ensures that the gradient vanishing rate is always less than 5% throughout the entire iteration process, maintaining an effective gradient signal and significantly improving gradient optimization efficiency.

[0163] Differences in multi-objective and constraint adaptability: Traditional EI transforms a single objective through weighted summation, which cannot effectively balance multiple mutually constraining optimization objectives. Furthermore, the penalty function method is prone to numerical instability of the constraint boundary, resulting in a feasible solution rate of less than 60%. In contrast, the present invention qLogCEHVI directly performs multi-objective optimization based on hypervolume indices. By fusing constraints through logarithmic probability, the feasible solution rate exceeds 90%, and the Pareto front found has a wider coverage, which can simultaneously meet the multi-objective requirements of strength, neutral pH, water retention, and low cost.

[0164] Stability and robustness differences: In 32 repeated experiments, traditional EI experienced optimization stagnation in nearly 60% of the experiments, with large dispersion of results, a coefficient of variation close to 20%, and unstable performance; the qLogCEHVI of this invention did not experience optimization stagnation once, and the coefficient of variation of the results was less than 5%. It can stably find high-quality optimal solutions under different random initial conditions, and its robustness is significantly better than that of traditional EI.

[0165] Parallel scenario adaptability differences: When the parallel batch size q increases from 4 to 8 or 16, the performance of traditional EI decreases significantly, optimization stalls earlier, and the HV value continues to decrease as the batch size increases; the qLogCEHVI of this invention maintains stable high performance under different batch sizes, and the larger the batch size, the more obvious the advantage over traditional EI, perfectly adapting to the parallel R&D needs of high-throughput material experiments.

[0166] Based on the above embodiments, the core beneficial effects of the present invention are as follows:

[0167] (1) This invention achieves integrated synergy between slope mechanical reinforcement and ecological revegetation. The blocks of this invention adopt a shell-core double-layer composite structure. The outer shell layer meets the structural strength requirements of industry standards and can achieve long-term stable support for steep slopes with a design service life of ≥50 years. The inner core layer provides a neutral and highly water-retaining plant growth environment. Combined with built-in biodegradable plant seed slow-release capsules, it enables the natural germination and growth of endophytic plants, completely solving the industry pain point of "green desert" in traditional concrete slope protection. The natural survival rate of plants is ≥85%.

[0168] (2) This invention constructs a full-gradient water retention, water conduction, drainage and anti-loss system, which greatly improves the plant's stress resistance and slope stability. The three-level biomimetic interconnected pore system inside the block realizes rapid infiltration, uniform conduction and long-term storage of rainfall. During the dry season, it can slowly release water and the water holding period can reach more than 30 days. In areas with an annual rainfall of more than 400 mm, it can enable plants to survive naturally without artificial irrigation. The inverted filter groove structure on the back of the block and the drainage channel of the splicing joint form a drainage network that runs through the entire slope, which can quickly drain the water accumulated inside the slope and prevent soil loss, thereby improving the slope's anti-sliding stability by more than 40%.

[0169] (3) The modular self-locking structure of the present invention enables rapid construction in complex terrain and significantly reduces engineering costs. The cross-locking tenon and mortise structure of the present invention enables dry assembly of blocks without the need for mortar masonry and large construction machinery. Manual labor can quickly construct on steep mine slopes with complex terrain and inconvenient transportation, improving construction efficiency by more than 50% compared with traditional processes. At the same time, the blocks use coal gangue as the main raw material, with a total solid waste content of ≥70%, realizing on-site disposal of mine solid waste and reducing raw material costs by more than 30% compared with traditional concrete blocks.

[0170] (4) The present invention can stably control the pH value of the block leachate in the neutral range suitable for plant growth through a composite pH regulation system of rapid neutralization and long-term buffering, without affecting the hydration process and strength development of the geopolymer. Through standardized quality control of the bio-fermentation process and solidification of molding process parameters, the problem of large batch performance fluctuation of coal gangue-based materials is solved, and the batch performance variation coefficient of the product is ≤8%, which meets the quality requirements of large-scale engineering applications.

[0171] (5) The qLogCEHVI Bayesian optimization system innovatively constructed in this invention completely solves the problems of numerical underflow, gradient vanishing and optimization stagnation of traditional EI-type acquisition functions by reconstructing the logarithmic domain and smoothing the operator. It is suitable for multi-objective, multi-constraint and parallel experimental scenarios in mixed variable formulation space. It can quickly find the Pareto optimal formulation within 15 to 20 iterations and a total of no more than 100 sets of experiments, shortening the material development cycle by more than 60% and significantly reducing the experimental cost. It achieves the optimal balance of multiple objectives such as block strength, pH neutrality, water retention, environmental protection and low cost.

[0172] Example 3

[0173] Embodiment 3 of the present invention provides a method for preparing modified coal gangue-based modular assembled ecological restoration blocks, used to prepare blocks as described in Embodiment 1, such as... Figure 5 As shown, the method includes:

[0174] S31: Coal gangue was subjected to calcination modification and bio-fermentation modification pretreatment, and biodegradable plant seed sustained-release capsules were prepared.

[0175] Preferably, the pre-calcination modification of coal gangue includes:

[0176] The raw coal gangue is crushed to a particle size of ≤5mm by a jaw crusher, then transferred to a muffle furnace and calcined at 600~800℃ for 2 hours. After natural cooling, it is ground with a ball mill to a specific surface area of ​​400~600m² / kg to obtain calcined coal gangue powder, which is then sealed and stored for later use.

[0177] Preferably, the bio-fermentation modification of coal gangue (batch stability quality control system) includes the following steps:

[0178] Raw material pretreatment: Crush raw coal gangue powder to a particle size ≤0.15mm, crush corn stalks to 20 mesh, and mix them evenly at a mass ratio of 8:2 to obtain fermentation base material;

[0179] Inoculation with microbial agents: Add compound microbial agents (composed of Bacillus subtilis, Aspergillus niger, Bacillus mucilaginosus, and Azotobacter chrysogenum in a mass ratio of 3:2:2:1, with an effective viable count ≥5×10⁸ CFU / g), the inoculation amount is 2% of the total mass of solid materials, and mix evenly;

[0180] Initial condition control: Add deionized water to adjust the material moisture content to 45%±2%, and use citric acid to adjust the initial pH value to 6.5~7.0;

[0181] Controlled solid-state fermentation: Windlass fermentation is adopted, with a windlass height of 0.8~1.0m and a width of 1.2~1.5m. The fermentation environment temperature is controlled at 20~30℃ and the relative humidity is ≥60%. When the temperature at the center of the windlass rises above 55℃, the windlass is turned over. The windlass is turned over once every 7 days. The high temperature period (≥55℃) is maintained for no less than 10 days in total to achieve the harmlessness and maturity of the material.

[0182] Process quality control and endpoint determination: Samples are taken every 7 days to test the material moisture content, pH value, organic matter degradation rate, and seed germination index. When the material pH value is stable at 6.5~7.5, the organic matter degradation rate is ≥30%, and the seed germination index is ≥80%, it is determined as the fermentation endpoint. The fermentation cycle is 25~30 days.

[0183] Finished product processing: After fermentation, the material is dried at 60℃ and pulverized to a specific surface area of ​​300~500m² / kg to obtain bio-modified coal gangue powder, which is then sealed and stored for later use.

[0184] Preferably, the core components of the plant seed sustained-release capsules are weighed according to the formula, mixed evenly under sterile conditions, and then encapsulated into a biodegradable PLA shell using a capsule filling machine to obtain seed sustained-release capsules with a diameter of 5-8 mm, which are then stored at 4°C for later use.

[0185] S32: A step-by-step pressing molding process is adopted. First, a shell layer blank is formed. Then, the core layer mixture containing bio-fermented modified coal gangue is filled into the mold, and the seed slow-release capsule is pre-embedded in a preset position. The mold is closed and pressure is applied to form the shell layer.

[0186] Preferably, the step-by-step pressing molding process specifically includes the following steps:

[0187] Step S321: Preparation of shell layer slurry

[0188] Weigh the raw materials according to the formula, first add calcined coal gangue powder, quartz sand, composite acid buffer regulator and water reducing agent to a twin-shaft mixer and dry mix for 2 minutes, then add alkali activator and mixing water, and wet mix for 3 minutes to obtain geopolymer slurry with a fluidity of 120~140mm.

[0189] Step S322: Preparation of core layer mixture

[0190] Weigh the raw materials according to the formula, first add the bio-modified coal gangue powder, coal gangue ceramsite, peat soil, slow-release fertilizer, binder, and water-retaining agent to the mixer and dry mix for 3 minutes, then add the mixing water and hydrogen peroxide foaming agent, and stir at low speed for 2 minutes to obtain a porous mixture.

[0191] Step S323: Block pressing and molding

[0192] Using a special steel mold, the shell layer slurry is first evenly injected into the cavity around the mold, and then vibrated to form a shell layer blank with uniform thickness. Then, the core layer mixture is filled into the cavity inside the mold, and seed slow-release capsules are evenly embedded in the preset position. After the mold is closed, it is pressed with a molding pressure of 10~20MPa and held for 30~60s to obtain the green block blank.

[0193] Step S324: Maintenance Treatment

[0194] After the blocks are demolded, they are cured in a standard curing room at a temperature of 20±2℃ and a relative humidity of ≥95% for 28 days; or cured with constant temperature steam at 60℃ for 24 hours and then transferred to a standard curing room for 7 days to ensure that the hydration process is fully completed.

[0195] Step S325: Finished Product Inspection

[0196] After curing, the appearance, dimensions, mortise and tenon structure tolerances, compressive strength, water absorption, permeability and pH value of the blocks are inspected before leaving the factory. Unqualified products are rejected and the final product is obtained after passing the inspection.

[0197] Based on the above embodiments, the core beneficial effects of the present invention are as follows:

[0198] (1) This invention achieves integrated synergy between slope mechanical reinforcement and ecological revegetation. The blocks of this invention adopt a shell-core double-layer composite structure. The outer shell layer meets the structural strength requirements of industry standards and can achieve long-term stable support for steep slopes with a design service life of ≥50 years. The inner core layer provides a neutral and highly water-retaining plant growth environment. Combined with built-in biodegradable plant seed slow-release capsules, it enables the natural germination and growth of endophytic plants, completely solving the industry pain point of "green desert" in traditional concrete slope protection. The natural survival rate of plants is ≥85%.

[0199] (2) This invention constructs a full-gradient water retention, water conduction, drainage and anti-loss system, which greatly improves the plant's stress resistance and slope stability. The three-level biomimetic interconnected pore system inside the block realizes rapid infiltration, uniform conduction and long-term storage of rainfall. During the dry season, it can slowly release water and the water holding period can reach more than 30 days. In areas with an annual rainfall of more than 400 mm, it can enable plants to survive naturally without artificial irrigation. The inverted filter groove structure on the back of the block and the drainage channel of the splicing joint form a drainage network that runs through the entire slope, which can quickly drain the water accumulated inside the slope and prevent soil loss, thereby improving the slope's anti-sliding stability by more than 40%.

[0200] (3) The modular self-locking structure of the present invention enables rapid construction in complex terrain and significantly reduces engineering costs. The cross-locking tenon and mortise structure of the present invention enables dry assembly of blocks without the need for mortar masonry and large construction machinery. Manual labor can quickly construct on steep mine slopes with complex terrain and inconvenient transportation, improving construction efficiency by more than 50% compared with traditional processes. At the same time, the blocks use coal gangue as the main raw material, with a total solid waste content of ≥70%, realizing on-site disposal of mine solid waste and reducing raw material costs by more than 30% compared with traditional concrete blocks.

[0201] (4) The present invention can stably control the pH value of the block leachate in the neutral range suitable for plant growth through a composite pH regulation system of rapid neutralization and long-term buffering, without affecting the hydration process and strength development of the geopolymer. Through standardized quality control of the bio-fermentation process and solidification of molding process parameters, the problem of large batch performance fluctuation of coal gangue-based materials is solved, and the batch performance variation coefficient of the product is ≤8%, which meets the quality requirements of large-scale engineering applications.

[0202] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0203] The above embodiments are merely illustrative examples and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A multi-objective optimization design method for modified coal gangue-based ecological restoration blocks, characterized in that, include: Construct a hybrid search space for formulation-process parameters that includes both continuous and discrete variables; Multiple initial samples were collected in the search space, and block preparation and performance testing were performed on each initial sample to obtain an initial experimental dataset containing formula-process parameters and their corresponding block performance. A pre-built surrogate model is trained based on the initial experimental dataset, and a multi-objective optimization function is set that includes at least block strength, pH value, water retention rate and cost. Based on the trained surrogate model and the improved acquisition function using logarithmic expectation, the process iteratively executes candidate formulation recommendation, experimental verification, and surrogate model update until a Pareto optimal formulation set that satisfies multiple objectives is obtained.

2. The method according to claim 1, characterized in that, The log-expectation improved acquisition function is constructed based on the log-expectation improved framework and includes: The classic expectation-improved EI is reconstructed in the logarithmic field to obtain the basic LogEI acquisition function; By integrating constraint processing capabilities into the basic LogEI acquisition function, a LogCEI acquisition function for single-objective constrained scenarios is obtained. Log-domain reconstruction and operator smoothing approximation of the classical expected hypervolume improved EHVI are performed to obtain the qLogEHVI acquisition function for multi-objective unconstrained parallel scenarios. By combining the multi-objective parallel optimization capability of the qLogEHVI acquisition function with the constraint handling capability of the LogCEI acquisition function, a qLogCEHVI acquisition function for multi-objective constrained parallel scenarios is obtained.

3. The method according to claim 2, characterized in that, Based on the trained surrogate model and the improved acquisition function using log-expectation, candidate formulation recommendation, experimental verification, and surrogate model update are iteratively performed until a Pareto-optimal formulation set satisfying multiple objectives is obtained, including: A hybrid optimization strategy is used to solve for the maximum value of the qLogCEHVI acquisition function in the multi-objective constrained parallel scenario, generating an experimental batch consisting of multiple candidate formulations; The candidate formulation batches were prepared into blocks and their performance was tested. The experimental verification results were added to the initial dataset to complete the dataset update. Determine whether the preset termination condition is met. If it is met, stop the iteration and output the Pareto optimal formula set. Otherwise, return to the surrogate model training step based on the updated dataset and continue the iteration. The block material formulation and process parameters with the best overall performance were selected from the Pareto optimal formulation set.

4. The method according to claim 3, characterized in that, The method further includes: preparing modified coal gangue-based modular ecological restoration blocks according to the optimal block material formula and process parameters, wherein the block includes a block body, the block body adopts a shell-core double-layer composite structure, consisting of an outer shell layer and an inner core layer, wherein... The shell layer is a high-strength modified coal gangue geopolymer material, and its leachate after hydration and hardening is weakly alkaline and meets the preset strength requirements. The core layer is a bio-modified porous water-retaining material made of coal gangue, which is pre-filled with biodegradable plant seed slow-release capsules and has a biomimetic interconnected pore system that includes plant growth pores, capillary water conduction channels and water storage pores. The side of the block body is provided with a tenon and mortise locking structure for realizing dry assembly and two-way interlocking between blocks; The back of the block body that contacts the slope soil is provided with an inverted filter structure, which is connected to the drainage channel at the block joint to form a drainage and anti-loss system.

5. The method according to claim 4, characterized in that, The shell layer contains a composite acidic buffer regulator composed of organic acids, inorganic phosphates and humic acids, used to synergistically regulate the alkaline environment of the polymer system.

6. The method according to claim 4, characterized in that, The bio-modified porous water-retaining material for coal gangue is obtained by solid-state fermentation modification of coal gangue with composite microbial agents, and its leachate is neutral or weakly acidic.

7. The method according to claim 4, characterized in that, The biodegradable seed slow-release capsule comprises a biodegradable polymer shell, and plant seeds, slow-release fertilizer, and microbial agents encapsulated within the shell.

8. The method according to claim 4, characterized in that, Modified coal gangue-based modular assembled ecological restoration blocks are prepared according to the optimal block material formula and process parameters, including the following steps: Coal gangue was subjected to calcination modification and bio-fermentation modification pretreatment, and biodegradable plant seed sustained-release capsules were prepared. The process employs a step-by-step pressing molding process. First, a shell layer blank is formed. Then, a core layer mixture containing bio-fermented modified coal gangue is filled into the mold, and the seed slow-release capsules are pre-embedded in a preset position. The mold is then closed and pressed to form the final product.

9. The method according to claim 8, characterized in that, The bio-fermentation modification includes: Using coal gangue powder and organic matter as the base material, solid-state fermentation was carried out by inoculating compound microbial agents, and the pH value of the material, the degradation rate of organic matter and the seed germination index were used as quality control indicators to determine the fermentation endpoint.

10. The method according to claim 8, characterized in that, In the step-by-step pressing process, the seed-release capsule is pre-embedded in the area surrounding the plant growth pores of the core layer.

Images