Proportion design method and system for making building materials based on ocean dredged mud

Abstract

This application relates to the field of data processing technology and discloses a method and system for designing mix ratios for building materials based on marine dredged mud. The method includes: collecting dredged mud samples, measuring characteristic parameters to form a parameter matrix; calculating the similarity between the parameter matrix and a standard library to determine the appropriate building material type; pre-treating the dredged mud, measuring strength and plotting a curve; constructing a convolutional neural network to predict strength; adjusting mix ratio parameters for iterative prediction to form a strength database; screening the optimal mix ratio, preparing samples for testing, and determining the final mix ratio data. This application rapidly determines the appropriate building material type and optimal mix ratio parameters based on the characteristics of the dredged mud, reducing experimental workload, improving the accuracy and efficiency of mix ratio design, and maximizing the resource utilization of the dredged mud.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method and system for designing a proportion of building materials based on ocean dredged mud. Background Art

[0002] Marine dredged mud is a large amount of bottom mud material generated during the construction of marine engineering projects. Traditionally, it is mainly disposed of by dumping it into the sea or piling it on land, which not only occupies a large amount of land resources, but also may cause environmental pollution. In recent years, with the increasingly stringent environmental protection requirements and the promotion of the concept of resource utilization, the conversion of marine dredged mud into building materials has become a research hotspot. At present, there have been research and application cases at home and abroad on the use of dredged mud for the preparation of building materials such as bricks, expanded clay, cement concrete, and roadbed fillers. These methods usually use traditional orthogonal experimental methods or single-factor experimental methods for ratio design, and determine the optimal ratio through a large number of test sample preparation and performance testing. Some researchers have also tried to use mathematical models such as response surface method to assist in ratio optimization.

[0003] However, existing methods for designing mix ratios for building materials using marine dredged mud have significant shortcomings. First, traditional testing methods require the preparation of a large number of test samples, which is time-consuming, labor-intensive, and inefficient. Second, the physical and chemical properties of dredged mud are complex and variable, with significant differences in characteristics between dredged mud from different sources and treatment methods, making it difficult for existing methods to quickly adapt to such changes. Third, most existing mathematical models only consider a limited number of influencing factors and fail to fully reflect the complex nonlinear relationship between dredged mud properties, mix parameters, and final performance. Fourth, existing methods often focus on optimizing a single performance indicator, lacking comprehensive consideration and balancing of multiple performance indicators. Finally, mix ratio designs in existing technologies are mostly preliminary settings based on experience and specifications. The mix ratio screening process lacks systematicity and specificity, resulting in inefficient resource utilization. Summary of the Invention

[0004] The present application provides a method and system for designing the proportion of building materials based on marine dredged mud, which is used to quickly determine the appropriate type of building materials and the optimal proportion parameters according to the characteristics of the dredged mud, reduce the experimental workload, improve the accuracy and efficiency of the proportion design, and maximize the resource utilization of the dredged mud.

[0005] In a first aspect, the present application provides a method for designing a mix ratio for building materials based on marine dredged mud, the method comprising: collecting marine dredged mud samples, measuring the particle size distribution, moisture content, organic matter content, and heavy metal content of the marine dredged mud samples, and generating a dredged mud characteristic parameter matrix; performing similarity calculation on the dredged mud characteristic parameter matrix and a standard parameter library for building materials to determine the type of building materials suitable for the dredged mud; pre-processing the dredged mud, measuring the compressive strength values ​​at different cement dosages, and plotting a strength-cement dosage curve; constructing a three-layer convolutional neural network structure based on the strength-cement dosage curve, wherein the input layer includes dredged mud parameters and admixture variables, and the output layer is a predicted strength value; gradually adjusting the mix ratio parameters of cement, aggregate, and admixture through the convolutional neural network, recording the strength prediction results of each iteration, and forming a mix ratio-strength database; screening the top three mix ratios with the highest strength from the mix ratio-strength database, preparing physical samples and performing a 28-day strength test, and determining the building material mix ratio data.

[0006] In a second aspect, the present application provides a system for designing a proportion of building materials based on marine dredged mud, the system comprising:

[0007] A collection module is used to collect marine dredged mud samples, measure the particle size distribution, moisture content, organic matter content and heavy metal content of the marine dredged mud samples, and generate a dredged mud characteristic parameter matrix;

[0008] A calculation module, configured to perform similarity calculation between the dredged mud characteristic parameter matrix and a standard parameter library of building materials to determine the type of building materials suitable for the dredged mud;

[0009] a processing module, configured to pre-process the dredged mud, measure the compressive strength values ​​at different cement dosages, and draw a strength-cement dosage curve;

[0010] A construction module is used to construct a three-layer convolutional neural network structure based on the strength-cement dosage curve, wherein the input layer includes dredged mud parameters and admixture variables, and the output layer is the predicted strength value;

[0011] a recording module for gradually adjusting the mix parameters of cement, aggregate, and admixture through the convolutional neural network, recording the strength prediction results of each iteration, and forming a mix ratio-strength database;

[0012] The proportioning module is used to screen the top three proportions with the highest strength from the proportion-strength database, prepare physical samples and conduct 28-day strength tests to determine the proportioning data of building materials.

[0013] A third aspect of the present invention provides a computer device comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the computer device executes the above-mentioned method for designing a proportion of building materials based on marine dredged mud.

[0014] A fourth aspect of the present invention provides a computer-readable storage medium having instructions stored therein, which, when executed on a computer, enables the computer to execute the above-mentioned method for designing a proportion of building materials based on marine dredged mud.

[0015] In the technical solution provided in this application, the key characteristic indicators of dredged mud are comprehensively collected and quantified through a systematic method for generating a matrix of characteristic parameters of dredged mud. Secondly, with the help of similarity calculation technology, intelligent matching of dredged mud and building material types is achieved, avoiding blind trial and error and improving resource allocation efficiency. Thirdly, by constructing a strength-cement dosage curve, the correlation between cement dosage and strength development is revealed, providing directional guidance for mix ratio optimization. Fourthly, a three-layer convolutional neural network structure is applied to process multidimensional feature data, giving full play to the algorithmic advantages of convolutional neural networks in nonlinear mapping and feature extraction, enabling the model to accurately capture the complex relationship between dredged mud characteristics, mix ratio parameters and strength. This deep learning-based mix ratio prediction model has stronger generalization ability and prediction accuracy than traditional statistical methods. Fifthly, through the iterative optimization mechanism of the neural network, efficient adjustment and optimization of the mix ratio parameters are achieved, which greatly reduces the experimental workload and saves time and cost. Sixth, by integrating the preparation of physical samples and the strength test results, a method system combining theoretical prediction with practical verification is established to ensure the reliability and practicality of the mix ratio design results. Especially in the application of three-layer convolutional neural networks, this method fully considers the matching between algorithm characteristics and actual application needs. By rationally designing the network structure and parameters, the model can not only process the high-dimensional characteristic parameters of dredged mud, but also accurately predict the strength performance under different ratios, solving the problem of high-dimensional nonlinear relationship modeling that is difficult to handle with traditional methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 This is a schematic diagram of an embodiment of a method for designing a proportion of building materials based on ocean dredged mud in an embodiment of the present application;

[0018] Figure 2 This is a schematic diagram of an embodiment of a system for designing a proportion of building materials based on ocean dredged mud in an embodiment of the present application;

[0019] Figure 3 It is a schematic block diagram of the structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The present application embodiment provides a kind of ratio design method and system for making building materials based on ocean dredged mud. The terms "first", "second", "third", "fourth" etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments described here can be implemented in an order other than the content illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or equipment comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0021] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the method for designing the proportion of building materials based on ocean dredged mud includes:

[0022] Step S101: Collect marine dredged mud samples, measure the particle size distribution, moisture content, organic matter content, and heavy metal content of the marine dredged mud samples, and generate a dredged mud characteristic parameter matrix;

[0023] Step S102: Calculate the similarity between the dredged mud characteristic parameter matrix and the building material standard parameter library to determine the type of building materials suitable for the dredged mud;

[0024] Step S103: pre-treating the dredged mud, measuring the compressive strength values ​​at different cement dosages, and drawing a strength-cement dosage curve;

[0025] Step S104: Based on the strength-cement dosage curve, a three-layer convolutional neural network structure is constructed, where the input layer contains dredged mud parameters and admixture variables, and the output layer is the predicted strength value;

[0026] Step S105: gradually adjust the mix parameters of cement, aggregate, and admixtures through a convolutional neural network, record the strength prediction results of each iteration, and form a mix ratio-strength database;

[0027] Step S106: Filter the top three mix ratios with the highest strength from the mix ratio-strength database, prepare physical samples and conduct 28-day strength tests to determine the building material mix ratio data.

[0028] It is understandable that the execution subject of this application can be a system for designing a proportion of building materials based on marine dredged mud, or a terminal or a server, which is not limited here. The embodiment of this application is described by taking the server as the execution subject as an example.

[0029] Specifically, marine dredged mud samples were collected from various depths and locations, and the sampling point locations were recorded using a positioning system. Comprehensive physical and chemical properties of the collected dredged mud samples were measured. Particle size distribution was determined using a laser particle size analyzer, yielding D10, D50, and D90 values, representing the particle sizes corresponding to the 10%, 50%, and 90% percentiles on the cumulative volume distribution curve, respectively. Moisture content was determined using a drying-and-weighing method, organic matter content was determined using potassium dichromate oxidation, and heavy metal content was determined using atomic absorption spectrometry. These measurement results were standardized to eliminate dimensional differences between parameters, and a dredged mud characteristic parameter matrix was constructed. For example, a dredged mud sample from a certain sea area was measured to have a D50 value of 0.075 mm, a moisture content of 42%, an organic matter content of 3.8%, and a lead content of 32 mg / kg. These data were organized into rows and columns to form a characteristic parameter matrix, with rows representing different samples and columns representing different characteristic parameters.

[0030] A standard parameter library for building materials, including cement concrete, building blocks, permeable bricks, ceramsite, and anti-seepage materials, was constructed to record the raw material requirements for each type of building material. Key eigenvalues ​​were extracted from the dredged mud characteristic parameter matrix and Euclidean distance calculations were performed against the parameters of each building material type in the standard building material parameter library to generate a similarity score table. Euclidean distance calculation measures the similarity between data sets by comparing the square root of the sum of the differences in each dimension between the two sets of data, with smaller values ​​indicating greater similarity. The values ​​in the similarity score table were normalized, and a hierarchical analysis decision tree was designed to screen building material types with high compatibility. A gradient library for the suitability of dredged mud building materials was established, containing primary and alternative types. For example, after similarity calculations, a dredged mud sample showed a compatibility of 0.86 with ceramsite and 0.72 with permeable bricks, confirming that the dredged mud was primarily suitable for ceramsite production. The dredged mud was centrifugally dehydrated and dried to control the moisture content below 25%. The mineral and chemical composition of pretreated dredged mud was determined using an X-ray fluorescence analyzer, recording the content of calcium, silicon, aluminum, and iron. Using the pretreated dredged mud as the substrate, cement was added in gradients of 5%, 10%, 15%, 20%, 25%, and 30% to prepare standard cubic specimens measuring 100 mm x 100 mm x 100 mm. The standard cubic specimens were cured at a temperature of 20 ± 2°C and a relative humidity of at least 95%. Samples were taken and tested after 3, 7, and 28 days. A pressure tester was used to apply loads to the standard cubic specimens until failure, and the compressive strength values ​​of the specimens at different cement dosages were recorded at different ages. The compressive strength values ​​were paired with the corresponding cement dosage data, and a strength-cement dosage curve was plotted, with cement dosage as the horizontal axis and compressive strength as the vertical axis. This curve visually illustrates the strength trends of dredged mud mixed with different cement ratios, providing a basis for subsequent mix optimization. Data points were extracted from the strength-cement content curve, and the cement content and corresponding compressive strength values ​​were combined to form a training dataset. Key parameters from the dredged mud characteristic parameter matrix were combined to construct a neural network input feature vector, including the dredged mud particle size distribution, moisture content, organic matter content, heavy metal content, and cement content. A three-layer convolutional neural network architecture was designed. The first convolutional layer used 16 3×3 convolutional kernels for feature extraction, the second convolutional layer used 32 3×3 convolutional kernels for feature deepening, and the third convolutional layer used 64 3×3 convolutional kernels for feature integration. The output features of the convolutional layers were mapped through a fully connected layer, connected to a hidden layer of 256 neurons, and then connected to a single output neuron, representing the predicted strength value. The three-layer convolutional neural network was trained using the training dataset, employing a mean squared error loss function and the Adam optimizer to adjust the network weights. The training process yielded a fully trained three-layer convolutional neural network architecture, with the input layer containing dredged mud parameters and admixture variables, and the output layer representing the predicted strength value.

[0031] A mix parameter matrix was constructed, with cement content ranging from 5% to 35%, aggregate ratios ranging from 40% to 70%, and admixture ratios ranging from 5% to 15%. The mix parameter matrix was partitioned into multiple mix points with a step size of 1%, generating a set of candidate mix points. Each mix point contained specific cement, aggregate, and admixture ratios. Mix points were sequentially extracted from the candidate mix set and assembled into an input data vector, combined with dredged mud parameters. The input data vector was passed into a convolutional neural network to obtain the corresponding strength prediction output value, and the correspondence between the mix point and the strength value was recorded. Based on the strength prediction output value, the mix parameters were adjusted, gradually exploring towards high-strength areas, generating new mix points and recording their strength prediction values. All tested mix points and their strength prediction values ​​were organized into a structured data table, indexed by mix number and strength value, to form a mix-strength database. The mix-strength database was sorted in descending order by strength prediction value, and the top three mix parameter groups with the highest strength were extracted and marked as the preferred mix groups. According to the parameter ratios in the preferred mix ratio group, the corresponding weights of pre-treated dredged mud, cement, aggregate and admixture were weighed to prepare three groups of physical test samples. The three groups of physical test samples were cured in a standard environment with a temperature of 20±2°C and a relative humidity of more than 95%, and samples were taken for strength testing after 3 days, 7 days and 28 days. The actual compressive strength values ​​of the three groups of physical test samples at each age were recorded and compared with the strength values ​​predicted by the convolutional neural network, and the prediction error was calculated. Based on the actual compressive strength values ​​and prediction errors, the three groups of mix ratios were comprehensively scored, taking into account the 28-day strength, strength growth rate and prediction accuracy factors. The mix ratio with the highest comprehensive score was selected from the three groups of mix ratios to form a standardized mix ratio table to determine the building material mix ratio data.

[0032] In the embodiments of the present application, a systematic method for generating a matrix of characteristic parameters of dredged mud is used to comprehensively collect and quantify the key characteristic indicators of dredged mud. Secondly, with the help of similarity calculation technology, intelligent matching of dredged mud and building material types is achieved, avoiding blind trial and error and improving resource allocation efficiency. Thirdly, by constructing a strength-cement dosage curve, the correlation between cement dosage and strength development is revealed, providing directional guidance for mix optimization. Fourthly, a three-layer convolutional neural network structure is applied to process multidimensional feature data, giving full play to the algorithmic advantages of convolutional neural networks in nonlinear mapping and feature extraction, enabling the model to accurately capture the complex relationship between dredged mud characteristics, mix parameters and strength. This deep learning-based mix prediction model has stronger generalization ability and prediction accuracy than traditional statistical methods. Fifthly, through the iterative optimization mechanism of the neural network, efficient adjustment and optimization of the mix parameters are achieved, which greatly reduces the experimental workload and saves time and cost. Sixth, by integrating the preparation of physical samples and the strength test results, a method system combining theoretical prediction with practical verification is established to ensure the reliability and practicality of the mix design results. Especially in the application of three-layer convolutional neural networks, this method fully considers the matching between algorithm characteristics and actual application needs. By rationally designing the network structure and parameters, the model can not only process the high-dimensional characteristic parameters of dredged mud, but also accurately predict the strength performance under different ratios, solving the problem of high-dimensional nonlinear relationship modeling that is difficult to handle with traditional methods.

[0033] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0034] (1) Collect marine dredged mud samples at different depths and in different areas, record the location data of the sampling points through a positioning system, and form a sampling record sheet;

[0035] (2) Use a laser particle size analyzer to measure the particle size of marine dredged mud samples, obtain the D10, D50, and D90 particle size values ​​and distribution curves, and record the particle size distribution of marine dredged mud;

[0036] (3) The moisture content of marine dredged mud samples was determined by the drying-weighing method, the organic matter content was determined by the potassium dichromate oxidation method, and the heavy metal content was determined by atomic absorption spectrometry;

[0037] (4) Standardize the particle size distribution, moisture content, organic matter content, and heavy metal content of marine dredged mud to remove unit differences and obtain standardized data;

[0038] (5) Arrange and organize the standardized data according to sample number and parameter type, and construct a data table containing all samples and parameters;

[0039] (6) Identify the main characteristic parameters in the data table through correlation analysis and integrate them to form a dredged mud characteristic parameter matrix.

[0040] Specifically, marine dredged mud samples were collected at different depths and in different areas. A dedicated mud sampler was used to collect samples from the surface (0-30cm), middle (30-60cm), and bottom (60-100cm) layers of the marine dredging project. At least three sample points were collected from each depth area to ensure representative samples. At the same time, the precise coordinates of each sampling point were recorded using a GPS positioning system to form a sampling record sheet containing information such as the sample number, sampling depth, latitude and longitude coordinates of the sampling point, and sampling date.

[0041] After collection, samples were sent to the laboratory for physical and chemical property measurements. Marine dredged mud samples were measured for particle size using a laser particle size analyzer. Samples were pretreated prior to measurement, including organic matter removal and dispersion. The treated samples were then placed in the laser particle size analyzer for measurement. This instrument utilizes the principle of laser diffraction to accurately measure particle size distribution. The D10, D50, and D90 particle size values ​​are obtained, representing the particle sizes corresponding to the 10%, 50%, and 90% points on the cumulative distribution curve, respectively. A complete particle size distribution curve is also recorded, showing the distribution of particles of different sizes within the dredged mud. The moisture content of marine dredged mud samples was determined using the drying-weighing method. The wet mud was first weighed, then dried to a constant weight at 105±5°C, and weighed again. The moisture content was calculated based on the difference in mass before and after drying. Organic matter content was determined using the potassium dichromate oxidation method. The sample was mixed with potassium dichromate and concentrated sulfuric acid, and the organic matter content was calculated based on the amount of potassium dichromate consumed in the oxidation reaction. The determination of heavy metal content is completed by atomic absorption spectrometry. This method first digests the sample and then uses an atomic absorption spectrometer to determine the content of heavy metal elements such as lead, cadmium, chromium, and copper.

[0042] After obtaining the above data, the particle size distribution, moisture content, organic matter content, and heavy metal content of the marine dredged mud were standardized. The purpose of standardization is to eliminate differences in dimensions and numerical ranges between different indicators, making them comparable. Standardization uses the Z-score method: the raw data is subtracted from the mean and divided by the standard deviation to transform it into a standard distribution with a mean of 0 and a standard deviation of 1. For example, for moisture content data, the mean and standard deviation of the moisture content of all samples are calculated. The mean is then subtracted from the moisture content of each sample and divided by the standard deviation to obtain the standardized moisture content value. The same standardization process is performed for other parameters, such as particle size characteristics, organic matter content, and heavy metal content, to obtain standardized data. The standardized data is then organized by sample number and parameter type to construct a data table containing all samples and parameters. The rows of the table represent different samples, and the columns represent different parameters, including standardized D10, D50, and D90 particle size values, moisture content, organic matter content, and various heavy metal contents. This structured data table facilitates subsequent data analysis and processing.

[0043] Correlation analysis identifies key characteristic parameters in a data table. Correlation coefficients are calculated between each parameter to identify groups of parameters with high correlations and key indicators that significantly impact building material performance. For example, analysis may reveal a strong correlation between particle size (D50) and material strength, while certain heavy metal content is closely related to material durability. These key parameters are integrated into a dredged mud characteristic parameter matrix, which serves as a crucial basis for subsequent building material selection and mix design. For example, during a dredging project in a certain sea area, 15 dredged mud samples were collected from various locations and depths. Analysis revealed that: Sample 1 had a D50 value of 0.063 mm (fine silt grade), a moisture content of 48%, an organic matter content of 4.2%, and a lead content of 28 mg / kg; Sample 2 had a D50 value of 0.125 mm (fine sand grade), a moisture content of 35%, an organic matter content of 2.8%, and a lead content of 18 mg / kg. After standardization, all sample data were integrated into a 15-row × 12-column characteristic parameter matrix. Correlation analysis showed that particle size distribution and organic matter content were the most critical factors affecting the subsequent performance of building materials. These parameters were given higher weights to form a dredged mud characteristic parameter matrix.

[0044] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0045] (1) Extract key eigenvalues ​​from the dredged mud characteristic parameter matrix, including average particle size, organic matter content range, and heavy metal index, to form a dredged mud characteristic index set;

[0046] (2) Build a standard parameter library for building materials including cement concrete, building blocks, permeable bricks, ceramsite, and anti-seepage layer materials, and record the physical and chemical index requirements of various building materials;

[0047] (3) Calculate the Euclidean distance between the dredged mud characteristic index set and the parameters of each building material type in the building material standard parameter library to obtain a similarity score table;

[0048] (4) Normalize the values ​​in the similarity score table, assign applicability weights to each building material type, and generate a weight matrix;

[0049] (5) Based on the weight matrix, a hierarchical analysis decision tree is designed to select building materials with high adaptability based on the characteristics of dredged mud, processing difficulty, and material performance requirements;

[0050] (6) Sort the selected building material types by their suitability, establish a gradient library of dredged mud building material applicability, including the main applicable types and alternative types, determine the building material types that are applicable to dredged mud, and retain the alternative scheme data.

[0051] Specifically, key eigenvalues ​​are extracted from the dredged mud characteristic parameter matrix. This is achieved through data dimensionality reduction and importance analysis, primarily extracting parameters that have the greatest impact on building material performance, such as average particle size, particle size distribution uniformity, organic matter content range, and heavy metal indicators. The average particle size is directly calculated using the D50 value, while the particle size distribution uniformity is calculated using the nonuniformity coefficient. The organic matter content range includes minimum, maximum, and average values, and the heavy metal index focuses on the content of common harmful elements such as lead, cadmium, and mercury. These extracted key eigenvalues ​​constitute a set of dredged mud characteristic indicators with low dimensionality but high information density.

[0052] A standard parameter library for various building materials has been constructed. This library covers common building material types such as cement concrete, building blocks, permeable bricks, expanded clay, and anti-seepage materials. Each building material has corresponding raw material requirements, including particle size requirements, upper organic matter content limits, and heavy metal content limits. These indicators are primarily formulated with reference to the national or industry standards for relevant building materials in the "Compilation of Building Material Standards" to ensure that the final building materials meet quality requirements. For example, for permeable bricks, the raw material particle size requirements are 0.15-4.75mm, the organic matter content should be less than 3%, and the lead content should be less than 50mg / kg. These indicators are digitized and structured to form a standard parameter library that can be processed by computers.

[0053] After completing the construction of the above two data sets, the similarity calculation began. The Euclidean distance calculation was performed between the dredged mud characteristic index set and the parameters of each building material type in the building material standard parameter library to obtain a similarity score table. The Euclidean distance calculation uses the following formula:

[0054]

[0055] Among them, D(P,Q) represents the distance between the dredged mud feature set P and a certain building material standard parameter Q, p a and q a Represents the value of the ath indicator in the feature set and standard parameter respectively, k is the total number of feature indicators, ω a is the weight coefficient of the ath characteristic index. The weight coefficient is set according to the degree of influence of each index on the performance of the building material. For example, the particle size distribution has a greater impact on permeable bricks. When calculating the similarity of permeable bricks, the weight of the particle size index will be increased accordingly. The smaller the distance value, the more the characteristics of the dredged mud match the requirements of the building material. In the calculated similarity score table, each column represents a type of building material, each row represents a dredged mud sample, and the values ​​in the table are the corresponding Euclidean distance values. Due to the different dimensions and ranges of different indicators, these distance values ​​need to be normalized. The normalization process uses the following formula:

[0056]

[0057] Among them, N(d ij ) represents the normalized applicability score, d ij represents the Euclidean distance between the i-th dredged mud sample and the j-th building material type, d max and d min The normalized scores range from 0 to 100%, with higher scores indicating better suitability. These normalized scores form a weight matrix, where each element represents the weight of a particular dredged mud sample's suitability for a particular building material type.

[0058] Based on the weight matrix, a hierarchical analysis decision tree was designed to conduct the final screening of building material types. The factors at each level of the decision tree include: applicability weight (first level), processing difficulty (second level), and material performance requirements (third level). The processing difficulty considers the complexity and cost of the pre-treatment process that the dredged mud needs to go through, while the material performance requirements focus on indicators such as the strength and durability of the final building materials. The decision tree scores each building material type by comprehensively evaluating the factors at these three levels, and selects the building material type with the highest adaptability.

[0059] Finally, the screening results are sorted by suitability to establish a gradient library of dredged mud construction material suitability. This library not only contains the primary suitable type but also retains suboptimal alternative types and their scoring data, providing flexible selection options for actual production. For example, after the above analysis process, a dredged mud sample was ultimately determined to have ceramsite as the primary suitable type, with a suitability score of 88 points; the alternative type was permeable brick, with a suitability score of 76 points. This multi-option result design enables practical selection between the primary and alternative options based on factors such as market demand and equipment conditions, enhancing the practicality and flexibility of the method.

[0060] Taking dredged mud from a marine engineering project as an example, a set of characteristic indicators (D50 = 0.12 mm, organic matter content = 3.5%, lead content = 30 mg / kg) was extracted and Euclidean distances were calculated with standard parameters for five building materials. The particle size indicator was weighted 0.5, the organic matter content was weighted 0.3, and the heavy metal content was weighted 0.2. The results showed that the Euclidean distance between this dredged mud and ceramsite was the smallest (0.28), followed by permeable bricks (0.42), with the distances being greater than those for other building materials. After normalization, the applicability weights for ceramsite were 85%, permeable bricks were 65%, and all other materials were below 40%. Considering the relatively low processing difficulty of ceramsite production (requiring simple dehydration and screening) and its moderate performance requirements, this dredged mud was determined to be primarily suitable for ceramsite production, with permeable bricks considered as an alternative.

[0061] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0062] (1) centrifugally dewatering and drying the dredged mud to control the moisture content below 25% to obtain pretreated dredged mud;

[0063] (2) Determine the mineral composition and chemical composition of the pre-treated dredged mud by X-ray fluorescence analyzer, and record the content data of calcium, silicon, aluminum, and iron;

[0064] (3) Using pretreated dredged mud as the base material, cement was added in a gradient of 5%, 10%, 15%, 20%, 25%, and 30% to prepare standard cubic test blocks of 100 mm × 100 mm × 100 mm;

[0065] (4) Curing the standard cubic specimens at a temperature of 20±2°C and relative humidity above 95%, and taking samples for testing after 3 days, 7 days, and 28 days;

[0066] (5) Use a pressure testing machine to apply load to the standard cubic specimen until it is destroyed, and record the compressive strength values ​​of the specimens with different cement dosages at different ages;

[0067] (6) Pair the compressive strength value with the corresponding cement usage data, and draw a strength-cement usage curve with cement usage as the horizontal axis and compressive strength as the vertical axis.

[0068] Specifically, dredged mud undergoes pretreatment to reduce its moisture content. Marine dredged mud typically has a moisture content of 60-80%. Direct use can lead to problems such as insufficient strength and large shrinkage in building materials. Therefore, dehydration is essential. This pretreatment utilizes a two-stage dehydration process: First, the dredged mud undergoes mechanical dehydration using a centrifugal dewatering machine at a controlled speed of 1500-2000 rpm for 10-15 minutes. This stage reduces the moisture content to approximately 40%. The centrifuged dredged mud is then placed in a drying machine for thermal dehydration at a controlled temperature of 60-80°C to prevent excessive temperatures from decomposing organic matter in the dredged mud. The drying time is adjusted based on the batch and moisture content until the moisture content is reduced to below 25%. The moisture content is tested using a sampling and drying method to ensure it meets subsequent processing requirements. After moisture content control, the mineral and chemical composition of the pretreated dredged mud is determined using an X-ray fluorescence analyzer. Prior to testing, dredged mud samples are ground to a mesh size of less than 200 mesh, pressed into 32mm diameter discs, and then placed in an analyzer for compositional analysis. X-ray fluorescence analysis uses the principle that each element produces characteristic fluorescence when excited by X-rays. The intensity of this fluorescence is measured to determine the elemental content. The concentrations of elements such as calcium, silicon, aluminum, and iron, which have a significant impact on cement hydration and strength development, are specifically focused on. These elemental content data are recorded and archived, serving as a crucial basis for subsequent mix design.

[0069] After obtaining basic property data for the dredged mud, test sample preparation began. Using pretreated dredged mud as the base material, cement was added in six gradients of 5%, 10%, 15%, 20%, 25%, and 30% to create test mixes with varying cement dosages. General-purpose Portland cement was used, and mix design was calculated using mass ratios. For example, for a 15% cement dosage, 15 kg of cement was added for every 100 kg of pretreated dredged mud. According to the standard requirements for cement concrete test methods, the mixed materials were stirred with water (controlling the water-cement ratio between 0.4 and 0.5), poured into a standard 100 mm x 100 mm x 100 mm cube mold, compacted using a vibrating table, and smoothed to produce standard cube test blocks. Nine test blocks were prepared for each cement dosage gradient and used for compressive strength testing at three ages: 3 days, 7 days, and 28 days, with three replicates for each age.

[0070] Prepared standard cube specimens must undergo standard curing according to specifications. The specimens are placed in a standard curing chamber, maintained at a temperature of 20±2°C and a relative humidity above 95% to ensure sufficient cement hydration reaction. Curing is divided into three cycles: 3 days, 7 days, and 28 days, representing early strength, mid-term strength, and design strength, respectively. At each designated age, the corresponding batch of specimens is removed from the curing chamber for surface treatment and testing preparation. At each age, the standard cube specimens are subjected to compressive strength testing using a pressure testing machine. Before testing, the specimen surface is inspected for flatness and, if necessary, ground flat. The specimens are placed in the pressure testing machine and uniformly loaded at the loading rate specified in the cement concrete test method standard until failure. The maximum load at failure is recorded, and the compressive strength is calculated. For each cement content and age, the average of three replicate specimens is calculated as the final compressive strength value, and the standard deviation is recorded to assess data reliability.

[0071] After completing compressive strength tests for all ages and cement content gradients, the compressive strength values ​​were paired with the corresponding cement content data. Data for ages 3, 7, and 28 days were plotted, with cement content as the horizontal axis and compressive strength as the vertical axis, to form three strength-cement content curves. These curves intuitively demonstrate the impact of different cement content levels on the strength development of dredged mud-based construction materials, providing foundational data for subsequent neural network modeling.

[0072] For example, dredged mud from a marine engineering project had its moisture content reduced to 23% after pretreatment. Composition analysis revealed that its primary components were silicon dioxide, aluminum oxide, iron oxide, and calcium oxide. Test blocks were prepared and cured according to six cement dosage gradients. Compressive strength tests at 28 days showed the following: 5% cement dosage resulted in a strength of 4.2 MPa, 10% to 8.5 MPa, 15% to 13.7 MPa, 20% to 19.2 MPa, 25% to 22.8 MPa, and 30% to 24.5 MPa. A strength-cement dosage curve was plotted, revealing a significant slowdown in strength growth above 25%, indicating diminishing returns from further increases in cement dosage.

[0073] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0074] (1) Extract data points from the strength-cement content curve and form a training data set with cement content and corresponding compressive strength values;

[0075] (2) Combining the key parameters in the dredged mud characteristic parameter matrix, a neural network input feature vector is constructed, which includes the dredged mud particle size distribution characteristics, moisture content, organic matter content, heavy metal content and cement dosage;

[0076] (3) Design a three-layer convolutional neural network structure. The first convolutional layer uses 16 3×3 convolution kernels to extract features, the second convolutional layer uses 32 3×3 convolution kernels to deepen features, and the third convolutional layer uses 64 3×3 convolution kernels for feature integration.

[0077] (4) The output features of the convolutional layer are mapped through a fully connected layer, connected to a hidden layer of 256 neurons, and then connected to a single output neuron, representing the predicted intensity value;

[0078] (5) Use the training data set to train the three-layer convolutional neural network, use the mean square error loss function and Adam optimizer to adjust the network weights;

[0079] (6) The three-layer convolutional neural network structure is obtained through the training process. The input layer contains dredged mud parameters and admixture variables, and the output layer is the predicted intensity value.

[0080] Specifically, data points in the strength-cement content curve are extracted and converted into a training data set. The specific operation is to digitize the strength-cement content curve obtained in the previous step, extract the compressive strength values ​​of 3 days, 7 days and 28 days corresponding to each cement content (5%, 10%, 15%, 20%, 25%, 30%), and form a structured data table. This data table contains cement content as an independent variable, compressive strength as a dependent variable, and records age information as a conditional variable. For example, a set of data may be recorded as: cement content 20%, age 28 days, compressive strength 19.2MPa. In this way, 18 data points (6 cement content × 3 age periods) are extracted from the curve, and these data points constitute the basic training data set. Combined with the key parameters in the dredged mud characteristic parameter matrix, the input feature vector of the neural network is constructed. Dredged mud parameters such as particle size distribution (D10, D50, and D90 values), moisture content, organic matter content, and heavy metal content are closely related to cement hydration reactions and ultimate strength development. To construct a comprehensive input feature vector, these parameters are integrated with cement dosage to form a multidimensional feature vector. The input feature vector includes: dredged mud particle size distribution (three values), moisture content (one value), organic matter content (one value), heavy metal content (typically three to five values ​​depending on the metal being tested), cement dosage (one value), and cement age (one value), totaling approximately 10-12 feature parameters. These parameters are standardized and used as the input layer of the neural network.

[0081] Based on the dimensionality of the input feature vector and the complexity of the intensity prediction task, a three-layer convolutional neural network architecture was designed. Convolutional neural networks excel at processing data with spatial correlation, and since the various characteristic parameters of dredged mud exhibit certain correlations, convolutional neural networks were chosen for feature extraction and intensity prediction. The first convolutional layer uses 16 3×3 convolutional kernels. This layer primarily extracts low-level feature patterns from the raw input features. The 3×3 convolutional kernel size is a common size in computer vision and is suitable for capturing local feature correlations. The second convolutional layer uses 32 3×3 convolutional kernels to further deepen the feature representation based on the basic features extracted in the first layer, increasing the level of feature abstraction. The third convolutional layer uses 64 3×3 convolutional kernels for higher-level feature integration and abstraction, generating feature representations that are highly relevant to the prediction task. Each convolutional layer is followed by an activation function (such as ReLU) to introduce nonlinearity. Batch normalization is also used to reduce internal covariate shift and accelerate training.

[0082] The features extracted by the convolutional layer need to be mapped through a fully connected layer and converted into strength predictions. After flattening the output of the third convolutional layer, it is connected to a hidden layer containing 256 neurons, which has sufficient capacity to represent complex nonlinear mapping relationships. The hidden layer uses the ReLU activation function to ensure that the network has sufficient nonlinear expression capabilities. The hidden layer is then connected to a single output neuron, which uses a linear activation function (or no activation function) to directly output the predicted compressive strength value. This mapping structure from multidimensional features to a single strength value can effectively capture the combined impact of each input parameter on strength.

[0083] After the network structure is established, the network is trained using the previously prepared training dataset. This training utilizes a supervised learning approach, with the input feature vector serving as the network input and the actual measured compressive strength serving as the label. The loss function chosen is the mean squared error (MSE), which calculates the squared difference between the predicted and actual strength values ​​and is a commonly used loss function in regression problems. The optimizer chosen is Adam (Adaptive Moment Estimation), an adaptive learning rate optimization algorithm that combines the advantages of momentum and RMSProp, effectively handling gradient sparsity and accelerating convergence. During training, the network weights are gradually adjusted as the gradient descends until the loss function reaches a stable, low value, indicating that the network has effectively learned the mapping between input features and strength. After sufficient training, the final three-layer convolutional neural network model is obtained. This model is capable of predicting compressive strength under different conditions based on the characteristic parameters of dredged mud and cement content. To validate the model's performance, a portion of the data is typically retained as a test set. The prediction error on this test set is then calculated to assess the model's generalization ability. Using marine dredged mud as an example, the neural network model built using this method achieved an average relative error of less than 5% on a test set, demonstrating excellent predictive accuracy. This trained neural network became a core tool for subsequent mix optimization design. It rapidly predicted material strength under varying mix parameters, eliminating extensive experimental work and significantly improving the efficiency and accuracy of mix design.

[0084] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0085] (1) Set the cement content range to 5%-35%, the aggregate ratio range to 40%-70%, and the admixture ratio range to 5%-15%, and construct the mix parameter matrix;

[0086] (2) Divide the mix parameter matrix into multiple mix points with a step size of 1% to generate a mix candidate set, where each mix point contains a specific ratio value of cement, aggregate, and admixture;

[0087] (3) Sequentially extract the matching points from the matching candidate set, combine them with the dredged mud parameters, and assemble them into the input data vector;

[0088] (4) Pass the input data vector into the convolutional neural network, obtain the corresponding intensity prediction output value, and record the corresponding relationship between the matching point and the intensity value;

[0089] (5) According to the intensity prediction output value, adjust the direction of the ratio parameters, gradually explore the high-intensity area, generate new ratio points and record their intensity prediction values;

[0090] (6) All tested ratio points and their strength prediction values ​​are organized into a structured data table, indexed by ratio number and strength value, and a ratio-strength database is formed.

[0091] Specifically, reasonable ranges for mix parameters are set based on building material design specifications. The cement content range is set between 5% and 35%. The lower limit of this range is based on the need to ensure minimum bond strength, while the upper limit takes into account cost control and the potential shrinkage caused by excessive cement content. The aggregate ratio range is set between 40% and 70%, based on the fundamental requirement of aggregate as a skeletal material in building materials. The admixture ratio range is set between 5% and 15%, primarily considering the impact of admixtures on workability, strength development, and durability. Using these three parameter ranges, a three-dimensional mix parameter matrix is ​​constructed, with each dimension of the matrix corresponding to cement content, aggregate ratio, and admixture ratio.

[0092] To ensure refined mix design, the mix parameter matrix is ​​partitioned into 1% increments. This means that starting with 5% cement content, the matrix is ​​divided into 31 increments, with increments of 1% up to 35%; starting with 40% aggregate content, the matrix is ​​divided into 31 increments, with increments of 1% up to 70%; and starting with 5% admixture content, the matrix is ​​divided into 11 increments, with increments of 1% up to 15%. This partitioning method theoretically generates 31 × 31 × 11 = 10,571 mix parameters, each containing specific cement, aggregate, and admixture ratios. However, given that the sum of all component ratios in a mix must be 100%, a constraint is added: cement content + aggregate ratio + admixture ratio + dredged mud ratio = 100%. These constraints are then selected to form the final set of candidate mix parameters. This step essentially discretizes the continuous mix parameter space, facilitating subsequent systematic strength prediction and optimization.

[0093] From the generated set of candidate mixes, mix points are extracted according to a pre-set order (e.g., cement content from low to high). Each extracted mix point is then combined with the dredged mud's characteristic parameters to form a complete input data vector. This vector contains not only the mix information (cement content, aggregate ratio, admixture ratio), but also the dredged mud characteristic parameters obtained in the previous steps (e.g., particle size distribution, moisture content, organic matter content, heavy metal content, etc.). Thus, a complete input data vector may contain 10-15 parameters, comprehensively describing the material composition and basic properties.

[0094] The assembled input data vector is passed into a trained convolutional neural network, which directly outputs the corresponding strength prediction value. This process is called forward propagation. The input data first passes through three convolutional layers for feature extraction and transformation, and then is mapped to the final strength prediction value through a fully connected layer. For each mix point, the recipe parameters and the corresponding strength prediction value are recorded, establishing a correspondence between the mix point and the strength value. This approach avoids the extensive preparation and testing of experimental specimens, greatly improving the efficiency of mix design.

[0095] Based on the preliminary intensity prediction results, the ratio parameters are adjusted in a targeted manner to explore higher intensity areas. This adjustment process follows the gradient ascent strategy, and the calculation formula is:

[0096]

[0097] Among them, P new Represents the new ratio parameter vector, P old represents the current ratio parameter vector, η is the step coefficient, Represents the gradient of the intensity function at the current matching point. The gradient calculation is approximated by the finite difference method:

[0098]

[0099] Here, δ is a small perturbation value (typically 0.5%-1%), e1, e2, and e3 are unit vectors in the directions of cement content, aggregate ratio, and admixture ratio, respectively, and S(P) represents the predicted strength value corresponding to the mix parameter vector P. By calculating the rate of change of strength in all directions around a specific mix point, the direction of fastest strength growth is determined, and a new mix point is generated along that direction. The neural network is then used to continuously predict the strength of each newly generated mix point, and the prediction results are recorded. This iteration continues until a preset termination condition is met (e.g., the strength growth rate falls below a threshold).

[0100] All tested mix points and their predicted strength values ​​are organized into a structured data table. This table contains multiple fields, such as mix number, cement content, aggregate ratio, admixture ratio, dredged mud ratio, and predicted strength value. A primary index is created based on mix number, while a secondary index is created for strength values, facilitating data query and analysis from various perspectives. This resulting mix-strength database serves as the foundation for subsequent screening of optimal mixes.

[0101] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0102] (1) Arrange the ratio-intensity database in descending order according to the intensity prediction value, extract the top three groups of ratio parameters with the highest intensity, and mark them as the preferred ratio groups;

[0103] (2) According to the parameter ratios in the preferred mix ratio group, the corresponding weights of pre-treated dredged mud, cement, aggregate and admixture were weighed to prepare three groups of physical test samples;

[0104] (3) Three groups of physical test samples were cured in a standard environment with a temperature of 20±2°C and a relative humidity of 95% or higher, and strength tests were performed after 3 days, 7 days, and 28 days respectively;

[0105] (4) Record the actual compressive strength values ​​of the three groups of physical test samples at various ages, compare them with the predicted strength values ​​of the convolutional neural network, and calculate the prediction error;

[0106] (5) Based on the actual compressive strength values ​​and prediction errors, the three groups of mixes were scored comprehensively, taking into account the 28-day strength, strength growth rate, and prediction accuracy factors;

[0107] (6) Select the ratio with the highest comprehensive score from the three groups of ratios, form a standardized ratio table, and determine the building material ratio data.

[0108] Specifically, through the database query statement, all records in the proportion-strength database are sorted in descending order of predicted strength values, so that the proportion with the highest strength will be at the front. From the sorted database, the top three groups of proportion parameters with the highest strength prediction values ​​are extracted. These three groups of proportion parameters represent the proportion schemes that are most likely to achieve high strength based on the neural network prediction results. These three groups of proportion parameters are marked as preferred proportion groups, and their parameters such as cement dosage, aggregate ratio, admixture ratio and dredged mud ratio are accurately recorded. This preferred proportion screening method based on strength prediction results avoids a lot of trial and error processes in traditional proportion design and improves design efficiency. According to the parameter ratios in the preferred proportion group, physical test samples are prepared. For each group of proportions, the corresponding weights of pretreated dredged mud, cement, aggregate and admixture are accurately weighed according to their parameter ratios. For example, for a test batch weighing 10kg, if the mix ratio is 20% cement, 58% aggregate, 10% admixture, and 12% dredged mud, 2kg of cement, 5.8kg of aggregate, 1kg of admixture, and 1.2kg of pre-treated dredged mud are required. After weighing the materials, they are mixed and formed according to the standard process for preparing building materials. First, the dry powder materials (cement and admixture) are mixed evenly, then the pre-treated dredged mud and aggregate are added and mixed. Finally, water is added to adjust the workability to the appropriate level. The entire mixing process typically takes 3-5 minutes to ensure uniform mixing of the materials. The mixed materials are poured into a standard 100mm×100mm×100mm cube mold, compacted on a vibrating table, and the surface smoothed to complete the preparation of three sets of physical test samples. At least nine test blocks are prepared for each mix ratio, and strength tests are performed at three ages: 3 days, 7 days, and 28 days. The three sets of prepared physical test samples are placed in a standard curing chamber for curing. Curing conditions are strictly controlled within a standard environment with a temperature of 20±2°C and a relative humidity of 95% or higher, which is conducive to the normal progress of the cement hydration reaction. During the curing process, corresponding test blocks are removed for strength testing after 3 days, 7 days, and 28 days. Before strength testing, the test blocks are surface treated to ensure a flat loading surface. Then, a pressure tester is used to apply load to the test blocks at the specified loading rate until failure. The failure load is recorded and the compressive strength is calculated.

[0109] The actual compressive strength values ​​of the three groups of physical test samples at various ages were recorded in detail. These measured strength values ​​were then compared with the strength values ​​predicted by the convolutional neural network under the same mix conditions. The prediction error for each mix at each age was calculated by dividing the difference between the actual and predicted strength values ​​by the actual strength value, and then multiplying by 100% to obtain the relative error percentage. The magnitude of the prediction error can be used to assess the accuracy of the neural network model's predictions for different mixes and will serve as an important factor in the subsequent comprehensive scoring. A comprehensive score was then assigned to the three mixes based on the measured compressive strength values ​​and the prediction error. This comprehensive score takes into account several factors: first, the 28-day strength value, a key performance indicator for building materials; second, the strength growth rate. Strength development curves were calculated using strength values ​​at 3, 7, and 28 days to assess the material's early strength and potential for strength growth; and third, prediction accuracy, which measures the degree of agreement between the actual and predicted strengths. This metric reflects the reliability of the mix design method. The three factors were weighted separately (e.g. 28-day strength accounted for 60%, strength growth rate accounted for 25%, and prediction accuracy accounted for 15%), and the weighted total score was calculated to obtain the comprehensive score of each ratio.

[0110] The mix with the highest overall score among the three groups was selected as the final building material mix. The mix parameters (cement content, aggregate ratio, admixture ratio, and dredged mud ratio) were organized according to a standard format to form a standardized mix ratio table. The corresponding performance parameters (such as expected strength, density, and cement content) were also recorded to complete the final building material mix data. This method for determining the final mix ratio based on actual test results not only leverages the efficient screening capabilities of the neural network model but also ensures the reliability of the results through physical verification. It represents a mix design method that combines theory with practice.

[0111] The above describes the ratio design method for making building materials based on ocean dredged mud in the embodiment of the present application. The following describes the ratio design system for making building materials based on ocean dredged mud in the embodiment of the present application. Figure 2 In one embodiment of the present application, a system for designing a proportion of building materials based on ocean dredged mud includes:

[0112] The acquisition module is used to collect marine dredged mud samples, measure the particle size distribution, moisture content, organic matter content and heavy metal content of the marine dredged mud samples, and generate a dredged mud characteristic parameter matrix;

[0113] A calculation module, configured to perform similarity calculation between the dredged mud characteristic parameter matrix and a standard parameter library of building materials to determine the type of building materials suitable for the dredged mud;

[0114] a processing module, configured to pre-process the dredged mud, measure the compressive strength values ​​at different cement dosages, and draw a strength-cement dosage curve;

[0115] A construction module is used to construct a three-layer convolutional neural network structure based on the strength-cement dosage curve, wherein the input layer includes dredged mud parameters and admixture variables, and the output layer is the predicted strength value;

[0116] a recording module for gradually adjusting the mix parameters of cement, aggregate, and admixture through the convolutional neural network, recording the strength prediction results of each iteration, and forming a mix ratio-strength database;

[0117] The proportioning module is used to screen the top three proportions with the highest strength from the proportion-strength database, prepare physical samples and conduct 28-day strength tests to determine the proportioning data of building materials.

[0118] Through the collaborative efforts of the aforementioned components and a systematic method for generating a dredged mud characteristic parameter matrix, the key characteristic indicators of dredged mud were comprehensively collected and quantified. Second, similarity calculation technology was used to achieve intelligent matching of dredged mud with building material types, avoiding blind trial and error and improving resource allocation efficiency. Third, a strength-cement dosage curve was constructed to reveal the correlation between cement dosage and strength development, providing directional guidance for mix ratio optimization. Fourth, a three-layer convolutional neural network structure was applied to process multidimensional feature data, fully leveraging the algorithmic advantages of convolutional neural networks in nonlinear mapping and feature extraction, enabling the model to accurately capture the complex relationship between dredged mud characteristics, mix ratio parameters, and strength. This deep learning-based mix ratio prediction model has stronger generalization ability and prediction accuracy than traditional statistical methods. Fifth, through the iterative optimization mechanism of the neural network, efficient adjustment and optimization of mix ratio parameters were achieved, significantly reducing experimental workload, saving time and cost. Sixth, by integrating physical sample preparation and strength test results, a method system combining theoretical prediction with practical verification was established, ensuring the reliability and practicality of the mix ratio design results. Especially in the application of three-layer convolutional neural networks, this method fully considers the matching between algorithm characteristics and actual application needs. By rationally designing the network structure and parameters, the model can not only process the high-dimensional characteristic parameters of dredged mud, but also accurately predict the strength performance under different ratios, solving the problem of high-dimensional nonlinear relationship modeling that is difficult to handle with traditional methods.

[0119] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3As shown. The computer device includes a processor, memory, display screen, input device, network interface and database connected via a system bus. The processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the above method is implemented.

[0120] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0121] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which implements the above-described method when executed by a processor. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0122] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Among them, any reference to memory, storage, database, or other media provided by the present invention and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAM bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM.

[0123] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0124] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc., various media that can store program code.

[0125] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for designing a proportion of building materials based on ocean dredged mud, characterized in that: The method for designing the proportion of building materials based on ocean dredged mud comprises: Collecting marine dredged mud samples, measuring particle size distribution, moisture content, organic matter content, and heavy metal content of the marine dredged mud samples, and generating a dredged mud characteristic parameter matrix; Performing similarity calculation on the dredged mud characteristic parameter matrix and the building material standard parameter library to determine the type of building materials suitable for the dredged mud; Pre-treating the dredged mud, measuring the compressive strength values ​​at different cement dosages, and drawing a strength-cement dosage curve; Based on the strength-cement dosage curve, a three-layer convolutional neural network structure is constructed, wherein the input layer contains dredged mud parameters and admixture variables, and the output layer is the predicted strength value; The mix ratio parameters of cement, aggregate and admixture are gradually adjusted by the convolutional neural network, and the strength prediction results of each iteration are recorded to form a mix ratio-strength database; The top three mix ratios with the highest strengths are screened from the mix ratio-strength database, and physical samples are prepared and subjected to 28-day strength tests to determine building material mix ratio data.

2. The method for designing a proportion of building materials based on ocean dredged mud according to claim 1, characterized in that: The method comprises collecting marine dredged mud samples, measuring the particle size distribution, moisture content, organic matter content and heavy metal content of the marine dredged mud samples, and generating a dredged mud characteristic parameter matrix, including: Collect marine dredged mud samples at different depths and in different areas, record the location data of the sampling points through the positioning system, and form a sampling record sheet; Using a laser particle size analyzer to measure the particle size of the marine dredged mud sample, obtain D10, D50, D90 particle size values ​​and distribution curves, and record the marine dredged mud particle size distribution; The moisture content of the marine dredged mud sample is determined by a drying-weighing method, the organic matter content is determined by a potassium dichromate oxidation method, and the heavy metal content is determined by atomic absorption spectrometry; Standardizing the particle size distribution, moisture content, organic matter content, and heavy metal content of the marine dredged mud to remove unit differences and obtain standardized data; Arrange and organize the standardized data according to sample number and parameter type, and construct a data table containing all samples and parameters; The main characteristic parameters in the data table are identified through correlation analysis and integrated to form the dredged mud characteristic parameter matrix.

3. The method for designing a proportion of building materials based on ocean dredged mud according to claim 1, characterized in that: The similarity calculation between the dredged mud characteristic parameter matrix and the building material standard parameter library is performed to determine the type of building materials suitable for the dredged mud, including: Extracting key characteristic values ​​from the dredged mud characteristic parameter matrix, including average particle size, organic matter content range, and heavy metal index, to form a dredged mud characteristic index set; Build a standard parameter library for building materials including cement concrete, building blocks, permeable bricks, ceramsite, and anti-seepage layer materials, and record the physical and chemical index requirements of various building materials; Performing Euclidean distance calculation on the dredged mud characteristic index set and the parameters of each building material type in the building material standard parameter library to obtain a similarity score table; Normalizing the values ​​in the similarity score table, assigning an applicability weight to each building material type, and generating a weight matrix; Based on the weight matrix, a hierarchical analysis decision tree is designed to select building materials with high adaptability based on the characteristics of dredged mud, processing difficulty and material performance requirements; The selected building material types are sorted by suitability, and a gradient library of dredged mud building material applicability is established, which includes the main applicable types and alternative types, determines the building material types that are applicable to dredged mud, and retains the alternative plan data.

4. The method for designing a proportion of building materials based on ocean dredged mud according to claim 1, characterized in that: The pre-processing of the dredged mud, measuring the compressive strength values ​​under different cement dosages, and drawing a strength-cement dosage curve include: The dredged mud is subjected to centrifugal dehydration and drying treatment to control the moisture content below 25% to obtain pretreated dredged mud; The mineral composition and chemical composition of the pre-treated dredged mud are measured by X-ray fluorescence analyzer, and the content data of calcium, silicon, aluminum and iron elements are recorded; Using the pretreated dredged mud as a base material, cement was added in a gradient of 5%, 10%, 15%, 20%, 25%, and 30% to prepare a 100 mm × 100 mm × 100 mm standard cubic test block; The standard cube test block is cured at a temperature of 20±2°C and a relative humidity of 95% or higher, and samples are taken for testing after 3 days, 7 days, and 28 days respectively; Using a pressure testing machine, a load is applied to the standard cubic test block until it is destroyed, and the compressive strength values ​​of the test blocks with different cement dosages at different ages are recorded; The compressive strength value is paired with the corresponding cement usage data, and a strength-cement usage curve is drawn with cement usage as the horizontal axis and compressive strength as the vertical axis.

5. The method for designing a proportion of building materials based on ocean dredged mud according to claim 1, characterized in that: Based on the strength-cement dosage curve, a three-layer convolutional neural network structure is constructed, wherein the input layer includes dredged mud parameters and admixture variables, and the output layer is the predicted strength value, including: Extracting data points from the strength-cement dosage curve, and forming a training data set with cement dosage and corresponding compressive strength values; Combining the key parameters in the dredged mud characteristic parameter matrix, constructing a neural network input feature vector, which includes the dredged mud particle size distribution characteristics, moisture content, organic matter content, heavy metal content and cement dosage; A three-layer convolutional neural network structure is designed. The first convolutional layer uses 16 3×3 convolution kernels to extract features, the second convolutional layer uses 32 3×3 convolution kernels to deepen features, and the third convolutional layer uses 64 3×3 convolution kernels for feature integration. The output features of the convolutional layer are mapped through a fully connected layer, connected to a hidden layer of 256 neurons, and then connected to a single output neuron, representing the predicted intensity value; The three-layer convolutional neural network is trained using the training data set, and the network weights are adjusted using the mean square error loss function and the Adam optimizer; The three-layer convolutional neural network structure is trained through the training process. The input layer contains dredged mud parameters and admixture variables, and the output layer is the predicted intensity value.

6. The method for designing a proportion of building materials based on ocean dredged mud according to claim 1, characterized in that: The convolutional neural network is used to gradually adjust the mix parameters of cement, aggregate, and admixture, and record the strength prediction results of each iteration to form a mix ratio-strength database, including: Set the cement dosage range to 5%-35%, the aggregate ratio range to 40%-70%, and the admixture ratio range to 5%-15% to construct a mix parameter matrix; Divide the mix parameter matrix into multiple mix points according to a step size of 1% to generate a mix candidate set, where each mix point contains a specific ratio value of cement, aggregate, and admixture; Sequentially extracting matching points from the matching candidate set, combining them with the dredged mud parameters, and assembling them into an input data vector; Passing the input data vector into the convolutional neural network, obtaining the corresponding intensity prediction output value, and recording the corresponding relationship between the matching point and the intensity value; According to the intensity prediction output value, the direction of the ratio parameters is adjusted, and the high intensity area is gradually explored to generate a new ratio point and record its intensity prediction value; All tested mix points and their strength prediction values ​​are organized into a structured data table, indexed by mix number and strength value, to form a mix-strength database.

7. The method for designing a proportion of building materials based on ocean dredged mud according to claim 6, characterized in that: The method of screening the top three mix ratios with the highest strength from the mix ratio-strength database, preparing physical samples and conducting 28-day strength tests to determine building material mix ratio data includes: Arrange the ratio-strength database in descending order according to the strength prediction value, extract the top three groups of ratio parameters with the highest strength, and mark them as the preferred ratio groups; According to the parameter ratios in the preferred mix ratio group, corresponding weights of pre-treated dredged mud, cement, aggregate and admixture are weighed to prepare three groups of physical test samples; The three groups of physical test samples were cured in a standard environment with a temperature of 20±2°C and a relative humidity of 95% or higher, and samples were taken for strength testing after 3 days, 7 days, and 28 days respectively; Record the actual compressive strength values ​​of the three groups of physical test samples at various ages, compare them with the predicted strength values ​​of the convolutional neural network, and calculate the prediction error; Based on the actual compressive strength values ​​and prediction errors, the three mix ratios were comprehensively scored, taking into account the 28-day strength, strength growth rate, and prediction accuracy factors; The proportion with the highest comprehensive score is selected from the three groups of proportions to form a standardized proportion table and determine the proportion data of building materials.

8. A system for designing the proportion of building materials based on marine dredged mud, for implementing the method for designing the proportion of building materials based on marine dredged mud as claimed in any one of claims 1 to 7, characterized in that: The ratio design system for making building materials based on ocean dredged mud includes: A collection module is used to collect marine dredged mud samples, measure the particle size distribution, moisture content, organic matter content and heavy metal content of the marine dredged mud samples, and generate a dredged mud characteristic parameter matrix; A calculation module, configured to perform similarity calculation between the dredged mud characteristic parameter matrix and a standard parameter library of building materials to determine the type of building materials suitable for the dredged mud; a processing module, configured to pre-process the dredged mud, measure the compressive strength values ​​at different cement dosages, and draw a strength-cement dosage curve; A construction module is used to construct a three-layer convolutional neural network structure based on the strength-cement dosage curve, wherein the input layer includes dredged mud parameters and admixture variables, and the output layer is the predicted strength value; a recording module for gradually adjusting the mix parameters of cement, aggregate, and admixture through the convolutional neural network, recording the strength prediction results of each iteration, and forming a mix ratio-strength database; The proportioning module is used to screen the top three proportions with the highest strength from the proportion-strength database, prepare physical samples and conduct 28-day strength tests to determine the proportioning data of building materials.

9. A computer device, characterized in that: The invention comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and is characterized in that when the processor executes the computer program, the method for designing the proportion of building materials based on marine dredged mud according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor is caused to execute the method for designing a proportion of building materials based on marine dredged mud according to any one of claims 1 to 7.

Images