Soil stabilizer production monitoring method and system based on Internet of Things

Abstract

The invention relates to the technical field of the Internet of Things, and discloses a soil stabilizer production monitoring method and system based on the Internet of Things. The method comprises the following steps: deploying an Internet of Things sensor network at the positions of a raw material mixer, a reaction kettle and a compaction forming machine of the soil stabilizer production line, and collecting original monitoring data in the production process; creating a staged monitoring strategy according to the original monitoring data of the production process; differential sampling frequency control is carried out on the Internet of Things sensor network, and a production process target parameter set is obtained; identifying abnormal fluctuation of electrolyte concentration change, micellar formation state and crystal water fixing efficiency based on the production process target parameter set to obtain production quality risk early warning information; and a corresponding production process parameter adjustment scheme is generated by combining the demand difference of the performance of the soil stabilizer in different application scenes. According to the invention, the problems of signal interference and data acquisition blind areas in traditional monitoring are solved, and intelligent and accurate regulation and control of process parameters are realized.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, and in particular to an IoT-based method and system for monitoring the production of soil stabilizers. Background Technology

[0002] Soil stabilizers, as a new type of energy-saving and environmentally friendly engineering material, are widely used in road base reinforcement, foundation treatment, and water conservancy seepage prevention. Their quality directly affects the safety and service life of the project. Traditional monitoring of soil stabilizer production mainly relies on manual experience and fixed-interval sampling methods, which have problems such as many monitoring blind spots, serious data delays, and limited processing capacity. It is difficult to accurately capture changes in key parameters during the production process, such as electrolyte concentration, micelle current, and crystallinity, thus affecting the stability and consistency of the final product quality.

[0003] The application of existing IoT monitoring systems in the production of specialized chemical materials such as soil stabilizers still faces numerous challenges. Particularly under complex operating conditions, achieving low-power, high-efficiency operation of sensor networks, dynamically adjusting sampling strategies to adapt to the monitoring needs of different production stages, and conducting intelligent analysis and early warning based on massive amounts of monitoring data are all pressing technical problems that need to be solved. Furthermore, soil stabilizers have varying performance requirements due to different application scenarios (road base courses, foundation treatment, water conservancy seepage prevention). Traditional production methods struggle to adjust process parameters in real time according to the specific needs of different application scenarios, resulting in poor traceability of the production process. Quality control relies mainly on finished product testing, failing to achieve precise intervention and regulation during the process, leading to large fluctuations in product qualification rates and significant waste of energy and resources. Summary of the Invention

[0004] This invention provides a method and system for monitoring the production of soil stabilizers based on the Internet of Things. This invention solves the problems of signal interference and data acquisition blind spots in traditional monitoring, and realizes intelligent and precise control of process parameters.

[0005] In a first aspect, the present invention provides a method for monitoring the production of soil stabilizers based on the Internet of Things (IoT), the method comprising: An IoT sensor network is deployed at the raw material mixer, reaction vessel, and compaction molding machine locations in the soil stabilizer production line to collect raw monitoring data of the production process, including temperature, humidity, pressure, conductivity, and characteristic spectral data. Based on the original monitoring data of the production process, the soil stabilizer production process is divided into the raw material ratio stage, the chemical reaction stage, and the finished product quality stage, and a staged monitoring strategy is created. According to the phased monitoring strategy, differentiated sampling frequency control is implemented on the Internet of Things sensor network to obtain the target parameter set of the production process; Based on the set of target parameters for the production process, abnormal fluctuations in electrolyte concentration, micelle formation state, and crystal water fixation efficiency are identified to obtain early warning information on production quality risks. Based on the aforementioned production quality risk warning information, and combined with the differences in the performance requirements of soil stabilizers in different application scenarios, corresponding production process parameter adjustment schemes are generated.

[0006] Secondly, the present invention provides an Internet of Things (IoT)-based soil stabilizer production monitoring system, the IoT-based soil stabilizer production monitoring system comprising: The data acquisition module is used to deploy an IoT sensor network at the raw material mixer, reactor, and compaction molding machine locations in the soil stabilizer production line, and to collect raw monitoring data of the production process, including temperature, humidity, pressure, conductivity, and characteristic spectral data. A module is created to divide the soil stabilizer production process into three stages: raw material proportioning, chemical reaction, and finished product quality, based on the original monitoring data of the production process, and to create a phased monitoring strategy. The control module is used to implement differentiated sampling frequency control on the Internet of Things sensor network according to the phased monitoring strategy, and to obtain the target parameter set of the production process; The identification module is used to identify abnormal fluctuations in electrolyte concentration, micelle formation state, and crystal water fixation efficiency based on the set of production process target parameters, and to obtain production quality risk warning information. The generation module is used to generate corresponding production process parameter adjustment schemes based on the production quality risk warning information and the differences in the performance requirements of soil stabilizers in different application scenarios.

[0007] The technical solution provided by this invention deploys modular IoT sensors with dustproof and waterproof performance at key locations on the production line. Employing a tree-like topology and multi-protocol compatible communication parameter settings, a comprehensive, stable, and reliable monitoring network is formed, solving the problems of signal interference and data acquisition blind spots in traditional monitoring. Based on the original monitoring data, production process stages are divided, and a parameter sensitivity assessment model and differentiated monitoring frequency configuration are established to achieve accurate monitoring of key parameters at different production stages, avoiding data redundancy and omissions of key information caused by traditional fixed-interval sampling methods. By implementing a four-level sampling strategy (ultra-high frequency, high frequency, standard, and low frequency) combined with a dynamic sampling frequency adjustment mechanism, high-precision monitoring of key process points is ensured while significantly reducing sensor network energy consumption and extending equipment lifespan. Using a multi-level parameter decomposition and anomaly detection model, combined with multivariate correlation analysis, abnormal fluctuations in electrolyte concentration changes, micelle formation state, and crystal water fixation efficiency are accurately identified, allowing for early detection of potential quality risks and changing the traditional passive monitoring mode that relies on experience-based judgment. Based on production quality risk early warning information and combined with the performance requirements of different application scenarios, the optimal process parameter adjustment scheme that balances the strength, energy consumption and environmental performance of finished products is generated through multi-objective evaluation and Bayesian optimization algorithm, realizing intelligent and precise control of process parameters. Attached Figure Description

[0008] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a schematic diagram of an embodiment of the soil stabilizer production monitoring method based on the Internet of Things in this invention; Figure 2 This is a schematic diagram of one embodiment of the Internet of Things-based soil stabilizer production monitoring system of the present invention. Detailed Implementation

[0010] This invention provides a method and system for monitoring the production of soil stabilizers based on the Internet of Things (IoT). The terms "first," "second," "third," "fourth," etc. (if applicable) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0011] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the soil stabilizer production monitoring method based on the Internet of Things in this invention includes: Step S101: Deploy an Internet of Things (IoT) sensor network at the locations of the raw material mixer, reactor, and compaction molding machine in the soil stabilizer production line, and collect raw monitoring data of the production process, including temperature, humidity, pressure, conductivity, and characteristic spectral data. It is understood that the executing entity of this invention can be an IoT-based soil stabilizer production monitoring system, or it can be a terminal or a server; no specific limitation is made here. This embodiment of the invention will be described using a server as an example.

[0012] Specifically, on-site process layout analysis is conducted at each key location on the production line. Based on the process characteristics of different stages such as mixing, reaction, and molding, diverse sensor arrays are deployed, and each sensor is assigned a unique identifier and physical location marker to obtain an initial sensor deployment plan, ensuring the accuracy of subsequent data traceability and fault location. All deployed sensors are uniformly configured with communication protocols to ensure stable network access in industrial environments with strong electromagnetic fields or complex structures. Protocols such as NB-IoT, LoRa, or Wi-Fi are selected based on the actual production environment to flexibly support different bandwidth and coverage requirements. Based on the sensor communication network, the entire production line is logically divided into raw material, reaction, and molding areas. A high-performance gateway node is configured in each area to handle the aggregation, preliminary edge processing, and uploading of various sensor data within that area, thus constructing a regional network topology. During the data acquisition phase, the raw data streams from all low-noise sensors are acquired in real time according to the regional network topology. To improve data utilization efficiency and reduce redundancy, the rate of change and signal quality indicators of real-time data at each monitoring point are dynamically calculated. By analyzing the trend, volatility, and signal-to-noise ratio of each data stream, a sampling weight table reflecting process sensitivity is dynamically generated. In the weight table, the weight of each acquisition point is adjusted in real time based on the rate of change of its monitored variables, anomaly probability, and process criticality. This allows relevant sensors to automatically increase sampling frequency when critical process conditions change drastically, while appropriately reducing sampling density during stable phases. This optimizes network resources and energy consumption while ensuring process safety and accurate monitoring. Based on the aforementioned dynamic sampling weight table, the IoT sensor network automatically performs differentiated data acquisition and simultaneously records raw monitoring information of the production process, including multi-dimensional process data such as temperature, humidity, pressure, conductivity, and characteristic spectra. All raw data is accompanied by an acquisition timestamp, a unique sensor identifier, and a physical location marker, facilitating subsequent data traceability, anomaly tracking, and big data modeling.

[0013] Step S102: Based on the original monitoring data of the production process, the soil stabilizer production process is divided into the raw material ratio stage, the chemical reaction stage, and the finished product quality stage, and a phased monitoring strategy is created. Specifically, a systematic time-series analysis is performed on the raw data, including temperature, humidity, pressure, conductivity, and characteristic spectra, collected in real time by the IoT sensor network. Using data mining and statistical methods, key characteristic points in the soil stabilizer production process are captured, including typical process nodes such as raw material addition, stirring start-up, initial reaction, micelle formation, crystal water stabilization, and molding and compaction. This yields a complete characteristic point set for the entire soil stabilizer production process. Based on the obtained full-process characteristic point set, and considering the time-series data and physicochemical change patterns, combined with set process thresholds and change rates, intelligent algorithms such as segmented clustering or breakpoint detection are used to automatically delineate the boundaries of each stage, clarifying the start and end times or process nodes of the raw material proportioning stage, chemical reaction stage, and finished product quality stage. According to the key process requirements of each stage, a stage parameter mapping table is established, assigning conductivity and stirring torque monitoring parameters to the raw material proportioning stage, characteristic spectra and temperature monitoring parameters to the chemical reaction stage, and humidity and compressive strength monitoring parameters to the finished product quality stage. For monitoring parameters allocated at different stages, parameter sensitivity assessments are conducted. Using historical data or online change characteristics, the impact weight of each parameter on final product quality and process stability is quantitatively evaluated, generating a parameter sensitivity matrix. This matrix reflects the monitoring priority and sensitivity of each parameter at different stages, facilitating the optimal allocation of monitoring resources. Based on the parameter sensitivity matrix, a hierarchical monitoring mechanism is used to assign higher sampling frequencies and monitoring densities to parameters with high sensitivity and high impact, while appropriately reducing the monitoring frequency for parameters with slow changes and limited impact, resulting in differentiated and hierarchical monitoring frequency configuration parameters. These differentiated monitoring frequency configuration parameters are intelligently correlated with the ideal process parameter ranges for three types of soil stabilizers: road base layer, foundation treatment, and water conservancy seepage prevention. By searching and matching the optimal monitoring intervals and parameter requirements for different product types in the knowledge base, the system automatically pushes the most suitable phased monitoring strategy for the current production task and product type, achieving precise adaptation between the process flow and the performance of the end product.

[0014] In this embodiment, a sliding window analysis technique is applied to the feature point set of the entire production process. The data streams output by various sensors, such as temperature, conductivity, and humidity, are segmented and statistically calculated within the window to dynamically extract the local change feature sequences of parameters at each moment, capturing abrupt changes, extreme points, and inflection points in the process. Based on the sliding window, the parameter change feature sequences are further classified and refined. For conductivity data, a first-order differential method is used to calculate the rate of change of conductivity over time in real time, revealing the moments of abrupt changes in the physical or chemical state of the mixture. When a significant peak or abrupt change occurs in the first-order differential result, the system identifies it as a conductivity abrupt change point. The time intervals between adjacent conductivity abrupt change points are marked as conductivity abrupt change intervals, corresponding to key production stages such as raw material addition and the start of the mixing reaction. Simultaneously, for temperature curve data, wavelet transform is used to perform multi-scale analysis of the time series, effectively filtering out measurement noise and extracting local extreme points and key inflection points in the temperature change process. Local extreme points reflect the heating or exothermic limits in the reaction process, while inflection points correspond to nodes with significant physical or chemical meaning, such as abrupt changes in reaction rate or a shift in reaction mechanism. Based on these extremes and inflection points, the inflection point intervals of the temperature curve are marked, thereby narrowing the error range for process stage division. For humidity sensor data, stationarity testing techniques are used to analyze the data. By judging the rate of change and statistical volatility of the data sequence, stable and unstable periods in the humidity signal are distinguished, and intervals where the rate of change in humidity significantly decreases and tends to stabilize are identified. Integrating the above multi-parameter analysis results, based on three key characteristic segments—conductivity abrupt change intervals, temperature curve inflection point intervals, and humidity stable intervals—a time-series pattern recognition algorithm based on time distribution is used to construct a process stage transition model. This model integrates the physical logic of parameter change trends with actual empirical rules of the production process, and can automatically determine the timing of transitions between stages. According to the model, the conductivity abrupt change point is determined as the boundary between the raw material proportioning stage and the chemical reaction stage, and the time from the last inflection point of the temperature curve to the beginning of the stable humidity data interval is set as the boundary between the chemical reaction stage and the finished product quality stage, generating production stage boundary markers.

[0015] Step S103: According to the phased monitoring strategy, implement differentiated sampling frequency control on the Internet of Things sensor network to obtain the target parameter set of the production process; Specifically, based on the pre-defined process flow of raw material proportioning, chemical reaction, and finished product quality stages, the monitoring focus of each stage is linked to different key physical or chemical parameters. For example, the raw material proportioning stage focuses on the rate of change of electrical conductivity; the chemical reaction stage mainly focuses on changes in temperature curves and reaction kinetics; and the finished product quality stage involves intensive sampling and evaluation of parameters such as humidity and compressive strength. Combining the mapping relationship of the staged parameters, differentiated sampling frequencies are set for different stages. The electrical conductivity change rate sampling in the raw material proportioning stage is set to a high-frequency mode to capture instantaneous fluctuations in mixing uniformity and raw material proportioning accuracy. Temperature monitoring in the chemical reaction stage is set to a medium-high frequency mode to respond promptly to changes in reaction rate and process anomalies. The finished product quality stage uses a dynamically adjusted sampling frequency at critical moments of molding and performance finalization to ensure data coverage of the entire process's quality judgment window. Based on the above staged sampling strategy, a staged sampling control instruction set is generated and distributed to sensor terminals in each section of the production line via gateway nodes in the IoT architecture. Each gateway node adjusts the data acquisition rate of its subordinate sensors as needed based on received control commands, achieving coordinated distribution of multi-level sampling frequencies among different monitoring points within the same time period. This differentiated sampling control mode improves the system's data utilization efficiency and monitoring sensitivity, while effectively reducing data redundancy and bandwidth pressure. After actual data acquisition, a raw sensor dataset covering various production stages, different parameters, and multi-level sampling frequencies is obtained. For the raw data obtained under multi-level sampling frequencies, and considering the type of soil stabilizer for the current production task (e.g., road base, foundation treatment, or water conservancy seepage prevention), different product types have their own requirements for target performance indicators of production process parameters. The system has a built-in database of corresponding target performance indicators and parameter mapping templates, automatically filtering and comparing the collected raw data to extract process parameters closely related to the target product type (e.g., specific requirements for crystallinity, strength, and durability for different types), forming a set of production process target parameters specific to the current production task.

[0016] It should be noted that after the soil stabilizer is formed, it will undergo a wet-dry cycle in the actual application environment, which has a significant impact on the durability and engineering performance of the stabilizer. The monitoring system of this invention can also monitor and analyze the changes in the apparent properties and physicochemical parameters of the stabilizer under wet-dry cycle conditions through an extended monitoring module. Under water immersion conditions, the soil stabilizer exhibits the following characteristics: surface water absorption rate rises rapidly within the initial 30 minutes, then stabilizes, reaching an equilibrium water absorption rate of 5-8%; electrical conductivity shows a trend of first increasing and then stabilizing, with the ratio of equilibrium value to original electrical conductivity being 1.2-1.5; microstructure observed by electron microscopy reveals the formation of intercrystalline water films, but the crystal skeleton remains stable; compressive strength decreases by 10-15% after 24 hours of water immersion, but still remains above 85% of the design strength. Under dry conditions, the soil stabilizer exhibits the following characteristics: the surface shows a network of fine cracks, with crack density positively correlated with the drying rate; when the moisture content drops to 1-2%, the volume shrinkage rate is controlled within the range of 0.5-1%; infrared spectroscopy analysis shows a stable bond between crystal water molecules and colloidal structure; surface hardness increases by 15-20%, but the elastic modulus decreases slightly, exhibiting a certain degree of increased brittleness; the thermal conductivity under dry conditions is 25-30% lower than that under wet conditions. The system can collect real-time data on various parameter changes during wet-dry cycles, including apparent characteristics, strength changes, microstructure evolution, and fluctuations in physicochemical properties. It automatically identifies key inflection points and abnormal fluctuations during wet-dry cycles, such as the critical reduction point of compressive strength, the critical moisture content for microcrack formation, and the limiting number of cycles for structural stability. Based on the needs of different application scenarios (road base courses, foundation treatment, or hydraulic seepage prevention), it generates optimized formulations and process parameter adjustment schemes to improve the durability and stability of soil stabilizers under wet-dry cycle conditions. For road base course applications, the system focuses on monitoring changes in rebound modulus and fatigue characteristics under wet-dry cycles; for foundation treatment applications, it emphasizes monitoring the cyclic response of volumetric stability and shear strength; and for hydraulic seepage prevention applications, it mainly focuses on the dynamic evolution of permeability coefficient and micropore structure. Through these customized monitoring schemes, the system ensures that soil stabilizers maintain stable engineering performance under various extreme environments and cyclic loading conditions. Furthermore, this system can monitor the ion migration rate and redistribution characteristics within the solidifier during wet-dry cycles. By analyzing conductivity scanning spectra, it can predict and quantitatively assess the structural degradation mechanism by examining the dynamic changes in salt accumulation and dissolution regions. When a structural degradation trend caused by wet-dry cycles is detected, the system automatically generates reinforcement formulation suggestions, such as adjusting the proportion of hydrated calcium silicate, adding specific slow-release waterproofing agents, or introducing nano-reinforcing phases, thereby improving the material's resistance to cyclic damage from the source. Through monitoring and analyzing wet-dry cycle characteristics, this system achieves comprehensive quality assurance from production process monitoring to the entire product lifecycle, further enhancing the application reliability and engineering adaptability of soil stabilizers.

[0017] Step S104: Based on the set of production process target parameters, identify abnormal fluctuations in electrolyte concentration, micelle formation state, and crystal water fixation efficiency to obtain production quality risk warning information; Specifically, the set of target parameters for the production process is divided into three main types according to parameter properties and process logic: physical layer, chemical layer, and quality layer. Physical layer parameters include temperature curves and pressure distribution, reflecting macroscopic thermodynamic and mechanical changes during production. Chemical layer parameters include electrolyte concentration and micelle current, focusing on the molecular-level dynamic behavior of material mixing and reaction. Quality layer parameters focus on characteristic indicators directly related to the final product performance, such as crystallinity and humidity change rate. Targeted analysis is performed on key parameters at different levels to obtain anomaly measurement indicators. For example, for electrolyte concentration change curves, time series analysis is used to calculate their trend, fluctuation amplitude, and periodic characteristics, capturing abrupt changes, instability, or abnormally slow reaction dynamics. For micelle formation, spectral matching analysis is used to match real-time acquired characteristic spectra with a template library of ideal micelle formation processes, identifying phenomena such as absorption peak drift, disappearance, or abnormal intensity, thereby determining whether the micelle aggregation state is abnormal. Monitoring of crystal water fixation efficiency involves analyzing the humidity change rate, automatically extracting the stability range of the humidity reduction rate in the later stages of the crystallization reaction, and identifying periods of efficiency reduction or abnormal fluctuations. Anomaly detection is performed on various parameter metrics. Using statistical thresholds, trend analysis, and clustering algorithms, abrupt changes in electrolyte concentration, abnormal intervals in micelle formation, and significant decreases in water of crystallization fixation efficiency are automatically identified. These identified anomalies constitute quality risk warning points in the production process. To assess multivariate coupled risks, based on the anomaly identification results and considering the historical correlations and process logic between parameters at each level, the interrelationships between electrolyte concentration, micelle formation, and water of crystallization fixation efficiency are calculated, generating a parameter correlation impact map. This dynamically reflects the comprehensive risks arising from the coupling of multiple parameter changes and clearly identifies which combinations of process fluctuations are most likely to lead to quality hazards. Based on the parameter correlation impact map and the identified anomaly markers, and according to a pre-set risk assessment model, combined with real-time monitoring data and product target performance requirements, the production quality risk of the current batch is graded and assessed, providing a risk level assessment result. If the risk level exceeds the set safety threshold, a production quality risk warning is automatically triggered, generating complete risk information including the type of abnormal parameter, fluctuation range, impact relationship, and response suggestions.

[0018] Based on the anomaly labeling results, cross-correlation analysis was performed on three sets of time series data representing core variables of the chemical and mass layers: electrolyte concentration changes, micelle formation state, and water of crystallization fixation efficiency. This analysis considered the consistency of fluctuations between synchronous variables and focused on the response relationships between variables under time lags. Specifically, by constructing multiple lag step windows, the cross-correlation function between each pair of parameters was calculated, and the lag effect strength and direction of each variable on the other variable were extracted, generating a parameter time-lag effect strength table. This table records the lag point where the peak influence of different parameters on each other is located, the positive or negative correlation, and the significance level. Based on this time-lag effect strength table, a two-way causality test was conducted. Techniques such as Granger causality test, Transfer Entropy, or structural equation modeling were used to verify whether there is a statistically significant causal effect between each pair of parameters. For example, it was used to determine whether the electrolyte concentration change is "Granger causally" related to the micelle formation state, or vice versa, and whether there is a directionality of information flow. This process outputs a parameter causal relationship network, organizing the three types of parameters and their directional connections into a directed graph structure. Nodes represent process variables, edges represent paths with causal effects, and edge weights represent causal strength. This network structure reveals the potential regulatory mechanisms and mutual driving characteristics among the parameters. Based on this, sensitivity analysis is performed on the entire parameter causal network to assess the impact of each parameter node on the overall system output or stability. During the analysis, perturbation analysis and sensitivity propagation techniques are introduced. Within a set perturbation range, the values ​​of individual variables are changed one by one, and the changes in the responses of other variables are observed. Sensitivity scores are generated through gradient estimation or Monte Carlo simulation, forming a parameter sensitivity heatmap. This heatmap uses color intensity to represent the influence weight of variables on the system output, revealing which variables are in a core controlling position and which are in a responding or controlled position. The parameter sensitivity heatmap is integrated with the causal relationship network to form a structured parameter correlation influence map. During the fusion process, the influence weights in the sensitivity heatmap are mapped to the node strength and edge weight values ​​of the causal network, achieving a unified expression from structural logic to quantitative influence. The resulting parameter correlation influence map shows the causal paths and directions between key process parameters, and strengthens the influence strength of each path through sensitivity gradient.

[0019] Step S105: Based on the production quality risk warning information and combined with the differences in the performance requirements of soil stabilizers in different application scenarios, generate corresponding production process parameter adjustment schemes.

[0020] Specifically, based on the anomalies and fluctuation trends of parameters marked in the real-time triggered production quality risk warning information, and combined with the application scenario to which the current task belongs, the built-in target performance index database is invoked to extract the expected performance thresholds of this type of soil stabilizer in terms of strength, stability, durability, and permeability. A scenario-based quality requirement matching table, including performance requirements, parameter deviation directions, and response priorities, is constructed to provide rigid constraints and flexible guidance for subsequent optimization. Based on the target performance items defined in this matching table, instance and relational layer data regarding the influence relationship between parameters and quality are invoked from the soil stabilizer production knowledge graph to automatically generate a parameter-quality mapping network for the current working condition. This network uses key process parameters in the production process as nodes, including raw material ratios, reaction temperature, micelle formation rate, crystallization time, and compaction density, while using target quality attributes as correlation ports to map the intensity and directionality of the influence of each parameter on different performance indicators, forming a multivariate coupled control model. Based on this, key process parameters are set as optimization variables, constructing a multi-constraint optimization problem with the objective function of "maximizing the quality indicator satisfaction rate of the current application scenario." Efficient search strategies such as Bayesian optimization, genetic algorithms, or particle swarm optimization are employed to perform iterative calculations in the parameter space. This evaluates the performance response and risk mitigation capabilities under different parameter combinations, generating a set of candidate parameter adjustment schemes that meet fundamental constraints. Based on this set, a multi-objective non-dominated optimization mechanism is introduced to evaluate the comprehensive performance of each scheme within the candidate set under three core objectives: maximizing finished product strength, minimizing production energy consumption, and optimizing environmental performance. By constructing a Pareto front, the interrelationships between objectives are solved non-dominatedly, selecting the optimal multi-objective balance optimization scheme that balances performance and sustainability. The corresponding production process parameter adjustment scheme is automatically generated based on the multi-objective balance optimization results. This includes specific operational instructions such as fine-tuning of raw material ratios, correcting gradients of temperature curves, setting values ​​for stirring rates, and adjusting the range of compaction pressure and time. Safety boundary constraint logic is embedded to ensure the adjustment scheme is executable within physical safety and equipment capacity limits. The adjustment scheme is distributed to the corresponding equipment execution units via the industrial control network, and the parameter response effect is fed back in real time by an IoT sensor network, achieving dynamic closed-loop control of the process.

[0021] In this embodiment, a target evaluation index system is defined based on the parameter adjustment candidate scheme set. The finished product strength dimension is refined into compressive strength, resilient modulus, and shear strength; production energy consumption is refined into multiple energy consumption factors such as electricity consumption, heat consumption, and water resource consumption; and environmental performance is refined into environmental impact parameters such as exhaust gas emissions, wastewater discharge, and solid waste generation. Based on the target evaluation index system and the actual application scenarios of soil stabilizers (such as road base, foundation treatment, or water conservancy seepage prevention), weights are set for each index to form a weight vector, taking into account the different emphases of performance, energy consumption, and environmental protection in different industries or projects. According to the weights, the quantitative performance of all parameter adjustment candidate schemes on each evaluation index is weighted and scored, and a scheme scoring matrix is ​​generated. Based on the scheme scoring matrix, a classic non-dominated ranking algorithm (such as NSGA-II) is used to rank all schemes according to non-dominated relationships. The core of non-dominated ranking is to divide all candidate schemes into several layers according to whether they are superior to other schemes on certain objectives and not inferior to other schemes on the remaining objectives. Schemes within each layer cannot dominate each other, thus forming a dominance relationship graph. Based on the dominance graph, a multi-level Pareto front structure is extracted, where each front represents a set of efficient solutions that are not interchangeable under the objective trade-offs. This multi-level Pareto front structure reflects the gradient of merits and demerits of solutions and the trade-off space under multi-objective optimization. A congestion ranking algorithm is executed based on the multi-level Pareto front structure to analyze the density and sparsity of each solution in the solution space, thereby encouraging solution diversity and avoiding local optima. Specifically, combined with the congestion quantification results, the distances from each solution to the ideal solution (i.e., the set of all optimal objective values) and the negative ideal solution (the set of all worst objective values) are normalized. By calculating the ratio of each solution's distance to the ideal solution and the negative ideal solution, its overall proximity in the multi-objective optimal solution space is evaluated. The candidate solution with the largest distance ratio is selected for parameter adjustment, which is the multi-objective balanced optimization solution that achieves the optimal balance between the three objectives of maximizing product strength, minimizing energy consumption, and optimizing environmental performance.

[0022] In this embodiment of the invention, modular IoT sensors with dustproof and waterproof properties are deployed at key locations on the production line. A tree-like topology and multi-protocol compatible communication parameter settings form a comprehensive, stable, and reliable monitoring network, solving the problems of signal interference and data acquisition blind spots in traditional monitoring. Based on the raw monitoring data, production process stages are divided, and a parameter sensitivity assessment model and differentiated monitoring frequency configuration are established to achieve accurate monitoring of key parameters at different production stages, avoiding data redundancy and omissions of key information caused by traditional fixed-interval sampling methods. By implementing a four-level sampling strategy (ultra-high frequency, high frequency, standard, and low frequency) combined with a dynamic sampling frequency adjustment mechanism, high-precision monitoring of key process points is ensured while significantly reducing sensor network energy consumption and extending equipment lifespan. Using a multi-level parameter decomposition and anomaly detection model, combined with multivariate correlation analysis, abnormal fluctuations in electrolyte concentration changes, micelle formation state, and crystal water fixation efficiency are accurately identified, allowing for early detection of potential quality risks and changing the traditional passive monitoring mode that relies on experience-based judgment. Based on production quality risk early warning information and combined with the performance requirements of different application scenarios, the optimal process parameter adjustment scheme that balances the strength, energy consumption and environmental performance of finished products is generated through multi-objective evaluation and Bayesian optimization algorithm, realizing intelligent and precise control of process parameters.

[0023] In one specific embodiment, the process of performing step S101 may specifically include the following steps: An IoT sensor network is deployed at the locations of the raw material mixer, reaction vessel, and compaction molding machine in the soil stabilizer production line, and a unique identifier and physical location marker are assigned to each sensor in the IoT sensor network to obtain an initial sensor deployment scheme; Configure the communication protocol for each sensor in the initial sensor deployment scheme to obtain the sensor communication network; Based on the sensor communication network, the production line is divided into raw material area, reaction area and molding area, and a gateway node is configured for each area to collect data, resulting in a sub-regional network topology. Low-noise sensor data streams are collected based on regional network topology maps, and the data change rate and signal quality index of each monitoring point are calculated based on the low-noise sensor data streams to generate a dynamic sampling weight table. Data acquisition is performed based on a dynamic sampling weight table and an IoT sensor network to record raw monitoring data of the production process, including temperature, humidity, pressure, conductivity, and characteristic spectral data.

[0024] Specifically, a systematic IoT sensor network deployment is implemented at key nodes in the production line, such as the raw material mixer, reactor, and compaction molding machine. Considering the process characteristics and environmental variables of each stage of the production line, various types of sensors, including those for temperature, humidity, pressure, conductivity, and characteristic spectra, are distributed at these nodes. This ensures coverage of the core process while also accommodating environmental complexity and ease of maintenance. Each sensor is assigned a unique identifier and a physical location marker accurate to its actual position on the production line through the system software platform, resulting in an initial sensor deployment plan. A unified communication protocol is configured for all types of sensors. Due to the strong electromagnetic interference and complex spatial obstacles in the soil stabilizer production environment, appropriate industrial IoT communication protocols, such as NB-IoT, LoRa, Wi-Fi, and Ethernet, are selected based on the actual site conditions. Parameters such as the access method, data format, encryption mechanism, communication frequency band, and transmission rate for each sensor are configured to ensure seamless connection and interoperability between sensors from different manufacturers and of different types. Simultaneously, the communication protocol configuration must be compatible with the access standards of the backend edge computing and data cloud platform to ensure uninterrupted data flow. Through standardized communication protocols, sensor data achieves functions such as second-level reporting, event-level response, and breakpoint resumption. Based on the sensor communication network, the entire production line is divided into typical process areas such as raw material area, reaction area, and forming area according to the process flow and spatial distribution. Each area is configured with a gateway node according to the sensor distribution and data flow characteristics. The gateway node uses a high-performance ARM or x86 architecture industrial computing platform, which has local edge computing capabilities and multi-protocol data aggregation functions, and can perform real-time acquisition, preliminary cleaning, and compression of the raw data stream uploaded by the sensors in its area. The gateways in each area are interconnected through high-bandwidth industrial Ethernet, wireless mesh network, or enterprise dedicated lines to form a self-organizing and self-recovering regional network topology. Low-noise sensor data streams are collected based on the regional network topology. Low-noise sensors are high-precision monitoring devices that have undergone rigorous calibration and hardware anti-interference design. The fluctuations in their output signals mainly originate from actual process changes rather than environmental or electronic noise. The raw data is uploaded to the central monitoring platform through the regional gateways, and the data change rate and signal quality indicators are calculated in real time based on the data streams of each monitoring point. The data change rate reflects the dynamic fluctuations of the production process on a specific process variable. Signal quality indicators are measured using statistical methods such as signal-to-noise ratio, missing value rate, and outlier percentage to determine the current health status of the sensors and data availability. Based on the analysis results of the data change rate and signal quality, an adaptive dynamic sampling algorithm is adopted to automatically generate sampling weights for each monitoring point, forming a dynamic sampling weight table. This weight table prioritizes monitoring points with drastic data changes, critical processes, or high signal risks, allocating them higher sampling frequencies and resources, while appropriately reducing sampling density for less sensitive points. Based on the dynamic sampling weight table and the IoT sensor network, differentiated and refined data acquisition is implemented.Each sensor uploads raw monitoring data of the production process, including temperature, humidity, pressure, conductivity, and characteristic spectrum, according to the assigned sampling strategy at regular intervals, and adds a unique sensor identifier, physical location, timestamp, and signal quality label to each data point.

[0025] In one specific embodiment, the process of performing step S102 may specifically include the following steps: Time series analysis was performed on the original monitoring data of the production process to obtain the feature point set of the entire production process of soil stabilizer; Based on the feature point set of the entire production process, the process stages are divided to obtain the boundary markers of the production stages; Based on the boundary markers of the production stages, the soil stabilizer production process is divided into the raw material proportioning stage, the chemical reaction stage, and the finished product quality stage. Electrical conductivity and stirring torque monitoring parameters are assigned to the raw material proportioning stage, characteristic spectrum and temperature monitoring parameters are assigned to the chemical reaction stage, and humidity and compressive strength monitoring parameters are assigned to the finished product quality stage, resulting in a stage parameter mapping table. Based on the stage parameter mapping table, parameter sensitivity is evaluated, a parameter sensitivity matrix is ​​generated, and the parameter sensitivity matrix is ​​monitored in a hierarchical manner to obtain differentiated monitoring frequency configuration parameters. By associating the differentiated monitoring frequency configuration parameters with the ideal parameter ranges of three types of soil stabilizers—road base type, foundation treatment type, and water conservancy seepage prevention type—a phased monitoring strategy is generated.

[0026] Specifically, IoT sensors distributed across key nodes of the production line collect multi-dimensional raw monitoring data in real time and continuously, including temperature, humidity, pressure, conductivity, stirring torque, characteristic spectra, and compressive strength. Time series analysis is then performed on this raw monitoring data. Using sliding window statistics and dynamic segmentation techniques, the entire production process is divided into several data segments. Within each segment, characteristic points such as extreme points, abrupt changes in the rate of change, periodic inflection points, and abnormal fluctuations are extracted, forming a set of characteristic points for the entire production process. For example, in the raw material proportioning stage, sudden rises or falls in the conductivity curve identify the instantaneous points of raw material input, and peak points in the stirring torque identify the moment when the mixture reaches a homogeneous state. In the chemical reaction process, absorption peak shifts in the characteristic spectra identify the reaction progression stages, and changes in the slope of the temperature curve capture exothermic and endothermic reaction nodes. In the finished product quality stage, the key process stages of water of crystallization fixation and strength formation are identified based on the stable range of humidity change rates and the stable rise points in compressive strength data. Based on the feature point set of the entire production process, combined with process logic and expert knowledge, the entire production process is precisely divided into stages, resulting in boundary markers for each stage. Using intelligent algorithms based on breakpoint detection or segmented clustering, feature points of raw material input and mixing are defined as the start and end markers of the raw material proportioning stage. The characteristic spectra and temperature curve inflection points of the chemical reaction process are used as the boundaries of the chemical reaction stage. The stable humidity data zone and the critical point of compressive strength growth are used as markers for the finished product quality stage, thus achieving a structured division in time sequence. Based on these stage boundary markers, a stage-specific parameter mapping table is established. The raw material proportioning stage is assigned key monitoring parameters such as conductivity and stirring torque to strengthen control over mixing uniformity and raw material ratio accuracy. The chemical reaction stage is assigned characteristic spectra and temperature curves to focus on micelle formation and the thermodynamic process of the reaction. The finished product quality stage primarily monitors humidity and compressive strength to reflect the physical properties and curing effect of the finished product. Based on the monitoring parameters assigned to each stage, historical data and online volatility analysis are used to assess parameter sensitivity. This assessment employs statistical analysis and machine learning models to quantify the impact of each monitoring parameter on final product quality, process stability, and energy consumption indicators, forming a parameter sensitivity matrix. The parameter sensitivity matrix is ​​then categorized: for highly sensitive parameters, such as conductivity in the raw material proportioning stage, temperature curves in the chemical reaction stage, and compressive strength in the finished product quality stage, high sampling frequencies and anomaly monitoring weights are assigned; for medium- and low-sensitivity parameters, sampling frequencies are appropriately reduced to save system energy and data bandwidth. Through this categorized monitoring mechanism, differentiated monitoring frequency configuration parameters are obtained. These differentiated monitoring frequency configuration parameters are then correlated with the ideal process parameter ranges for three typical soil stabilizers: road base layer type, foundation treatment type, and water conservancy seepage prevention type.By mapping the performance requirements of different application scenarios (such as the requirements for strength and durability for the base type, the requirements for permeability and uniformity for the foundation type, and the requirements for density and water stability for the impermeable type) to the parameter monitoring priority and frequency configuration, the system automatically adjusts the monitoring strategy and pushes the most suitable phased monitoring scheme for the current product type and process flow.

[0027] In one specific embodiment, the process of dividing the production process into stages based on the feature point set of the entire production process to obtain the boundary markers of the production stages can specifically include the following steps: A sliding window analysis was performed on the feature point set of the entire production process to obtain the parameter change feature sequence; The conductivity data in the parameter variation feature sequence is processed by first-order differentiation to identify conductivity abrupt change points, and the region between adjacent conductivity abrupt change points is divided into conductivity abrupt change intervals. Wavelet transform is performed on the temperature curve data in the parameter change feature sequence to extract the local extreme points and inflection points of the temperature curve data, and the inflection point interval of the temperature curve is generated based on the local extreme points and inflection points. The stationarity of humidity sensor data in the parameter change characteristic sequence is tested to obtain the humidity change rate, and the humidity stability interval is determined based on the humidity change rate. Based on the temporal pattern recognition of the execution time distribution in the conductivity abrupt change interval, temperature curve inflection point interval, and humidity stable interval, a process stage transition model is constructed. Based on the process stage transition model, the conductivity mutation point is taken as the boundary between the raw material proportioning stage and the chemical reaction stage, and the area between the last inflection point of the temperature curve data and the starting point of the humidity stability range is taken as the boundary between the chemical reaction stage and the finished product quality stage, thus generating production stage boundary markers.

[0028] Specifically, a sliding window analysis method is applied to the feature point set of the entire production process. This analysis is based on multivariate time-series data collected by sensors, dividing the time series into a series of sliding windows of equal length, overlapping or non-overlapping. Within each window, various features such as the mean, variance, extreme values, and rate of change of parameters are statistically analyzed and extracted, thus transforming the raw data stream into a parameter change feature sequence. This method enhances the detection sensitivity of subtle but significant changes in the production process and provides a high-density information foundation for subsequent refined segmentation. For example, the original conductivity curve over time, after sliding window processing, reveals localized areas of dramatic fluctuation; the temperature curve clearly shows the dynamic processes of local heating and cooling within each stage; and humidity data reveals the stable sections of finished product crystallization and drying. Based on this, the conductivity data in the parameter change feature sequence is processed using the first derivative to reveal the rate of change of the conductivity curve, enabling the identification of conductivity abrupt changes caused by external disturbances or chemical changes in stages such as raw material addition, mixing, and preliminary reaction. When the first derivative exceeds a threshold or exhibits a maximum or minimum value, it is identified as a conductivity abrupt change point. By continuously tracking the time or process regions between adjacent conductivity abrupt change points, these time periods are defined as conductivity abrupt change intervals. This partition corresponds to the dynamic transition stage of raw material mixing or the phase transition moment in the reaction system, which is a core sensitive node in the production process. For temperature curve data, wavelet transform is used for multi-scale decomposition to separate local extrema and trend inflection points from complex, non-stationary temperature curves. Through wavelet decomposition, the influence of environmental noise on temperature measurement is filtered out, and drastic temperature fluctuations caused by exothermic reactions, endothermic processes, and external heating regulation are captured. The system automatically extracts all local maxima, minima, and curvature abrupt change points, and divides the temperature data into multiple inflection point intervals accordingly. Each interval reflects the transformation of physical or chemical mechanisms in the production process. Temperature inflection point intervals are highly correlated with key chemical reaction nodes such as micelle formation, reaction acceleration, or termination, and are the main basis for identifying the dynamic characteristics of chemical reaction stages. When analyzing humidity sensor data, stationarity testing methods, such as unit root tests and analysis of variance, are introduced to determine the changing trend and fluctuation amplitude of the humidity time series in different intervals. The rate of change in humidity is a crucial parameter for determining the efficiency of water of crystallization fixation and the degree of dryness. When the rate of change in humidity data gradually decreases and tends to stabilize, it indicates that the system has entered a stable range for finished product quality. Therefore, the continuous, minimally fluctuating segments in the data are designated as the humidity stability range, representing the termination of the crystallization process and the stable formation of physical properties. Based on the temporal pattern recognition of the execution time distribution of conductivity abrupt change ranges, temperature curve inflection point ranges, and humidity stability ranges, a process stage transition model is constructed by comprehensively comparing the sequence, duration, and overlap characteristics of each range in the production process. This model, based on the distribution characteristics of different parameter ranges, the dynamic correlation between parameters, and the physical logic between process steps, achieves automated division of raw material ratios, chemical reactions, and finished product quality stages.In practice, the point of sudden change in conductivity is used as the boundary between the raw material proportioning stage and the chemical reaction stage. The time period between the last inflection point of the temperature curve and the starting point of the humidity stabilization range is set as the boundary between the chemical reaction stage and the finished product quality stage, thereby generating production stage boundary markers.

[0029] In one specific embodiment, the process of executing step S103 may specifically include the following steps: Based on the phased monitoring strategy, sampling frequencies are allocated for the conductivity change rate in the raw material ratio stage, the temperature curve monitoring in the chemical reaction stage, and the parameter monitoring in the finished product quality stage, resulting in a phased sampling control instruction set. In the Internet of Things sensor network, each gateway node issues a set of phased sampling control instructions and performs differentiated data acquisition on each sensor to obtain raw sensor data with multiple sampling frequencies. Based on raw sensor data with multi-level sampling frequencies, and combined with the target performance requirements of soil stabilizers for road base, foundation treatment, or water conservancy seepage prevention, a set of production process target parameters matching the current production type is selected.

[0030] Specifically, in the phased monitoring strategy design phase, based on the typical process flow of soil stabilizer production, the production process is divided into three stages: raw material proportioning, chemical reaction, and finished product quality. Each stage focuses on different physical or chemical variables, therefore, the most representative monitoring parameters that have the greatest impact on product performance are configured for each stage. In the raw material proportioning stage, the most critical aspect is the uniformity and accuracy of the raw material mixing ratio; the rate of change in electrical conductivity, as a high-sensitivity indicator, is given the highest monitoring priority. In the chemical reaction stage, the temperature curve changes of the reaction system best reflect the reaction rate, micelle formation, and molecular structure transformation; temperature monitoring parameters and their dynamic change rates become the focus. In the finished product quality stage, physical indicators such as humidity, water of crystallization fixation, and compressive strength are the main targets, reflecting the stability and compliance rate of the final physicochemical properties of the stabilizer. Based on the phased objectives and according to real-time process status and historical sensitivity assessment results, differentiated sampling frequencies are allocated. All sampling frequency allocations are generated into a phased sampling control instruction set through an automatic parameter mapping table. Each gateway node in the IoT sensor network issues the phased sampling control instruction set. Each gateway node is responsible for the dynamic management of the sampling frequency and the synchronous adjustment of the acquisition strategy for the sensor groups within its jurisdiction or stage. After receiving the control command set from the upper-level platform or algorithm module, the gateway distributes the corresponding sampling configuration parameters to various types of sensors, including settings for sampling interval, upload cycle, and data preprocessing parameters. The sensor network performs differentiated data acquisition based on its respective sampling tasks: monitoring points in high-dynamic fluctuation zones enter high-frequency data acquisition mode, while monitoring points of lower importance or with slow fluctuations enter standard or low-frequency sampling mode. The acquired multi-level sampling frequency raw sensor data undergoes preliminary preprocessing and tagging through the gateway nodes, forming a high-dimensional data stream with sampling timestamps, parameter types, stage attributes, and signal quality, and is archived in the central database or edge computing nodes. Based on data content, acquisition stage, and actual process status, target data segments that conform to the current process mainline are dynamically selected. Examples include conductivity change curves due to abnormal raw material ratios, temperature fluctuation ranges with abnormal reactions, and trends in finished product humidity and strength compliance. Based on data screening, and to achieve precise adaptation to actual engineering application scenarios, target performance index requirements are introduced for various soil stabilizer application types, such as road base course, foundation treatment, and water conservancy seepage prevention. These indicators include multiple quantitative and qualitative constraints such as strength grade, durability, permeability coefficient, and environmental adaptability. Based on production orders, batch identifiers, or on-site application settings, the ideal process parameter range for the current product type is automatically associated and retrieved. The collected multi-level raw data is matched, filtered, and normalized with the target parameters to extract the production process target parameter set that best matches the process requirements.

[0031] In one specific embodiment, the process of executing step S104 may specifically include the following steps: The production process target parameter set is divided into physical layer parameters, chemical layer parameters, and quality layer parameters. The physical layer parameters include temperature curves and pressure distribution, the chemical layer parameters include electrolyte concentration and micelle current, and the quality layer parameters include crystallinity and humidity change rate, thus obtaining a layered parameter structure. Based on the hierarchical parameter structure, time series analysis was performed on the electrolyte concentration change curve, spectral matching analysis was performed on the characteristics of micelle formation state, and humidity change rate was calculated for the crystal water fixation efficiency to obtain parameter anomaly measurement index. Parameter anomaly detection is performed on the parameter anomaly measurement index to identify abrupt changes in electrolyte concentration, abnormal ranges in micelle formation, and ranges in which crystal water fixation efficiency decreases, thus obtaining parameter anomaly labeling results. Based on the results of parameter anomaly labeling, the interaction between electrolyte concentration changes, micelle formation state and crystal water fixation efficiency is calculated, and a parameter correlation influence map is generated. A quality risk assessment is conducted based on the parameter correlation impact map and parameter anomaly labeling results to obtain the risk level assessment results. Threshold judgments are then made on the risk level assessment results to generate production quality risk early warning information.

[0032] Specifically, the set of production process target parameters is divided into a three-layer structure based on different physical, chemical, and final quality attributes. The physical layer parameters focus on reflecting the macroscopic state of raw materials, mixing, and molding processes; temperature curves reflect the dynamic evolution of thermodynamic processes; and pressure distribution reveals real-time changes in physical compaction and rheology. The chemical layer parameters characterize the molecular-level reactions and transformations within the material; electrolyte concentration reflects ion migration, solution homogeneity, and the adequacy of material blending; and micelle current represents the formation, destruction, and structural changes of surfactants or polymer clusters in the dispersion system. The quality layer parameters focus on the performance of the final product; crystallinity reflects the degree of material densification and phase transformation; and humidity change rate quantifies key processes affecting durability and mechanical properties, such as moisture removal and crystal water fixation. Based on this layered parameter structure, different data mining and intelligent analysis techniques are applied to each layer of parameters to form parameter anomaly measurement indicators. For electrolyte concentration change curves, time series analysis was performed. Methods such as autoregressive moving average models, sliding window statistics, or dynamic time warping were used to identify typical characteristics such as sharp fluctuations, plateaus, and abrupt changes in the curves. The change process was modeled by calculating higher-order statistics such as mean, variance, slope, and sharpness. For micelle formation, online feature spectral matching was used to perform correlation analysis between the current spectrum and ideal spectral types in the process template library, detecting absorption peak drift, intensity reduction, and spectral anomalies during micelle formation or depolymerization. For water of crystallization fixation efficiency, humidity change rate was continuously monitored. Piecewise linear fitting and stationarity tests were performed on the humidity curve to extract high-variance intervals, inflection points, and stabilization segments, quantifying them as water fixation and water loss efficiency indicators. All abnormal parameter metrics were output as structured data streams. Parameter anomaly detection was performed on these metrics. Statistical thresholding, cluster anomaly detection, and machine learning classification methods were integrated to uniformly identify anomalies in all parameter metrics. For electrolyte concentration, abrupt changes exceeding the historical distribution mean by multiple standard deviations are detected to identify scenarios such as raw material blending, uncontrolled mixing processes, and abnormal ion migration. For micelle formation, indicators such as spectral template mismatch, abnormal disappearance of absorption peaks, or drastic changes in peak area are used to identify instability, incomplete reaction, or excessive dispersion during micelle formation. For water of crystallization fixation efficiency, a significant decrease or persistent failure to meet humidity change rate is monitored to pinpoint risks of excessive moisture content or unstable crystal structure in the finished product. All detected anomalies are assigned timestamps, parameter categories, severity levels, and process stages, forming parameter anomaly labeling results. Based on these results, correlation analysis, causal inference, and network modeling are used to calculate the interrelationships between electrolyte concentration changes, micelle formation status, and water of crystallization fixation efficiency. For example, Granger causality tests or dynamic Bayesian networks are used to determine whether a sudden change in electrolyte concentration at a given moment necessarily leads to a decrease in micelle formation efficiency, thereby causing water of crystallization fixation failure.The time-lag influence strength, bidirectional causal network, and sensitivity heatmap of all parameters are automatically integrated into a parameter correlation influence map. This map visually displays the direct, indirect, and multi-level influences between variables in the form of nodes and directed edges. Based on the parameter correlation influence map and anomaly labeling results, the system proceeds to the quality risk assessment and early warning stage. The system designs a multi-objective risk assessment model that considers key parameter anomalies, correlation path anomalies, and their weighted impact on the final product performance. It employs various methods such as fuzzy comprehensive discrimination, threshold grading, and machine learning prediction to assess the risk level for each production batch and each monitoring window. If the correlation probability and influence strength between a certain type of parameter anomaly and the target performance indicator (such as compressive strength, durability, and permeability coefficient) exceed a threshold, the risk level is automatically increased, and the system immediately generates a production quality risk early warning message.

[0033] In one specific embodiment, the process of performing the step of calculating the interaction between electrolyte concentration change, micelle formation state, and water of crystallization fixation efficiency based on parameter anomaly labeling results, and generating a parameter correlation influence map, can specifically include the following steps: Based on the results of parameter anomaly labeling, cross-correlation analysis was performed on electrolyte concentration changes, micelle formation state and crystal water fixation efficiency to obtain a table of parameter time lag influence intensity. A bidirectional causal relationship test was conducted based on the parameter time delay influence intensity table to obtain the parameter causal relationship network; Sensitivity analysis is performed based on the parameter causal relationship network to generate a parameter sensitivity heatmap; By fusing parameter sensitivity heatmaps with parameter causal relationship networks, a parameter correlation influence map is generated.

[0034] Specifically, based on high-frequency raw monitoring data collected at each stage of the soil stabilizer production process, combined with previous parameter anomaly detection and labeling results, the time series sequences of three parameters with the greatest impact on the final product performance under different stages and application scenarios—electrolyte concentration changes, micelle formation state, and crystal water fixation efficiency—were selected. To reveal the temporal mutual responses and influences among these key parameters, cross-correlation analysis was employed to systematically calculate the full-time-domain correlation between each pair of parameters (e.g., electrolyte concentration and micelle formation state, micelle formation state and crystal water fixation efficiency, electrolyte concentration and crystal water fixation efficiency). The cross-correlation function can quantitatively characterize the time lag, intensity, and direction of the response of one parameter to another. For example, if the electrolyte concentration change at the current moment has a strong positive correlation with the micelle formation state 5 minutes later, it indicates that the change in the former, at this time lag, has a stimulating or driving effect on the latter. Similarly, the multi-time-lag influences of all parameters on themselves and each other were calculated, forming a parameter time-lag influence intensity table. Each item in the table includes the parameter pair, time lag window, correlation coefficient, and significance test results. A bidirectional causal relationship test is performed based on the parameter time-delay influence intensity table. Methods such as Granger causality test, Transfer Entropy, and dynamic Bayesian networks are used to perform bidirectional causal inference on the influence path between each pair of parameters. This not only determines whether the change in electrolyte concentration is a "Granger factor" in the micelle formation state, but also whether micelle formation has a reciprocal effect on electrolyte concentration, ensuring the omnidirectionality and integrity of the parameter relationship network. The causal test requires the parameter sequence to be stationary or stationary after differential processing, and considers multiple time delays and noise effects, outputting a parameter causal relationship network. This network consists of nodes (the parameters themselves) and directed edges (causal action paths). Each edge has a direction (causal driving direction) and a weight (causal influence intensity), illustrating the driving chain and feedback path of the core parameters throughout the production process. For example, if the change in electrolyte concentration not only affects the micelle formation state but also indirectly affects the crystal water fixation efficiency through the latter, the system will automatically construct this multi-level causal path, achieving a digital reconstruction of the complex process mechanism. Sensitivity analysis is then performed based on the parameter causal relationship network. Sensitivity analysis assesses the impact of changes in various parameters within a network on the overall output or key quality indicators. It employs methods such as local perturbation (e.g., observing the response of other parameters to minor perturbations of a single parameter), global sensitivity analysis (e.g., Sobol decomposition), or edge-weighted cumulative analysis based on causal networks. Through sensitivity analysis, the weight, controllability, and key role of each core parameter (e.g., electrolyte concentration, micelle formation state, and water of crystallization fixation efficiency) on the final process quality are quantitatively revealed. For example, if the sensitivity heatmap shows that perturbations in the micelle formation state have a very high weight on the water of crystallization fixation efficiency and final crystallinity, while electrolyte concentration mainly affects the finished product quality through indirect pathways, then process optimization will focus more on process control of the micelle formation stage.The generated parameter sensitivity heatmap uses parameters as coordinates and employs color gradients to represent the intensity of each parameter's influence on the target variable or the overall system output; darker colors indicate higher sensitivity. The sensitivity heatmap is then fused with a parameter causal relationship network to generate a parameter correlation influence map. The sensitivity scores of each parameter in the heatmap are mapped to the node weights of the causal network nodes, while the edge weights of the causal relationship network are used as path strengths, resulting in a multi-level, multi-attribute parameter correlation influence map.

[0035] In one specific embodiment, the process of executing step S105 may specifically include the following steps: Based on the production quality risk early warning information and the target performance index requirements of three types of soil stabilizers—road base type, foundation treatment type, and water conservancy seepage prevention type—a scenario-based quality demand matching table is constructed. Based on the scenario-based quality requirement matching table, the parameter influence rules under the current production scenario are queried from the soil stabilizer production knowledge graph to generate a parameter-quality mapping network. Based on the parameter-quality mapping network, multiple process parameters are set as optimization variables. The objective function is the quality index satisfaction rate of the current application scenario. Iterative optimization calculations are performed to obtain a set of candidate solutions for parameter adjustment. The candidate solution set is adjusted according to the parameters, and a non-dominated solution is performed to maximize the strength of the finished product, minimize the production energy consumption, and optimize the environmental performance, so as to obtain an optimal solution with multi-objective balance. Based on the multi-objective balance optimization scheme, a corresponding production process parameter adjustment scheme is generated.

[0036] Specifically, the raw monitoring data collected by IoT sensors during the production process is fused and analyzed with real-time quality risk warning information to identify key risk points affecting the quality stability or process consistency of the current batch. Combining process knowledge base and empirical rules, these risk points are matched with the target performance indicators of various soil stabilizer products. For road base courses, the focus is on compressive strength, durability, and rapid setting; for foundation treatment, the focus is on impermeability, long-term stability, and construction adaptability; and for water conservancy seepage prevention, the focus is on density, permeability coefficient, and environmental indicators. Through parameter database and historical process data analysis, a multi-dimensional set of target performance indicators is established for different product types, and these indicators are refined to specific numerical ranges, levels, weights, etc., forming a structured, scenario-based quality requirement matching table. Based on the scenario-based quality requirement matching table, parameter influence rules are retrieved from the soil stabilizer production knowledge graph. The knowledge graph, as a big data semantic network of the production process, contains causal and correlational relationships between numerous parameters such as raw material ratios, reaction temperature, pressure, stirring rate, and additive ratios and finished product performance indicators. By leveraging product scenarios and target performance indicators in a matching table, relevant nodes and edges in the knowledge graph are intelligently retrieved, automatically aggregating a mapping network centered on "parameter-quality." This mapping network not only reveals the direct and indirect impact paths of various production parameters on end-product performance such as strength, density, impermeability, and durability, but also dynamically adjusts edge weights and node priorities based on current process risks and historical batch data, achieving real-time adaptive knowledge updates. Based on the parameter-quality mapping network, key process parameters affecting product performance (such as raw material water-cement ratio, binder content, reaction temperature, stirring speed, molding pressure, and curing time) are used as optimization variables to establish an intelligent iterative optimization model with the satisfaction rate of various quality indicators in the current application scenario as the objective function. This model employs artificial intelligence optimization algorithms such as Bayesian optimization, genetic algorithms, and multi-objective particle swarm optimization to simulate and evaluate the target performance indicators mapped to each set of parameters in the knowledge graph through continuous or discrete search of the process parameter space. In each optimization iteration, the model automatically evaluates the multi-dimensional impact of parameter combinations on finished product strength, density, energy consumption, and environmental indicators, and uses the satisfaction rate or deviation as a fitness function to automatically select high-quality parameter combinations, gradually approaching the optimal solution. Through adaptive iteration, the system can converge efficiently, forming a set of candidate parameter adjustment schemes. Based on the set of candidate parameter adjustment schemes, non-dominated solutions are performed to maximize finished product strength, minimize production energy consumption, and optimize environmental performance. The performance of each parameter scheme under different objectives is mapped to a multi-objective evaluation space through methods such as normalization, weighting, and clustering. The system automatically identifies all Pareto optimal solutions that are not mutually dominant. By analyzing the feasibility of each equilibrium solution under actual production conditions, equipment capacity, and management strategies, the optimal multi-objective parameter solution is finally selected. The optimal multi-objective equilibrium parameter solution is automatically transformed into a structured and operable production process parameter adjustment scheme.The plan specifies the adjustment range, target value, execution sequence, and control scope for each key process variable, and includes targeted intervention suggestions, emergency plans, and automatic closed-loop feedback mechanisms for each quality risk point.

[0037] In one specific embodiment, the process of adjusting the candidate solution set according to the parameters and performing non-dominated solutions to maximize finished product strength, minimize production energy consumption, and optimize environmental performance to obtain a multi-objective balanced optimal solution can specifically include the following steps: The target evaluation index system is defined based on the parameter adjustment candidate scheme set. The target evaluation index system includes the compressive strength, resilient modulus, and shear strength of the finished product strength factors; the production energy consumption factors include electricity consumption, heat consumption, and water consumption; and the environmental performance factors include exhaust gas emissions, wastewater emissions, and solid waste generation. Based on the target evaluation index system and the weight settings of the execution index according to the characteristics of soil stabilizer application scenarios, the scheme scoring matrix is ​​obtained; The non-dominated sorting algorithm is executed based on the scheme scoring matrix to obtain the dominance graph, and a multi-level Pareto front structure is generated based on the dominance graph. The congestion ranking algorithm is executed based on the multi-level Pareto front structure to obtain the distribution characteristic quantification result. Combined with the distribution characteristic quantification result, the distance ratio from each parameter adjustment candidate to the ideal solution and the negative ideal solution is calculated in the parameter adjustment candidate set. The parameter adjustment candidate with the largest distance ratio is selected as the multi-objective balance optimization scheme to ensure that the multi-objective balance optimization scheme achieves the best balance point among the three objectives of maximizing intensity, minimizing energy consumption and optimizing environmental performance.

[0038] Specifically, a target evaluation index system is defined based on the parameter adjustment of the candidate scheme set. The first layer of the index system focuses on the strength factors of the finished product, including compressive strength, resilient modulus, and shear strength. These three factors jointly determine the load-bearing capacity, deformation response, and shear stability of the soil stabilizer in actual engineering, and can quantitatively evaluate the structural performance and engineering adaptability of the stabilizer. The second layer of indicators focuses on production energy consumption factors, covering electricity consumption, heat consumption, and water consumption. This dimension reflects the energy utilization efficiency and water resource intensification of the entire production process. The third layer is the environmental performance factor, including exhaust gas emissions, wastewater discharge, and solid waste generation, reflecting the multidimensional impact on the environment during the production process. The index weights are set based on the target evaluation index system and the characteristics of the soil stabilizer application scenario. The differences in scenario requirements are considered. For example, road base courses have higher strength requirements, while water conservancy seepage prevention courses have greater weights on environmental protection and durability. The index weight vector is obtained through expert experience, historical data analysis, or the analytic hierarchy process. The performance of all parameter-adjusted candidate solutions on each metric is multiplied by their corresponding weights to form a solution scoring matrix. Each row of this matrix corresponds to a solution, and each column corresponds to a weighted metric score. Based on the solution scoring matrix, a non-dominated ranking algorithm (such as NSGA-II) is introduced to evaluate the merits of each parameter-adjusted candidate solution in the multi-objective space. The core idea of ​​non-dominated ranking is that for any two solutions, if one solution is no worse than the other in all objectives and is better in at least one objective, then it is said to dominate the latter. By iteratively traversing all solutions, a multi-level dominance graph is obtained: the first level is all optimal solutions that cannot be dominated by any other solution (i.e., the first-order Pareto front), the second level is the remaining undominated solutions after removing the first level, and so on, forming a multi-level Pareto front structure. To avoid the phenomenon of overly concentrated "clustering" in the optimal solution space, the system executes a crowding ranking algorithm based on the multi-level Pareto front structure. Crowding density is used to evaluate the distribution density of a solution within the objective space relative to its neighboring solutions. A high crowding density indicates a dense distribution of solutions within that interval, representing limited choice diversity. A low crowding density suggests the solution is located at the edge or in a sparse region of the solution space, helping to preserve global diversity. By calculating the distance interval of each solution across all objectives, a crowding density ranking is obtained, and solutions with low crowding density and high value are prioritized for retention, ensuring that the optimization results are comprehensive and representative across different objective directions. Based on the crowding density ranking, the Euclidean distances to the ideal solution (i.e., the theoretically optimal value for all objectives) and the negative ideal solution (i.e., the theoretically worst value for all objectives) are calculated for all parameter-adjusted candidate solutions. A small distance to the ideal solution and a large distance to the negative ideal solution indicate that the solution is close to the optimum across all objectives; conversely, a small distance indicates mediocre performance or a weakness. For comprehensive ranking, the distance ratio (distance to the ideal solution / distance to the negative ideal solution) is used as the criterion for solution quality; a larger distance ratio indicates that the solution is closer to the global optimum.By traversing all the solutions, the candidate solution with the largest distance ratio is finally selected as the optimal solution for multi-objective balance.

[0039] The above describes the IoT-based soil stabilizer production monitoring method in the embodiments of the present invention. The following describes the IoT-based soil stabilizer production monitoring system in the embodiments of the present invention. Please refer to [link / reference]. Figure 2 One embodiment of the IoT-based soil stabilizer production monitoring system of the present invention includes: The data acquisition module 201 is used to deploy an Internet of Things sensor network at the locations of the raw material mixer, reactor and compactor in the soil stabilizer production line, and to collect raw monitoring data of the production process, including temperature, humidity, pressure, conductivity and characteristic spectral data. Create module 202 to divide the soil stabilizer production process into raw material ratio stage, chemical reaction stage and finished product quality stage based on the original monitoring data of the production process, and create a staged monitoring strategy; Control module 203 is used to implement differentiated sampling frequency control of the Internet of Things sensor network according to the phased monitoring strategy, and to obtain the set of target parameters for the production process; The identification module 204 is used to identify abnormal fluctuations in electrolyte concentration, micelle formation state, and crystal water fixation efficiency based on the set of production process target parameters, and to obtain production quality risk warning information. The generation module 205 is used to generate corresponding production process parameter adjustment schemes based on production quality risk warning information and the different performance requirements of soil stabilizers in different application scenarios.

[0040] Through the collaborative efforts of the aforementioned components, modular IoT sensors with dustproof and waterproof properties are deployed at key locations on the production line. A tree-structured topology and multi-protocol compatible communication parameter settings form a comprehensive, stable, and reliable monitoring network, resolving signal interference and data acquisition blind spots inherent in traditional monitoring methods. Based on raw monitoring data, production process stages are divided, and a parameter sensitivity assessment model and differentiated monitoring frequency configurations are established to achieve accurate monitoring of key parameters at different production stages, avoiding data redundancy and missing critical information caused by traditional fixed-interval sampling methods. By implementing a four-level sampling strategy (ultra-high frequency, high frequency, standard, and low frequency) combined with a dynamic sampling frequency adjustment mechanism, high-precision monitoring of key process points is ensured while significantly reducing sensor network energy consumption and extending equipment lifespan. Employing a multi-level parameter decomposition and anomaly detection model, combined with multivariate correlation analysis, abnormal fluctuations in electrolyte concentration, micelle formation state, and crystal water fixation efficiency are accurately identified, allowing for early detection of potential quality risks and changing the traditional passive monitoring mode that relies on experience-based judgment. Based on production quality risk early warning information and combined with the performance requirements of different application scenarios, the optimal process parameter adjustment scheme that balances the strength, energy consumption and environmental performance of finished products is generated through multi-objective evaluation and Bayesian optimization algorithm, realizing intelligent and precise control of process parameters.

[0041] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0042] If the integrated unit is implemented as 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, in essence, 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. This computer software product is stored in a storage medium and includes several instructions to cause an Internet of Things-based soil stabilizer production monitoring device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0043] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring the production of soil stabilizers based on the Internet of Things, characterized in that, include: An IoT sensor network is deployed at the raw material mixer, reaction vessel, and compaction molding machine locations in the soil stabilizer production line to collect raw monitoring data of the production process, including temperature, humidity, pressure, conductivity, and characteristic spectral data. Based on the original monitoring data of the production process, the soil stabilizer production process is divided into the raw material ratio stage, the chemical reaction stage, and the finished product quality stage, and a staged monitoring strategy is created. According to the phased monitoring strategy, differentiated sampling frequency control is implemented on the Internet of Things sensor network to obtain the target parameter set of the production process; Based on the set of target parameters for the production process, abnormal fluctuations in electrolyte concentration, micelle formation state, and crystal water fixation efficiency are identified to obtain early warning information on production quality risks. Based on the aforementioned production quality risk warning information, and combined with the differences in the performance requirements of soil stabilizers in different application scenarios, corresponding production process parameter adjustment schemes are generated.

2. The method for monitoring soil stabilizer production based on the Internet of Things according to claim 1, characterized in that, The aforementioned deployment of an IoT sensor network at the raw material mixer, reaction vessel, and compaction molding machine locations in the soil stabilizer production line collects raw process monitoring data, including temperature, humidity, pressure, conductivity, and characteristic spectral data, including: An IoT sensor network is deployed at the locations of the raw material mixer, reaction vessel, and compaction molding machine in the soil stabilizer production line, and a unique identifier and physical location marker are assigned to each sensor in the IoT sensor network to obtain an initial sensor deployment scheme; Configure the communication protocol for each sensor in the initial sensor deployment scheme to obtain the sensor communication network; Based on the sensor communication network, the production line is divided into a raw material area, a reaction area, and a molding area, and a gateway node is configured for each area to aggregate data, resulting in a regional network topology. Based on the regional network topology, low-noise sensor data streams are collected, and the data change rate and signal quality index of each monitoring point are calculated according to the low-noise sensor data streams to generate a dynamic sampling weight table. Data acquisition is performed based on the dynamic sampling weight table and the Internet of Things sensor network to record raw monitoring data of the production process, including temperature, humidity, pressure, conductivity and characteristic spectrum data.

3. The method for monitoring soil stabilizer production based on the Internet of Things according to claim 1, characterized in that, Based on the original monitoring data of the production process, the soil stabilizer production process is divided into three stages: raw material proportioning, chemical reaction, and finished product quality. A phased monitoring strategy is then created, including: Time series analysis was performed on the original monitoring data of the production process to obtain the feature point set of the entire production process of soil stabilizer; Based on the set of feature points of the entire production process, process stages are divided to obtain production stage boundary markers; Based on the production stage boundary markers, the soil stabilizer production process is divided into the raw material proportioning stage, the chemical reaction stage, and the finished product quality stage. Electrical conductivity and stirring torque monitoring parameters are assigned to the raw material proportioning stage, characteristic spectrum and temperature monitoring parameters are assigned to the chemical reaction stage, and humidity and compressive strength monitoring parameters are assigned to the finished product quality stage, resulting in a stage parameter mapping table. Based on the stage parameter mapping table, parameter sensitivity is evaluated, a parameter sensitivity matrix is ​​generated, and the parameter sensitivity matrix is ​​monitored in a hierarchical manner to obtain differentiated monitoring frequency configuration parameters. The differentiated monitoring frequency configuration parameters are correlated with the ideal parameter ranges of three types of soil stabilizers: road base type, foundation treatment type, and water conservancy seepage prevention type, to generate a phased monitoring strategy.

4. The method for monitoring soil stabilizer production based on the Internet of Things according to claim 3, characterized in that, The process stage division based on the feature point set of the entire production process, resulting in production stage boundary markers, includes: A sliding window analysis is performed on the feature point set of the entire production process to obtain the parameter change feature sequence; The conductivity data in the parameter change feature sequence is processed by first-order differentiation to identify conductivity abrupt change points, and the region between adjacent conductivity abrupt change points is divided into conductivity abrupt change intervals. Wavelet transform is performed on the temperature curve data in the parameter change feature sequence to extract the local extreme points and inflection points of the temperature curve data, and the inflection point interval of the temperature curve is generated based on the local extreme points and inflection points. The humidity sensor data in the parameter change feature sequence is subjected to a stationarity test to obtain the humidity change rate, and the humidity stability range is determined based on the humidity change rate. Based on the time-series pattern recognition of the conductivity abrupt change interval, the temperature curve inflection point interval, and the humidity stable interval, a process stage transition model is constructed. Based on the process stage transition model, the conductivity mutation point is taken as the boundary between the raw material proportioning stage and the chemical reaction stage, and the area between the last inflection point of the temperature curve data and the starting point of the humidity stable range is taken as the boundary between the chemical reaction stage and the finished product quality stage, thus generating production stage boundary markers.

5. The method for monitoring soil stabilizer production based on the Internet of Things according to claim 1, characterized in that, The step of implementing differentiated sampling frequency control on the IoT sensor network according to the phased monitoring strategy to obtain a set of production process target parameters includes: Based on the aforementioned phased monitoring strategy, sampling frequencies are allocated for the conductivity change rate in the raw material proportioning stage, the temperature curve monitoring in the chemical reaction stage, and the parameter monitoring in the finished product quality stage, resulting in a phased sampling control instruction set. Each gateway node in the IoT sensor network issues the phased sampling control instruction set and performs differentiated data acquisition on each sensor to obtain raw sensor data with multiple sampling frequencies. Based on the raw sensor data from the multi-level sampling frequency, and combined with the target performance requirements of soil stabilizers for road base, foundation treatment, or water conservancy seepage prevention, a set of production process target parameters matching the current production type is selected.

6. The method for monitoring soil stabilizer production based on the Internet of Things according to claim 1, characterized in that, The process identifies abnormal fluctuations in electrolyte concentration, micelle formation state, and crystal water fixation efficiency based on the set of production process target parameters to obtain production quality risk warning information, including: The set of target parameters for the production process is divided into physical layer parameters, chemical layer parameters, and quality layer parameters. The physical layer parameters include temperature curves and pressure distribution, the chemical layer parameters include electrolyte concentration and micelle current, and the quality layer parameters include crystallinity and humidity change rate, thus obtaining a layered parameter structure. Based on the layered parameter structure, time series analysis was performed on the electrolyte concentration change curve, spectral matching analysis was performed on the micelle formation state characteristics, and humidity change rate was calculated for the crystal water fixation efficiency to obtain parameter anomaly measurement index. The parameter anomaly measurement index is used to detect parameter anomalies, identify abrupt changes in electrolyte concentration, abnormal ranges in micelle formation, and ranges in which crystal water fixation efficiency decreases, and obtain parameter anomaly marking results. Based on the abnormal labeling results of the parameters, the interaction between electrolyte concentration change, micelle formation state and crystal water fixation efficiency is calculated, and a parameter correlation influence map is generated. A quality risk assessment is performed based on the parameter correlation impact map and the parameter anomaly labeling results to obtain a risk level assessment result. A threshold judgment is then made on the risk level assessment result to generate production quality risk early warning information.

7. The method for monitoring soil stabilizer production based on the Internet of Things according to claim 6, characterized in that, Based on the anomaly labeling results of the parameters, the interaction between electrolyte concentration changes, micelle formation state, and crystal water fixation efficiency is calculated, and a parameter correlation influence map is generated, including: Based on the anomaly labeling results of the parameters, cross-correlation analysis was performed on electrolyte concentration changes, micelle formation state and crystal water fixation efficiency to obtain a table of parameter time lag influence intensity. A bidirectional causal relationship test is performed based on the parameter time delay influence intensity table to obtain the parameter causal relationship network; Sensitivity analysis is performed based on the parameter causal relationship network to generate a parameter sensitivity heatmap; The parameter sensitivity heatmap is fused with the parameter causal relationship network to generate a parameter correlation influence map.

8. The method for monitoring soil stabilizer production based on the Internet of Things according to claim 1, characterized in that, Based on the production quality risk warning information and considering the differences in soil stabilizer performance requirements across different application scenarios, a corresponding production process parameter adjustment plan is generated, including: Based on the production quality risk early warning information and the target performance index requirements of three types of soil stabilizers—road base type, foundation treatment type, and water conservancy seepage prevention type—a scenario-based quality demand matching table is constructed. Based on the scenario-based quality requirement matching table, the parameter influence rules under the current production scenario are queried from the soil stabilizer production knowledge graph to generate a parameter-quality mapping network. Based on the parameter-quality mapping network, multiple process parameters are set as optimization variables. The objective function is the quality index satisfaction rate of the current application scenario. Iterative optimization calculations are performed to obtain a set of candidate solutions for parameter adjustment. The candidate scheme set is adjusted according to the parameters, and a non-dominated solution is performed to maximize the strength of the finished product, minimize the production energy consumption, and optimize the environmental performance, so as to obtain an optimal scheme with multi-objective balance. Based on the multi-objective balance optimization scheme, a corresponding production process parameter adjustment scheme is generated.

9. The method for monitoring soil stabilizer production based on the Internet of Things according to claim 8, characterized in that, The process of adjusting the candidate solution set according to the parameters, and performing non-dominated solutions to maximize finished product strength, minimize production energy consumption, and optimize environmental performance, yields a multi-objective balanced optimization scheme, including: The target evaluation index system is defined by adjusting the candidate scheme set according to the parameters. The target evaluation index system includes the compressive strength, resilient modulus, and shear strength of the finished product strength factors; the production energy consumption factors include electricity consumption, heat consumption, and water consumption; and the environmental performance factors include exhaust gas emissions, wastewater emissions, and solid waste generation. Based on the target evaluation index system and the weight settings of the execution index according to the characteristics of soil stabilizer application scenarios, a scheme scoring matrix is ​​obtained; A non-dominated sorting algorithm is executed based on the scheme scoring matrix to obtain a dominance graph, and a multi-level Pareto front structure is generated based on the dominance graph. The congestion ranking algorithm is executed based on the multi-level Pareto front structure to obtain the distribution characteristic quantification result. Combined with the distribution characteristic quantification result, the distance ratio from each parameter adjustment candidate scheme to the ideal solution and the negative ideal solution is calculated. The parameter adjustment candidate scheme with the largest distance ratio is selected as the multi-objective balance optimization scheme to ensure that the multi-objective balance optimization scheme achieves the best balance point among the three objectives of maximizing intensity, minimizing energy consumption, and optimizing environmental performance.

10. A soil stabilizer production monitoring system based on the Internet of Things, characterized in that, For implementing the IoT-based soil stabilizer production monitoring method as described in any one of claims 1-9, the IoT-based soil stabilizer production monitoring system comprises: The data acquisition module is used to deploy an IoT sensor network at the raw material mixer, reactor, and compaction molding machine locations in the soil stabilizer production line, and to collect raw monitoring data of the production process, including temperature, humidity, pressure, conductivity, and characteristic spectral data. A module is created to divide the soil stabilizer production process into three stages: raw material proportioning, chemical reaction, and finished product quality, based on the original monitoring data of the production process, and to create a phased monitoring strategy. The control module is used to implement differentiated sampling frequency control on the Internet of Things sensor network according to the phased monitoring strategy, and to obtain the target parameter set of the production process; The identification module is used to identify abnormal fluctuations in electrolyte concentration, micelle formation state, and crystal water fixation efficiency based on the set of production process target parameters, and to obtain production quality risk warning information. The generation module is used to generate corresponding production process parameter adjustment schemes based on the production quality risk warning information and the differences in the performance requirements of soil stabilizers in different application scenarios.

Images