Sludge organic nutrient soil production detection method and system

Abstract

The application provides a sludge organic nutrient soil production detection method and system, relates to sludge resource utilization and production detection technical field, and the method comprises the following steps: after product crushing and screening, multi-point synchronous sampling is carried out at a preset frequency in the conveying link before automatic packaging, and the samples at all points are collected to form a mixed sample representing the composition distribution characteristics of the batch; an intelligent analysis model based on multi-source data fusion is constructed, the spectral response characteristics of the mixed sample are coupled and calculated with a preset composition correlation characteristic library, and an equivalent detection data set is synchronously generated; on the basis of obtaining the equivalent detection data set, according to the geographical space coordinates of each point in the multi-point synchronous sampling process and the material property parameters of the corresponding point, a composition concentration spatial distribution matrix of the product on the cross section of the conveying belt is constructed. The application improves the spatial representativeness and composition detection accuracy of sludge organic nutrient soil online sampling, realizes composition distribution visualization and accurate identification of abnormal areas.

Description

Technical Field

[0001] This invention relates to the field of sludge resource utilization and production testing technology, and in particular to a method and system for testing the production of sludge organic nutrient soil. Background Technology

[0002] The resource utilization of sewage sludge to prepare organic nutrient soil is an important technical route for the harmless, reduced-volume, and resource-based utilization of sewage sludge from urban wastewater treatment plants, and it has been widely used in landscaping, soil improvement, and large-scale planting. As the industry continues to develop towards continuous, automated, and large-scale production, online quality testing and process control of the finished product after crushing and screening and before automatic packaging directly determine the product uniformity, application safety, and market compliance of the sludge organic nutrient soil.

[0003] Existing online detection and quality control technologies generally suffer from the following technical deficiencies: they cannot perform spatially representative sampling and visual analysis of the component distribution of finished materials on conveyor belts. In actual continuous production line scenarios, because materials are naturally piled up on conveyor belts, existing sampling methods cannot cover spatial feature points such as the toe, surface, and crown of the cross-section. The obtained mixed samples cannot accurately reflect the batch component distribution, leading to deviations in spectral detection and component analysis results from reality. This makes it impossible to accurately determine whether the product meets the requirements for indicators such as organic matter, moisture content, heavy metals, and fecal coliforms. Furthermore, they cannot identify the component concentration gradient and heterogeneous regions in the cross-section of the material, nor can they reverse-map spatial distribution anomalies to upstream process links such as fermentation temperature and mixing uniformity. Quality control is lagging and passive, making it difficult to achieve closed-loop control from detection and traceability to process adjustment. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for the production and testing of sludge organic nutrient soil, which can realize the closed-loop control of the entire process of sludge organic nutrient soil from online sampling, rapid component analysis, spatial distribution visualization, reverse deduction of the causes of anomalies to dynamic adjustment of the front-end process.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] In a first aspect, a method for testing the production of sludge-based organic nutrient soil, the method comprising:

[0007] After the finished product is crushed and screened and before automatic packaging, multi-point synchronous sampling is carried out at a preset frequency, and samples from each point are collected to form a mixed sample that characterizes the batch composition distribution characteristics.

[0008] A smart analytical model based on multi-source data fusion is constructed to obtain the spectral response characteristics of mixed samples and couple them with a preset component association feature library to generate an equivalent detection dataset simultaneously.

[0009] Based on the equivalent detection dataset, according to the geographic spatial coordinates of each point and the corresponding material property parameters during the multi-point synchronous sampling process, a spatial distribution matrix of component concentration on the cross-section of the conveyor belt is constructed. After spatial interpolation fitting of the distribution matrix, a virtual component distribution cloud map is generated. By identifying the component concentration gradient change pattern in the cloud map, homogeneous and heterogeneous regions are divided.

[0010] The boundary coordinates of the heterogeneous region in the component distribution cloud map are mapped onto the conveyor belt running trajectory. Spatial reverse deduction is performed in combination with the conveyor belt running speed parameters to calculate the upstream process disturbance period that causes component heterogeneity. Based on this, dynamic adjustment values ​​for fermentation process parameters are generated and fed back to the temperature control and mixing processes in the front-end fermentation stage.

[0011] During the feedback process, the equivalent test dataset is compared with the preset quality standard threshold. If the evaluation is qualified, the package is released; if it is unqualified, the diversion and return of materials is triggered and the corresponding cloud map and process adjustment record are sealed.

[0012] The pre-stored test data, cloud map feature values, and production time information of the current batch of finished products are associated and bound with the packaging batch to generate a unique traceability code to achieve quality traceability.

[0013] Secondly, a sludge organic nutrient soil production and testing system includes:

[0014] The multi-point synchronous sampling module is used to perform multi-point synchronous sampling at a preset frequency during the conveying process after the finished product is crushed and screened and before automatic packaging, and to collect samples from each point to form a mixed sample that characterizes the batch composition distribution characteristics.

[0015] The multi-source spectral intelligent analysis module is used to build an intelligent analysis model based on multi-source data fusion, obtain the spectral response characteristics of mixed samples and couple them with a preset component association feature library to generate an equivalent detection dataset simultaneously.

[0016] The component spatial distribution reconstruction module is used to construct a component concentration spatial distribution matrix on the cross-section of the conveyor belt based on the geographic spatial coordinates of each point and the corresponding material property parameters during the multi-point synchronous sampling process, based on the equivalent detection dataset. After spatial interpolation fitting of the distribution matrix, a virtual component distribution cloud map is generated. The homogeneous region and heterogeneous region are divided by identifying the component concentration gradient change pattern in the cloud map.

[0017] The process reverse deduction and dynamic control module is used to map the boundary coordinates of heterogeneous regions in the component distribution cloud map to the conveyor belt running trajectory. Combined with the conveyor belt running speed parameters, it performs spatial reverse deduction to calculate the upstream process disturbance period that causes component heterogeneity. Based on this, it generates dynamic adjustment values ​​for fermentation process parameters and feeds them back to the temperature control and mixing processes in the front-end fermentation stage.

[0018] The online quality assessment and diversion control module is used to compare the equivalent test dataset with the preset quality standard threshold during the feedback process. If the assessment is qualified, the package is released; if it is unqualified, the diversion and return of materials is triggered and the corresponding cloud map and process adjustment record are sealed.

[0019] The end-to-end quality traceability module is used to associate and bind the pre-stored test data, cloud map feature values, and production time information of the current batch of finished products with the packaging batch, and generate a unique traceability code to achieve quality traceability.

[0020] The above-described solution of the present invention has at least the following beneficial effects:

[0021] By employing multi-point synchronous sampling during the finished product conveying process, the problems of insufficient sampling representativeness and inability to accurately reflect material distribution are overcome, ensuring reliable testing foundation and accurate component analysis. Through multi-source data fusion and spectral coupling calculation, the problems of slow speed, single indicator, and offline delays in traditional testing are solved, enabling simultaneous and rapid testing of multiple indicators and improving testing efficiency and real-time performance. By constructing a spatial distribution matrix and cloud map of component concentration, the problems of invisible component distribution and inability to determine uniformity are overcome, enabling intuitive identification of uniformity and accurate differentiation between homogeneous and heterogeneous regions. Through coordinate mapping of heterogeneous regions and spatial inverse deduction, the problems of untraceable quality anomalies and passively delayed process adjustments are solved, enabling precise location of anomaly causes and dynamic closed-loop control of front-end processes. Through online quality judgment and diversion and return material control, non-conforming products are prevented from entering the packaging stage, improving product qualification rate and safety. By binding test data with packaging batches and generating unique traceability codes, the problem of untraceable product quality throughout the entire process is solved, enabling information to be traceable, verifiable, and verifiable throughout the entire process of production, testing, adjustment, and packaging. Attached Figure Description

[0022] Figure 1 This is a schematic flowchart of a method for producing and testing organic nutrient soil from sludge, provided by an embodiment of the present invention.

[0023] Figure 2 This is a schematic diagram of a sludge organic nutrient soil production and testing system provided in an embodiment of the present invention. Detailed Implementation

[0024] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0025] like Figure 1 As shown in the figure, an embodiment of the present invention proposes a method for testing the production of sludge organic nutrient soil, the method comprising the following steps:

[0026] Step 1: After the finished product is crushed and sieved and before automatic packaging, multi-point synchronous sampling is carried out at a preset frequency, and samples from each point are collected to form a mixed sample that characterizes the batch composition distribution.

[0027] Step 2: Construct an intelligent analytical model based on multi-source data fusion, obtain the spectral response characteristics of the mixed sample and couple them with the preset component association feature library for calculation, and simultaneously generate an equivalent detection dataset;

[0028] Step 3: Based on the equivalent detection dataset, construct the spatial distribution matrix of component concentration on the cross-section of the conveyor belt according to the geographic spatial coordinates of each point and the corresponding material property parameters during the multi-point synchronous sampling process. After spatial interpolation fitting of the distribution matrix, generate a virtual component distribution cloud map. By identifying the component concentration gradient change pattern in the cloud map, divide the homogeneous region and the heterogeneous region.

[0029] Step 4: Map the boundary coordinates of the heterogeneous region in the component distribution cloud map onto the conveyor belt running trajectory, and perform spatial reverse deduction in combination with the conveyor belt running speed parameters to calculate the upstream process disturbance period that causes component heterogeneity. Based on this, generate dynamic adjustment values ​​for fermentation process parameters and feed them back to the temperature control and mixing processes in the front-end fermentation stage.

[0030] Step 5: During the feedback process, the equivalent test dataset is compared with the preset quality standard threshold. If the evaluation is qualified, the package is released. If it is unqualified, the diversion and return of materials is triggered and the corresponding cloud map and process adjustment record are sealed.

[0031] Step 6: Associate and bind the pre-stored test data, cloud map feature values, and production time information of the current batch of finished products with the packaging batch to generate a unique traceability code to achieve quality traceability.

[0032] In this embodiment of the invention, by performing simultaneous sampling at multiple points during the finished product transportation process, a truly representative mixed sample can be obtained, improving the reliability of the detection basis. Utilizing multi-source data fusion and spectral coupling calculation, an accurate equivalent detection dataset can be quickly obtained, enabling simultaneous online detection of multiple indicators. Constructing a component distribution matrix based on spatial coordinates and generating a cloud map can intuitively reflect the material uniformity and accurately distinguish between homogeneous and heterogeneous regions. Through spatial reverse deduction of heterogeneous regions, the source of upstream process disturbances can be located, enabling dynamic adjustment of fermentation and mixing processes. Combined with online quality judgment and diversion control, unqualified products can be effectively intercepted, improving the finished product qualification rate and safety of use. Binding detection data, cloud map features, and packaging batches to generate a unique traceability code enables full-process quality traceability from production to packaging, comprehensively improving the intelligence, precision, and standardization of sludge organic nutrient soil production and testing.

[0033] In a preferred embodiment of the present invention, step 1 above may include:

[0034] Step 1.1: Based on the conveyor belt's operating speed and the finished product discharge rate, a sampling time interval is set, thus forming periodically occurring sampling time points. According to the natural angle of repose cross-sectional shape formed by the material accumulation on the conveyor belt, characteristic sampling points with spatially representative distribution are determined on the cross-section. These include boundary reference points located at the left and right slope toes, the crown reference point at the top of the cross-section, and the midpoint of the slope located in the middle of the line connecting each side boundary reference point and the crown reference point. Specifically, in the continuous production process of sludge organic nutrient soil, to ensure that the sampling operation is precisely matched with the production rhythm and that the collected samples are truly representative... This method accurately reflects the compositional distribution characteristics of batch materials. By combining the actual operating parameters of the production line, the sampling time interval is calculated and set. Specifically, it takes into account the stable operating speed of the conveyor belt, the continuous discharge rate of finished materials, the residence time of materials on the conveyor belt, and the batch production scale. Through comprehensive calculation, a fixed sampling time interval is determined, so that the sampling action is repeatedly executed according to the set periodic pattern, forming periodic sampling time points. This ensures that each sampling time point corresponds to different batch material segments conveyed on the conveyor belt, avoiding sample omissions due to excessively long sampling intervals and sampling redundancy and material waste due to excessively short intervals.

[0035] Meanwhile, considering the particle and accumulation characteristics of the sludge organic nutrient soil, and based on the natural angle of repose cross-sectional morphology formed after the material naturally accumulates on the conveyor belt, characteristic sampling points with comprehensive spatial distribution representativeness were precisely selected and determined on the transverse cross-section of the conveyor belt. Among them, the boundary reference point was selected at the toe of the slope on both sides of the material accumulation body, where it contacts the surface of the conveyor belt. This point can reflect the compositional state of the material accumulation edge area. The crown reference point was selected at the center of the top of the material accumulation cross-section. This point can reflect the compositional state of the highest point of the material accumulation. The slope midpoint was selected at the midpoint of the line connecting the left boundary reference point and the crown reference point, and the midpoint of the line connecting the right boundary reference point and the crown reference point. These two points can reflect the compositional state of the material accumulation slope area. By selecting these four types of characteristic sampling points, the entire area of ​​material accumulation in the cross-section of the conveyor belt was covered, laying the foundation for collecting batch-representative mixed samples.

[0036] Step 1.2: At each sampling time point, the sampling execution units positioned directly above each characteristic sampling point are synchronously triggered, causing each sampling execution unit to simultaneously extend downwards to a preset depth within its corresponding material accumulation layer to capture the material sample flowing instantaneously through the longitudinal section of that point. Specifically, when the production line reaches any pre-set sampling time point, the sampling execution units distributed directly above each characteristic sampling point are synchronously triggered in the same sequence, ensuring that all sampling execution units start their actions and complete sampling synchronously at exactly the same time. The sampling execution unit is a specialized sampling structure adapted to the particle characteristics and accumulation state of sludge organic nutrient soil. It is made entirely of a hard, inert material that is corrosion-resistant and does not cause secondary pollution, avoiding any reaction between the material and the sample during sampling that could affect the test results. Its structure mainly includes a sampling probe, a telescopic drive mechanism, and a quantitative sampling chamber. The sampling probe is cone-shaped with an arc-shaped sampling port at the tip, allowing it to smoothly penetrate the material accumulation layer without damaging the original accumulation morphology of the material. The telescopic drive mechanism is used to control the extension of the probe. The depth of the sampling chamber is fixed, ensuring that the volume of material collected by each sampling unit remains consistent. After each sampling unit is activated, it is simultaneously and smoothly extended downwards into the material accumulation layer at the corresponding feature point, driven by the telescopic drive mechanism. The preset depth is determined based on the actual thickness of the material accumulation on the conveyor belt, set according to the intermediate thickness from the surface to the bottom layer. This ensures the probe can penetrate the surface, middle, and bottom layers of the material, achieving uniform collection of material across the entire thickness range of the longitudinal section at that point. At this time, the arc-shaped sampling port of the sampling probe conforms to the longitudinal section of the material, performing directional and quantitative grabbing of the material flowing instantaneously through the longitudinal section at that point without disturbing the surrounding material. After grabbing, each sampling unit simultaneously stops collecting material and keeps the sampling chamber sealed to prevent spillage or mixing. This ensures that samples obtained at different points at the same time are completely consistent in terms of collection time, depth, quantity, and state, thus enabling each sample to accurately reflect the actual composition and properties of the material at that spatial location.

[0037] Step 1.3 involves synchronously collecting the instantaneous material samples captured by each sampling execution unit into the same collection container. This allows material samples from different spatial characteristic points to undergo physical mixing within the collection container, forming a mixed sample that characterizes the overall compositional distribution across the conveyor belt. Specifically, after each sampling execution unit completes material capture, it maintains a sealed sampling chamber and synchronously transports the instantaneous material samples captured by each unit to the same dedicated collection container. This ensures that samples from different characteristic sampling points enter the collection container at the same time, avoiding uneven sample mixing due to differences in transport timing. The collection container is made of the same corrosion-resistant, non-secondary-pollution, hard, inert material as the sampling execution units. The container is designed with a smooth inner wall and no dead corners to prevent material adsorption and residue. The container capacity is adapted to the total amount of material in a single sampling, ensuring that all samples can be contained while avoiding insufficient mixing due to excessive capacity. After all samples are put into the collection container, the materials in the container are physically mixed by gentle stirring. During the stirring process, the stirring speed and stirring time are controlled to ensure that the material samples from different spatial feature points of the conveyor belt cross-section are fully mixed and there is no local material aggregation. The final result is a mixed sample that can truly and comprehensively characterize the material composition distribution characteristics of the entire cross-section of the conveyor belt, ensuring that each part of the mixed sample can represent the overall composition state of the batch of materials.

[0038] Step 1.4: Record the sampling time point corresponding to the mixed sample, the spatial coordinate code of each characteristic sampling point, and the identification mark of the sample collection container as the basic traceability information for this batch of samples. Specifically, this includes: while completing the preparation of the mixed sample, comprehensively and accurately recording and organizing all basic information related to the mixed sample to ensure that every piece of information is traceable and without omission; specifically, accurately recording the sampling time point corresponding to the mixed sample, accurate to the specific hour, minute, and second, to ensure complete correspondence with the operating sequence of the production line; and recording in detail the spatial coordinate code of each characteristic sampling point. This code is uniquely identified based on the specific location of each point on the cross-section of the conveyor belt (such as the left and right slope toes, crown top, and midpoint of the slope), clearly distinguishing different... The spatial location differences of sampling points are recorded; at the same time, the unique identification mark of the sample collection container is recorded. This mark is a unique code for each sample collection container, used to distinguish mixed samples from different batches and sampling times to avoid sample confusion. The above three types of information, namely the sampling time point, the spatial coordinate code of each characteristic sampling point, and the identification mark of the sample collection container, are linked and bound together, and uniformly organized as the basic traceability information of the batch of material samples. The information is stored and archived in a standardized recording method to ensure the integrity and accuracy of the information. This provides comprehensive and reliable basic data support for the subsequent component analysis of mixed samples, the construction of component spatial distribution matrix, the tracing of quality anomalies, and the whole process of quality traceability, ensuring that the entire testing process of each batch of samples is traceable and verifiable.

[0039] In a preferred embodiment of the present invention, step 2 above may include:

[0040] Step 2.1 involves performing a full-band spectral scan on the collected mixed sample to obtain the original spectral response curve of the mixed sample. Specifically, this includes ensuring that the surface of the prepared, uniformly mixed, and unblemished sludge-organic nutrient soil mixed sample is flat and free of protrusions or depressions to prevent local shadows from interfering with the detection results when the light is irradiated; performing a full-coverage full-band spectral scan on the mixed sample, covering the visible to near-infrared bands, to ensure that the characteristic spectral bands corresponding to various components in the sample (such as organic matter, moisture, heavy metals, microorganisms, etc.) can be captured; and controlling the light intensity to be uniform and the scanning speed to be stable during the scanning process to avoid spectral signal distortion caused by light intensity fluctuations or excessively fast or slow scanning speeds.

[0041] The raw spectral response curves obtained in this scan are continuous curves reflecting the absorption, reflection, and transmission characteristics of the mixed sample to different wavelengths of light. The horizontal axis represents the spectral band, covering the visible light band (400 to 760 nm) and the near-infrared band (760 to 2500 nm), with different bands corresponding to the characteristic responses of different components in the sample. The vertical axis represents the spectral response intensity (units of absorbance or reflectance). The intensity of the response is directly related to the content and properties of the corresponding component in the sample; the higher the content of the component, the more pronounced the response intensity of its corresponding characteristic wavelength band usually is. The raw spectral response curves can completely record... The study records the effects of various components in the mixed sample on different wavelengths of light. For example, water has a specific strong absorption peak in the near-infrared band, organic matter has a characteristic absorption range in the visible to near-infrared band, and heavy metal ions will produce a slight spectral response shift at a specific wavelength. All these characteristic information will be fully reflected in the original spectral response curve. At the same time, the original spectral response curve will also contain certain interference signals, such as slight fluctuations caused by ambient stray light, instrument fluctuations, and sample particle scattering. These interference signals will cause slight irregular fluctuations in the curve, laying the foundation for subsequent spectral preprocessing.

[0042] Step 2.2 involves preprocessing the original spectral response curve with background subtraction, noise filtering, and baseline correction to extract a set of characteristic spectral signals that can characterize the intrinsic components of the sample. Specifically, this includes performing three consecutive preprocessing operations on the acquired original spectral response curve: background subtraction, noise filtering, and baseline correction. Background subtraction removes background interference signals caused by ambient stray light, dark current of the detection equipment, and reflection from the sample stage, eliminating fixed spectral shifts not caused by the sample itself, and bringing the spectral curve back to a state that truly reflects the sample's component response. Noise filtering smooths high-frequency spikes and random fluctuations in the curve caused by instrument vibration, circuit interference, and uneven scattering of material particles, retaining effective spectral features related to components such as organic matter, moisture content, and heavy metals. Finally, baseline correction corrects the overall drift and tilt deviation of the curve caused by factors such as light source attenuation and changes in the optical path, bringing the spectral baseline to a uniform level. After the above progressive preprocessing, invalid interference information can be effectively eliminated, and a set of characteristic spectral signals that can truly, stably, and accurately reflect the composition and content differences of the sample's intrinsic components can be retained.

[0043] Step 2.3: Retrieve the pre-constructed component association feature library. This feature library stores a mapping matrix between different component indicators and characteristic spectral signals. Specifically, it includes: retrieving and using the pre-constructed component association feature library. This feature library is constructed by combining the actual production conditions of sludge organic nutrient soil with various detection indicators, through extensive standard sample testing, repeated data calibration, and standardization, and can accurately match the detection requirements of this invention. When constructing this feature library, standard samples consistent with the raw material source, fermentation process, and component range of sludge organic nutrient soil in actual production are selected, covering samples from different wastewater treatment plants, with different fermentation cycles, different moisture contents, different organic matter contents, and different heavy metal concentrations, ensuring the applicability and representativeness of the feature library. For each standard sample, its characteristic spectral signal is obtained through full-band spectral scanning, and pH value, moisture content, organic matter content, fecal coliform count, and heavy metal concentration are simultaneously measured. The precise determination of sludge content involves correlating the quantified values ​​of each detection indicator with their corresponding characteristic spectral signals. After data cleaning and normalization, a stable mapping matrix is ​​formed and stored in the feature library. This mapping matrix clearly records the correspondence between each common detection indicator in the sludge organic nutrient soil and the characteristic bands and response intensities of the corresponding characteristic spectral signals. For example, the characteristic spectral band ranges corresponding to organic matter, the correlation ratio between moisture content and spectral response intensity, and the spectral response shift characteristics corresponding to heavy metal ions are all standardized and stored in the matrix. In addition, the feature library will regularly supplement new standard sample data and update the mapping matrix according to the fluctuations in sludge composition in actual production to ensure its timeliness and accuracy. This provides accurate, reliable, and comprehensive comparison basis for the component coupling calculation of the multi-source data fusion intelligent analysis model, ensuring that the calculation results can truly reflect the actual composition state of the sample.

[0044] Step 2.4: Input the feature spectral signal set into the intelligent analytical model of multi-source data fusion. Through the coupling solution algorithm inside the model, match and inverse calculate the feature spectral signal and the mapping relationship matrix until the algorithm converges and completes the coupling solution. Specifically, the intelligent analytical model of multi-source data fusion is constructed by combining a convolutional neural network (CNN) as the core feature extraction architecture with a partial least squares regression (PLSR) model. This fusion architecture retains the ability of the CNN model to automatically mine and adaptively extract deep and subtle features in high-dimensional spectral data, effectively capturing weak features related to component indicators in the spectral signals of sludge organic nutrient soil. It also has the efficient fitting and stable prediction ability of the PLSR model for the linear and nonlinear correlation between spectral signals and component indicators. It can effectively solve the problems of incomplete feature extraction, insufficient fitting accuracy, and weak anti-interference ability that are prone to occur in complex spectral analysis by a single model. It is more suitable for the actual continuous production scenario where the raw materials of sludge organic nutrient soil are complex, the fermentation degree is different, the composition fluctuates greatly, and there are many spectral interference factors. It can effectively meet the needs of online rapid detection and accurate analysis.

[0045] During the model building phase, the input parameters were set as the preprocessed characteristic spectral signal set of the sludge-organic nutrient soil mixture (covering characteristic spectral data in different bands), and the output parameters were set as the pH value, moisture content, organic matter content, fecal coliform count, and heavy metal content (such as common harmful heavy metals like lead, cadmium, and chromium) of the sludge-organic nutrient soil. When building the overall structure of the fusion model, the bottom layer adopted a CNN architecture with 3 to 5 convolutional and pooling layers to perform layer-by-layer convolution and dimensionality reduction on the input characteristic spectral signals, automatically extracting deep features from the spectral signals and filtering out invalid interference features. The middle layer was a feature fusion layer that initially fused the deep spectral features extracted by the CNN with the mapping relationship matrix in the component-related feature library to achieve complementarity of multi-source features. The top layer connected to the PLSR model, inputting the fused feature data into the PLSR model to construct a precise mapping relationship between spectral features and various component indicators, completing the solution mapping from spectral signals to component indicators. Meanwhile, a regularization layer is added to the model architecture to suppress overfitting, improve the model's generalization ability in practical applications, and ensure that the model can run stably under different production conditions.

[0046] During model training, a comprehensive training sample system was constructed, collecting standard samples of sludge organic nutrient soil covering all actual production conditions. These included organic nutrient soil prepared from sludge from different wastewater treatment plants, finished product samples with different fermentation cycles, fermented samples with different turning frequencies, samples with different moisture contents and organic matter contents, and samples containing different concentrations of heavy metals and different quantities of fecal coliforms. This ensured the diversity and representativeness of the samples, covering all possible fluctuations in composition and differences in operating conditions during actual production. Each standard sample underwent authoritative laboratory testing, using standard testing methods to determine its pH value, moisture content, organic matter content, fecal coliform count, and heavy metal content. These measurements served as label data for model training. Simultaneously, each standard sample underwent full-band spectral scanning and preprocessing to obtain the corresponding characteristic spectral signal set, which served as input data for model training.

[0047] The training process employs supervised learning, dividing the constructed training samples into training, validation, and test sets in a 7:2:1 ratio. The training set is used for initial training and optimization of model parameters, the validation set is used to monitor the fitting accuracy and overfitting during model training, and the test set is used to ultimately validate the model's detection accuracy and stability. In the initial training phase, appropriate hyperparameters such as learning rate, number of iterations, and batch size are set. The feature spectral signal set from the training set is input into the constructed fusion model. Forward propagation within the model yields predicted values ​​for each component index, which are then compared with the label values ​​obtained from laboratory testing to calculate the prediction error. Through backpropagation, the convolutional kernel weights, bias parameters, and fitting coefficients of the PLSR model are continuously adjusted to gradually reduce the prediction error.

[0048] During training, considering the numerous interferences and complex composition of the spectral signals in sludge organic nutrient soil, slight perturbations (such as band shifts and intensity fine-tuning) were applied to the spectral data of the training set. This simulated spectral interference caused by stray light and uneven particle size in actual production, improving the model's anti-interference capability. Simultaneously, the model was periodically validated using a validation set. If overfitting occurred, optimization was performed by adjusting regularization parameters and increasing the training sample size to ensure the model could accurately fit the features of the training samples while also adapting to unseen test samples and actual production samples. Training was continuously iterated until the model's prediction accuracy on the validation set reached the preset requirements, the prediction accuracy on the test set stabilized within the set range, and the model's convergence speed and stability met the requirements for online detection. At this point, training was stopped, and the trained model parameters and structure were saved, completing the construction and training of the multi-source data fusion intelligent analysis model.

[0049] The preprocessed feature spectral signal set of the current mixed sample is input into the trained multi-source data fusion intelligent analysis model. Through the model's internal pre-set coupled solution algorithm, the CNN underlying architecture first performs deep feature extraction and interference filtering on the input feature spectral signal. Then, the feature fusion layer performs band-by-band matching and multi-source feature fusion on the mapping relationship matrix between the extracted spectral features and the component-related feature library. Finally, the top-level PLSR model performs reverse inference calculation on the fused feature data, continuously iterating and optimizing the calculation results until the calculation results tend to be stable and convergence conditions are met. Ultimately, it achieves accurate, fast, and synchronous solution of multiple component indicators, from spectral signals to pH value, water content, organic matter content, fecal coliforms, and heavy metals.

[0050] Step 2.5: After the coupled solution is completed, the values ​​of pH, moisture content, organic matter content, fecal coliforms, and heavy metals are output synchronously to form an equivalent detection dataset corresponding to the current mixed sample. Specifically, this includes: after the multi-source data fusion intelligent analysis model completes the coupled solution and the calculation results reach the preset convergence conditions (i.e., the prediction error tends to stabilize and the numerical fluctuation is within the allowable range), the quantitative values ​​of various detection indicators corresponding to the current sludge-organic nutrient soil mixed sample are output synchronously; among them, the pH value is accurate to two decimal places, which meets the accuracy requirements of sludge-organic nutrient soil acidity and alkalinity detection; the moisture content is presented as a percentage, accurately reflecting the moisture content of the sample; the organic matter content is also expressed as a percentage, clearly indicating the proportion of organic matter in the sample; fecal coliforms... The bacterial count is expressed as the number of colonies per gram of sample, conforming to the standard expression for microbial testing; the content of various heavy metals is expressed as milligrams per kilogram, strictly following the specifications for heavy metal testing; the quantitative values ​​of the above indicators are organized and integrated in a unified format, and the detection unit and corresponding characteristic spectral correlation information of each indicator are labeled to form an equivalent detection dataset that is completely uniquely corresponding to the current mixed sample, ensuring that every value in the dataset is traceable and verifiable; this equivalent detection dataset fully covers the core quality indicators of sludge organic nutrient soil, providing accurate data support for the construction of the spatial distribution matrix of component concentration and the generation of component distribution cloud maps in subsequent steps, and also providing clear quantitative basis for subsequent online quality judgment and non-conforming product diversion control.

[0051] In a preferred embodiment of the present invention, step 3 above may include:

[0052] Step 3.1: Extract the spatial coordinate codes of each feature sampling point from the basic traceability information, as well as the material property parameters synchronously collected at each point at the sampling time. Obtain the overall composition values ​​of the current batch of finished products from the equivalent detection dataset. Specifically, this includes: accurately extracting the core information of the five feature sampling points corresponding to this sampling from the recorded basic traceability information; extracting the unique spatial coordinate codes of the left and right boundary reference points, the crown reference point, and the two slope midpoints. This code contains the lateral distance (based on the conveyor belt centerline) and longitudinal height (based on the conveyor belt bearing surface) of the conveyor belt cross-section, which can accurately locate the physical position of each point on the conveyor belt cross-section; simultaneously extracting the material properties collected synchronously by online sensors at each feature sampling point at the same sampling time. Material property parameters such as stack thickness (accurate to millimeters) and material density (measured by mass per unit volume of material) are used to ensure that the property parameters are completely matched with the sampling time and location. From the standardized and integrated equivalent test dataset, the overall composition values ​​of pH value, moisture content (percentage), organic matter content (percentage), fecal coliform count (number of colonies per gram of sample), and heavy metal content (milligrams per kilogram) corresponding to the current batch of sludge organic nutrient soil are retrieved. These values ​​are then structured according to the correspondence between spatial coordinate coding, material property parameters, and overall composition values. A dedicated data entry is established for each characteristic sampling point, so that the location information, property information, and composition information form a complete data link with one-to-one correspondence, ensuring clear data traceability and unbiased matching in subsequent calculations.

[0053] Step 3.2: Input the material property parameters and overall composition values ​​of each sampling point into the spatial inversion model for calculation to obtain the estimated composition concentration of the local area corresponding to each feature sampling point. Specifically, this includes: inputting the material property parameters (including material accumulation thickness and material density) corresponding to each feature sampling point and the overall composition values ​​of the current batch of finished products provided by the equivalent detection dataset into the pre-constructed and trained spatial inversion model. The material accumulation thickness accurately corresponds to the actual accumulation height of each feature point (boundary reference point, crown reference point, and slope midpoint) on the cross-section of the conveyor belt, while the material density reflects the compactness of the material particles at each point. Together, they serve as the core parameters characterizing the local distribution state of the material. Combined with the overall composition values ​​such as pH value, moisture content, and organic matter content of the current batch of finished products given by the equivalent detection dataset, comprehensive basic data is provided for model calculation.

[0054] This model combines the actual accumulation state of sludge organic nutrient soil on the cross-section of the conveyor belt: taking into account the slope morphology formed by the natural angle of repose of the material, the differences in the accumulation thickness at each point, and the distribution pattern of density, while fully considering the spatial correlation between each characteristic sampling point, including the distance between points, relative positions, and component transfer patterns, to avoid estimation errors caused by ignoring point correlations. Through the model's internal solution process, a separate solution is performed on the local small area (the area centered on the point and covering a reasonable surrounding range) where each characteristic sampling point is located: with each characteristic sampling point as the center, a suitable sludge organic nutrient soil is delineated. A localized area with a particle size of nutrient soil (range controlled within 3-5cm, to avoid both excessively large areas leading to mixing of components in different areas and affecting estimation accuracy, and excessively small areas resulting in insufficient sample representativeness) is then identified. For this localized area, the material accumulation thickness and density parameters at the corresponding points are correlated and matched with the overall composition values ​​of the current batch of finished products. Combining the spatial relationship between this point and other feature sampling points, and the component transfer patterns, the nonlinear correlation between material property parameters and component values ​​is first captured using the underlying RBFNN architecture of the model, to preliminarily predict the concentration of various components in this localized area.

[0055] By using the Kriging interpolation module in the middle layer, combined with the spatial distribution differences of each feature point, the preliminary prediction results are spatially corrected to correct deviations caused by uneven local material accumulation and component correlation between points. Through the error correction layer at the top layer, the corrected results are fine-tuned to eliminate abnormal deviations. The estimated values ​​of various components such as pH value, moisture content, organic matter content, fecal coliform count, and heavy metal content in each local area are accurately calculated to ensure that the estimated values ​​can fit the distribution characteristics of sludge organic nutrient soil on the conveyor belt. This not only reflects the correlation between each local area and the overall composition of the batch, but also accurately reflects the component differences of materials near different points, presenting the actual component level of materials near the point in a true and accurate manner.

[0056] The spatial inversion model is constructed by integrating and improving the radial basis function neural network (RBFNN) with the Kriging interpolation algorithm. Compared with a single interpolation algorithm or neural network model, this improved architecture combines the strong fitting ability of RBFNN for nonlinear spatial relationships with the accurate interpolation advantage of Kriging interpolation for spatial data. It can effectively solve the problems of uneven accumulation of sludge organic nutrient soil on the cross-section of the conveyor belt, strong correlation between components at different points, and large local concentration fluctuations.

[0057] In the spatial inversion model construction stage, the input parameters of the model are set as the material property parameters of each feature sampling point and the overall composition values ​​of the corresponding batch of finished products. The output parameters are set as the estimated values ​​of the concentrations of various components in the local area where each feature sampling point is located, ensuring that the model input and output are fully matched with the detection process and data requirements of this invention. When building the overall model structure, the bottom layer adopts the RBFNN architecture, setting up an input layer, a hidden layer, and an output layer. The input layer receives the material property parameters and the overall composition values. The hidden layer captures the nonlinear correlation between parameters through radial basis functions. The output layer initially outputs the predicted values ​​of local component concentrations. A Kriging interpolation module is added to the middle layer to perform spatial correction and optimization on the predicted values ​​output by the RBFNN using the spatial coordinate information of each feature sampling point, making up for the deficiency of a single neural network in ignoring spatial location correlation. An error correction layer is set at the top layer to reduce estimation errors and ensure the accuracy of the output results.

[0058] During model training, a training sample system that closely matches actual production conditions is constructed. Sampling data of sludge organic nutrient soil with different production batches, different material stacking states, and different component distributions are collected. Each sample includes the spatial coordinates of each characteristic sampling point, material property parameters, the overall component value of the corresponding batch, and the actual component concentration data of each characteristic point in the local area obtained through multi-point detection in the laboratory. This ensures that the samples cover various scenarios of material stacking in the cross-section of the conveyor belt, including areas with thicker stacking and higher density, as well as edge areas with thinner stacking and lower density, thereby improving the applicability of the model.

[0059] The training process employs supervised learning, dividing the training samples into a training set and a validation set in an 8:2 ratio. The training set is used for training and optimizing model parameters, while the validation set is used to monitor the model's training accuracy and stability. In the initial training phase, appropriate hyperparameters such as the learning rate and number of iterations are set. The input parameters from the training set are fed into the model, and initial predictions are performed using an RBFNN. Spatial correction is then performed using a Kriging interpolation module, outputting estimated local component concentrations. These estimated values ​​are compared with actual label values ​​detected in the laboratory, and the error is calculated. The model's internal parameters are adjusted using a backpropagation algorithm, continuously optimizing the radial basis function parameters of the RBFNN and the weighting coefficients of the Kriging interpolation. During training, interference data from different production scenarios (such as uneven material accumulation and fluctuations in morphological parameters) are introduced to simulate complex situations in actual production, enhancing the model's anti-interference ability and generalization ability. The model's accuracy is periodically verified using a validation set until the model's estimation error falls below a preset threshold and its stability meets the requirements. At this point, training is stopped, and the model parameters are saved, completing the construction and training of the spatial inversion model.

[0060] Step 3.3: Using the spatial coordinate codes of each feature sampling point as the location reference and their corresponding component concentration estimates as the assignment variables, a discrete component concentration spatial distribution matrix of the finished product on the conveyor belt cross-section is constructed. Specifically, this includes using the spatial coordinate codes corresponding to each feature sampling point as the core location reference on the conveyor belt cross-section. This coordinate code accurately includes the geometric location information in the transverse direction (extending symmetrically to both sides from the conveyor belt centerline as the origin, in millimeters) and the longitudinal direction (extending upwards to the top surface of the material accumulation, based on the conveyor belt bearing surface, in millimeters). It can clearly locate the specific orientation of the left and right boundary reference points, the crown reference point, and the two slope midpoints within the cross-section, ensuring the spatial distribution of each point. The location is unbiased. At the same time, the estimated concentration values ​​of each component obtained by the spatial inversion model at each feature sampling point are used as assignment variables. Combined with the cross-sectional centroid-constrained spatial gridding algorithm, the discrete component concentration spatial distribution matrix is ​​constructed step by step, as follows: The core principle of the cross-sectional centroid-constrained spatial gridding algorithm is based on the geometric shape of the conveyor belt cross-section and the natural accumulation law of the material. With the geometric centroid of the cross-section as the core constraint anchor point, a non-uniform grid system that fits the actual distribution of the material is constructed through geometric coordinate calculation and adaptive grid division. This breaks the drawback of traditional uniform gridding that ignores the differences in material accumulation and causes waste or insufficient accuracy, and achieves accurate matching between grid density and material distribution.

[0061] The first step is to calculate the geometric centroid of the conveyor belt cross-section: extract the complete boundary coordinates of the conveyor belt cross-section, including the x-coordinate X1 of the left edge of the conveyor belt, the x-coordinate X2 of the right edge, the y-coordinate Y1 of the bearing surface (fixed at 0, with the bearing surface of the conveyor belt as the zero point of the y-coordinate), and the contour coordinates of the top surface of the material accumulation: including the y-coordinate Y of the crown datum point. max The ordinates of the highest point of the material accumulation, the midpoints of the two slopes (Y3 / Y4), and the left and right boundary reference points (Y5 / Y6) are calculated. Then, the coordinates of the centroid are calculated using a weighted average of geometric coordinates. The lateral centroid coordinates are also calculated. Because the cross-section of the conveyor belt is symmetrically distributed laterally, with the center line as the core, the longitudinal center of gravity coordinates are... By weighting the stacking height of each feature point and aligning it with the actual stacking center of gravity of the material, the coordinates of the geometric center of gravity of the cross-section are finally obtained. This center point serves as the core constraint benchmark for the entire grid division, ensuring that the grid distribution always revolves around the core area of ​​material accumulation, rather than simply using the physical center of the conveyor belt as the benchmark, which better reflects the actual accumulation state of the sludge organic nutrient soil.

[0062] The second step involves determining the density rules for grid division based on the slope geometry formed by the natural angle of repose of the material: by calculating the accumulation thickness of each feature sampling point = the vertical coordinate of the point - the vertical coordinate of the bearing surface Y1, the thickness gradient of the material accumulation is obtained. The grid density is set to be positively correlated with the accumulation thickness, that is, in the thicker crown and slope areas with a thickness ≥15cm, the grid density is increased (the grid side length is set to 1 to 2cm) to ensure the accuracy of the component concentration data in the core area; in the thinner boundary areas with a thickness <15cm, the grid is appropriately sparse (the grid side length is set to 3 to 4cm) to avoid data redundancy caused by invalid grids, while also taking into account the component characterization needs of the boundary areas.

[0063] The third step is to complete the non-uniform mesh generation: using the centroid point... Centered on the horizontal conveyor belt, with X1 as the left boundary and X2 as the right boundary along the width direction, and Y1 as the lower boundary and Y2 as the right boundary along the vertical material stacking height direction. max Using the upper boundary as the grid, a complete two-dimensional grid system is gradually divided according to the set grid density. The grid boundary strictly conforms to the geometric contour of the conveyor belt cross-section, not exceeding the range of the two side edges, the bearing surface, and the top surface of the material accumulation, ensuring that the grid system completely matches the actual material distribution. After completing the grid division, the detailed construction process of the discrete component concentration spatial distribution matrix begins: five feature sampling points are encoded according to their spatial coordinates and accurately mapped one by one to the corresponding grid nodes, ensuring that the horizontal and vertical coordinates of each point completely correspond to the coordinates of the grid nodes without any offset; the estimated values ​​of various component concentrations corresponding to each feature sampling point are assigned one by one to its mapped grid nodes, realizing a one-to-one correspondence between point coordinates, grid nodes, and component concentrations; then, for each Each grid cell is identified, with its corresponding grid node coordinates, feature sampling point type, and component concentration value marked to facilitate subsequent data tracing and interpolation fitting. Finally, all grid cells with component concentration assignments are arranged in an orderly manner according to the horizontal coordinate from left to right and the vertical coordinate from bottom to top (from the bearing surface to the top of the accumulation surface), forming a complete discrete component concentration spatial distribution matrix. This matrix not only clearly records the component concentration data of each grid cell, but also accurately fits the actual shape of the material accumulation through the design of non-uniform grids. It ensures the data accuracy of core areas such as the crown and slope, while avoiding redundant data. It provides a reliable data framework that is structurally standardized, data accurate, and fits reality for spatial interpolation fitting and component distribution cloud map generation.

[0064] Step 3.4 involves spatial interpolation fitting of the discrete component concentration spatial distribution matrix to generate a continuous virtual component distribution cloud map covering the entire conveyor belt cross-section. Specifically, this includes: performing spatial interpolation fitting on the constructed discrete component concentration spatial distribution matrix using an optimized and improved adaptive neighborhood interpolation algorithm based on centroid constraints. This algorithm retains the numerical fitting advantages of traditional interpolation algorithms while addressing edge deviations and detail loss through centroid constraints and adaptive neighborhood adjustment. Its principle and implementation process are naturally integrated into the entire interpolation process, ensuring that the interpolation results accurately and reliably match the actual material accumulation morphology. Before the interpolation operation, two preparatory steps need to be completed: extracting the core reference data from the discrete matrix, including the calculated geometric centroid coordinates (X, Y, X) of the conveyor belt cross-section. c Y c ), coordinates (X) of all assigned feature points i Y i ) and the corresponding component concentration value C i (i=1,2,…,5, corresponding to the left and right boundary reference points, the crown reference point, and the midpoints of the two slopes, respectively), while defining the geometric boundary range of the conveyor belt cross-section, namely the left boundary X1, the right boundary X2, the lower boundary Y1=0 bearing surface, and the upper boundary Y max The top surface of the material accumulation; in addition, interpolation judgment parameters are defined, and the component concentration fluctuation threshold ΔC is preset, where the pH value fluctuation threshold is set to 0.2 and the moisture content fluctuation threshold is set to 2%. At the same time, the initial value of the neighborhood range R0=5cm is set to provide a clear judgment basis for subsequent neighborhood adaptive adjustment.

[0065] After the preliminary preparations are completed, the unassigned blank grids in the discrete matrix are traversed one by one, and the center coordinates (X, Y, Z) of each blank grid are extracted. p Y p Next, the Euclidean distance from the mesh to the centroid of the cross section is calculated. Then calculate the Euclidean distance from the grid to all the assigned feature points. (i=1,2,…,5), and at the same time, feature points with a distance smaller than the initial neighborhood range R0 are selected and used as the initial candidate interpolation neighborhood set for the blank grid.

[0066] Based on the component concentration differences of feature points in the initial candidate neighborhood set, the neighborhood range is further adaptively adjusted to balance interpolation smoothness and detail preservation: first, the standard deviation of component concentration of feature points within the initial neighborhood set is calculated. ,in Let n be the average component concentration of feature points within the neighborhood set, and n be the number of feature points within the neighborhood set; if the calculated... This indicates that the composition of the region fluctuates significantly, and in this case, the neighborhood range should be expanded to... And the maximum size is no more than 8cm. By incorporating more surrounding feature points and increasing the number of reference samples, the smoothness of the interpolation results is ensured, and local concentration abrupt changes are avoided. This indicates that the compositional distribution in this region is stable, and the neighborhood range is then narrowed down to... The minimum value should be no less than 3cm to reduce the number of reference points and avoid loss of composition details due to excessive smoothing. It should be noted that the adjusted neighborhood range must be strictly limited to the geometric boundary of the conveyor belt cross-section. If the expanded neighborhood exceeds the boundary, the feature points in the neighborhood should be re-selected with the boundary as the limit to ensure that the neighborhood reference points are all within the actual distribution range of the material.

[0067] After the neighborhood is determined, interpolation weights are assigned to each feature point based on the centroid constraint. The key is to assign higher weights to points closer to the centroid and closer to blank grid lines to avoid excessive influence of edge points on the interpolation results of the core region: first, the distance weights of each feature point are calculated. Points closer to the blank grid have a greater distance weight; then the centroid association weight of each feature point is calculated. ,in This is the distance from the feature point to the centroid. The maximum distance from all feature points to the centroid is used; points closer to the centroid have a higher centroid association weight. Finally, the overall weight is calculated. ,in This is used to balance the influence of distance and centroid correlation, and the sum of the comprehensive weights of all feature points satisfies... To ensure reasonable weight allocation and accurate interpolation calculation, after weight allocation, calculate the component concentration interpolation of the blank grid based on the comprehensive weight. The calculation formula is as follows: Immediately after interpolation, perform boundary verification on the interpolation result to confirm the center coordinates of the blank grid. It does not exceed the geometric boundary of the conveyor belt cross-section, thus satisfying the condition. , If the result exceeds the boundary, the interpolation result is discarded, and only the valid interpolation data within the actual distribution range of the material is retained, thus preventing invalid interpolation data from affecting subsequent processes.

[0068] After completing the interpolation calculations for all blank grids, the numerical values ​​of the entire discrete matrix are smoothed using a local mean filter with a 3×3 grid window to eliminate minor noise generated during interpolation, ensuring a natural transition of component concentration values ​​in each grid cell and avoiding abrupt numerical fluctuations. The complete interpolated matrix data is then imported into a visualization engine. Using the conveyor belt cross-section as a base, color gradients are mapped according to component concentration values, with low-concentration areas mapped to light blue and high-concentration areas to dark red. This generates a continuous virtual component distribution cloud map that fully covers the entire conveyor belt cross-section. This cloud map features smooth color transitions, clearly presenting the component distribution trends in core areas such as the crown and slope, while also preserving local details in fluctuating areas such as boundaries. It intuitively reflects the differences and distribution patterns of various components such as pH, moisture content, and organic matter content across the conveyor belt cross-section, providing accurate and intuitive visualization support for subsequent component concentration gradient analysis and region division.

[0069] Step 3.5 involves performing gradient analysis on the continuous virtual component distribution cloud map to identify the gradient variation pattern of component concentrations. Based on a preset gradient threshold, the cloud map region is divided into a homogeneous region with relatively uniform component distribution and a heterogeneous region with significant component distribution fluctuations. Specifically, this includes: using a cross-sectional polar coordinate gradient partitioning algorithm to analyze the component concentration gradient in the generated continuous virtual component distribution cloud map. This algorithm overcomes the limitations of traditional rectangular coordinate gradient analysis, better reflects the natural accumulation morphology of sludge organic nutrient soil on the conveyor belt cross-section, and effectively avoids the gradient identification bias problem of traditional algorithms for special areas such as slopes and boundaries, improving the accuracy and rationality of gradient identification and region partitioning. During the analysis process, the direct... Using the calculated geometric centroid of the conveyor belt cross-section as the pole, the longest distance from the centroid to the edge of the material accumulation as the polar radius, and the line connecting the centroid and the crown datum point as the polar axis, the established polar coordinate analysis system is used without rebuilding the analysis framework, ensuring that the analysis process is consistent with the previous steps and the data is consistent. Gradient calculations are performed on all grid cells in the virtual component distribution cloud map, and the radial gradient and angular gradient are extracted synchronously for each grid cell. The radial gradient mainly reflects the rate of change of component concentration in the direction of the line connecting the grid cell and the centroid (from the center to the edge), while the angular gradient mainly reflects the fluctuation range of component concentration in the direction of the tangent of the material slope (extending along the polar angle). The two types of gradient values ​​are then weighted and fused. The formula for weighted fusion is: the comprehensive gradient change rate is equal to the radial gradient value multiplied by the radial gradient weight (0.6), plus the angular gradient value multiplied by the angular gradient weight (0.4), that is, the comprehensive gradient change rate = radial gradient value × 0.6 + angular gradient value × 0.4. This formula accurately calculates the comprehensive gradient change rate of each grid cell, which can comprehensively and objectively reflect the degree of spatial change of component concentration, taking into account the concentration change characteristics of the core area, and not ignoring the fluctuation details of special areas such as slopes and boundaries.

[0070] After obtaining the comprehensive gradient change rate of all grid units, it is compared point by point with the concentration gradient thresholds pre-set in the production process and verified under multiple batches of actual working conditions. These gradient thresholds are set separately for different detection indicators, such as a pH gradient threshold of 0.15 / cm and a moisture content gradient threshold of 1.5% / cm, to ensure the relevance and rationality of the comparison standard. Based on the comparison results, the internal areas of the cloud map are standardized and refined: for areas where the comprehensive gradient change rate is lower than the corresponding indicator gradient threshold, the component concentration values ​​are small, the spatial changes are gentle and the distribution is continuous, they are uniformly identified and marked as homogeneous areas with relatively uniform component distribution. In such areas, the various components of the sludge organic nutrient soil are uniformly mixed and meet the production quality standards. For areas where the comprehensive gradient change rate is higher than the corresponding indicator gradient threshold, the component concentration fluctuates drastically, the numerical differences exceed the production allowable range, and there are obvious concentration abrupt changes and no continuous distribution characteristics, they are identified and marked as heterogeneous areas with obvious component distribution fluctuations. Such areas need to be closely monitored, as there may be problems such as uneven material mixing and insufficient local fermentation.

[0071] After the regional division is completed, the division results are further refined: the clear boundaries between homogeneous and heterogeneous areas are accurately identified and marked, and the coordinate range and extension trend of the boundaries are clarified; at the same time, the specific location of each heterogeneous area on the cross-section of the conveyor belt (such as the left slope, right boundary, crown perimeter, etc.), distribution range (number of grid units covered, actual area), and concentration anomaly degree (concentration difference with homogeneous areas, the extent of exceeding the threshold) are located, and heterogeneous areas are graded and marked (slight anomaly, moderate anomaly, severe anomaly), forming a complete and detailed cross-sectional component distribution uniformity judgment report; the judgment result can intuitively present the overall compositional distribution of sludge organic nutrient soil on the cross-section of the conveyor belt, clearly distinguishing between qualified and abnormal areas, providing accurate, intuitive and stable reliable judgment basis for subsequent quality anomaly tracing (accurately locating the sampling points and production links corresponding to abnormal areas), optimizing fermentation process parameters (adjusting parameters such as stirring and fermentation time for heterogeneous areas), and implementing material homogenization control (targeted mixing treatment of heterogeneous areas), effectively ensuring the stability and consistency of the finished sludge organic nutrient soil quality.

[0072] In a preferred embodiment of the present invention, step 4 above may include:

[0073] Step 4.1: Extract the spatial coordinate set of the boundary contours of the heterogeneous regions on the cross-section of the conveyor belt as the boundary coordinate data of the heterogeneous regions. Specifically, this includes: first, confirming the number of heterogeneous regions (single or multiple) from the completed and marked component distribution heterogeneous regions; then, performing boundary contour coordinate extraction for each heterogeneous region separately to avoid confusion between coordinates of multiple heterogeneous regions; during the extraction process, based on a continuous virtual component distribution cloud map, along the boundary line between the heterogeneous and homogeneous regions, at a preset coordinate acquisition interval (not exceeding 1mm) to ensure accurate boundary contour restoration; extracting all boundary contour coordinate points of the heterogeneous region in the cross-sectional coordinate system of the conveyor belt point by point. Each coordinate point corresponds to both the horizontal and vertical coordinates in the cross-sectional coordinate system, and is consistent with the coordinates obtained in step [step 4.1]. The coordinate system used in sections 3.3, 3.4, and 3.5 remains completely consistent to ensure the continuity and uniformity of coordinate data. After extraction, all boundary contour coordinate points undergo two rounds of rigorous processing: The first round performs continuity verification by comparing the distance and angle between adjacent coordinate points to identify and eliminate breakpoint coordinates and abnormal offset coordinates caused by noise in the cloud map. For areas with breakpoints, linear interpolation is used to supplement the complete coordinates, ensuring that the boundary lines are smooth and continuous, and closely match the actual contour of the heterogeneous region. The second round removes redundant points by deleting redundant coordinate points with duplicate coordinate values ​​and spacing less than 0.5 mm to reduce invalid data usage. At the same time, the coordinates of key inflection points of the boundary contour (such as the intersection of the heterogeneous region with the slope and the boundary) are retained to ensure the integrity and accuracy of the boundary contour.

[0074] After processing, the selected boundary contour coordinate points are sorted clockwise and integrated into a standardized and ordered set of spatial coordinates. Each coordinate set is labeled with a unique identifier for the corresponding heterogeneous region (e.g., heterogeneous region 1, heterogeneous region 2) and the corresponding detection index (e.g., pH heterogeneity, moisture content heterogeneity), forming complete boundary coordinate data for heterogeneous regions. Finally, the coordinate data is compared and verified point by point with the virtual component distribution cloud map to confirm that the boundary range and location corresponding to the coordinate set completely match the heterogeneous regions marked in the cloud map, without deviation or omission. This provides accurate and reliable basic coordinate support for the spatial mapping, spatiotemporal transformation, and the entire anomaly tracing process in subsequent steps.

[0075] Step 4.2: Map the boundary coordinate data of the heterogeneous region onto the running trajectory of the conveyor belt, establishing the spatial interval between the start and end positions of the heterogeneous region along the length of the conveyor belt. Specifically, this includes: performing coordinate system dimensional transformation and spatial mapping operations on the processed and verified boundary coordinate data of the heterogeneous region. The core is to map the original two-dimensional cross-sectional coordinate system, which only contains horizontal and vertical coordinates, to a three-dimensional spatial coordinate system that covers the length of the conveyor belt, ensuring that the position of the heterogeneous region can be accurately located on the entire running trajectory of the conveyor belt and completely conforms to the actual working conditions of the production line. Before mapping, confirm the unified setting of the three-dimensional spatial coordinate system. This coordinate system takes the left end point of the conveyor belt bearing surface and the outlet position of the front fermentation stage as the coordinate origin, the width direction of the conveyor belt as the horizontal direction (consistent with the horizontal direction of the cross-sectional coordinate system), the material accumulation height direction as the vertical direction (consistent with the vertical direction of the cross-sectional coordinate system), and the running direction of the conveyor belt as the length direction. The three coordinate axes are perpendicular to each other and maintain the same parameters and units (millimeters) as the coordinate system used in steps 3.3, 3.4, 3.5 and step 4.1 to avoid mapping deviations due to differences in coordinate systems.

[0076] Based on the fixed installation position of the current inspection station in the entire production line, the length direction coordinates of the inspection station in the three-dimensional coordinate system are extracted, i.e., the distance between the inspection station and the origin, which is a fixed value and pre-stored in the system. The correspondence between the cross-section corresponding to the boundary coordinate data of the heterogeneous region and the inspection station in the length direction is determined. Simultaneously, considering the operating characteristics of the conveyor belt, the mapping rules for each boundary coordinate point within the cross-section along the conveyor belt length direction are confirmed: using the instantaneous position of the material in the heterogeneous region passing through the inspection station as a reference, and based on the continuous accumulation characteristics of the material on the conveyor belt, each boundary contour coordinate point within the two-dimensional cross-section is mapped to its corresponding length direction coordinate in the three-dimensional coordinate system, ensuring that each... The three-dimensional coordinates of each boundary point can accurately correspond to the actual spatial position of the heterogeneous material on the conveyor belt. After mapping, for each heterogeneous region (if there are multiple heterogeneous regions, they are processed separately), the minimum and maximum coordinate values ​​of its boundary coordinates in the length direction in three-dimensional space are selected. The corresponding coordinate points are the physical starting point and physical ending point of the batch of heterogeneous materials in the running direction of the conveyor belt. Among them, the physical starting point is the length direction coordinate corresponding to the front end of the heterogeneous material entering the detection station, and the physical ending point is the length direction coordinate corresponding to the end of the heterogeneous material leaving the detection station. The continuous range between the two points is the complete spatial range of the heterogeneous region in the length direction of the conveyor belt.

[0077] Finally, the validity of the constructed spatial interval is verified. Combining the actual width of the conveyor belt and the thickness of the material accumulation, it is confirmed that the length of the spatial interval is reasonable, does not exceed the actual running trajectory of the conveyor belt, and matches the range of the heterogeneous area within the cross-section, avoiding problems such as spatial interval offset, excessively large or small range. After the verification is qualified, the unique identifier and detection index of the spatial interval and the corresponding heterogeneous area are associated and stored to clarify the one-to-one correspondence between the spatial interval and the heterogeneous area. This provides an accurate and unique location basis for subsequent spatial and temporal reverse conversion and upstream process disturbance period tracing, ensuring the accuracy and consistency of subsequent traceability analysis.

[0078] Step 4.3: Retrieve the real-time operating speed parameters of the conveyor belt and perform spatial reverse deduction in conjunction with the aforementioned spatial interval to convert the spatial interval into a time interval. Calculate the upstream process disturbance period corresponding to the material forming this heterogeneous region. Specifically, this includes retrieving various real operating parameters of the current production line, focusing on core parameters such as the stable operating speed of the conveyor belt, cumulative running time, real-time belt displacement, and material conveying delay records. During the retrieval process, the parameters are verified in real time to eliminate abnormal data caused by fluctuations in equipment sensors (such as sudden speed changes, displacement deviations, etc.) to ensure that the retrieved parameters are real, stable, and effective. Furthermore, the units of all parameters are consistent with the coordinate system and spatial interval units mentioned above (all are millimeters and seconds), providing a reliable calculation basis for subsequent spatiotemporal conversion and reverse deduction.

[0079] After the parameters are retrieved, the actual length of the heterogeneous area is calculated first, taking the constructed and verified heterogeneous area space interval as the core. That is, the difference between the maximum and minimum coordinate values ​​in the length direction of the space interval, so as to obtain the actual physical length occupied by the heterogeneous material in the length direction of the conveyor belt. Combined with the retrieved stable running speed of the conveyor belt, a forward conversion from space to time is performed to calculate the time required for the batch of heterogeneous material to pass through the current detection station completely. That is, the passing time = space interval length ÷ stable running speed of the conveyor belt. The time interval for the heterogeneous material to pass through the detection station is determined in this way: the instantaneous time when the front end of the heterogeneous material enters the detection station is taken as the start time, and the instantaneous time when the end of the heterogeneous material leaves the detection station is taken as the end time, forming a complete detection passing time interval, and this time interval corresponds one-to-one with the space interval in step 4.2.

[0080] Next, a reverse time simulation was performed to accurately trace the upstream process disturbance period: Pre-stored fixed delay parameters for material transport were retrieved. These parameters refer to the fixed time required for material to flow from the outlet of the front-end fermentation and mixing stage, be transported by conveyor belt to the current detection station. This time was calibrated through multiple batches of production tests, covering the material's transmission time, turning time, etc., ensuring that the delay parameters accurately match actual production conditions. Using the time interval for heterogeneous material to pass through the detection station as a benchmark, the fixed delay for material transport was subtracted from the start and end times of this time interval to obtain the corresponding upstream time nodes: the start time minus the fixed delay is the start time for heterogeneous material to exit the front-end fermentation and mixing stage; the end time minus the fixed delay is the end time for heterogeneous material to exit the front-end fermentation and mixing stage.

[0081] The continuous time period between these two upstream time nodes is the upstream process disturbance period corresponding to the formation of this batch of heterogeneous materials. After the simulation is completed, the disturbance period is double-verified: on the one hand, the material conveying trajectory within the disturbance period is verified in reverse by combining the cumulative running time of the conveyor belt and the belt displacement, confirming that the time from the material discharge from the front fermentation stage to the detection station is completely matched with the fixed delay; on the other hand, the process operation record of the disturbance period is compared with that of the front fermentation stage to confirm that the fermentation stage is in normal operation during the period, without any abnormalities such as shutdown or maintenance, thus avoiding the tracing of invalid disturbance periods. After the verification is qualified, the upstream process disturbance period is associated with the unique identifier, spatial range, and detection index of the corresponding heterogeneous area and stored to clarify the one-to-one correspondence between the disturbance period and the heterogeneous area.

[0082] Step 4.4: Based on the upstream process disturbance period, locate the actual operating parameters of the front-end fermentation stage within this period in the pre-stored historical process parameter records, and compare the actual operating parameters with the preset standard process parameters to identify the deviation factors that cause component heterogeneity. Specifically, this includes: based on the verified and associated upstream process disturbance period, as well as the unique identifier and detection index of the corresponding heterogeneous region, performing precise retrieval and location operations in the system's pre-stored historical process parameter database to ensure that the retrieval scope is strictly limited to the disturbance period and to avoid parameter extraction errors caused by time period deviation; before the retrieval, confirm the types of core process parameters of the front-end fermentation stage, covering parameters directly related to the uniformity of material composition, such as fermentation temperature, mixing speed, material residence time, feeding parameters, humidity in the fermentation chamber, and aeration intensity (aeration flow rate during aerobic fermentation), to ensure that the extracted parameters are comprehensive and targeted.

[0083] During the retrieval process, actual operating data of various process parameters within the disturbance period are extracted second by second according to the timestamp, forming a complete parameter time-series dataset. Each data point is labeled with the corresponding time, parameter name, and actual value, and is associated with the detection indicators of the corresponding heterogeneous region (e.g., for pH heterogeneity, parameters related to acidity and alkalinity, such as fermentation temperature and feed ratio, are extracted). After extraction, the parameter time-series dataset is preprocessed: missing values ​​and abnormal abrupt changes caused by sensor failure and data transmission interruption are removed; missing data is corrected by using the average of adjacent time periods to ensure the integrity and continuity of the dataset; at the same time, the units of all parameters are uniformly converted to units consistent with standard process parameters to avoid the impact of unit differences on the comparison results; the preprocessed data is then processed. The actual operating parameters after processing are precisely compared item by item and time point by point with the built-in standard process parameter ranges that have been verified through multiple batches of production. The standard process parameter ranges are clearly set for different detection indicators and different fermentation stages. For example, the standard range for fermentation temperature corresponding to qualified pH value is 55 to 65℃, the standard range for mixing speed is 30 to 40 r / min, and the standard range for material residence time is 48 to 72h. Each standard range has upper and lower floating thresholds. If the floating threshold is exceeded, it is judged as parameter deviation. During the comparison process, the focus is on the core parameters related to the detection indicators of heterogeneous areas. For example, for heterogeneous moisture content, the focus is on comparing fermentation temperature, mixing speed, and residence time. For heterogeneous pH value, the focus is on comparing feeding ratio and fermentation temperature.

[0084] By comparison, parameters that continuously exceed the standard range, have fluctuation amplitudes exceeding the floating threshold, and are directly related to the abnormal characteristics of heterogeneous components during the disturbance period are screened out. These parameters are formally identified as deviation factors causing heterogeneity in material composition. At the same time, the specific deviation of each deviation factor is recorded, including the start time, end time, maximum deviation value, and average deviation value of deviation from the standard range. The duration and degree of influence of the deviation factor are clarified. The deviation factor is associated with the unique identifier of the corresponding heterogeneous region, the disturbance period, and the detection index and stored to form a complete deviation factor analysis report.

[0085] Step 4.5: Based on the identified deviation factors, generate dynamic adjustment values ​​for fermentation process parameters, including temperature compensation, mixing speed correction, and residence time adjustment. Specifically, based on the generated deviation factor analysis report, determine the type of each deviation factor (such as temperature deviation, speed deviation, residence time deviation, etc.), its specific deviation range, duration, and degree of influence on material composition. Combined with the actual process rules of sludge organic nutrient soil fermentation and mixing, as well as the mutual influence relationships between different process parameters, generate implementable dynamic adjustment values ​​for each deviation factor. This ensures that the adjustment values ​​accurately correspond to the deviation problem, conform to the actual operating capacity of the production line, and take into account the synergy between various parameters to avoid abnormalities in other process indicators caused by adjusting a single parameter.

[0086] Regarding temperature deviation factors, if the actual fermentation temperature remains consistently above the standard range, a corresponding negative temperature compensation amount, i.e., a cooling adjustment amount, is generated based on the deviation magnitude. The magnitude of the adjustment amount is positively correlated with the temperature deviation magnitude. For example, if the temperature exceeds the standard upper limit by 3°C, the cooling compensation amount is set to 3 to 4°C, while limiting the cooling rate, such as no more than 2°C per hour, to avoid sudden temperature drops leading to insufficient fermentation of the material. If the actual fermentation temperature remains consistently below the standard range, a positive temperature compensation amount, i.e., a heating adjustment amount, is generated. Similarly, the adjustment value is set according to the deviation magnitude to ensure that the temperature can steadily rise back to the standard range, while avoiding excessively rapid heating that could lead to carbonization of the material and loss of nutrients. Nutrient loss; Regarding the mixing speed deviation factor, if the actual mixing speed is lower than the standard range, resulting in uneven material mixing and component stratification, a positive speed correction amount is generated based on the deviation range, i.e., speed increase adjustment amount. For example, if the speed is 5 r / min lower than the standard lower limit, the speed increase correction amount is set to 5-6 r / min to ensure that the mixing intensity meets the material mixing requirements and promotes the full integration of sludge and auxiliary materials. If the actual mixing speed is higher than the standard range, resulting in excessive stirring of materials and particle breakage, affecting the properties of the finished product, a negative speed correction amount is generated, i.e., speed decrease adjustment amount. The adjustment value is reasonably set according to the deviation range to take into account both mixing uniformity and material particle integrity.

[0087] Regarding the residence time deviation factor, if the actual residence time of the material is shorter than the standard range, resulting in insufficient fermentation and failure to meet the qualified standards, a positive residence time adjustment is generated, i.e., extending the residence time. For example, if the residence time is 8 hours shorter than the lower limit of the standard, the adjustment is set to extend it by 8 to 10 hours to ensure that the material has sufficient time to complete the fermentation reaction. If the actual residence time is longer than the standard range, resulting in over-fermentation and component degradation, affecting the nutritional content of the finished product, a negative residence time adjustment is generated (i.e., shortening the residence time). The adjustment value is set according to the deviation range to avoid resource waste caused by over-fermentation. Quality decline; in addition, if there are other deviation factors such as feeding ratio, aeration intensity, etc., corresponding dynamic adjustment values ​​are generated simultaneously. For example, if the feeding ratio deviates, the sludge and auxiliary material feeding rate is adjusted, and if the aeration intensity deviates, the aeration flow rate is adjusted. All dynamic adjustment values ​​are quantified according to the executable accuracy of the production line, and the order and priority of adjustment are confirmed. For example, the temperature is adjusted first, then the rotation speed is adjusted, and finally the residence time is adjusted. A complete process parameter adjustment plan is generated, and the deviation factor, adjustment purpose and execution requirements corresponding to each adjustment value are marked to ensure that the adjustment plan is operable, targeted and reasonable.

[0088] Step 4.6: The dynamic adjustment values ​​are fed back in real time to the temperature control and mixing processes in the front-end fermentation stage to correct the current operating parameters. Specifically, this includes: integrating the generated dynamic adjustment values ​​of process parameters (temperature compensation, mixing speed correction, residence time adjustment, etc.) along with their adjustment priorities and execution requirements into standardized control commands, labeling them with corresponding heterogeneous zone identifiers and deviation factor information to ensure standardized command format and complete information; transmitting the control commands in real time via the production line's industrial Ethernet using encrypted transmission to the corresponding control units in the front-end fermentation stage, including the temperature control module, mixing and stirring control module, and material conveying control module, to prevent data loss or tampering and ensure accurate command delivery; after receiving the commands, each control module strictly corrects the current operating parameters in real time according to the execution priority and adjustment values. The temperature control module adjusts the heating or cooling devices according to the compensation amount to ensure that the fermentation temperature returns to the standard range smoothly; the mixing and stirring control module adjusts the motor frequency according to the speed correction amount to ensure uniform mixing of materials; the material conveying control module adjusts the feed and discharge rates according to the residence time adjustment amount to ensure that the materials ferment fully and without over-fermentation; during the calibration process, each module collects the actual value of the parameters in real time and feeds them back to the monitoring terminal to monitor the adjustment effect throughout the process; if the adjusted parameters do not meet the standard, the adjustment value is finely adjusted and calibrated again until the parameters meet the requirements; after the parameters stabilize, the current operating state is maintained, and the adjustment results, final parameter values ​​and corresponding heterogeneous regions, deviation factors and other information are associated and stored to form a complete process adjustment record, realizing closed-loop control from anomaly tracing to process calibration, reducing the occurrence of component heterogeneous regions from the source, and ensuring the stability of finished product quality.

[0089] In a preferred embodiment of the present invention, step 5 above may include:

[0090] Step 5.1: While feeding back the dynamic adjustment values ​​to the front-end fermentation stage in real time, retrieve the equivalent test dataset corresponding to the current batch of finished products, and extract the quality standard thresholds that the current batch of finished products should meet from the preset database. Specifically, this includes: while feeding back the generated dynamic adjustment values ​​such as temperature compensation, mixing speed correction, and residence time adjustment to the front-end fermentation stage in real time, simultaneously retrieving the equivalent test dataset collected throughout the entire testing process of the current batch of finished products. This dataset contains multiple continuous test indicators directly related to the quality of the finished products, such as pH value, moisture content, organic matter content, and component uniformity, and maintains a one-to-one correspondence with the component distribution cloud map and heterogeneous region location information. The preset database is a long-term accumulated production specifications and finished product acceptance index library, which includes the qualification judgment conditions under different raw material ratios and fermentation processes. The quality standard thresholds are the upper and lower limits of the qualified indicators set according to the requirements for the finished sludge organic nutrient soil to be shipped out, covering the allowable fluctuation range of various key indicators, and are used to directly determine whether the finished product meets the standards.

[0091] Step 5.2 involves comparing the values ​​of each indicator in the equivalent test dataset with the quality standard thresholds item by item. The comparison results are used to determine whether the current batch of finished products meets the qualification standard. Specifically, this includes: comparing each indicator value in the equivalent test dataset with the corresponding extracted quality standard thresholds item by item and point by point, and determining whether each indicator is within the qualified range specified by the standard. After completing the comparison of all indicators, a comprehensive judgment is made. Only when all key indicators meet the standard requirements is the current batch of finished products deemed qualified. If any indicator fails to meet the standard, it is directly determined to be unqualified. The evaluation process is objective and rigorous, and is carried out entirely based on real-time test data.

[0092] Step 5.3: If the evaluation result is qualified, the current batch of finished products is allowed to enter the quantitative weighing and automatic packaging process. Specifically, if the overall evaluation result is qualified, the test data and qualification evaluation results of the current batch of finished products are briefly recorded, and the batch number, test time and compliance status of each indicator are marked to ensure that the qualified batch is traceable. Then, the qualified batch release process is started to determine that the current batch of finished products can enter the subsequent quantitative weighing and automatic packaging stage. During the weighing process, the weight of each package is strictly controlled according to the finished product's factory specifications to ensure that the weight error is within the allowable range. In the packaging stage, the finished products are packaged and labeled according to the standardized process, and key information such as batch, production date and test qualification mark are marked to ensure that the qualified finished products meet the factory specifications and smoothly complete the subsequent factory related processes.

[0093] Step 5.4: If the evaluation result is unqualified, the current batch of finished products will be diverted to the return processing channel during transportation. Simultaneously, the data sealing procedure will be triggered. Specifically, if the overall evaluation result is unqualified, the unqualified batch processing flow will be immediately initiated, recording the batch number, unqualified indicators, and degree of unqualification of the current batch of finished products to avoid confusion between unqualified and qualified batches. Subsequently, the finished product transportation path will be adjusted, guiding the current batch of finished products away from the normal production process and into a dedicated return processing stage. Following this, based on the cause of the unqualification, the batch of finished products will undergo targeted treatment such as secondary fermentation and remixing to ensure rational resource utilization. Simultaneously, the data sealing procedure will be initiated. This procedure automatically locks and collects all process data related to this unqualified batch, prohibiting arbitrary modification, overwriting, or deletion to ensure the integrity, authenticity, and immutability of the abnormal information, providing a complete basis for subsequent abnormality analysis.

[0094] Step 5.5: The data sealing program packages and seals the generated component distribution cloud map, the generated dynamic adjustment values, and the identified deviation factors to form a quality anomaly record corresponding to this non-conforming batch. Specifically, this includes: after the data sealing program is started, sorting out all key data related to this non-conforming batch, including the generated continuous virtual component distribution cloud map, the identified deviation factors causing component heterogeneity, the generated dynamic adjustment values ​​of process parameters, and information such as non-conformity assessment results, non-conformity indicators, judgment time, and batch number; structuring and organizing this data, packaging it according to a unified standard, labeling the corresponding non-conforming batch number and sealing time, and classifying and storing it according to archiving requirements to form a complete and standardized quality anomaly record; this record enables full traceability, and can be used to review the causes of non-conformity, optimize front-end fermentation process parameters, avoid the recurrence of similar non-conformity problems, and provide reliable data support for production process improvement.

[0095] In a preferred embodiment of the present invention, step 6 above may include:

[0096] Step 6.1: Extract various test data of the current batch of finished products from the output equivalent test dataset, extract cloud map feature values ​​representing its distribution characteristics from the generated component distribution cloud map, and extract the production period information corresponding to the current batch from the recorded basic traceability information. Specifically, this includes: extracting all key test data of the current batch of finished products from the output equivalent test dataset, covering all indicators related to finished product quality such as pH value, moisture content, organic matter content, fecal coliform count, and component uniformity, while extracting auxiliary information such as the test time and test location for each indicator to ensure the completeness of the test data. Traceability; from the continuous virtual component distribution cloud map, cloud map feature values ​​that can characterize its distribution characteristics are extracted, including the area ratio of homogeneous and heterogeneous regions, the maximum concentration gradient, the distribution location and range of abnormal regions, etc., to accurately reflect the spatial distribution of components in the current batch of finished products; at the same time, from the recorded basic traceability information, complete production time information such as the front-end fermentation period, mixing period, transportation period and testing period corresponding to the current batch of finished products is extracted, clarifying the time nodes of each link, and ensuring that the extracted testing data, cloud map feature values ​​and production time information are accurately corresponding to the current batch of finished products, without overlap or omission.

[0097] Step 6.2 involves establishing a one-to-one correspondence between the packaging batch code corresponding to the current batch of finished products in the automatic packaging process and the extracted test data, cloud map feature values, and production time information. This establishes a multi-dimensional data association relationship. Specifically, this includes: confirming the packaging batch code generated in the automatic packaging process for the current batch of finished products. This code contains key information such as the production date and batch number, serving as the core identifier for the current batch of finished products. The packaging batch code is then precisely matched and bound to each extracted test data, cloud map feature value, and production time information to establish a multi-dimensional data association relationship. Specifically, binding the packaging batch code to the test data determines the quality indicator details for each batch of finished products; binding it to the cloud map feature value determines the component distribution characteristics for each batch of finished products; and binding it to the production time information determines the entire production process timeline for each batch of finished products. During the binding process, the correlation of each data point is strictly verified to ensure that each data point can be accurately assigned to the corresponding packaging batch code, forming a one-to-one correspondence between code, data, feature, and time period. This lays the foundation for traceability code generation and the establishment of traceability archives.

[0098] Step 6.3: Based on the established multi-dimensional data association, a unique traceability code corresponding to the batch of finished products is generated through coding calculation. Specifically, this includes: organizing, combining, and coding information such as packaging batch code, core nodes of production period, and key testing indicator characteristics according to the established multi-dimensional data association and the set unified coding rules; during the coding generation process, the unique characteristics of the current batch are fully combined, and information such as production date, batch number, and simplified identification of core testing indicators are integrated into the code to ensure that the generated traceability code has uniqueness and exclusivity, with each batch of finished products corresponding to a unique traceability code, without duplication or confusion; after generating the traceability code, the code is validated to confirm that the code format is standardized, the information is complete, and it can accurately associate with all relevant data of the current batch, serving as an exclusive identity identifier for the current batch of finished products for subsequent quality traceability, information query, and other related operations.

[0099] Step 6.4 involves uploading the unique traceability code and all associated data to the quality traceability platform for storage, forming a complete quality traceability archive that can be queried and accessed. Specifically, this includes: uniformly organizing the generated unique traceability code and all associated data, including various test data, cloud map feature values, production period information, packaging batch codes, packaging specifications, and all other relevant information, ensuring data completeness, accuracy, and error-free association; uploading the organized traceability code and associated data to a dedicated quality traceability platform via a secure and stable transmission method; the platform performs structured classification and standardized archiving of the uploaded data, establishing a dedicated data storage directory according to the traceability code, and confirming the data storage path and access permissions; simultaneously, the platform performs integrity verification on the uploaded data, checking for missing data, incorrect associations, and other issues, ensuring that the stored data is authentic, complete, searchable, and accessible, ultimately forming a complete quality traceability archive for subsequent quality verification, process traceability, and problem review, achieving full-process traceability for each batch of finished products.

[0100] Step 6.5: If the current batch of finished products is determined to be unqualified, the sealed quality anomaly record is simultaneously associated when generating the unique traceability code. This ensures that the traceability file of the current batch of finished products contains complete information on the cause and handling of the anomaly. Specifically, if the current batch of finished products is determined to be unqualified during quality assessment, the sealed quality anomaly record is simultaneously associated and bound when generating the unique traceability code. During the association process, it is ensured that the traceability code and the quality anomaly record correspond accurately. All information contained in the anomaly record, such as details of heterogeneous regions of components, deviation factors, dynamic process adjustment values, unqualified judgment results, and unqualified handling methods, is included in the traceability code association data of this batch. Ultimately, the quality traceability file of this batch of finished products not only contains conventional testing data, cloud map feature values, production time, etc., but also fully covers key content such as the cause and handling process of the anomaly. This achieves integrated storage of information on the entire process of production, testing, anomaly, and handling of unqualified batches, providing complete data support for reviewing the causes of anomalies, optimizing processes, and avoiding similar problems.

[0101] like Figure 2 As shown, embodiments of the present invention also provide a sludge organic nutrient soil production testing system, comprising:

[0102] The multi-point synchronous sampling module is used to perform multi-point synchronous sampling at a preset frequency during the conveying process after the finished product is crushed and screened and before automatic packaging, and to collect samples from each point to form a mixed sample that characterizes the batch composition distribution characteristics.

[0103] The multi-source spectral intelligent analysis module is used to build an intelligent analysis model based on multi-source data fusion, obtain the spectral response characteristics of mixed samples and couple them with a preset component association feature library to generate an equivalent detection dataset simultaneously.

[0104] The component spatial distribution reconstruction module is used to construct a component concentration spatial distribution matrix on the cross-section of the conveyor belt based on the geographic spatial coordinates of each point and the corresponding material property parameters during the multi-point synchronous sampling process, based on the equivalent detection dataset. After spatial interpolation fitting of the distribution matrix, a virtual component distribution cloud map is generated. The homogeneous region and heterogeneous region are divided by identifying the component concentration gradient change pattern in the cloud map.

[0105] The process reverse deduction and dynamic control module is used to map the boundary coordinates of heterogeneous regions in the component distribution cloud map to the conveyor belt running trajectory. Combined with the conveyor belt running speed parameters, it performs spatial reverse deduction to calculate the upstream process disturbance period that causes component heterogeneity. Based on this, it generates dynamic adjustment values ​​for fermentation process parameters and feeds them back to the temperature control and mixing processes in the front-end fermentation stage.

[0106] The online quality assessment and diversion control module is used to compare the equivalent test dataset with the preset quality standard threshold during the feedback process. If the assessment is qualified, the package is released; if it is unqualified, the diversion and return of materials is triggered and the corresponding cloud map and process adjustment record are sealed.

[0107] The end-to-end quality traceability module is used to associate and bind the pre-stored test data, cloud map feature values, and production time information of the current batch of finished products with the packaging batch, and generate a unique traceability code to achieve quality traceability.

[0108] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for testing the production of sludge organic nutrient soil, characterized in that, The method includes: After the finished product is crushed and screened and before automatic packaging, multi-point synchronous sampling is carried out at a preset frequency, and samples from each point are collected to form a mixed sample that characterizes the batch composition distribution characteristics. A smart analytical model based on multi-source data fusion is constructed to obtain the spectral response characteristics of mixed samples and couple them with a preset component association feature library to generate an equivalent detection dataset simultaneously. Based on the equivalent detection dataset, according to the geographic spatial coordinates of each point and the corresponding material property parameters during the multi-point synchronous sampling process, a spatial distribution matrix of component concentration on the cross-section of the conveyor belt is constructed. After spatial interpolation fitting of the distribution matrix, a virtual component distribution cloud map is generated. By identifying the component concentration gradient change pattern in the cloud map, homogeneous and heterogeneous regions are divided. The boundary coordinates of the heterogeneous region in the component distribution cloud map are mapped onto the conveyor belt running trajectory. Spatial reverse deduction is performed in combination with the conveyor belt running speed parameters to calculate the upstream process disturbance period that causes component heterogeneity. Based on this, dynamic adjustment values ​​for fermentation process parameters are generated and fed back to the temperature control and mixing processes in the front-end fermentation stage. During the feedback process, the equivalent test dataset is compared with the preset quality standard threshold. If the evaluation is qualified, the package is released; if it is unqualified, the diversion and return of materials is triggered and the corresponding cloud map and process adjustment record are sealed. The pre-stored test data, cloud map feature values, and production time information of the current batch of finished products are associated and bound with the packaging batch to generate a unique traceability code to achieve quality traceability.

2. The method for testing the production of sludge organic nutrient soil according to claim 1, characterized in that, In a continuous production line, during the conveying process after finished product crushing and screening but before automatic packaging, multi-point synchronous sampling is performed at a preset frequency. Samples from each point are collected to form a mixed sample characterizing the batch composition distribution, including: The sampling time interval is set by combining the running speed of the conveyor belt and the finished product discharge rate, thereby forming periodic sampling time points. Based on the natural angle of repose cross-sectional shape formed by the accumulation of materials on the conveyor belt, characteristic sampling points with spatial distribution representativeness are determined on the cross-section, including the boundary reference points located at the toes of the left and right slopes, the crown reference point located at the top of the cross-section, and the midpoint of the slope located in the middle of the line connecting each side boundary reference point and the crown reference point. When each sampling time point arrives, the sampling execution unit arranged directly above each feature sampling point is synchronously triggered, so that each sampling execution unit extends downward to the preset depth inside its corresponding material accumulation layer to capture the material sample that flows through the longitudinal section of this point instantaneously. The instantaneous material samples captured by each sampling execution unit are synchronously collected into the same sample collection container, so that the material samples from different spatial feature points are physically mixed in the sample collection container to form a mixed sample that can characterize the composition distribution characteristics of the entire cross-section of the conveyor belt. Record the sampling time point corresponding to the mixed sample, the spatial coordinate code of each characteristic sampling point, and the identification mark of the sample collection container as the basic traceability information of this batch of samples.

3. The method for testing the production of sludge organic nutrient soil according to claim 2, characterized in that, A smart analytical model based on multi-source data fusion is constructed to obtain the spectral response characteristics of mixed samples and couple them with a pre-defined component association feature library for calculation, simultaneously generating an equivalent detection dataset, including: The aggregated mixed sample was subjected to a full-band spectral scan to obtain the original spectral response curve of the mixed sample; The original spectral response curves were preprocessed by background subtraction, noise filtering and baseline correction to extract a set of characteristic spectral signals that can characterize the intrinsic components of the sample. Retrieve a pre-constructed component association feature library, which stores a mapping relationship matrix between different component indices and characteristic spectral signals; The feature spectral signal set is input into the intelligent analytical model of multi-source data fusion. Through the coupling solution algorithm inside the model, the feature spectral signal and the mapping relationship matrix are matched and inverted until the algorithm converges and the coupling solution is completed. After the coupling calculation is completed, the values ​​of pH, moisture content, organic matter content, fecal coliforms and heavy metals are output synchronously to form an equivalent detection dataset corresponding to the current mixed sample.

4. The method for testing the production of sludge organic nutrient soil according to claim 3, characterized in that, Based on the equivalent detection dataset, and according to the geographic coordinates of each point and the corresponding material property parameters during multi-point synchronous sampling, a spatial distribution matrix of the component concentration of the finished product on the cross-section of the conveyor belt is constructed. After spatial interpolation fitting of the distribution matrix, a virtual component distribution cloud map is generated. By identifying the component concentration gradient variation patterns in the cloud map, homogeneous and heterogeneous regions are delineated, including: Extract the spatial coordinate codes of each feature sampling point from the basic traceability information, as well as the material property parameters collected synchronously at each point at the sampling time, and obtain the overall composition value of the current batch of finished products from the equivalent detection dataset. The material property parameters and overall composition values ​​of each point are input into the spatial inversion model for calculation, so as to calculate the estimated composition concentration of the local area corresponding to each feature sampling point. Using the spatial coordinate codes of each feature sampling point as the location reference and the corresponding component concentration estimation value as the assignment variable, a discrete component concentration spatial distribution matrix of the finished product on the cross-section of the conveyor belt is constructed. Spatial interpolation fitting is performed on the discrete component concentration spatial distribution matrix to generate a continuous virtual component distribution cloud map covering the entire cross-section of the conveyor belt; Gradient analysis is performed on continuous virtual component distribution cloud maps to identify the gradient variation pattern of component concentration in the cloud map. Based on the preset gradient threshold, the cloud map region is divided into a homogeneous region with relatively uniform component distribution and a heterogeneous region with obvious component distribution fluctuations.

5. The method for testing the production of sludge organic nutrient soil according to claim 4, characterized in that, The boundary coordinates of heterogeneous regions in the component distribution cloud map are mapped onto the conveyor belt's running trajectory. Spatial inverse extrapolation is then performed using the conveyor belt speed parameters to calculate the upstream process disturbance periods that cause component heterogeneity. Based on this, dynamic adjustment values ​​for fermentation process parameters are generated and fed back to the temperature control and mixing processes in the front-end fermentation stage, including: Extract the set of spatial coordinates of the boundary contours of the divided heterogeneous regions on the cross-section of the conveyor belt, and use them as the boundary coordinate data of the heterogeneous regions. Map the boundary coordinate data of the heterogeneous region onto the running trajectory of the conveyor belt to establish the spatial interval between the start and end positions of the heterogeneous region along the length of the conveyor belt. The real-time operating speed parameters of the conveyor belt are retrieved and combined with the spatial interval to perform spatial reverse deduction, so as to convert the spatial interval into a time interval and calculate the upstream process disturbance period corresponding to the material that forms this heterogeneous region. Based on the upstream process disturbance period, locate the actual operating parameters of the front-end fermentation stage within this period in the pre-stored historical process parameter records, and compare the actual operating parameters with the preset standard process parameters to identify the deviation factors that cause component heterogeneity. Based on the identified deviation factors, dynamic adjustment values ​​for fermentation process parameters are generated, including temperature compensation, mixing speed correction, and residence time adjustment. The dynamic adjustment values ​​are fed back in real time to the temperature control and mixing processes in the front-end fermentation stage to correct the current operating parameters.

6. The method for testing the production of sludge organic nutrient soil according to claim 5, characterized in that, During the feedback process, the equivalent test dataset is compared with the preset quality standard threshold. If the assessment is satisfactory, the material is released for packaging; otherwise, it is redirected and returned, and the corresponding cloud map and process adjustment records are sealed, including: While feeding the dynamic adjustment value back to the front-end fermentation stage in real time, the equivalent test dataset corresponding to the current batch of finished products is retrieved, and the quality standard threshold that the current batch of finished products should meet is extracted from the preset database. The values ​​of each indicator in the equivalent test dataset are compared with the quality standard threshold item by item, and the comparison results are used to determine whether the current batch of finished products meets the qualified standard. If the evaluation result is qualified, the current batch of finished products is allowed to enter the quantitative weighing and automatic packaging process; If the evaluation result is unqualified, the current batch of finished products will be diverted to the return processing channel during the transportation process, and the data sealing procedure will be triggered at the same time. The data sealing program packages and seals the generated component distribution cloud map, the generated dynamic adjustment value, and the identified deviation factor to form a quality anomaly record corresponding to this non-conforming batch.

7. The method for testing the production of sludge organic nutrient soil according to claim 6, characterized in that, The pre-stored test data, cloud map feature values, and production period information of the current batch of finished products are associated and bound with the packaging batch to generate a unique traceability code for quality traceability, including: Extract various test data of the current batch of finished products from the output equivalent test dataset, extract cloud map feature values ​​that characterize its distribution characteristics from the generated component distribution cloud map, and extract the production period information corresponding to the current batch from the recorded basic traceability information; The packaging batch code corresponding to the current batch of finished products in the automatic packaging process is matched one-to-one with the extracted test data, cloud map feature values ​​and production time information to establish a multi-dimensional data association relationship; Based on the established multi-dimensional data association, a unique traceability code corresponding to the batch of finished products is generated through coding and calculation. The unique traceability code and all associated data are uploaded to the quality traceability platform for storage, forming a complete quality traceability archive that can be queried and accessed. If the current batch of finished products is determined to be unqualified, the sealed quality anomaly record will be linked simultaneously when a unique traceability code is generated, so that the traceability file of the current batch of finished products contains complete information on the cause and handling of the anomaly.

8. A sludge organic nutrient soil production and testing system, wherein the system implements the method as described in any one of claims 1 to 7, characterized in that, include: The multi-point synchronous sampling module is used to perform multi-point synchronous sampling at a preset frequency during the conveying process after the finished product is crushed and screened and before automatic packaging, and to collect samples from each point to form a mixed sample that characterizes the batch composition distribution characteristics. The multi-source spectral intelligent analysis module is used to build an intelligent analysis model based on multi-source data fusion, obtain the spectral response characteristics of mixed samples and couple them with a preset component association feature library to generate an equivalent detection dataset simultaneously. The component spatial distribution reconstruction module is used to construct a component concentration spatial distribution matrix on the cross-section of the conveyor belt based on the geographic spatial coordinates of each point and the corresponding material property parameters during the multi-point synchronous sampling process, based on the equivalent detection dataset. After spatial interpolation fitting of the distribution matrix, a virtual component distribution cloud map is generated. The homogeneous region and heterogeneous region are divided by identifying the component concentration gradient change pattern in the cloud map. The process reverse deduction and dynamic control module is used to map the boundary coordinates of heterogeneous regions in the component distribution cloud map to the conveyor belt running trajectory. Combined with the conveyor belt running speed parameters, it performs spatial reverse deduction to calculate the upstream process disturbance period that causes component heterogeneity. Based on this, it generates dynamic adjustment values ​​for fermentation process parameters and feeds them back to the temperature control and mixing processes in the front-end fermentation stage. The online quality assessment and diversion control module is used to compare the equivalent test dataset with the preset quality standard threshold during the feedback process. If the assessment is qualified, the package is released; if it is unqualified, the diversion and return of materials is triggered and the corresponding cloud map and process adjustment record are sealed. The end-to-end quality traceability module is used to associate and bind the pre-stored test data, cloud map feature values, and production time information of the current batch of finished products with the packaging batch, and generate a unique traceability code to achieve quality traceability.

Images