AI-based experimental data structured management method and system

Abstract

The present application relates to the technical field of experimental data intelligent management, and particularly relates to an experimental data structured management method and system based on AI. The method comprises the following steps: obtaining calcination experimental data, and performing batch processing according to the calcination experimental data to obtain color experiment batch data; performing reduction measure event analysis and cooling path matching on the color experiment batch data to obtain reduction event data and cooling path characteristic data respectively; performing chroma feature extraction on the calcination experimental data to obtain sample chroma data; and performing structured mapping according to the reduction event data, the cooling path characteristic data and the sample chroma data to obtain color control mapping data. The present application can establish traceable correlation between scattered experimental process data and final color results, improve the data consistency and reproducibility of the calcination clay color control experiment, and provide a data basis for experimental anomaly attribution, process parameter optimization and automatic report generation.

Description

Technical Field

[0001] This invention relates to the field of intelligent management technology for experimental data, and in particular to an AI-based method and system for structured management of experimental data. Background Technology

[0002] Calcined clay, as an important component of low-carbon cementitious materials, is affected by various factors such as calcination temperature, furnace atmosphere, CO injection, pulverized coal reduction, cooling method, and material residence time. In actual experiments, to obtain calcined clay samples with the target gray color or close to the target chromaticity, multiple rounds of testing with different reduction methods, cooling paths, and process parameters are usually required. However, existing experimental management methods largely rely on manual recording, table archiving, or separately exported test files from equipment, making it difficult to uniformly bind the material feeding and discharging information, reduction methods, cooling equipment status, temperature profiles, sample images, and final chromaticity results from a single calcination experiment to the same experimental batch.

[0003] While existing experimental data management systems can handle sample registration, equipment management, data storage of test samples, and report generation, they still fall short in experimental scenarios with strong process coupling characteristics, such as color control of calcined clay. On one hand, reduction measures such as CO injection and pulverized coal addition are often only documented through manual notes or operation records, lacking information on start and end times, location of action, dosage, temperature conditions, and effectiveness markings. This makes it difficult to determine whether color changes originate from the reduction measures themselves or are influenced by the cooling path or furnace atmosphere. On the other hand, the equipment parameters for different cooling paths, such as water cooling, rotary cooling, and cyclone cooling, vary significantly. Simply recording "a certain cooling method is used" fails to reflect process information such as cooling inlet temperature, outlet temperature, cooling rate, residence time, and abnormal heat release characteristics, making it difficult to analyze the actual contribution of the cooling process to sample color formation. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes an AI-based method and system for structured management of experimental data, thereby resolving at least one of the aforementioned technical issues.

[0005] This application provides an AI-based method for structured management of experimental data, including the following steps:

[0006] Acquire calcination experiment data and perform batch processing based on the calcination experiment data to obtain color experiment batch data;

[0007] The color experiment batch data was analyzed for restoration measures and matched for cooling paths to obtain restoration event data and cooling path feature data, respectively.

[0008] Colorimetric features were extracted from the calcination experimental data to obtain the sample colorimetric data;

[0009] Color control mapping data is obtained by performing structured mapping based on the restoration event data, cooling path feature data, and sample chromaticity data.

[0010] This invention processes calcination experimental data continuously by batch, reduction method, cooling path, and colorimetric results, creating a traceable data chain from previously scattered records of CO injection, coal powder reduction, cooling methods, and final color performance. The system can distinguish the impact of different reduction events and cooling processes on sample colorimetry, reducing reliance on manual experience to judge color control effectiveness. Simultaneously, by generating color control mapping data through structured mapping, it improves the data utilization and analytical consistency of calcined clay color control experiments.

[0011] Optionally, batch processing specifically includes:

[0012] Sample batch boundary identification is performed on the calcination experiment data to obtain batch interval data; sample identification is bound based on the batch interval data to obtain batch sample binding data; process window is segmented based on the batch sample binding data to obtain color experiment batch data.

[0013] This invention enables the division of continuously collected calcination experimental data into experimental batches with clearly defined start and end boundaries, preventing data from different samples, different feeding stages, or different discharging stages from being mixed together. By binding sample identification, process data such as temperature, atmosphere, reduction measures, and cooling paths can be accurately mapped to specific samples; furthermore, by segmenting the process window, continuous operating conditions within a batch are transformed into calculable time-series segments, improving data traceability and analytical reliability.

[0014] Optionally, the event analysis for restoration measures is as follows:

[0015] The restoration measure field is identified in the color experiment batch data to obtain restoration measure data; the restoration type is determined in the restoration measure data to obtain restoration type data; the event boundary is located based on the restoration type data to obtain restoration event interval data; the action conditions are extracted from the restoration event interval data to obtain restoration event feature data; and the event validity is marked based on the restoration event feature data to obtain restoration event data.

[0016] This invention can transform scattered records of CO injection, pulverized coal addition, and reducing atmosphere adjustment in batch data into event data with clear types, start and end boundaries, and operating conditions. Through field identification and type discrimination, the source and mode of action of different reduction measures can be distinguished; through event boundary positioning, the actual time interval of the reduction measure can be accurately determined; and by combining conditions such as temperature, atmosphere concentration, addition amount, and contact duration for validity marking, it can avoid the coarse management method of simply recording "whether reduction measures were taken".

[0017] Optionally, the cooling path matching is specifically as follows:

[0018] The cooling equipment status is identified from the batch data of the color experiment to obtain cooling equipment status data; the material cooling inlet is located based on the cooling equipment status data to obtain cooling inlet event data; cooling path type is matched based on the cooling inlet event data to obtain cooling path type data; cooling process features are extracted from the cooling path type data to obtain cooling process feature data; and path validity is verified based on the cooling process feature data to obtain cooling path feature data.

[0019] This invention transforms the cooling process of calcined materials from simple equipment records into traceable path feature data. By identifying the status of the cooling equipment and locating the material cooling inlet, the starting point of the material entering the cooling section can be accurately determined. Then, by matching path types, it distinguishes between water cooling, rotary cooling, cyclone cooling, or combined paths, and extracts process features such as temperature changes, cooling rates, residence times, and equipment operating status. Path validity verification eliminates interference from no-load operation, path mismatches, or abnormal temperature transitions that could affect data analysis.

[0020] Optionally, the cooling path type matching is specifically as follows:

[0021] The inlet location data of the cooling inlet event is parsed to obtain inlet location identification data; the material flow direction valve status is matched based on the inlet location identification data to obtain material flow direction matching data; the cooling equipment operation is verified based on the material flow direction matching data to obtain equipment verification data; the path type is determined based on the equipment verification data to obtain single path type data; and continuous path tracing is performed based on the single path type data to obtain cooling path type data.

[0022] This invention accurately extracts the actual cooling path of materials from the complex operating status of equipment. By analyzing the inlet position, the initial position of the material entering the cooling system can be determined; by matching the material flow direction with the valve status, misjudging the cooling path based solely on equipment start signals can be avoided; further, by verifying the operation of the cooling equipment, invalid paths under no-load, shutdown, or abnormal operating conditions are eliminated. The system performs single-path discrimination and continuous path tracing, identifying water cooling, rotary cooling, cyclone cooling, and combined cooling paths, establishing an accurate correspondence between the cooling method and subsequent colorimetric results, improving the traceability of color control experimental data and the reliability of path determination.

[0023] Optionally, the feature extraction of the cooling process specifically includes:

[0024] Cooling temperature sequences are constructed based on cooling path type data to obtain cooling curve data. Derivative curves are extracted from the cooling curve data, and standard Newtonian cooling kinetics decay is performed to obtain derivative curve data and benchmark cooling model data, respectively. Abnormal heat release feature points are identified from the derivative curve data to obtain heat release feature point data. Residual integrals are calculated from the cooling curve data and benchmark cooling model data to obtain residual integral data. The heat release feature point data and residual integral data are integrated to obtain cooling process feature data.

[0025] This invention transforms ordinary time-series records of temperature changes in the cooling path into analyzable thermal response characteristics. By constructing cooling curves and extracting derivative curves, abnormal phenomena such as cooling rate variations, cooling stagnation, and localized heat release can be identified. By introducing a standard Newtonian cooling kinetic model, a normal cooling attenuation benchmark can be established, quantifying the deviation between the actual and theoretical cooling processes. Combining abnormal heat release characteristic points and residual integral data, the system can determine whether there is thermal stagnation, phase change heat effects, or insufficient cooling contact during the cooling process, improving the reliability of color control mapping data.

[0026] Optionally, the identification of abnormal heat release feature points specifically includes:

[0027] Inflection points are selected from the derivative curve data to obtain inflection point data;

[0028] The material heat release temperature range is matched based on the cooling path type data to obtain pyrolysis range data; the inflection point data is filtered using the pyrolysis range data to obtain effective inflection point data; the effective inflection point data is truncated with a disturbance waveform to obtain disturbance waveform data; the disturbance waveform data is integrated with energy and decomposed in the frequency domain to obtain heat release disturbance characteristic data and disturbance frequency data, respectively; the feature points are determined based on the heat release disturbance characteristic data and disturbance frequency data to obtain heat release characteristic point data.

[0029] This invention identifies truly significant anomalous locations with thermal effects from the cooling derivative curve, rather than simply treating all temperature fluctuations as anomalies. By selecting inflection points, locations of abrupt changes in cooling rate can be initially captured; further, interval filtering based on the material's heat release temperature range eliminates ordinary disturbances unrelated to the material's thermal response. Energy integration and frequency domain decomposition of the disturbance waveform near effective inflection points simultaneously quantify the local heat release intensity and disturbance source characteristics, distinguishing between heat retention, phase change thermal effects, and pseudo-anomalies caused by equipment vibration or airflow fluctuations. The resulting heat release characteristic point data is more physically directional.

[0030] Optionally, the chromaticity feature extraction specifically involves:

[0031] The sample image data in the calcination experiment data is collected and standardized to obtain standard sample image data; the standard sample image data is segmented to obtain effective sample area data; the effective sample area data is converted to color space to obtain color space feature data; the color offset is calculated based on the color space feature data to obtain chromaticity offset data; and the chromaticity level is determined based on the chromaticity offset data to obtain sample chromaticity data.

[0032] This invention transforms the color evaluation of calcined clay samples from manual visual inspection into calculable and verifiable structured data. By standardizing the acquisition of sample images, interference from differences in light source, angle, and background can be reduced for color recognition. Sample region segmentation eliminates labels, shadows, reflections, and background areas, improving the accuracy of colorimetric calculations. Converting the effective region to a color space and calculating the color shift quantifies the degree of reddish, dark, or yellowish tint of the sample relative to the target gray or standard sample. Colorimetric level determination generates sample colorimetric data to provide data support.

[0033] Optionally, the structured mapping specifically refers to:

[0034] Indexing and matching the restoration event data, cooling path feature data, and sample chromaticity data yields batch-related data; restoration-cooling combined features are constructed based on the batch-related data to obtain combined feature data; chromaticity response mapping is performed based on the combined feature data and sample chromaticity data to obtain chromaticity response relationship data; and data encapsulation is performed based on the chromaticity response relationship data to obtain color control mapping data.

[0035] This invention transforms scattered data on reduction measures, cooling processes, and final chromaticity results into a color control data chain with batch-related relationships. Through index matching, it ensures that the characteristics of CO injection, pulverized coal reduction, and cooling paths belong to the same experimental batch as the corresponding sample chromaticity results, avoiding data mismatch. By constructing reduction-cooling combination features, the synergistic effects of different reduction measures and cooling methods can be expressed. Through chromaticity response mapping, the influence of combined operating conditions on sample color shift and chromaticity level changes can be quantified. The system encapsulates the resulting color control mapping data, providing a structured basis for color change attribution, experimental condition reproduction, and process parameter optimization.

[0036] Optionally, this application also provides an AI-based experimental data structuring management system for executing the AI-based experimental data structuring management method described above. The AI-based experimental data structuring management system includes:

[0037] The batch processing module is used to acquire calcination experiment data and perform batch processing based on the calcination experiment data to obtain color experiment batch data.

[0038] The color experiment batch feature extraction module is used to perform restoration measure event analysis and cooling path matching on the color experiment batch data, and obtain restoration event data and cooling path feature data respectively.

[0039] The colorimetric feature extraction module is used to extract colorimetric features from calcination experimental data to obtain sample colorimetric data.

[0040] The structured mapping module is used to perform structured mapping based on the restoration event data, cooling path feature data, and sample chromaticity data to obtain color control mapping data.

[0041] This invention uses batch processing to bind continuously collected calcination temperature, atmosphere parameters, material feeding and discharging records, and sample information to specific color experiment batches, avoiding data mixing between different samples or different operating conditions. Through reduction measure event analysis, operations such as CO injection and pulverized coal addition are transformed into reduction event data with start and end times, location of action, amount added, temperature conditions, and validity markers, moving beyond manual annotation of reduction effects. Cooling path matching identifies water cooling, rotary cooling, cyclone cooling, or combined cooling paths, and extracts process features such as cooling rate, residence time, and abnormal heat release characteristics, providing a basis for judging the impact of the cooling process on color formation. The system extracts chromaticity features from sample images, converting manually observed color differences into quantifiable chromaticity data. Through structured mapping, a correspondence is established between reduction events, cooling path features, and sample chromaticity results, forming color control mapping data, thus providing data support for color anomaly attribution, experimental condition reproduction, process parameter optimization, and automatic report generation. Attached Figure Description

[0042] Other features, objects, and advantages of this application will become more apparent from the following detailed description of the non-limiting embodiments, taken with reference to the accompanying drawings:

[0043] Figure 1 A flowchart illustrating the steps of an AI-based experimental data structure management method according to one embodiment is shown.

[0044] Figure 2 A flowchart illustrating the steps of a batch processing method according to one embodiment is shown.

[0045] Figure 3 A flowchart illustrating the steps of an event analysis method for reconstructing a response is shown in one embodiment.

[0046] Figure 4 A flowchart illustrating the steps of a chromaticity feature extraction method according to an embodiment is shown.

[0047] Figure 5A flowchart illustrating the steps of a structured mapping method according to one embodiment is shown. Detailed Implementation

[0048] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0049] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. Functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0050] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0051] A laboratory conducted a calcination color control experiment on iron-containing ceramic pigment samples. The system acquired 120 records of calcination experiments, operation records, and sample images over three consecutive days, and categorized them into 18 color experiment batches. For batch B006, the system identified its set temperature as 820℃, heating rate as 5℃ / min, holding time as 90min, calcination atmosphere as low-oxygen atmosphere, and raw material formula as F-02. The system detected that this batch involved the introduction of reducing gas and reduction of oxygen supply under exhaust gas treatment conditions, and matched the sample to the air-cooling section within 3 minutes after calcination, with the cooling path type being rapid cooling. Based on the Lab color space, the system extracted the sample's average brightness value as 62.4, average red-green component value as 18.7, average yellow-blue component value as 31.5, and color dispersion as 4.8.

[0052] For batch B006, the system identified its set temperature as 820℃, heating rate as 5℃ / min, holding time as 90min, calcination atmosphere as low-oxygen atmosphere, and raw material formula as F-02. The system analyzed the reduction measure events in the operator records for this batch, identifying two effective reduction measures: "introducing carbon monoxide at the 20th minute of holding" and "reducing oxygen supply at the 55th minute of holding," and merged them into reduction event data of reducing gas introduction type and oxygen supply reduction type. Simultaneously, the system matched the cooling path in the cooling records, identifying that this batch entered the air-cooling section 3 minutes after calcination, with the cooling path type being rapid cooling, and the cooling start time being 3 minutes after the experiment ended.

[0053] The system reads the sample image data corresponding to batch B006, performs standardization processing, sample region segmentation, and color space conversion on the sample image, obtaining an average brightness value of 62.4, an average red-green component value of 18.7, an average yellow-blue component value of 31.5, and a color dispersion of 4.8. Based on the batch identifier, the system performs structured mapping of the reduction event data, cooling path feature data, and sample chromaticity data to generate color control mapping data, namely, "F-02 formula—calcination at 820℃—low-oxygen atmosphere—introduction of carbon monoxide and reduction of oxygen supply—rapid air cooling—brightness value 62.4, red-green component 18.7, yellow-blue component 31.5".

[0054] Through this embodiment, the system can uniformly map data that was originally scattered across experimental records, operation logs, cooling records, and sample images into traceable color control relationships. Compared with manual processing, the system can not only reduce problems such as batch confusion, omissions in recording reduction measures, and misjudgments of cooling paths, but also clarify the correspondence between reduction measures, cooling methods, and the final colorimetric results, providing a structured data foundation for optimizing calcination atmosphere, cooling methods, and color stability.

[0055] Please see Figures 1 to 5 This application provides an AI-based method for structured management of experimental data, comprising the following steps:

[0056] S1. Obtain calcination experiment data and perform batch processing based on the calcination experiment data to obtain color experiment batch data;

[0057] In one embodiment, the system reads calcination experiment data, which includes experiment record number, sample number, experiment time, calcination furnace number, set temperature, heating rate, holding time, calcination atmosphere, raw material formula, and operator records. The experiment time includes the experiment start time, and the system determines the experiment end time based on the experiment start time and holding time. Alternatively, the calcination experiment data includes the experiment start time and experiment end time, and the operator record set includes the operation record time, operation content, and recorder information. The calcination experiment data also includes sample image data, which includes the image acquisition time, sample number, and sample surface image. The system first performs deduplication and temporal sorting according to the sample number, experiment record number, and experiment time. Then, it uses the same calcination furnace number, the same raw material formula, a set temperature difference not exceeding 5°C, a heating rate difference not exceeding 1°C / min, a holding time difference not exceeding 10min, and an experiment start time interval not exceeding 30min as batch merging conditions, grouping experiment records that meet these conditions into the same color experiment batch. For records lacking batch numbers, the system automatically generates batch identifiers based on sample number, calcining furnace number, raw material formula, and holding time. Based on this, the system obtains color experiment batch data, including batch identifier, experiment record number, sample number, calcining furnace number, set temperature, heating rate, holding time, calcination atmosphere, raw material formula, operator record, experiment start time, and experiment end time.

[0058] S2. Perform event analysis and cooling path matching on the color experiment batch data to obtain the event data and cooling path feature data, respectively.

[0059] In one embodiment, the system reads batch data of color experiments and extracts the calcination atmosphere, experiment start time, experiment end time, and operator record set for the corresponding batch using the batch identifier as an index. The operator record set includes the operation record time and operation content. The system performs keyword parsing on the operation content, identifying descriptions of reduction measures such as "introducing reducing gas," "reducing oxygen supply," "closing the air intake," "adding reducing agent," and "adjusting the atmosphere." For example, the system reads the operation content from the operator record set and binds each operation content with its corresponding operation record time to obtain the operation record to be parsed. The system matches the regularized operation content with a preset reduction measure terminology to obtain candidate reduction measure terms. The preset reduction measure terminology includes terms such as reducing gas introduction, oxygen supply reduction, air intake closure, reducing agent addition, and atmosphere adjustment. The terms for introducing reducing gas include "introduce reducing gas", "introduce carbon monoxide", and "introduce hydrogen"; terms for reducing oxygen supply include "reduce oxygen supply", "reduce oxygen", and "close the oxygen valve"; terms for sealing the air intake include "close the air intake", "close the air inlet", and "close the furnace door"; terms for adding reducing agent include "add reducing agent" and "add carbon powder"; and terms for adjusting the atmosphere include "adjust the atmosphere", "switch the atmosphere", and "change the calcination atmosphere". The system performs a negative semantic judgment on the regularization operation content containing candidate reduction measure terms. If the candidate reduction measure term contains negative or terminating expressions such as "not", "no", "cancel", "stop", "prohibit", or "not executed", the candidate reduction measure term is removed; if there are no negative or terminating expressions in the regularization operation content, the candidate reduction measure term is determined as a valid reduction measure term. The system generates reduction measure types based on the term category to which the effective reduction measure term belongs. Specifically, effective reduction measure terms belonging to the reducing gas introduction term generate a reducing gas introduction type; effective reduction measure terms belonging to the oxygen supply reduction term generate an oxygen supply reduction type; effective reduction measure terms belonging to the intake closure term generate an intake closure type; effective reduction measure terms belonging to the reducing agent addition term generate a reducing agent addition type; and effective reduction measure terms belonging to the atmosphere adjustment term generate an atmosphere adjustment type. The system then associates the reduction measure type, the corresponding operation record time, and the corresponding operation content to obtain the reduction measure analysis results.

[0060] The system determines whether the operation record time falls between the experiment start time and the experiment end time, and filters out valid reduction measures. If the same reduction measure appears consecutively within the same batch, and the time interval between adjacent operation records does not exceed 10 minutes, they are merged into the same reduction event. If there is an atmosphere adjustment record after the reduction measure, the most recent atmosphere adjustment operation time is used as the event end time; otherwise, the experiment end time is used as the event end time. The atmosphere adjustment record is an operation record in the operator's record set that includes descriptions of atmosphere changes such as "adjust atmosphere," "switch atmosphere," "restore air atmosphere," "stop introducing reducing gas," "increase oxygen supply," and "open air intake." The atmosphere adjustment record includes the atmosphere adjustment operation time and atmosphere adjustment operation content. The system uses the calcination atmosphere of this batch as the associated calcination atmosphere. The system encapsulates the batch identifier, reduction measure type, event start time, event end time, associated calcination atmosphere, and corresponding operation content to obtain the reduction event data.

[0061] Cooling path matching involves the system reading batch data of color experiments and extracting the experiment end time, calcination atmosphere, and operator record set using the batch identifier as an index. The operator record set includes the operation record time and operation content. The system performs cooling keyword parsing on the operation content, identifying descriptions of cooling methods such as "furnace cooling," "air cooling," "wind cooling," "water cooling," "in-furnace cooling," and "heat preservation and slow cooling." This includes reading the operation content from the operator record set and binding each operation content with its corresponding operation record time to obtain the cooling operation record to be parsed. The system then performs text normalization on the operation content in the cooling operation record to be parsed, removing invalid spaces, standardizing full-width and half-width characters, and converting synonyms such as "in-furnace cooling," "in-furnace cooling," "natural cooling," "room temperature placement," "fan cooling," "air blowing cooling," "water quenching," "water cooling," and "heat preservation and cooling" into standard cooling expressions, resulting in normalized cooling operation content. The system matches the standardized cooling operation content with a preset cooling method terminology to obtain candidate cooling method terms. The preset cooling method terminology includes furnace cooling, air cooling, wind cooling, water cooling, furnace-in-flight cooling, and heat preservation and slow cooling terms. Furnace cooling terms include "furnace cooling" and "in-furnace cooling"; air cooling terms include "air cooling," "natural cooling," and "room temperature cooling"; wind cooling terms include "wind cooling," "fan cooling," and "blowing cooling"; water cooling terms include "water cooling," "water quenching," and "cooling in water"; furnace-in-flight cooling terms include "furnace-in-flight cooling" and "furnace-in-flight cooling"; and heat preservation and slow cooling terms include "heat preservation and slow cooling" and "heat preservation and cooling". The system performs a negative semantic judgment on the regularized cooling operation content containing candidate cooling method terms. If a candidate cooling method term contains negative or terminating expressions such as "not," "no," "cancel," "stop," "prohibit," or "not executed," it is removed. If no negative or terminating expressions are found in the regularized cooling operation content, the candidate cooling method term is determined as a valid cooling method term. The system generates cooling methods based on the term category of the valid cooling method term. Specifically, valid cooling method terms belonging to the furnace cooling term category generate furnace cooling methods, those belonging to the air cooling term category generate air cooling methods, those belonging to the air-cooling term category generate air-cooling methods, those belonging to the water-cooling term category generate water-cooling methods, those belonging to the furnace cooling term category generate furnace cooling methods, and those belonging to the heat preservation and slow cooling term category generate heat preservation and slow cooling methods. The system associates the cooling method, the corresponding operation record time, and the corresponding operation content to obtain the cooling method parsing result, which is then determined as the cooling method description.

[0062] The system matches the operation record time in the cooling method description with the experiment end time. If the operation record time is not earlier than the experiment end time, the corresponding cooling method description is identified as a valid cooling record. Based on a preset cooling path mapping table, the system converts the cooling method in the valid cooling record into a cooling path type and determines the operation record time in the valid cooling record as the cooling start time. The preset cooling path mapping table is constructed during system initialization based on a preset cooling method vocabulary, mapping terms such as furnace cooling, air cooling, wind cooling, water cooling, furnace cooling, and heat preservation slow cooling to standard cooling methods. The system then determines the cooling path type based on the cooling medium and cooling behavior attributes of each standard cooling method. For example, water cooling and wind cooling correspond to rapid cooling paths, air cooling corresponds to natural cooling paths, and furnace cooling, furnace cooling, and heat preservation slow cooling correspond to slow cooling paths. If no valid cooling record is found, the experiment end time is used as the cooling start time, and the cooling method is marked as unrecorded cooling. The system encapsulates the batch identifier, cooling method, cooling path type, cooling start time, calcination atmosphere, and corresponding operation content to obtain cooling path feature data.

[0063] S3. Extract colorimetric features from the calcination experimental data to obtain sample colorimetric data;

[0064] In one embodiment, the system reads sample image data from the calcination experiment data. The sample image data includes image acquisition time, sample number, and sample surface image. The sample image data in the calcination experiment data is acquired manually or through a preset image acquisition device. After reading the sample image data, the system matches the sample surface image with the corresponding experimental record based on the sample number. If the same sample number corresponds to multiple experimental records, the corresponding batch is determined based on the time relationship between the image acquisition time and the experiment start and end times, and the matched batch identifier is written into the sample image data. The system matches the sample surface image with the corresponding experimental record based on the sample number and performs background separation on the sample surface image to extract the sample area image. The system performs color space conversion on the sample area image to obtain the brightness value, red-green component value, and yellow-blue component value of the sample area; then, it removes outlier pixels within the sample area, eliminating reflective, shadow, and edge background pixels. The system calculates the sample's average brightness, average red-green component, average yellow-blue component, and color dispersion based on the remaining valid pixels to obtain the sample's chromaticity data. The sample colorimetric data includes batch identifier, sample number, image acquisition time, average sample brightness, average red-green component, average yellow-blue component, and color dispersion.

[0065] Color dispersion is calculated based on the dispersion of the colorimetric distribution of the remaining valid pixels in the sample region image. The system denotes the set of valid pixels after outlier removal as follows: ,in The number of valid pixels; each valid pixel The corresponding brightness values, red-green component values, and yellow-blue component values ​​are denoted as follows: , , The system first calculates the average brightness, average red-green component, and average yellow-blue component of the effective pixel set: Then, the color dispersion is calculated based on the degree of deviation of each effective pixel from the average chromaticity value: ,in This represents the color dispersion. The larger the value, the more obvious the color difference in the sample area, indicating the presence of local color spots, uneven calcination, or changes in surface brightness. The smaller the value, the more uniform the color distribution in the sample area.

[0066] In one embodiment, the system reads sample image data from the calcination experiment data. The sample image data includes image acquisition time, sample number, and sample surface image. The sample image data in the calcination experiment data is acquired manually or through a preset image acquisition device. After reading the sample image data, the system matches the sample surface image with the corresponding experimental record based on the sample number. If the same sample number corresponds to multiple experimental records, the corresponding batch is determined based on the time relationship between the image acquisition time and the experiment start and end times, and the matched batch identifier is written into the sample image data. The system reads the sample image data from the calcination experiment data and binds the batch identifier, sample number, image acquisition time, and sample surface image to obtain the sample image record to be processed. The system performs size unification, brightness normalization, and noise removal on the sample surface image in the sample image record to be processed, obtaining a standardized sample image. The system inputs standardized sample images into a pre-trained chromaticity feature extraction model. The chromaticity feature extraction model includes an image input layer, a sample region recognition layer, a color feature encoding layer, and a chromaticity output layer. The chromaticity feature extraction model is trained using historical sample surface images and their corresponding calibrated chromaticity values. The calibrated chromaticity values ​​include calibrated brightness values, calibrated red-green component values, and calibrated yellow-blue component values. The system comprises the following layers: an image input layer receives standardized sample images; a sample region recognition layer identifies the sample region from the standardized sample images, generates a sample region mask, and extracts the sample region image based on the mask; a color feature encoding layer performs convolutional feature extraction on the color distribution, local brightness variations, and color boundary variations in the sample region image to obtain a chromaticity feature vector; a chromaticity output layer outputs the sample brightness value, red-green component value, and yellow-blue component value based on the chromaticity feature vector. This includes receiving the chromaticity feature vector output by the color feature encoding layer and converting it into three continuous chromaticity output channels through a fully connected mapping. These three chromaticity output channels correspond to the sample brightness value, red-green component value, and yellow-blue component value, respectively. The system normalizes the output results of each channel against a preset chromaticity value range to obtain the actual chromaticity values. Finally, the brightness value, red-green component value, and yellow-blue component value corresponding to the image patch are summarized as the chromaticity feature output result for that sample. The system then calculates the color dispersion based on the sample brightness value, red-green component value, and yellow-blue component value corresponding to each pixel in the same sample area image, and encapsulates the batch identifier, sample number, image acquisition time, sample brightness value, red-green component value, yellow-blue component value, and color dispersion to obtain sample colorimetric data.

[0067] S4. Perform structured mapping based on the reduction event data, cooling path feature data, and sample chromaticity data to obtain color control mapping data.

[0068] In one embodiment, the system reads reduction event data, cooling path feature data, and sample colorimetric data, and performs correlation matching using the batch identifier as the primary index. The reduction event data includes the batch identifier, reduction measure type, event start time, event end time, associated calcination atmosphere, and corresponding operation content set; the cooling path feature data includes the batch identifier, cooling method, cooling path type, cooling start time, calcination atmosphere, and corresponding operation content; the sample colorimetric data includes the batch identifier, sample number, image acquisition time, sample brightness value (which may be the sample average brightness), red-green component value (which may be the average red-green component), yellow-blue component value (which may be the average yellow-blue component), and color dispersion.

[0069] The system first merges the reduction event data and cooling path feature data within the same batch based on the batch identifier to obtain batch process control data. Then, based on the batch identifier in the sample colorimetric data, it writes the colorimetric result of the corresponding sample into the batch process control data, obtaining a correlation record between reduction measures, cooling paths, and sample colorimetry. If multiple sample numbers exist within the same batch, the system generates a color control mapping record for each sample number. The system encapsulates the batch identifier, sample number, reduction measure type, event start time, event end time, cooling method, cooling path type, cooling start time, sample brightness value, red-green component value, yellow-blue component value, and color dispersion to obtain color control mapping data.

[0070] Optionally, batch processing specifically includes:

[0071] S11. Perform sample batch boundary identification on the calcination experiment data to obtain batch interval data;

[0072] In one embodiment, the system sorts the calcination experiment data according to the experiment time, and uses the calcination furnace number, raw material formula, set temperature, heating rate, holding time and calcination atmosphere as boundary judgment fields. When the time interval between the start of adjacent experimental records does not exceed 30 minutes, and the difference in set temperature does not exceed 5°C, the difference in heating rate does not exceed 1°C / min, and the difference in holding time does not exceed 10 minutes, they are classified into the same batch interval to obtain batch interval data. The batch interval data includes batch interval identifier, set of experimental record numbers, interval start time and interval end time.

[0073] S12. Bind sample identifiers based on batch interval data to obtain batch sample binding data;

[0074] In one embodiment, the system reads batch interval data and extracts the corresponding sample number from the calcination experiment data based on the set of experimental record numbers in the batch interval data. The batch interval identifier, the set of experimental record numbers, and the set of sample numbers are then bound together to obtain batch sample binding data.

[0075] S13. Segment the process window based on the batch sample binding data to obtain the color experiment batch data.

[0076] In one embodiment, the system reads the batch sample binding data and generates a batch identifier based on the batch interval identifier; then, the batch identifier, the set of experimental record numbers, the set of sample numbers, the calcining furnace number, the set temperature, the heating rate, the holding time, the calcining atmosphere, the raw material formula, the set of operator records, the start time of the experiment, and the end time of the experiment are packaged to obtain the color experiment batch data.

[0077] Optionally, the event analysis for restoration measures is as follows:

[0078] S21. Identify the restoration measures field in the color experiment batch data to obtain restoration measures data;

[0079] In one embodiment, the system reads the batch identifier, experiment start time, experiment end time, and operator record set from the color experiment batch data, and extracts the operation record time and operation content from the operator record set. When the system performs text normalization and keyword matching on the operation content, it first reads the operation content and corresponding operation record time from the operator record set, and binds the two as the operation record to be parsed. The system performs text normalization on the operation content in the operation record to be parsed, deleting invalid spaces, unifying full-width and half-width characters, and converting synonyms such as "pass CO", "pass hydrogen", "reduce oxygen supply", "close oxygen valve", "close furnace door", and "add carbon powder" into standard reduced expressions, resulting in normalized operation content. Among them, the standard reduced expressions include "pass CO", "pass carbon monoxide", and "charge CO gas" are uniformly normalized to "pass carbon monoxide", "pass hydrogen", "charge hydrogen", and "pass H2", "pass hydrogen", "reduce oxygen supply", "close oxygen valve", and "reduce oxygen flow rate", "close furnace door", "close air inlet", and "close air inlet", "close air inlet", and "add carbon powder", "add reducing agent", and "add reducing powder" are uniformly normalized to "add reducing agent". The system matches the standardized operation content with a preset reduction measure terminology. If the standardized operation content matches the terms "reducing gas introduction," "oxygen supply reduction," "inlet sealing," or "reducing agent addition," the corresponding operation record to be parsed is identified as an operation record containing a reduction measure description, and candidate reduction measure terms are generated. The system encapsulates the batch identifier, experiment start time, experiment end time, operation record time, operation content, and candidate reduction measure terms to obtain reduction measure data.

[0080] S22. Determine the restoration type of the restoration measures data to obtain the restoration type data;

[0081] In one embodiment, the system reads reduction measure data and determines the reduction type based on the category to which the candidate reduction measure terms belong. If the candidate reduction measure terms belong to "introducing reducing gas", "introducing carbon monoxide", or "introducing hydrogen", a reducing gas introduction type is generated; if they belong to "reducing oxygen supply", "closing the air intake", or "adding a reducing agent", an oxygen supply reduction type, an air intake closure type, and a reducing agent addition type are generated respectively, obtaining reduction type data, including batch identifier, experiment start time, experiment end time, operation record time, operation content, and reduction type.

[0082] S23. Locate the event boundary based on the restored type data to obtain the restored event interval data;

[0083] In one embodiment, the system reads the restoration type data and determines whether the corresponding operation record time is between the experiment start time and the experiment end time. If it is within this time range, the operation record time is determined as the event start time. If the same restoration type appears consecutively in the same batch, and the time interval between adjacent operation records does not exceed 10 minutes, they are merged into the same restoration event interval to obtain restoration event interval data, including batch identifier, restoration type, event start time, event end time, and set of operation content.

[0084] S24. Extract the conditions for the restored event interval data to obtain the restored event feature data;

[0085] In one embodiment, the system reads reduction event interval data and extracts the calcination atmosphere, set temperature, holding time, and operation content set of the corresponding batch from the color experiment batch data as the conditions for the reduction event interval. The system associates the batch identifier, reduction type, event start time, event end time, calcination atmosphere, set temperature, holding time, and operation content set to obtain reduction event characteristic data.

[0086] S25. Mark the validity of the event based on the restored event feature data to obtain the restored event data.

[0087] In one embodiment, the system reads the characteristic data of the restoration event and determines whether the restoration event has a valid duration based on the event start time and event end time. If the event start time is not later than the event end time, and there are no negative or termination descriptions such as "cancel," "not executed," or "stopped" in the corresponding operation content set, it is marked as a valid restoration event; otherwise, it is marked as an invalid restoration event. The system encapsulates the valid restoration events into restoration event data.

[0088] Optionally, the cooling path matching is specifically as follows:

[0089] S26. Perform cooling equipment status identification on the color test batch data to obtain cooling equipment status data;

[0090] In one embodiment, the system reads the batch identifier, experiment end time, and operator record set from the color experiment batch data, and extracts the operation record time and operation content. The system identifies cooling-related terms in the operation content. If it contains descriptions such as "turn on the fan," "put in water," "cool with the furnace," "heat preservation and cooling," or "natural placement," it generates the corresponding cooling equipment status and obtains cooling equipment status data.

[0091] S27. Locate the material cooling inlet based on the cooling equipment status data and obtain cooling inlet event data;

[0092] In one embodiment, the system reads the status data of the cooling equipment and determines whether the operation record time corresponding to the status of the cooling equipment is not earlier than the end time of the experiment. If the condition is met, the operation record time is determined as the cooling entry time, and the batch identifier, cooling entry time, cooling equipment status and operation content are associated to obtain cooling entry event data.

[0093] S28. Match the cooling path type based on the cooling inlet event data to obtain the cooling path type data;

[0094] In one embodiment, the system reads cooling inlet event data and matches the cooling path type according to the cooling equipment status. If the cooling equipment status is fan on or water cooling is in operation, it is matched as a fast cooling path; if it is placed naturally, it is matched as a natural cooling path; if it is furnace cooling or heat preservation cooling, it is matched as a slow cooling path, thus obtaining cooling path type data.

[0095] S29. Extract cooling process features from the cooling path type data to obtain cooling process feature data;

[0096] In one embodiment, the system reads the batch identifier, cooling inlet time, cooling equipment status, cooling path type, and operation content from the cooling path type data, and determines the cooling inlet time as the cooling start time. The system determines the cooling medium based on the cooling equipment status: air for fan-on status, water for water-cooled status, air for naturally placed status, residual heat environment for furnace-in-furnace cooling status, and insulation environment for heat preservation and cooling status. The system then generates the heat exchange method and cooling intensity level based on the cooling path type: rapid cooling path corresponds to forced heat exchange and high cooling intensity; natural cooling path corresponds to natural heat dissipation and medium cooling intensity; and slow cooling path corresponds to residual heat delayed heat dissipation or insulation delayed heat dissipation and low cooling intensity. The system encapsulates the batch identifier, cooling start time, cooling equipment status, cooling medium, heat exchange method, cooling intensity level, cooling path type, and operation content to obtain cooling process characteristic data.

[0097] S210. Validate the path validity based on the cooling process characteristic data to obtain the cooling path characteristic data.

[0098] In one embodiment, the system reads cooling process feature data and verifies the path validity based on the cooling start time and the experiment end time corresponding to the cooling inlet event data. If the cooling start time is not earlier than the experiment end time and the cooling path type is not empty, it is marked as a valid cooling path; otherwise, it is marked as an unconfirmed cooling path; thus obtaining the cooling path feature data.

[0099] Optionally, the cooling path type matching is specifically as follows:

[0100] S281. Parse the cooling inlet event data to obtain the inlet location identifier data;

[0101] In one embodiment, the system reads the batch identifier, cooling inlet time, cooling equipment status, and operation content from the cooling inlet event data, and identifies location terms such as "furnace outlet", "air-cooled section inlet", "water-cooled tank inlet", and "furnace insulation zone" in the operation content to obtain inlet location identifier data, including batch identifier, cooling inlet time, cooling equipment status, operation content, and inlet location identifier.

[0102] S282. Match the material flow direction valve status according to the inlet location identification data to obtain material flow direction matching data;

[0103] In one embodiment, the system reads the inlet location identifier data and matches it with the preset flow direction valve correspondence. If the inlet location is the air-cooled section inlet, the air-cooled valve opening flow direction is matched; if it is the water-cooled tank inlet, the water-cooled flow direction is matched. The sample flow direction state is jointly determined by the inlet location identifier and the cooling equipment status: when the inlet location identifier is the air-cooled section inlet and the cooling equipment status is fan on, an air-cooled flow direction state is generated; when the inlet location identifier is the water-cooled tank inlet and the cooling equipment status is water-cooled, a water-cooled flow direction state is generated; when the inlet location identifier is the natural placement area, an air-cooled flow direction state is generated; when the inlet location identifier is the furnace insulation area or the operation content includes "cooling with the furnace", a furnace stagnation state is generated; when the inlet location identifier is the insulation area or the operation content includes "insulation and cooling", an insulation and slow cooling flow direction state is generated. The sample flow direction matching data is obtained, including batch identifier, cooling inlet time, cooling equipment status, inlet location identifier, sample flow direction state, and operation content.

[0104] S283. Perform cooling equipment operation verification based on material flow direction matching data to obtain equipment verification data;

[0105] In one embodiment, the system reads sample flow direction matching data and cooling equipment status to determine whether the corresponding cooling equipment is consistent with the sample flow direction. For example, if the sample flow direction is air-cooled and the cooling equipment status is fan on, the verification passes and a corresponding verification pass mark is generated; otherwise, an abnormal verification mark is generated; the equipment verification data is obtained, including batch identifier, cooling inlet time, inlet position identifier, sample flow direction status, cooling equipment status, equipment verification result, and operation content.

[0106] S284. Determine the path type based on the device verification data to obtain single path type data;

[0107] In one embodiment, the system reads equipment calibration data and determines the path type based on the sample flow direction and cooling equipment status after successful calibration. Air-cooled or water-cooled flow direction corresponds to a rapid cooling path, natural placement / air-cooled flow direction corresponds to a natural cooling path, and furnace insulation zone / furnace stagnation state corresponds to a slow cooling path. This yields single path type data, including batch identifier, cooling inlet time, cooling equipment status, sample flow direction, equipment calibration result, and single path type.

[0108] S285. Perform continuous path tracing based on single path type data to obtain cooling path type data.

[0109] In one embodiment, the system reads single path type data and arranges it chronologically according to the cooling inlet time of the same batch. If there are multiple consecutive cooling inlet events in the same batch, the corresponding path types are concatenated in chronological order to obtain cooling path type data, including batch identifier, cooling path type, path type sequence, cooling inlet time sequence, and cooling device status sequence.

[0110] Optionally, the feature extraction of the cooling process specifically includes:

[0111] S291. Construct a cooling temperature sequence based on the cooling path type data to obtain cooling curve data;

[0112] In one embodiment, the system reads the batch identifier, cooling start time, cooling path type, and temperature acquisition record from the cooling path type data. The temperature acquisition record is acquired by a preset temperature acquisition device and includes sampling time, sample temperature, and ambient temperature. The system filters the temperature acquisition records whose sampling time is no earlier than the cooling start time and arranges them in ascending order of sampling time. The sampling time, sample temperature, and ambient temperature are combined to form a cooling temperature sequence, thus obtaining cooling curve data.

[0113] S292. Based on the cooling curve data, the derivative curve is extracted and the standard Newtonian cooling dynamics decay is performed to obtain the derivative curve data and the reference cooling model data, respectively.

[0114] In one embodiment, the system reads cooling curve data and calculates the rate of temperature change based on the sample temperature difference and time difference corresponding to adjacent sampling times, obtaining derivative curve data. The system uses the sample temperature corresponding to the cooling start time as the initial sample temperature, the ambient temperature as the cooling ambient temperature, and extracts the path attenuation coefficient from a preset cooling parameter table according to the cooling path type to construct baseline cooling model data. This is represented as: ,in The reference cooling temperature, The initial sample temperature, For ambient temperature, This is the path attenuation coefficient, for furnace cooling or heat preservation slow cooling paths. Can be set to Natural air cooling path Can be set to air-cooled path Can be set to Water cooling path Can be set to If the sample volume is large, the amount of material is large, or the heat preservation conditions are strong, then the smaller value within the corresponding range should be used; if the sample is thin, the mass is small, or the heat exchange conditions are strong, then the larger value within the corresponding range should be used. This is the cooling start time.

[0115] S293. Identify abnormal heat release feature points in the derivative curve data to obtain heat release feature point data;

[0116] In one embodiment, the system reads derivative curve data, and if the temperature change rate of the previous sampling interval in the continuous cooling phase is... Less than The temperature change rate in the current sampling interval Greater than and less than or equal to Or, the current sampling temperature Higher than the previous sampling temperature And the temperature rise If the temperature rise is not less than 1°C and the duration of the temperature rise phenomenon does not exceed two consecutive sampling intervals, then the sampling time corresponding to the sampling interval is determined as the abnormal heat release time. The abnormal heat release time, the corresponding sample temperature, and the temperature change rate are encapsulated to obtain heat release characteristic point data.

[0117] S294. Perform residual integral calculation on the cooling curve data and the benchmark cooling model data to obtain residual integral data;

[0118] In one embodiment, the system reads cooling curve data and reference cooling model data, and calculates the temperature residual between the measured sample temperature and the reference cooling temperature under the same sampling time; then, the temperature residuals are accumulated according to the sampling time sequence to obtain the residual integral data. This is represented as: ,in The value of the residual integral. Let be the measured sample temperature at the i-th sampling time. The reference cooling temperature corresponding to the sampling time. The time interval between adjacent sampling.

[0119] S295. Integrate the heat release characteristic point data and residual integral data to obtain the cooling process characteristic data.

[0120] In one embodiment, the system associates the abnormal heat release time, corresponding sample temperature, and temperature change rate in the heat release feature point data with the residual integral value in the residual integral data, and encapsulates them with batch identifier and cooling path type to obtain cooling process feature data. The cooling process feature data includes batch identifier, cooling path type, abnormal heat release time, temperature change rate, residual integral value, and cooling start time.

[0121] Optionally, the identification of abnormal heat release feature points specifically includes:

[0122] a. Select inflection points from the derivative curve data to obtain inflection point data;

[0123] In one embodiment, the system reads the batch identifier, sampling time, sample temperature, and temperature change rate from the derivative curve data. When the difference in temperature change rate between adjacent sampling points exceeds a preset change rate threshold (set according to the cooling path type, where rapid cooling path is 1℃ / min to 3℃ / min, natural cooling path is 0.3℃ / min to 1℃ / min, and slow cooling path is 0.05℃ / min to 0.3℃ / min), and the direction of change changes from continuous cooling (i.e., the temperature change rate of at least 3 consecutive sampling intervals before the current sampling point is less than 0) to cooling slowdown (i.e., the current temperature change rate is still less than or equal to 0, but its absolute value is more than 50% lower than the previous sampling interval, or the current temperature change rate is within the range of 0℃ / min to 0℃ / min), the system detects the temperature change rate when the difference in temperature change rate between adjacent sampling points exceeds a preset change rate threshold (set according to the cooling path type, where rapid cooling path is 1℃ / min to 3℃ / min, natural cooling path is 0.3℃ / min to 1℃ / min, and slow cooling path is 0.05℃ / min to 0.3℃ / min), and the direction of change changes from continuous cooling (i.e., the temperature change rate of the current sampling point is less than or equal to 0, but its absolute value is more than 50% lower than the previous sampling interval, or the current temperature change rate is within the range of 0℃ / min to 0). to ) or short-term heating (i.e., the current sample temperature increases by at least 1°C compared to the previous sampling time, or the current temperature change rate is greater than 1°C). When the duration does not exceed two consecutive sampling intervals, the corresponding sampling time is determined as the inflection point time, and the inflection point data is obtained.

[0124] b. Match the material heat release temperature range with the cooling path type data to obtain the pyrolysis range data;

[0125] In one embodiment, the system reads the batch identifier and cooling path type from the cooling path type data, and backfills the raw material formula from the color experiment batch data according to the batch identifier. The system queries the preset material heat release temperature range table according to the raw material formula to obtain the heat release range data. The preset material heat release temperature range table refers to the temperature range mapping table pre-established by the system for different raw material formulas during the initialization phase, used to record the temperature range in which abnormal heat release, phase change heat release, or residual reaction heat release may occur during the cooling process of the sample. The preset material heat release temperature range table includes the raw material formula identifier, material composition category, cooling path type, heat release start temperature, heat release end temperature, and range source identifier. Specifically, during construction, the system reads historical calcination experiment data and historical cooling curve data. The historical calcination experiment data includes the raw material formula, batch identifier, and sample number, and the historical cooling curve data includes the sampling time, sample temperature, and temperature change rate. The system extracts the sample temperature corresponding to the abnormal heat release time marked in the historical cooling curve to obtain the historical heat release temperature set. For a set of historical heat release temperatures under the same raw material formulation and the same cooling path type, the system takes the lowest effective temperature as the heat release start temperature and the highest effective temperature as the heat release end temperature, thus generating the corresponding material heat release temperature range. For example, if the historical abnormal heat release temperatures of a certain raw material formulation under the air-cooled path are concentrated between 420℃ and 510℃, then the material heat release temperature range for the corresponding air-cooled path can be defined as 420℃ to 510℃. If the heat release temperatures of the same raw material formulation under the furnace-cooled path are concentrated between 380℃ and 470℃, then a separate heat release temperature range of 380℃ to 470℃ is generated for the corresponding furnace-cooled path.

[0126] c. Use pyrolysis interval data to perform interval filtering on inflection point data to obtain effective inflection point data;

[0127] In one embodiment, the system reads inflection point data and heat release interval data, and determines whether the sample temperature in the inflection point data falls within the temperature range corresponding to the heat release interval data; if it does, the inflection point is retained, and valid inflection point data is obtained.

[0128] d. Extract the perturbation waveform from the effective inflection point data to obtain the perturbation waveform data;

[0129] In one embodiment, the system extracts the sample temperature and temperature change rate within a preset time length before and after the inflection point from the cooling curve data based on the batch identifier and inflection point time in the effective inflection point data, and obtains the disturbance waveform data.

[0130] e. Perform energy integration and frequency domain decomposition on the disturbance waveform data to obtain thermal release disturbance characteristic data and disturbance frequency data, respectively;

[0131] In one embodiment, the system reads disturbance waveform data, integrates the temperature deviation in the disturbance waveform over the sampling time to obtain heat release disturbance characteristic data, and performs frequency domain decomposition on the disturbance waveform to extract the main disturbance frequency. When performing frequency domain decomposition on the disturbance waveform data, the system first reads the sampling time sequence and temperature disturbance sequence from the disturbance waveform data. The temperature disturbance sequence is the difference between the sample temperature near the inflection point and the reference cooling temperature for the corresponding time period. If the sampling time intervals are inconsistent, the system first resamples the temperature disturbance sequence according to a preset uniform sampling interval to obtain an equally spaced disturbance sequence. Subsequently, the system performs mean removal processing on the equally spaced disturbance sequence to remove the overall temperature shift trend during the cooling process, obtaining the disturbance sequence to be decomposed. The system performs a Fast Fourier Transform on the disturbance sequence to be decomposed to obtain the disturbance amplitude spectrum at different frequencies. This can be expressed as: ,in Let j be the perturbation value to be decomposed at the j-th sampling point. For frequency The corresponding frequency domain amplitude results, For Fast Fourier Transform (FFT), the system searches for the frequency point with the largest amplitude within a preset effective frequency range and determines this frequency point as the main disturbance frequency. The system encapsulates the main disturbance frequency, its corresponding frequency domain amplitude, and the frequency search range to obtain the disturbance frequency data.

[0132] f. Based on the heat release disturbance characteristic data and disturbance frequency data, characteristic points are determined to obtain heat release characteristic point data.

[0133] In one embodiment, the system reads heat release disturbance characteristic data and disturbance frequency data. If the temperature deviation integral value exceeds a preset disturbance intensity threshold, and the main disturbance frequency is within a preset frequency range for the corresponding cooling path type, then the corresponding effective inflection point is determined as a heat release characteristic point, and heat release characteristic point data is obtained. The preset disturbance intensity threshold for the rapid cooling path is... to The preset frequency range is 0.02Hz to 0.20Hz, and the preset disturbance intensity threshold for the natural cooling path is [missing value]. to The preset frequency range is 0.005Hz to 0.08Hz, and the preset disturbance intensity threshold for the slow cooling path is [missing value]. to The preset frequency range is 0.001Hz to 0.03Hz.

[0134] Optionally, the chromaticity feature extraction specifically involves:

[0135] S31. Collect and standardize the sample image data in the calcination experiment data to obtain standard sample image data;

[0136] In one embodiment, the system reads the sample number, image acquisition time, and sample surface image from the sample image data, and matches the batch identifier based on the sample number and image acquisition time. The system performs size standardization, exposure correction, and noise removal on the sample surface image to obtain standard sample image data, which includes the batch identifier, sample number, image acquisition time, and standard sample image.

[0137] S32. Perform sample region segmentation on the standard sample image data to obtain the effective sample region data;

[0138] In one embodiment, the system reads standard sample image data and performs grayscale processing on the standard sample image to obtain a grayscale sample image. The system counts the number of pixels corresponding to each grayscale value in the grayscale sample image, generating a grayscale histogram; then, based on the grayscale histogram, the pixels are divided into low-grayscale candidate regions and high-grayscale candidate regions, and the inter-class difference between the two types of pixels is calculated under different candidate thresholds. The system selects the grayscale value that maximizes the difference between the low-grayscale candidate region and the high-grayscale candidate region as the foreground segmentation threshold. The candidate threshold is denoted as... Gray values ​​less than or equal to The pixels with gray values ​​greater than 0 are divided into the first candidate regions. The pixels are divided into second candidate regions; the system calculates the pixel proportion and average gray value of the two candidate regions respectively, and calculates the inter-class variance: ,in and These represent the pixel percentages of the first and second candidate regions, respectively. and These are the average grayscale values ​​of the first and second candidate regions, respectively. The system iterates through the candidate thresholds within the grayscale value range and then... The largest candidate threshold is determined as the foreground segmentation threshold. If the sample region appears as a brighter area in the image, pixels with gray values ​​greater than the foreground segmentation threshold are identified as candidate sample pixels; if the sample region appears as a darker area, pixels with gray values ​​less than or equal to the foreground segmentation threshold are identified as candidate sample pixels, thus obtaining candidate foreground regions. The system performs connected component analysis on the candidate foreground regions, extracts the largest connected region as the candidate sample body region, and removes scattered regions with areas less than 0.5% to 3% of the total number of pixels in the standard sample image. The system performs edge smoothing and hole filling on the candidate sample body region, removes broken areas, and generates a sample region mask. The system extracts corresponding pixels from the standard sample image based on the sample region mask to obtain the sample body region. The system removes pixels within the sample body region that are higher than the average brightness value plus 2.5 standard deviations and pixels that are lower than the average brightness value minus 2.5 standard deviations, treating them as reflective pixels and shadow pixels, respectively; at the same time, it removes residual background pixels outside the boundary of the sample region mask, retaining the remaining pixels as the effective region pixel set. The system encapsulates the batch identifier, sample number, image acquisition time, sample area mask, and effective area pixel set to obtain sample effective area data.

[0139] In one embodiment, the system reads standard sample image data, which includes batch identifier, sample number, image acquisition time, and standard sample image. The system inputs the standard sample image into a pre-trained sample region segmentation model, which includes an image input layer, an encoded feature extraction layer, a decoding segmentation layer, and a mask output layer. The image input layer receives the standard sample image and adjusts it to a preset input size, such as 512×512 pixels, while normalizing the pixel values ​​to the range of 0 to 1, thus obtaining the model input image. The encoded feature extraction layer performs multi-scale feature extraction on the model input image. The system extracts image features sequentially through a first convolutional unit, a second convolutional unit, and a third convolutional unit, where each convolutional unit includes convolution operations, normalization processing, and nonlinear activation processing. The first convolutional unit extracts sample edge and color transition features, the second convolutional unit extracts sample texture and brightness distribution features, and the third convolutional unit extracts sample main body shape and background difference features. The encoding feature extraction layer outputs a multi-scale encoded feature map. The decoding segmentation layer is used to recover the spatial position of the sample region based on the multi-scale encoded feature map. The system upsamples the deep encoded feature map output by the third convolutional unit and concatenates it with the shallow encoded feature maps output by the second and first convolutional units, combining the sample main body contour information with edge detail information. A sample region probability map is generated through decoding convolution. Each sample region probability map... Each pixel location corresponds to a sample region probability value, representing the likelihood that the pixel belongs to a sample region. The mask output layer converts the sample region probability map into a sample region mask. The system compares the sample region probability values ​​in the probability map with a preset mask threshold. When the sample region probability value at a pixel location is not lower than the preset mask threshold (e.g., 0.5), the pixel is marked as a sample pixel. When the sample region probability value is lower than the preset mask threshold, the pixel is marked as a background pixel, thus obtaining the initial sample region mask. The system performs connected component filtering, hole filling, and boundary smoothing on the initial sample region mask to obtain the final sample region mask. Based on the sample region mask, the system extracts the main sample pixels from the standard sample image and smooths the edges of the sample region mask to obtain the effective region pixel set. The system encapsulates the batch identifier, sample number, image acquisition time, sample region mask, and effective region pixel set to obtain the sample effective region data. The sample region segmentation model is trained using historical standard sample images and their corresponding manually labeled sample region masks. The historical standard sample images serve as the model input, and the manually labeled sample region masks serve as the model output labels, enabling the model to learn the boundary differences between the sample region and the background region.

[0140] S33. Perform color space conversion based on the effective area data of the sample to obtain color space feature data;

[0141] In one embodiment, the system reads the set of effective region pixels from the sample's effective region data and converts the effective region pixels from the RGB color space to the Lab color space to obtain the luminance value L, red-green component value a, and yellow-blue component value b for each effective pixel. The system then summarizes the L, a, and b components of the effective region pixels to obtain color space feature data.

[0142] The color offset is calculated based on the color space feature data to obtain the chromaticity offset data;

[0143] In one embodiment, the system reads color space feature data and backfills the raw material formula from the color experiment batch data according to the batch identifier. The system calculates the average brightness value, average red-green component value, and average yellow-blue component value of the sample based on the brightness value, red-green component value, and yellow-blue component value in the effective area pixel set, respectively; then, according to the raw material formula, it reads the standard brightness value, standard red-green component value, and standard yellow-blue component value from the preset standard colorimetric table, calculates the colorimetric offset, and obtains the colorimetric offset data. ,in This is the chromaticity offset. These represent the average brightness value, average red-green component value, and average yellow-blue component value of the sample, respectively. These are the standard luminance values, standard red-green component values, and standard yellow-blue component values ​​in the preset standard chromaticity table. The preset standard chromaticity table is constructed based on the calibrated chromaticity values ​​of historical qualified sample images and includes raw material formulas, standard luminance values, standard red-green component values, and standard yellow-blue component values.

[0144] S34. Determine the colorimetric level based on the colorimetric offset data to obtain the sample colorimetric data.

[0145] In one embodiment, the system reads chromaticity offset data and compares the chromaticity offset with a preset chromaticity level threshold to generate a chromaticity level. The system encapsulates the batch identifier, sample number, image acquisition time, average sample brightness value, average red-green component value, average yellow-blue component value, chromaticity offset, and chromaticity level to obtain sample chromaticity data.

[0146] Optionally, the structured mapping specifically refers to:

[0147] S41. Index and match the reduction event data, cooling path feature data and sample colorimetric data to obtain batch association data;

[0148] In one embodiment, the system reads reduction event data, cooling path feature data, and sample colorimetric data, and performs matching using the batch identifier as the primary index. If the sample colorimetric data does not contain a batch identifier, the system retrieves the correspondence between the batch identifier and the sample number from the color experiment batch data and backfills the batch identifier into the sample colorimetric data. The system associates the reduction events, cooling paths, and sample colorimetric results within the same batch to obtain batch association data.

[0149] S42. Construct the combined feature of restoration and cooling based on batch association data to obtain combined feature data;

[0150] In one embodiment, the system reads batch-related data, extracts the type of restoration measure, event start time, event end time, cooling path type, cooling start time, and path validity marker from the same batch, and combines the above fields into a restoration-cooling combined feature to obtain combined feature data.

[0151] S43. Perform chromaticity response mapping based on the combined feature data and sample chromaticity data to obtain chromaticity response relationship data;

[0152] In one embodiment, the system reads the combined feature data and correlates the reduction-cooling combined features corresponding to the same batch and the same sample number with the sample brightness value, red-green component value, yellow-blue component value, chromaticity offset and chromaticity level to obtain chromaticity response relationship data.

[0153] S44. Encapsulate the data based on the chromaticity response relationship data to obtain color control mapping data.

[0154] In one embodiment, the system reads the chromaticity response relationship data and encapsulates the batch identifier, sample number, type of restoration measure, event start time, event end time, cooling path type, cooling start time, path validity marker, sample brightness value, red-green component value, yellow-blue component value, chromaticity offset, and chromaticity level in a structured manner to obtain color control mapping data.

[0155] Optionally, this application also provides an AI-based experimental data structuring management system for executing the above-described AI-based experimental data structuring management method. The AI-based experimental data structuring management system includes:

[0156] The batch processing module is used to acquire calcination experiment data and perform batch processing based on the calcination experiment data to obtain color experiment batch data.

[0157] The color experiment batch feature extraction module is used to perform restoration measure event analysis and cooling path matching on the color experiment batch data, and obtain restoration event data and cooling path feature data respectively.

[0158] The colorimetric feature extraction module is used to extract colorimetric features from calcination experimental data to obtain sample colorimetric data.

[0159] The structured mapping module is used to perform structured mapping based on the restoration event data, cooling path feature data, and sample chromaticity data to obtain color control mapping data.

[0160] Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended application documents rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of the equivalents of the application documents be incorporated into the invention.

[0161] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. An AI-based method for structured management of experimental data, characterized in that, Includes the following steps: Acquire calcination experiment data and perform batch processing based on the calcination experiment data to obtain color experiment batch data; The color experiment batch data was analyzed for restoration measures and matched for cooling paths to obtain restoration event data and cooling path feature data, respectively. Colorimetric features were extracted from the calcination experimental data to obtain the sample colorimetric data; Color control mapping data is obtained by performing structured mapping based on the restoration event data, cooling path feature data, and sample chromaticity data.

2. The AI-based experimental data structured management method according to claim 1, characterized in that, Batch processing specifically refers to: Sample batch boundary identification is performed on the calcination experiment data to obtain batch interval data; sample identification is bound based on the batch interval data to obtain batch sample binding data; process window is segmented based on the batch sample binding data to obtain color experiment batch data.

3. The AI-based experimental data structured management method according to claim 1, characterized in that, The specific analysis of the restoration measures event is as follows: The restoration measure field is identified in the color experiment batch data to obtain restoration measure data; the restoration measure data is then classified by restoration type to obtain restoration type data; and event boundaries are located based on the restoration type data to obtain restoration event interval data. The conditions for the restored event intervals are extracted to obtain the restored event feature data; The event validity is marked based on the restored event feature data to obtain restored event data.

4. The AI-based experimental data structured management method according to claim 1, characterized in that, The specific cooling path matching is as follows: The cooling equipment status is identified from the color test batch data to obtain cooling equipment status data; the material cooling inlet is located based on the cooling equipment status data to obtain cooling inlet event data; the cooling path type is matched based on the cooling inlet event data to obtain cooling path type data; and cooling process features are extracted from the cooling path type data to obtain cooling process feature data. The path validity is verified based on the cooling process characteristic data to obtain the cooling path characteristic data.

5. The AI-based experimental data structured management method according to claim 4, characterized in that, Cooling path type matching specifically refers to: The inlet location of the cooling inlet event data is parsed to obtain inlet location identifier data; the material flow direction valve status is matched based on the inlet location identifier data to obtain material flow direction matching data; The cooling equipment operation is verified based on the material flow direction matching data to obtain equipment verification data; the path type is determined based on the equipment verification data to obtain single path type data. Continuous path tracing is performed based on single path type data to obtain cooling path type data.

6. The AI-based experimental data structured management method according to claim 4, characterized in that, The specific features of the cooling process are as follows: Cooling temperature sequences are constructed based on cooling path type data to obtain cooling curve data; derivative curves are extracted and standard Newtonian cooling kinetic decay is performed based on the cooling curve data to obtain derivative curve data and benchmark cooling model data, respectively; abnormal heat release feature points are identified on the derivative curve data to obtain heat release feature point data. Residual integral data is obtained by performing residual integral calculation on the cooling curve data and the benchmark cooling model data; By integrating the heat release characteristic point data and the residual integral data, the cooling process characteristic data are obtained.

7. The AI-based experimental data structured management method according to claim 6, characterized in that, The identification of abnormal heat release feature points is specifically as follows: Inflection points are selected from the derivative curve data to obtain inflection point data; the material heat release temperature range is matched according to the cooling path type data to obtain pyrolysis range data; the inflection point data is filtered using the pyrolysis range data to obtain effective inflection point data; the effective inflection point data is truncated with a disturbance waveform to obtain disturbance waveform data. Energy integration and frequency domain decomposition are performed on the disturbance waveform data to obtain thermal release disturbance characteristic data and disturbance frequency data, respectively; feature points are determined based on the thermal release disturbance characteristic data and disturbance frequency data to obtain thermal release feature point data.

8. The AI-based experimental data structured management method according to claim 1, characterized in that, The specific process of chromaticity feature extraction is as follows: The sample image data in the calcination experiment data is collected and standardized to obtain standard sample image data; the standard sample image data is segmented to obtain effective sample area data; and color space conversion is performed based on the effective sample area data to obtain color space feature data. The color offset is calculated based on the color space characteristic data to obtain the chromaticity offset data; the chromaticity level is determined based on the chromaticity offset data to obtain the sample chromaticity data.

9. The AI-based experimental data structured management method according to claim 1, characterized in that, Structured mapping specifically refers to: Index matching is performed on the reduction event data, cooling path feature data, and sample chromaticity data to obtain batch association data; reduction-cooling combined features are constructed based on the batch association data to obtain combined feature data; and chromaticity response mapping is performed based on the combined feature data and sample chromaticity data to obtain chromaticity response relationship data. Data is encapsulated based on chromaticity response relationship data to obtain color control mapping data.

10. An AI-based experimental data structured management system, characterized in that, For executing the AI-based experimental data structured management method as described in claim 1, the AI-based experimental data structured management system includes: The batch processing module is used to acquire calcination experiment data and perform batch processing based on the calcination experiment data to obtain color experiment batch data. The color experiment batch feature extraction module is used to analyze the restoration measures event and match the cooling path of the color experiment batch data to obtain restoration event data and cooling path feature data, respectively. The colorimetric feature extraction module is used to extract colorimetric features from calcination experimental data to obtain sample colorimetric data. The structured mapping module is used to perform structured mapping based on the restoration event data, cooling path feature data, and sample chromaticity data to obtain color control mapping data.

Images