A copper concentrate multi-source fusion grade and moisture content rapid estimation system

Abstract

The present application relates to the technical field of data processing and analysis, and further relates to a copper concentrate multi-source fusion grade and moisture content rapid estimation system, which comprises: a collection and fusion unit, which is used for collecting element response sequences, spectral reflection sequences, mass flow sequences and location sequences, forming aligned time indexes on a material trajectory grid, and obtaining aligned multi-source observation sequences; a decision estimation unit, which is used for running a ternary deterministic intersection decision algorithm for each material segment unit in the aligned multi-source observation sequences and the material trajectory grid; and a result publishing unit, which is used for receiving batch records of a testing system, positioning a batch window according to the aligned time indexes, generating parameter snapshots and traceability logs, and publishing process control results and trade settlement results.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of data processing and analysis, and particularly relates to a copper concentrate multi-source fusion grade and moisture content rapid estimation system. BACKGROUND

[0002] As an important non-ferrous metal raw material, the accurate determination of the grade and moisture content of copper concentrate is not only related to the process control of the mining enterprise, but also directly affects the trade settlement between the port and the smelting enterprise. The technical routes widely used in the industry at present mainly fall into two categories: laboratory chemical analysis and online detection means. Laboratory chemical analysis can obtain relatively accurate grade and moisture content results through sampling, drying, chemical decomposition and spectral analysis steps, but has problems such as insufficient sampling representativeness, long detection period, and inability to reflect material flow changes in real time. When bulk cargo is transported by a belt to the port or smelting plant, this delay can lead to a lag in settlement basis, and enterprises face risks in inventory management and financial settlement.

[0003] To make up for the shortcomings of laboratory analysis, the industry has gradually introduced online detection technology, including X-ray fluorescence detection, near-infrared spectroscopy detection and belt scale monitoring. X-ray fluorescence detection can obtain the response characteristics of the surface elements of the material in a relatively short time, and is suitable for rapid estimation of grade. However, due to the influence of ore surface roughness, particle size distribution and moisture content variation, single X-ray fluorescence detection often has problems such as large signal fluctuation and unstable results. Near-infrared spectroscopy detection is sensitive to moisture content and part of the mineral composition, and can achieve rapid prediction through the absorption characteristics of specific bands. However, near-infrared spectroscopy is sensitive to the accumulation state of the powder and the lighting conditions, and is easily disturbed by uneven accumulation thickness or dust covering on the conveyor belt, making it difficult to stably identify the absorption band. As a traditional method, belt scale monitoring can accurately obtain mass flow information and provide reliable basis for process measurement, but it cannot provide grade and moisture content information and can only provide reference at the material flow level. SUMMARY

[0004] The main purpose of the present application is to provide a copper concentrate multi-source fusion grade and moisture content rapid estimation system, which can effectively solve the problems of single detection method being easily disturbed by outliers, different observation sources lacking unified alignment mechanism and lacking closed-loop calibration, significantly improve the stability and reliability of the estimation results, and meet the real-time requirements of process control and provide representative and traceable basis for port bulk cargo trade settlement.

[0005] To solve the above problems, the technical scheme of the present application is as follows:

[0006] A copper concentrate multi-source fusion grade and moisture content rapid estimation system, the system comprising:

[0007] a sequence collection and fusion unit for collecting the element response sequence, the spectral reflectance sequence, the mass flow sequence and the mileage sequence, performing delay compensation and mixture alignment, forming an aligned time index on the material track grid, and obtaining an aligned multi-source observation sequence;

[0008] a decision estimation unit for running a ternary deterministic intersection decision algorithm on each material segment unit in the aligned multi-source observation sequence and the material track grid; establishing a candidate interval library and a bias table, the candidate interval library including an element peak mapping table, an absorption band mapping table and a feasibility table; generating an element response candidate interval set after peak position correction, background deduction and associated peak consistency test according to the spectral peak record; generating a spectral reflectance candidate interval set after scattering correction, baseline flattening and absorption band positioning according to the spectral record; generating a mass uniformization candidate interval set by calling the feasibility table after state inspection according to the delivery record; sequentially performing intersection of the three, intersection of two according to the preset priority and boundary shrinkage to obtain an estimated interval set that meets the preset minimum interval width; registering the center and endpoints of the estimated interval set as grade estimate, moisture content estimate and estimated range, and writing them into the estimation result set together with the equipment state code; saving the candidate interval library and the bias table;

[0009] a result publishing unit for receiving batch records of the assay system and positioning a batch window according to the aligned time index; calculating grade average estimate and moisture content average estimate from the estimation result set within the batch window, comparing the two with batch assay values to obtain difference markers, updating the bias table using the difference markers and shifting the entire candidate interval library by one step in the positive or negative direction of the difference, generating a parameter snapshot and a traceability log, and publishing process control results and trade settlement results.

[0010] Further, the element response sequence is a set of spectral peak records arranged in chronological order, each spectral peak record containing a peak position index list, a peak area list, a peak width list, a background estimate index and an equipment state code, generated by an online X-ray fluorescence device at a fixed sampling period and time stamped; the spectral reflectance sequence is a set of spectral records arranged in chronological order, each spectral record containing a waveband index list, a reflectance list, an absorption band marker list, a noise estimate index and a light source state code, generated by an online near-infrared device at a fixed sampling period and time stamped; the mass flow sequence is a set of delivery records arranged in chronological order, each delivery record containing instantaneous mass increment, conveying speed, stable state code and cumulative mileage, generated by a belt scale and speedometer combination and time stamped.

[0011] Further, the mileage sequence is a set of displacement records arranged in time sequence, each displacement record contains an encoder count, a mileage conversion value and a sensor installed odometer, which are generated by the position encoder and timestamped; a material track grid is established by the mileage sequence, the material track grid is composed of a series of material segment units, each material segment unit covers a fixed length and saves an alignment time index; according to the sensor installed odometer and the conveying speed, the timestamps of the element response sequence and the spectral reflection sequence are mapped to the target material segment unit in the material track grid to form an initial alignment.

[0012] Further, the process of establishing the candidate interval library includes: setting an element peak mapping table, an absorption band mapping table and a feasibility table, which are collectively referred to as the candidate interval library; the element peak mapping table takes the preset peak position combination and the associated peak combination as the retrieval key and returns the grade interval and the moisture content interval; the absorption band mapping table takes the preset absorption band position combination, the band depth sequence relationship and the band width sequence relationship as the retrieval key and returns the grade interval and the moisture content interval; the feasibility table takes the state combination composed of the conveying state code, the instantaneous mass increment and the conveying speed as the retrieval key and returns the grade interval and the moisture content interval.

[0013] Further, the decision estimation unit generates three types of candidate interval sets for each material segment unit according to the aligned multi-source observation sequence; the generation rule of the element response candidate interval set is that after the peak position correction, background deduction, spectral line list matching and associated peak consistency pass, the corresponding interval is returned using the element peak mapping table; the generation rule of the spectral reflection candidate interval set is that after the scattering correction, baseline flattening, absorption band positioning and envelope consistency pass, the corresponding interval is returned using the absorption band mapping table; the generation rule of the mass uniformization candidate interval set is that when the stable state code is established, the corresponding interval is returned using the stable entry in the feasibility table, and when the start-stop state code is established, the corresponding interval is returned using the start-stop entry in the feasibility table.

[0014] Further, the decision estimation unit sequentially performs three-set intersection, two-set intersection and boundary contraction on each material segment unit to form an estimated interval set. The three-set intersection is an interval intersection operation on the element response candidate interval set, the spectral reflectance candidate interval set and the quality uniformization candidate interval set according to interval endpoints and a calculation of intersection width. When the three-set intersection width reaches a preset minimum interval width, the intersection is output as the estimated interval set. When the three-set intersection width is less than the preset minimum interval width, two-set intersections of element response and quality uniformization, spectral reflectance and quality uniformization, and element response and spectral reflectance are sequentially calculated according to a fixed priority. When the width of any two-set intersection reaches the preset minimum interval width, the two-set intersection is output as the estimated interval set. When the widths of the above intersections are all less than the preset minimum interval width, a boundary contraction rule is performed. The boundary contraction rule is to simultaneously move the two end boundaries each by one interval step towards the center within the common coverage of the three types of candidate interval sets and to perform the rule in a loop until the interval width reaches the preset minimum interval width and the interval is output as the estimated interval set.

[0015] Further, the decision estimation unit also processes abnormal segment units in the material segment units. Specifically, when the device state code shows calibration, start-stop switching or impact detection is valid, or when the spectral peak record has a saturation mark, or when the spectral record has a light source failure mark, the material segment unit is marked as an abnormal segment unit. The abnormal segment unit generates a placeholder record and is stored in the abnormal partition of the estimation result set.

[0016] Further, for each material segment unit, the decision estimation unit records the center value of the estimated interval set as the grade estimate and the moisture content estimate, and records the endpoint pair of the estimated interval set as the estimation range, and writes them into the estimation result set together with the material segment unit position and the device state code.

[0017] Further, the result publishing unit receives the batch record returned by the assay system. The batch record includes a batch time range and a batch assay value. According to the alignment time index, the batch time range is mapped to a continuous material segment unit set in the material track grid to form a batch window. In the batch window, the arithmetic mean of the center values of the grade estimate and the moisture content estimate of the estimation result set is calculated to obtain the batch window average estimate. The difference value mark is obtained by comparing the batch window average estimate with the batch assay value. The bias table is updated according to the difference value mark. The update method is to add the difference value mark to the bias items of the grade channel and the moisture content channel. At the same time, the element peak mapping table and the absorption band mapping table are each shifted by one step in the positive and negative directions of the difference value. The element peak mapping table has priority over the absorption band mapping table to form an updated candidate interval library. The bias table and the updated candidate interval library are written into the online estimation engine to generate a parameter snapshot and a traceability log.

[0018] Further, the result publishing unit continuously outputs process control results and trade settlement results according to the estimated result set; wherein the process control results include: grade estimation, moisture content estimation and estimation range arranged in order of material segment units; the trade settlement results include: grade average estimation, moisture content average estimation and estimation range obtained by batch window statistics, with parameter snapshot identification and traceability log identification.

[0019] The copper concentrate multi-source fusion grade and moisture content rapid estimation system has the following beneficial effects: by introducing delay compensation and mixed alignment on the material track grid, the element response sequence, the spectral reflection sequence and the mass flow sequence can be integrated under a unified alignment time index, thereby significantly reducing the misalignment problem caused by sampling phase difference and speed fluctuation of different detection methods, and effectively improving the consistency between multi-source observations. In the candidate interval generation and processing process, the regular operations such as peak correction, background subtraction, absorption band positioning and state inspection are adopted, the traditional experience-dependent fusion method is replaced by deterministic mapping and interval operation, and stable candidate interval set can be obtained without relying on historical data and artificial weight, and through the intersection of the three, the intersection of the two according to the priority and the boundary shrinkage, the estimation interval set meeting the minimum interval width is gradually converged, and the interference of extreme abnormal values on the result is significantly suppressed. For the material segment unit with abnormal device state code display, the invention adopts the placeholder recording strategy to store it separately in the estimated result set, which does not damage the continuity of normal data, thereby enhancing the robustness and traceability of the result. At the batch window level, the invention accurately corresponds the batch record returned by the assay system to the alignment time index, and the calculated grade average estimation and moisture content average estimation can be directly compared with the assay value to form a difference mark and update the bias table in real time, while the interval level in the candidate interval library is shifted as a whole, a closed-loop self-calibration mechanism is established, and the estimation result is ensured not to produce systematic deviation due to device drift or ore source fluctuation in long-term operation. In this way, the invention not only provides continuous and stable grade and moisture content estimation for process control, but also outputs settlement basis with statistical representativeness and traceability in the trade settlement scenario, thereby significantly improving the technical reliability, commercial acceptability and application value. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 A system structure schematic diagram of a copper concentrate multi-source fusion grade and moisture content rapid estimation system provided for an embodiment of the invention;

[0021] Figure 2 A near-infrared spectrum absorption band positioning and moisture content estimation schematic diagram provided for an embodiment of the invention. DETAILED DESCRIPTION

[0022] In order to make the person skilled in the art better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor should belong to the scope of protection of the present application.

[0023] Reference Figure 1 A copper concentrate multi-source fusion grade and moisture content rapid estimation system, the system comprises:

[0024] A sequence acquisition and fusion unit is configured to acquire an element response sequence, a spectral reflectance sequence, a mass flow sequence and a position sequence, perform delay compensation and mixed alignment, form an aligned time index on a material track grid, and obtain an aligned multi-source observation sequence.

[0025] Specifically, in the acquisition process, the online X-ray fluorescence device outputs spectral peak records according to a fixed sampling period and writes them into the element response sequence, each spectral peak record containing a peak position index list, a peak area list, a peak width list, a background estimation index and a device state code, and carrying a time stamp of the sampling time. The online near-infrared device outputs spectral records according to a fixed sampling period and writes them into the spectral reflectance sequence, each spectral record containing a waveband index list, a reflectance list, an absorption band label list, a noise estimation index and a light source state code, and carrying a time stamp of the sampling time. The belt scale and the speed meter generate conveying records with the same time reference and write them into the mass flow sequence, each conveying record containing an instantaneous mass increment, a conveying speed, a steady state code and a cumulative distance, and carrying a time stamp of the sampling time. The position encoder generates displacement records with the same time reference and writes them into the position sequence, each displacement record containing an encoder count, a distance conversion value and a sensor installation distance table, and carrying a time stamp of the sampling time. The time stamps of the four types of sequences are kept consistent through the same time reference, avoiding systematic time drift in subsequent alignment.

[0026] In the alignment procedure, a material track grid is established by a sequence of material segment units. Each material segment unit covers a fixed length and is associated with a unique alignment time index. The fixed length is set with reference to a stable interval of the conveying speed, so that the average advancing distance of the belt within a sampling period is completely covered by a material segment unit. The benefit of this setting is that observations from different devices have the same physical origin within the same material segment unit, which can reduce the misalignment introduced by different sampling phases. When determining the target material segment unit for each sensor record, the relative distance between the device and the reference sampling position is first confirmed according to the sensor installation odometer, and then the record is mapped to a candidate material segment unit of the material track grid by combining the conveying speed at that time. The reason for using the combination of distance and speed mapping is that belt conveying is a displacement-dominated process, and the presentation order of the same material under different sensors is determined by displacement, and speed changes only affect the time of arrival but do not change the relative order of the material.

[0027] After completing the candidate mapping, delay compensation is performed, which is completed by double-pass offset re-registration. The first pass is forward re-registration, which processes the material segment units from upstream to downstream according to the direction of material travel. For each material segment unit, a offset gear is selected from the preset offset gear table, and the record of the element response sequence and the spectral reflectance sequence is moved to the position closest to the material segment unit along the direction of material travel. The selection of the offset gear follows the homogeneity determination rule: when the spectral line list and the absorption band marker appear at the same time and correspond to each other within the range of adjacent material segment units, select the offset gear that makes the two types of records meet in the same material segment unit; when only the spectral line list is stable and the absorption band marker changes greatly, select the offset gear that makes the spectral line list remain coherent in consecutive material segment units; when only the absorption band marker is stable, select the offset gear that makes the absorption band marker remain coherent in consecutive material segment units. The reason for using this determination rule is that the element response sequence is sensitive to metal elements, and the spectral reflectance sequence is sensitive to moisture content and mineralogical changes. When the stable characteristics of the two coincide within the same material segment unit, the same batch of material can be uniquely identified, thereby achieving reliable alignment. The second pass is backward re-registration, which is processed in reverse from downstream to upstream. The purpose is to eliminate the small deviations accumulated in the first pass, so that the alignment boundaries at both ends of the upstream and downstream satisfy the homogeneity determination rule at the same time. The double-pass sequence ensures that the alignment follows the direction of material travel and can correct the residual errors of the forward processing in the opposite direction.

[0028] In the mixed alignment procedure, considering that the material on the belt will form an extended platform and a tail phenomenon along the direction of travel when falling, accumulating and vibrating, which will cause the same feature to appear repeatedly in adjacent material segment units. Therefore, shape de-mixing is performed, including three steps of expansion, contraction and single peak maintenance. Expansion enables the main feature label of each material segment unit to be passed to an adjacent material segment unit, and the determination of the main feature label is based on the simultaneous stability of the peak index and the absorption band marker. The role of expansion is to fill the gap caused by the sampling phase difference, so that the same material continuously passing through will not be cut into unrelated fragments. Contraction is aimed at each local window composed of three adjacent material segment units, and only the main feature label with the highest occurrence frequency is retained within the window, and other labels are replaced with the main feature label. The role of contraction is to compress the excessive spread caused by expansion and mixing, and to restore the compact expression of the material feature. Single peak maintenance is aimed at platform regions with limited length, and only the main feature label at the center of the platform is retained, and the repeated labels on both sides of the platform are removed. The role of single peak maintenance is to eliminate the tail, so that the same material corresponds to the observation only occupying the necessary number of material segment units on the material trajectory grid, thereby obtaining an expression matching the physical passing length. Through the combination of the three steps of expansion filling, contraction compression and single peak maintenance, both the holes caused by the sampling phase difference and the repetitions caused by the mixed diffusion can be covered, and finally a stable shape and clear boundary alignment expression is obtained in the material trajectory grid.

[0029] After completing the shape de-mixing, an alignment time index is generated for each material segment unit. The alignment time index is a monotonically increasing time identifier that records the time when the center position of the material segment unit passes through the reference position, and also records the effective record number of the element response sequence, spectral reflectance sequence and mass flow sequence at that time. Through the alignment time index, the batch window and the corresponding estimated value can be accurately located in the subsequent steps, avoiding window boundary inconsistencies caused by device sampling phase difference and speed fluctuations.

[0030] At the same time of generating the alignment time index, the records from the element response sequence, spectral reflectance sequence and mass flow sequence are aggregated in units of material segment units to form the aligned multi-source observation sequence. For records marked as stable state code, they are directly written into the corresponding material segment unit. For records marked as start-stop or impact state code, the record is written into the abnormal record area of the same material segment unit, and the alignment time index of the material segment unit is kept unchanged to prevent repeated counting caused by start-stop. The reason for this processing is that start-stop and impact will change the sampling phase but not the physical identity of the material, and writing them into the abnormal record area can preserve the information without interfering with normal alignment.

[0031] The fixed length of the material trajectory grid is set as the average advancing distance corresponding to the larger sampling period of the online X-ray fluorescence device and the online near-infrared device. This setting enables complete coverage of both types of spectral data in one material segment unit, thereby improving the reliability of homogeneity determination. To adapt to speed fluctuations, the fixed length can be periodically re-estimated within the speed stable interval, but remains unchanged during any re-estimation to ensure the continuity of the aligned time index. The fixed length of the material trajectory grid is set as the average advancing distance corresponding to the larger sampling period of the online X-ray fluorescence device and the online near-infrared device. This setting enables complete coverage of both types of spectral data in one material segment unit, thereby improving the reliability of homogeneity determination. To adapt to speed fluctuations, the fixed length can be periodically re-estimated within the speed stable interval, but remains unchanged during any re-estimation to ensure the continuity of the aligned time index.

[0032] The unimodal preservation of deconvolution can be replaced by double-end truncation. After detecting the platform region, the double-end truncation truncates from both ends of the platform to the center to the shortest coverage length of the continuous segment of the main feature label inside the platform. Double-end truncation is suitable for working conditions where the conveying speed remains stable for a long time, and can represent the same material with fewer material segment units, thereby improving the compactness of the aligned multi-source observation sequence. The aligned time index is represented by the binary composition of absolute time and material segment unit sequence number, and the source number of each type of record is recorded at the same time when it is generated. The benefit of using binary composition representation is that it can directly switch between the time domain and the material segment unit domain, making it easy to access the same batch of material with the same identifier in subsequent batch window positioning and statistical calculations.

[0033] A decision estimation unit is used to run a ternary deterministic intersection decision algorithm for each material segment unit in the aligned multi-source observation sequence and the material trajectory grid; a candidate interval library and a bias table are established, including an element peak mapping table, an absorption band mapping table, and a feasibility table; after peak position correction, background subtraction, and consistency test of associated peaks, the element response candidate interval set is generated; after scattering correction, baseline flattening, and absorption band positioning, the spectral reflectance candidate interval set is generated; after state inspection, the quality consistent candidate interval set is generated by calling the feasibility table; the intersection of the three, the intersection of the two according to the preset priority, and the boundary shrinkage are sequentially executed to obtain the estimated interval set that meets the preset minimum interval width; the center and endpoints of the estimated interval set are registered as grade estimates, moisture content estimates, and estimated ranges, and are written into the estimation result set together with the equipment state code; the candidate interval library and the bias table are saved.

[0034] Specifically, the decision estimation unit implements a process of generating a candidate interval set of element response in each material segment unit in the aligned multi-source observation sequence and the material trajectory grid, which includes six continuous links of sequence reading, peak position correction, background deduction, spectrum line list matching, associated peak consistency inspection and interval mapping synthesis. The input is the spectrum peak record associated with the material segment unit and the element peak mapping table in the candidate interval library, and the output is the candidate interval set of element response, and the placeholder record is generated under abnormal conditions and written into the abnormal partition of the estimation result set. In the sequence reading link, the target material segment unit is located according to the alignment time index, and the spectrum peak record corresponding to the material segment unit is read. The spectrum peak record includes a peak index list, a peak area list, a peak width list, a background estimation index and a device state code. The device state code is used to indicate the states of calibration, stability, start-stop and saturation. When the device state code is stable, the peak position correction is entered; when the device state code is calibration or saturation, the placeholder record is generated and the abnormal segment unit is marked in the material segment unit, and the alignment time index of the material segment unit is retained to maintain the continuity with the subsequent material segment unit. In the peak position correction link, the element peak mapping table provides a reference peak group and a position tolerance range. By performing a local maximum value search on each reference peak position in the peak index list, the most prominent energy response is located. When multiple local maximum values appear, the peak near the center of the energy segment and with a medium width in the peak width list is selected as the alignment peak, because the response near the center of the energy segment is more stable, and the relative error from the instrument temperature drift and the sample surface effect is smaller. The index difference between each alignment peak and the corresponding reference peak is recorded as a correction offset, and all the peak index in the spectrum peak record is updated synchronously with the same offset. The use of a unified offset can keep the relative peak spacing unchanged within the same scan, avoiding additional error accumulation on the subsequent associated peak consistency inspection. After the correction is completed, the correction quality is judged according to the peak width list and the background estimation index. When the peak width falls within the width range given by the element peak mapping table and the background estimation index is at a normal level, it is determined as valid correction, and the background deduction link is entered.

[0035] In the background deduction step, the background estimation index and the peak width list are used to build a continuous baseline for the spectrum peak record. Specifically, a fixed-length sliding window is used on the peak position index list, and the local valley values at the low and high energy ends of each window are taken to form a low-frequency envelope. The envelope is then smoothed to obtain a monotonically changing background curve. The values at the corresponding positions in the peak area list are subtracted from the values of the background curve at those positions to obtain a net peak area list after background subtraction. Using the valley values at the ends of the sliding window to build the envelope can effectively suppress scattered and continuous background radiation, making the peak area more closely reflect the element characteristic response. If a small number of items after subtraction are less than zero, they are set to zero to avoid false peaks in subsequent matching. In the spectrum line list matching step, the element peak mapping table defines a preset peak group, a set of accompanying confirmation peaks, and a set of prohibited peaks for each element. Peak detection is performed on the net peak area list after background subtraction, and the detection rule is that the net peak area is higher than the noise level and within the allowed range of the peak width list. For each preset peak group, check if the mandatory main peak of the group appears. If the mandatory main peak appears, check if the accompanying confirmation peaks also appear, and record the order of the peak position indexes and the size order of the peak areas. The order of the peak position indexes should be consistent with the order registered in the element peak mapping table, and the size order of the peak areas should be consistent within a certain range, thereby proving that the peaks in the group come from the same element's stable emission process. If any of the prohibited peak set is significantly detected, the matching of the preset peak group is terminated to prevent false positives caused by spectral overlap or matrix interference from entering the candidate.

[0036] In the associated peak consistency verification step, the preset peak group of the same element is cross-verified with the preset peak group of its associated element. The associated element is given by the element peak mapping table, which is used to identify the element combination that often appears together or restricts each other in mineralogy. Cross-verification includes co-occurrence verification and exclusive verification. Co-occurrence verification is used to confirm that when the main peak of an element appears, the marker peak of its associated element also appears in a weaker but stable manner. This co-occurrence indicates that the material comes from a mineral source section with fixed mineral combination, which can improve the credibility of the candidate. Exclusive verification is used to confirm that when the strong peaks of some elements appear significantly, the marker peaks of other elements should be at a low level. This exclusive relationship reflects the matrix absorption or masking effect, which can eliminate the mixed pseudo-signal. Through the combination of co-occurrence verification and exclusive verification, ambiguities that are difficult to eliminate by a single peak group can be compressed to a smaller range, making subsequent interval mapping more focused. In the interval mapping synthesis step, the preset peak group that passes the associated peak consistency verification is used as a search key to query the element peak mapping table to obtain the grade interval and water content interval associated with the preset peak group. Each entry in the element peak mapping table is indexed by the preset peak group, the associated confirmation peak combination, and the prohibited peak combination, and the grade interval and water content interval are given in the form of interval levels. For multiple effective preset peak groups in the same material section unit, the interval endpoint alignment rule is used for synthesis. The interval endpoint alignment rule is to align the interval endpoints to the nearest interval level boundary, and then perform merging on the overlapping or adjacent intervals to obtain a set of interval entries with fewer numbers but stable coverage. When the intervals corresponding to different preset peak groups have a containing relationship, the narrower interval is retained, because the narrower interval represents a result defined by more characteristics, which can faster reach the preset minimum interval width in subsequent three-way intersection and two-way intersection. After synthesis, the interval entry set is registered as the element response candidate interval set, and the location and equipment state code of the material section unit are associated to generate a deterministic input for subsequent joint interval construction.

[0037] The peak position correction module prioritizes peaks near the center of the energy bin and with a medium width in the peak width list as the alignment peaks. The peaks near the center of the energy bin have a small change in position under the conditions of temperature fluctuations and slight drift of the high-voltage power supply, and the peaks with a medium width are not sensitive to noise and remain stable in shape after background subtraction. Taking such peaks as the alignment reference can reduce the mismatch probability of subsequent peak matching. For spectral peak records that lack a central peak, two reference peaks with symmetrical distribution on both sides can be selected to complete local search and offset estimation, respectively, and then the average offset of the two is used to update the peak position index, so that the overall consistency is still maintained in the absence of reference peaks. The background subtraction module adopts a two-stage envelope generation. The first stage constructs a coarse envelope with a fixed window, and the second stage refines the coarse envelope using a shorter window for peak-dense regions, so that the falling section next to the peak is more accurately fitted. The two-stage envelope can obtain a moderate background estimate in both peak-dense and peak-sparse regions, avoiding excessive subtraction in peak-dense regions and insufficient subtraction in peak-sparse regions. After adopting this path, the identification rate of weak peaks in the spectral line list matching is improved, the number of interval sets in the element response candidate interval set is reduced, and interval merging is more direct.

[0038] The consistency check of the associated peak introduces a time proximity check, and only when the associated relationship continues to exist in the adjacent two material bin units will the preset peak group be included in the interval mapping synthesis. The time proximity check uses the alignment time index to perform double-sample confirmation on the physically continuous passing materials, which can suppress occasional co-occurrence caused by single random disturbance, so that the element response candidate interval set more stably reflects the true composition of the material bin unit. The element response candidate interval set after interval mapping synthesis preferentially adopts an inward convergence strategy in endpoint processing. The inward convergence strategy moves the interval endpoints to the center of the set by one interval step, to offset the small residual offset of the peak position index in extreme environments. This strategy can make the candidate interval more easily form an effective intersection with the spectral reflectance candidate interval set and the quality consistency candidate interval set without relying on historical data and weights, thereby improving the achievement rate of the subsequent intersection of the three.

[0039] In the aligned multi-source observation sequence and the material trajectory grid, after locating the target material bin unit, the spectral record and the conveying record associated with the material bin unit are read. The spectral record includes a waveband index list, a reflectance list, an absorption band marker list, a noise estimation indicator, and a light source status code; the conveying record includes instantaneous mass increment, conveying speed, stable state code, and cumulative mileage. If the light source status code indicates that the light source is faulty or self-checking is in progress, an occupancy record is generated for the material bin unit and the alignment time index is retained, and it does not enter the subsequent processing of this step; if the stable state code indicates that the start-stop or impact detection is established, it is still processed but uses the feasibility items corresponding to start-stop or impact in the subsequent quality consistency candidate interval set.

[0040] The stable reference band set at both ends of the band index list and the non-absorption region set in the absorption band marker list are selected, and the reflectance list entries in these sets are used to establish a proportional correction relationship. The reference bands at both ends are used to limit the overall scale, and the non-absorption region is used to limit the local flat section. The reflectance list of all bands is adjusted simultaneously through the proportional correction relationship, so that the fluctuation of the non-absorption region is compressed to a range commensurate with the noise estimation index. The reason for using proportional correction is that the strength of powder surface scattering mainly changes the overall scale and local tilt. By the joint constraint of fixing both ends and flattening the middle section, the reflectance can be pulled back to a unified scale without introducing weights and historical data, and the relative shape of the absorption band is maintained. A fixed-length sliding window is slid on the band index list, and the reflectance entries at the local low percentile at both ends of the window are taken as the window endpoints, and the connection forms a rough baseline. A shorter window is used to refine the absorption band dense area to obtain a flattened baseline that changes monotonously and slowly. The flattened baseline is subtracted from the reflectance list point by point to obtain a flattened reflectance list. This rough-to-fine flattening path can take into account both the sparse and dense areas of the absorption band: the sparse area avoids excessive subtraction, and the dense area avoids under-subtraction, thereby preserving the true shape of the absorption band and facilitating subsequent positioning.

[0041] According to the target band position combination given by the absorption band mapping table, around each target band position, find the reflectance change section with continuous descent in the front section and continuous rise in the rear section, and take the lowest point at the junction of the descent section and the rise section as the band position center. When multiple candidate lowest points appear, prefer the candidate with more symmetric shoulder distance between the two sides and more balanced rise amplitude on both sides of the lowest point, because symmetric and balanced absorption bands often correspond to stable chemical bond vibration signals, which are less affected by noise and particle size. After positioning, the band depth, band width and band shape marker of each absorption band are determined. The band depth is discretized to the interval level of the absorption band mapping table by the reflectance difference of the lowest point relative to the shoulder on both sides, the band width is discretized to the interval level by the band distance between the shoulders on both sides, and the band shape marker is registered by comparing the slope discretization indication of the descent section and the rise section. Through this discretization, the continuous measurement of the absorption band is converted to interval level representation, which is consistent with the search key in the absorption band mapping table.

[0042] The located absorption band and the corresponding band depth, band width and band shape markers are used as the search keys to retrieve the matching entries in the absorption band map. Each matching entry returns the grade interval and moisture interval associated with the band position combination, and is presented in the form of interval bins. The returned results of multiple matching entries within the same material section unit are synthesized according to the interval end point alignment rule. The interval end point alignment rule is: aligning the end points of all intervals to the nearest interval bin boundary; performing merging for intervals that overlap with each other or are adjacent to each other with only one interval bin; preserving the narrower interval when there is a containing relationship. The resulting synthesized result is registered as the spectral reflectance candidate interval set, and is accompanied by the light source status code and noise estimation index, which are reserved for stability judgment in subsequent intersection and boundary shrinkage.

[0043] When the steady state code is valid and the instantaneous mass increment and the conveying speed are within the nominal range and the normal operation interval of the equipment, the stable entry in the feasibility table is retrieved; when the start-stop state code is valid, the start-stop entry in the feasibility table is retrieved; when the impact detection is valid, the impact entry in the feasibility table is retrieved. The feasibility table uses the state combination as the search key, and returns the grade interval and moisture interval compatible with the state combination. The stable entry usually gives a narrower feasible range, because the filling of the conveying cross section is stable, the layer thickness and the angle of repose change less, and extreme moisture and extreme grade are less likely to occur; the start-stop entry gives a wider range to cover the short-term inhomogeneous passage caused by loading and cleaning; the impact entry is between the two, limiting the instantaneous accumulation and dilution caused by the impact of the material wave. The retrieval result is registered as the mass uniformization candidate interval set, and is accompanied by the stable state code, so as to be uniformly referenced with the spectral reflectance candidate interval set in subsequent processing.

[0044] In the same material section unit, the element response candidate interval set, the spectral reflectance candidate interval set and the quality consistent candidate interval set are obtained. The three sets are sequentially subjected to intersection, intersection according to preset priority and boundary contraction to generate the estimated interval set. The intersection of the three sets is performed as follows: all intervals in the three candidate interval sets are sorted in ascending order according to the interval endpoints, the larger endpoint in the three sets is taken as the intersection start point, and the smaller endpoint in the three sets is taken as the intersection end point; when the intersection start point is not later than the intersection end point, the number of covered interval levels is calculated, and when the coverage number reaches the preset minimum interval width, the intersection is directly output as the estimated interval item, and the registration process is entered. When the coverage number is insufficient, the two-set intersection is entered. The two-set intersection is performed according to a fixed priority, with the priority of element response and quality consistency, followed by spectral reflectance and quality consistency, and then element response and spectral reflectance. The execution of each two-set intersection is the same as that of the three-set intersection, and when the coverage number reaches the preset minimum interval width, the two-set intersection is output as the estimated interval item. If the coverage number of all two-set intersections is still insufficient, boundary contraction is entered. The boundary contraction is limited within the common coverage range of the three candidate interval sets, and at the same time, the endpoints of the common coverage range are moved towards the center by one interval level, and the coverage number is recalculated; the movement and calculation are repeated until the coverage number reaches the preset minimum interval width. The boundary contraction ensures that the common coverage is not damaged, and its effect is to eliminate unstable levels on the edges of each candidate interval set and only keep the core levels with consistent information pointing in the same direction, thereby forming a more reliable estimated interval item.

[0045] After the estimated interval items are collected into the estimated interval set, each estimated interval item is registered in turn. The center value adopts the interval level of the middle position of the interval covered interval level, and when the number of covered levels is even, the middle level close to the lower end is adopted to ensure that the same algorithm obtains stable center values when repeatedly executed. The endpoints are represented by the two boundary levels of the interval. For each estimated interval item, the center value is registered as the grade estimate and the moisture content estimate, and the endpoint pair is registered as the estimated range, and is written into the normal partition of the estimated result set together with the position of the material section unit, the alignment time index, the equipment state code and the light source state code. If the material section unit is marked as an abnormal section unit due to the light source state code or the stable state code at the beginning of this step, after the writing into the normal partition is completed, an occupancy record is written into the abnormal partition at the same time, which contains the same position and alignment time index and abnormal marker to support traceability in subsequent statistics.

[0046] After the estimation interval items are gathered into an estimation interval set, each estimation interval item is registered in turn. The center value is represented by the middle position of the interval scale covered by the interval, and when the number of covered scales is even, the middle scale close to the lower end is used to ensure that the same algorithm gets stable center values when repeatedly executed. The endpoints are represented by the two boundary scales of the interval. For each estimation interval item, the center value is registered as a grade estimate and a moisture content estimate, and the endpoint pair is registered as an estimation range, and is written into the normal partition of the estimation result set together with the position of the material segment unit, the alignment time index, the equipment state code, and the light source state code. If the material segment unit is marked as an abnormal segment unit due to the light source state code or the stable state code at the beginning of this step, after the writing into the normal partition is completed, a placeholder record is simultaneously written into the abnormal partition, containing the same position and alignment time index and an abnormal marker, to support traceability during subsequent statistics.

[0047] In an optional implementation, the scattering correction uses a double-reference comparison. When the light source state code is stable, in the vicinity of the alignment time index at which each spectrum is recorded, a front and back non-absorption region is selected from the multi-source observation sequence after alignment to serve as a local reference, and together with the two end reference wavebands, a three-point reference is formed to complete the simultaneous correction of scale and tilt through the proportional relationship between the three points. The three-point reference comparison has stronger local adaptation ability than using only the two end reference wavebands, and can still maintain the shape stability of the absorption band when there is slight dust coverage or illumination fluctuation. The absorption band positioning introduces shoulder confidence checking. After the lowest point is determined, the reflectivity change of the shoulder position on both sides is detected to see if it remains monotonic within the adjacent several wavebands, and if the shoulder appears repeated fluctuations in a short distance, the candidate lowest point is discarded and the next candidate is selected. The shoulder confidence checking can exclude false lowest points caused by noise spikes, making the positioning more reliable. The boundary contraction performs an interval endpoint alignment rule before and after each movement to avoid additional dispersion errors caused by the endpoints falling outside the interval scale boundary. This processing can make the step size of the boundary contraction consistent with the interval scale, reducing the number of loops required to reach the preset minimum interval width. The registration of the center value introduces adjacent consistency checking. When the estimation interval items of two consecutive material segment units are completely consistent at the interval endpoints, the same center value is written into the estimation result set, even if the number of covered scales is even, the middle scale close to the lower end is not selected alternately between the two adjacent material segment units. The adjacent consistency checking can avoid meaningless jitter in the estimation time sequence, and facilitate more stable results when calculating the batch window average estimate in step three.

[0048] The result publishing unit is configured to receive the batch record of the assay system and locate the batch window according to the alignment time index; calculate the grade average estimate and the moisture content average estimate from the estimated result set within the batch window, compare the two with the batch assay value to obtain a difference marker, update the bias table using the difference marker and shift the candidate interval library by one interval in the positive or negative direction of the difference, generate a parameter snapshot and a traceability log, and publish the process control result and the trade settlement result.

[0049] Specifically, after receiving the batch record returned by the assay system, first read the batch time range and batch assay value in the batch record, and check the integrity and sequence of the start and end time of the batch time range. Convert the timestamp of the batch time range to the same time reference as the alignment time index, maintain the same time zone, the same time format and the same precision, and ensure that the one-to-one correspondence with the material track grid does not shift. After completing the time reference unification, generate a unique identifier for the batch record, and record the state identifier of the current candidate interval library and bias table at the beginning of this processing, as the reference starting point of the subsequent parameter snapshot and traceability log. When locating the batch window, use the material track grid as the search space and sequentially scan all material segment units of the alignment time index. The material segment unit whose alignment time index is between the start and end time of the batch time range is directly included in the batch window. When the alignment time index is equal to the start and end time, the half-open interval principle is adopted, including the start point and not including the end point, to avoid adjacent batch windows from being repeatedly counted into the same material segment unit at the boundary. For the boundary material segment unit that only partially overlaps with the batch time range, the close-in principle is used for inclusion judgment: when the time distance between the alignment time index of the material segment unit and the boundary of the batch time range is not more than half the length of the material segment unit, the material segment unit is included in the batch window; when it exceeds, the material segment unit is not included, so that the boundary processing and the spatial scale of the material passing are consistent, and the misentry or missed counting caused by the single sampling phase difference is reduced.

[0050] When the batch time range contains start-stop or impact periods, the batch window is still positioned according to the above rules, but the material segment units in these periods are marked with start-stop or impact flags. These material segment units are kept because start-stop and impact do not change the material identity, only the sampling phase. By keeping the flags instead of discarding, consistent data sources can be used in subsequent calculations of grade and moisture average estimates, and complete traceability is provided for trade settlement. When there is no time index alignment at all within the batch time range, an empty batch window is generated and the fact is logged in the traceability log to avoid subsequent miscomputation caused by empty sets. When the time ranges of two batch records overlap each other, a split is performed. The midpoint of the overlapping interval is calculated, and the material segment units before the midpoint are assigned to the batch record with the earlier start time, and the material segment units after the midpoint are assigned to the batch record with the later start time. By using the midpoint split without modifying either batch time range, it is guaranteed that the two batch windows do not overlap in space, and each material segment unit belongs to only one batch window, facilitating subsequent calculations and reconciliation. If the overlap of the two batch time ranges is only within one material segment unit, it is naturally resolved by the half-open interval principle, and no split needs to be performed.

[0051] After the batch window is positioned, a structured window description is generated for the batch window. The window description contains the sequence of consecutive material segment units, the alignment time index, position, equipment state code, and light source state code of each material segment unit, and the unique identifier of the window. The window description also records the first and last alignment time indexes within the window as the actual coverage of the batch window, and records the positions of start-stop or impact flags within the window to ensure that subsequent calculations can directly access the required information without rescanning the material track grid. To improve the direct usability between the batch window and the estimate result set, a consistency check is performed immediately after the window description is generated. The contents of the check include three items: the first item is whether each material segment unit within the window has a corresponding record in the normal partition of the estimate result set; the second item is whether there is a placeholder record in the abnormal partition with the same position and alignment time index as the material segment units within the window; the third item is whether the candidate interval library and bias table state identifier of all records within the window are consistent with the state identifier recorded at the beginning of this process. When any check fails, the number and position of the failed items are recorded and written into the traceability log, and the window description is kept unchanged to ensure that the determination of window positioning and data consistency is completely saved.

[0052] In an alternative embodiment, the batch record only contains the sampling start time and the sampling duration without the end time, in which case the end time is calculated as the sum of the start time and the duration, and the batch time range is formed by the end time and the start time, and then the above positioning step is performed. This method is suitable for scenarios where the assay system uses sampling duration to express the record time, and can complete the consistent positioning of the batch window without introducing additional fields. The batch record is accompanied by a sampling location identifier. If the sampling location identifier corresponds to a sensor installation odometer that is not synchronized with the reference position of the material track grid, after the time reference is unified, the batch time range is first corrected by the installation mileage difference corresponding to the sampling location identifier, and the corrected time range is used for window positioning. By completing the mileage correction before positioning, systematic time misalignment caused by different sampling locations can be avoided, and the batch window is closer to the actual sampling queue. For long-term unloading or loading operations, the batch record may cover multiple consecutive stop and start cycles. In order to reduce the overhead caused by repeated boundary judgments, the running section table can be generated according to the alignment time index, and the continuously running material section unit is divided into several running sections, and then the window positioning is performed on the running section table. Window positioning is only performed in the running section that intersects with the batch time range, which not only preserves the boundary accuracy, but also shortens the scanning range. There are multiple conveying channels in the field, and the assay system returns batch records separately for each channel. At this time, a material track grid and an alignment time index are established for each channel, and when receiving a batch record, the channel identifier is read to limit the window positioning within the material track grid of the corresponding channel. This method ensures that the material section unit within the window and the batch record come from the same physical channel, avoiding the estimation distortion caused by cross-channel mixing.

[0053] Let the belt speed be (unit length per unit time), the sampling period of the online X-ray fluorescence device be (unit time), the sampling period of the online near-infrared device be (unit time), and the belt metering sampling period be (unit time); the installation mileage difference of the online X-ray fluorescence device to the reference position be (unit length), the installation mileage difference of the online near-infrared device to the reference position be (unit length), and the length of a single material section unit in the material track grid be (unit length). The interval bin width is (dimensionless), the preset minimum interval width be (dimensionless). The grade estimate of a certain material section unit is (dimensionless), and the moisture content estimate is (dimensionless). There are a material segment unit, whose grade and moisture content central values are denoted as and (dimensionless), respectively. The batch grade and moisture content of laboratory assays are denoted as and (dimensionless), respectively. The grade bias and moisture content bias in the bias table are denoted as and (dimensionless), respectively. The difference values are denoted as and (dimensionless), respectively.

[0054] Let , , , , , , . Discretize the continuous running section into a sequence of material segment units, and there are material segment units in a one-minute window. Align the time stamps of the online X-ray fluorescence device and the online near-infrared device according to the time difference of the arrival at the reference location, and the alignment time difference is After shifting the two types of time stamps backward by the above values, perform the two-pass offset re-registration and the shaping solution mixing to obtain the aligned multi-source observation sequence and the monotonically increasing alignment time index. Take the material segment unit number as an example, the spectral peak record and the spectrum record, and the conveying record corresponding to the alignment time index of the material segment unit number are aggregated in the same material segment unit.

[0055] Perform peak position correction, background deduction, spectral line list matching, and associated peak consistency test on the spectral peak record numbered , and use the element peak mapping table to retrieve the element response candidate interval set. In order to facilitate calculation, let the interval bin width be , and the preset minimum interval width be . The element response candidate interval set obtains in the grade and moisture content dimensions, respectively. Perform scattering correction, baseline flattening, and absorption band positioning on the spectrum record numbered , and use the absorption band mapping table to retrieve the spectrum reflection candidate interval set: Perform state test on the conveying record of the same material segment unit, and the steady state code is true. Call the feasibility table to obtain the quality uniformization candidate interval set: Perform the intersection of the three, the intersection of the two according to the preset priority, and the boundary contraction in sequence. First, calculate the intersection of the three. The grade direction intersection of the three is The interval width is , which is directly used as the estimated interval item. The moisture content direction intersection of the three is The interval width is ., as the estimation interval item. The two items are combined to form the estimation interval set. According to the registration rule, the center value takes the midpoint of each interval, and the endpoint is recorded along the interval level boundary, so the center value and the endpoint of the material section unit are The above center value, estimation range, alignment time index and equipment state code are written into the estimation result set.

[0056] The assay system returns a batch record, whose time range is mapped to the material section unit number covered by the alignment time index to (half-open rule positioning from inclusive to exclusive is to ). To demonstrate the calculation, select the center value of representative material section units in the batch window to participate in the statistics (the same calculation is performed on all material section units in the window in the field implementation). Assuming that the center value given by the estimation result set is ; The arithmetic mean of the batch window average estimation , The laboratory assay value given by the assay system for this batch is set to The difference marker calculation is , The bias table is updated using the difference marker: At the same time, the candidate interval library is shifted by one level in the positive and negative direction of the difference value. Assuming that the interval level width is , the grade direction indicates that the online estimation is high, so all interval endpoints related to grade in the element peak mapping table and the absorption band mapping table are simultaneously moved to the lower end by ; for the moisture content direction , the same is true, moving to the lower end by . The example after shifting: the original grade interval is updated to , and the original moisture content interval is updated to . After completion, a parameter snapshot is generated, recording , , , , , , , and the version identifier of the candidate interval library, written into the traceability log. The process control result is published in the order of the material section unit and the estimation range; the trade settlement result is published in the batch window and the statistical boundary of the estimation range in the window, accompanied by the parameter snapshot and traceability log identifier of this time.

[0057] Reference Figure 2 , Figure 2 The absorption band locating technique and moisture content estimation algorithm in the spectral reflectance sequence processing procedure are illustrated in detail. The horizontal axis represents the near-infrared wavelength range from 1000 nm to 4000 nm, and the vertical axis represents the reflectance percentage ranging from 0% to 100%. Different line types are used in the figure to distinguish the processing stages: the gray solid line is the original NIR spectral data, the black solid line is the spectrum after scattering correction, the black dashed line is the spectrum after baseline flattening, and the black dotted line marks the key absorption band position. The near-infrared detection system uses a halogen tungsten lamp light source and an InGaAs detector, with a spectral resolution of 4 nm, an integration time of 100 ms, an average of 32 samples to improve the signal-to-noise ratio, and a ±0.1°C temperature compensation function to ensure measurement stability. The spectral data processing procedure includes four key steps: first, collect the original reflectance spectrum in the range of 1000-4000 nm; then perform scattering correction to eliminate the effects of particle size and surface roughness; then perform baseline flattening, and the gray shaded area indicates the reference range of baseline fitting; finally, implement absorption band positioning and envelope consistency test. The system identifies three main water feature absorption bands: the OH combined water absorption band at 1450 nm, with an absorption depth of 0.18, a bandwidth of 45 nm; the H2O molecular vibration absorption band at 1940 nm, with an absorption depth of 0.25, a bandwidth of 60 nm; the OH stretching vibration absorption band at 2950 nm, with an absorption depth of 0.32, a bandwidth of 85 nm. The black shaded area indicates the effective detection range of each absorption band. The absorption band mapping table uses the preset band position combination [1450, 1940, 2950], the band depth sequence relationship (2950>1940>1450), and the bandwidth sequence relationship (2950>1940>1450) as the retrieval key, and returns the corresponding grade interval and moisture content interval. The envelope consistency test ensures that the relative intensity of the three absorption bands meets the spectral feature rules of water molecules. Based on the quantitative analysis of spectral feature parameters, the system calculates the total moisture content of 6.8%, of which the combined water is 4.2% and the free water is 2.6%, with an estimation accuracy of ±0.3%, meeting the accuracy requirements of industrial online detection.

[0058] The above-described embodiments are merely used to illustrate the technical solutions of the present application, but not limit the present application; even though the foregoing embodiments of the present application have been described in detail, those skilled in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A system for rapid estimation of grade and moisture content of copper concentrate multi-source fusion, characterized in that, The system comprises: a sequence acquisition and fusion unit, configured to acquire an element response sequence, a spectral reflectance sequence, a mass flow sequence and a mileage sequence, perform delay compensation and mixed alignment, form an aligned time index on a material track grid, and obtain an aligned multi-source observation sequence; the element response sequence is a set of spectral peak records arranged in chronological order, each spectral peak record comprising a peak position index list, a peak area list, a peak width list, a background estimation index and a device status code; the spectral reflectance sequence is a set of spectral records arranged in chronological order, each spectral record comprising a waveband index list, a reflectance list, an absorption band label list, a noise estimation index and a light source status code; the mass flow sequence is a set of delivery records arranged in chronological order, each delivery record comprising an instantaneous mass increment, a delivery speed, a steady state code and a cumulative mileage; a decision estimation unit, configured to run a ternary deterministic intersection decision algorithm for each material segment unit in the aligned multi-source observation sequence and the material track grid; establish a candidate interval library and a bias table, the process of establishing the candidate interval library comprising: setting an element peak mapping table, an absorption band mapping table and a feasibility table, which are collectively referred to as the candidate interval library; the element peak mapping table uses preset peak position combinations and associated peak combinations as retrieval keys and returns grade intervals and moisture content intervals; the absorption band mapping table uses preset absorption band position combinations, band depth sequence relationships and band width sequence relationships as retrieval keys and returns grade intervals and moisture content intervals; the feasibility table uses state combinations composed of delivery state codes, instantaneous mass increments and delivery speeds as retrieval keys and returns grade intervals and moisture content intervals; for each material segment unit, generate three types of candidate interval sets according to the aligned multi-source observation sequence: the generation rule of the element response candidate interval set is that after peak position correction, background deduction, spectral line list matching and associated peak consistency pass, the element peak mapping table is used to return the corresponding interval; the generation rule of the spectral reflectance candidate interval set is that after scattering correction, baseline flattening, absorption band positioning and envelope consistency pass, the absorption band mapping table is used to return the corresponding interval; the generation rule of the mass consistency candidate interval set is that when the steady state code is valid, the feasibility table is used to return the corresponding interval, and when the start-stop state code is valid, the feasibility table is used to return the corresponding interval; The three-step intersection, pairwise intersection and boundary contraction are sequentially performed on each material segment unit to form an estimated interval set; the three-set intersection is an interval intersection operation on the element response candidate interval set, the spectral reflection candidate interval set and the quality consistent candidate interval set according to interval endpoints and the intersection width is calculated; when the three-set intersection width reaches the preset minimum interval width, the intersection is output as the estimated interval set; when the three-set intersection width is less than the preset minimum interval width, the pairwise intersection of the element response and the quality consistency, the pairwise intersection of the spectral reflection and the quality consistency, and the pairwise intersection of the element response and the spectral reflection are sequentially calculated according to the fixed priority, and any pairwise intersection width reaching the preset minimum interval width is output as the estimated interval set; when the above intersection widths are all less than the preset minimum interval width, the boundary contraction rule is executed, the boundary contraction rule is to simultaneously move the two end boundaries each by one interval step towards the center within the common coverage range of the three types of candidate interval sets and is cyclically executed until the interval width reaches the preset minimum interval width and the interval is output as the estimated interval set; The center and endpoints of the estimated interval set are registered as grade estimation, moisture content estimation and estimated range, and are written into the estimation result set together with the equipment state code; the candidate interval library and the bias table are saved; The result publishing unit is used for receiving the batch record of the assay system and positioning the batch window according to the alignment time index; the grade average estimation and the moisture content average estimation are calculated from the estimation result set in the batch window, and the difference value mark is obtained by comparing the two with the batch assay value, the bias table is updated using the difference value mark, and the candidate interval library is shifted by one step in the positive or negative direction of the difference value, a parameter snapshot and a traceability log are generated, and the process control result and the trade settlement result are published.

2. The system for rapid estimation of grade and moisture of copper concentrate multi-source fusion according to claim 1, wherein, Each spectrum peak record is generated by an online X-ray fluorescence device at a fixed sampling period and is time-stamped; each spectrum record is generated by an online near-infrared device at a fixed sampling period and is time-stamped; each conveying record is generated by a belt scale and a speed meter jointly and is time-stamped.

3. The system for rapid estimation of grade and moisture of copper concentrate multi-source fusion according to claim 2, wherein, The mileage sequence is a set of displacement records arranged in chronological order, each displacement record contains an encoder count, a mileage conversion value and a sensor installation odometer, and is generated by a position encoder and time-stamped; the material track grid is established by the mileage sequence, the material track grid is composed of continuously arranged material segment units, each material segment unit covers a fixed length and saves an alignment time index; according to the sensor installation odometer and the conveying speed, the timestamps of the element response sequence and the spectral reflection sequence are mapped to the target material segment unit in the material track grid to form an initial alignment.

4. The system for rapid estimation of grade and moisture of copper concentrate multi-source fusion according to claim 1, wherein, The decision estimation unit also processes the abnormal segment unit in the material segment unit, specifically including: when the device state code shows that calibration is in progress, start-stop switching or impact detection is established, or when the spectrum peak record has a saturation mark, or when the spectrum record has a light source failure mark, the material segment unit is marked as an abnormal segment unit; the abnormal segment unit generates a placeholder record and is stored in the abnormal partition of the estimation result set.

5. The system for rapid estimation of grade and moisture of copper concentrate multi-source fusion according to claim 4, wherein, For each material segment unit, the decision estimation unit records the center value of the interval set as the grade estimate and moisture content estimate, and records the endpoint pairs of the interval set as the estimation range, and writes them into the estimation result set together with the material segment unit position and the equipment state code.

6. The system for rapid estimation of grade and moisture of copper concentrate multi-source fusion according to claim 5, characterized in that, The result publishing unit receives the batch record returned by the assay system, the batch record containing the batch time range and the batch assay value; maps the batch time range to the continuous material segment unit set in the material track grid according to the alignment time index, forming a batch window; within the batch window, calculates the arithmetic mean of the center values of the grade estimate and the moisture content estimate in the estimation result set respectively to obtain the batch window average estimate; compares the batch window average estimate with the batch assay value to obtain the difference marker; updates the bias table according to the difference marker, and the update method is to add the difference marker to the bias items of the grade channel and the moisture content channel correspondingly; At the same time, each of the element peak mapping table and the absorption band mapping table is shifted by one step in the positive and negative directions of the difference, and the element peak mapping table is prior to the absorption band mapping table, forming the updated candidate interval library; the bias table and the updated candidate interval library are written into the online estimation engine and generate the parameter snapshot and the traceability log.

7. The system for rapid estimation of grade and moisture of copper concentrate multi-source fusion according to claim 6, wherein, The result publishing unit continuously outputs the process control result and the trade settlement result according to the estimation result set; wherein, the process control result includes: the grade estimate, the moisture content estimate and the estimation range arranged in the order of the material segment unit; the trade settlement result includes: the grade average estimate, the moisture content average estimate and the estimation range obtained by batch window statistics, and is accompanied by the parameter snapshot identifier and the traceability log identifier.

Images