A composite intelligent production method and system for liquid polysulfide aluminum chloride

Abstract

This invention discloses a composite intelligent production method and system for liquid polysulfide aluminum chloride, belonging to the field of liquid polysulfide aluminum chloride production technology. The method specifically includes: mixing sodium aluminate solution with hydrochloric acid to generate an aluminum hydroxide precursor solution, constructing an initial pH trajectory model, adding a sulfur source solution dropwise to the aluminum hydroxide precursor solution, obtaining a first reaction lag time based on the initial pH trajectory model, identifying the structure of the sulfur-aluminum coordination ions formed in the reaction and extracting their dynamic evolution characteristics, obtaining a second reaction lag time, adjusting the sulfur source addition acceleration rate and alkalinity based on the first and second reaction lag times, and producing liquid polysulfide aluminum chloride based on the adjusted sulfur source addition acceleration rate and alkalinity. This application improves the production efficiency and quality of liquid polysulfide aluminum chloride.

Description

Technical Field

[0001] This invention belongs to the field of liquid polysulfide aluminum chloride production technology, specifically a composite intelligent production method and system for liquid polysulfide aluminum chloride. Background Technology

[0002] Liquid polysulfide aluminum chloride is a novel, highly efficient inorganic polymeric flocculant, synthesized primarily by introducing sulfur-containing compounds into an aluminum salt reaction system. It possesses molecular characteristics including high charge density and a multidentate sulfur coordination structure. This material demonstrates superior performance compared to traditional flocculants in water treatment, heavy metal capture, and the treatment of highly challenging organic wastewater.

[0003] Unlike ordinary liquid polyaluminum chloride, the synthesis of liquid polysulfide aluminum chloride involves complex coordination reactions between aluminum ions and polyvalent sulfur anions, forming various bridging or chelating structures such as Al-S, Al-SO4, and Al-S-OH. This type of sulfur coordination complex system endows it with stronger charge neutralization and bridging adsorption capabilities, but it also has the following problems: the sulfur-aluminum coordination reaction has high uncertainty, and the reaction time misalignment leads to sensor data lag. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes a composite intelligent production method and system for liquid polysulfide aluminum chloride.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A composite intelligent production method for liquid polysulfide aluminum chloride includes:

[0007] Sodium aluminate solution was mixed with hydrochloric acid to generate aluminum oxychloride precursor solution, and an initial pH trajectory model was constructed.

[0008] A sulfur source solution is added dropwise to the aluminum hydroxide precursor solution, and the first reaction lag time is obtained according to the initial pH trajectory model.

[0009] Identify the structure of the sulfur-aluminum coordination ions formed in the reaction and extract their dynamic evolution characteristics to obtain the second reaction lag time;

[0010] The sulfur source injection acceleration rate and alkalinity are adjusted based on the first reaction lag time and the second reaction lag time, and liquid polysulfide aluminum chloride is produced based on the adjusted sulfur source injection acceleration rate and alkalinity.

[0011] Specifically, the process of mixing sodium aluminate solution with hydrochloric acid to generate an aluminum oxychloride precursor solution and constructing an initial pH trajectory model includes:

[0012] Hydrochloric acid is injected intermittently with sodium aluminate solution in the reaction tank through the acid injection interface. The pulse interval is adjusted according to the pH change rate of the solution in the previous cycle.

[0013] The pH signal of the reaction liquid in the reaction tank was obtained by multi-point time series sampling, and a pH response matrix was established.

[0014] Based on the pH response matrix and the evolution patterns under different initial ratios and temperature conditions, multiple trajectory curves are established, each trajectory curve corresponding to an evolution pattern.

[0015] The initial ratio and temperature conditions are matched with the initial ratio and temperature conditions corresponding to each evolution law, and the trajectory curve corresponding to the current evolution law is selected from the multiple trajectory curves based on the matching results, and used as the initial pH trajectory model.

[0016] Specifically, adding a sulfur source solution dropwise to the aluminum hydroxide precursor solution and obtaining the first reaction lag time includes:

[0017] Establish a perturbation flow window, identify the liquid phase shear response boundary, and determine the initial addition point of the sulfur source solution by combining the initial pH trajectory model;

[0018] Using delayed triggering for dripping, the timing pulses loaded with random noise perturbation are used for dripping, so that the dripping behavior presents a non-periodic waveform;

[0019] Hydrogen sulfide gas was guided through a gas chromatography-mass spectrometry dual-mode analysis in a one-step circulation manner at the top of the reaction tank, and a gas concentration-time response curve was established.

[0020] Based on the preset reference response envelope, the timestamp corresponding to the gas mutation inflection point is set as the first reaction lag time t1.

[0021] Specifically, the identification of the sulfur-aluminum coordination ion structure formed in the reaction and the extraction of its dynamic evolution characteristics to obtain the second reaction lag time includes:

[0022] In the reaction solution of aluminum hydroxide precursor solution and sulfur source solution, a multi-wavelength scanning sequence was excited to generate a set of transient response spectra of aluminum coordination structure, and a snapshot data package was built after each scan.

[0023] The snapshot data packets are converted into two-dimensional spectral trajectory diagrams in chronological order. Sulfur-aluminum coordination peak clusters with absorption shift behavior are extracted, and their initial appearance time, maximum intensity time, and disappearance time are marked to generate candidate coordination state time node clusters.

[0024] The candidate coordination state time node clusters are compared with the preset sulfur-aluminum coordination model to identify whether there are trans-state abrupt change signals or reconstruction signals, and the evolution path of the coordination structure is deduced.

[0025] The time point at which the first spectral switching behavior occurs in the evolutionary path is defined as the second reaction lag time t2, and the coordination state category corresponding to this time point is marked.

[0026] Specifically, adjusting the sulfur source addition acceleration rate and alkalinity based on the first reaction lag time and the second reaction lag time includes:

[0027] The time-series data streams of Raman spectroscopy, electrochemical impedance spectroscopy, and microwave dielectric changes were acquired, and independent parameter sequence matrices were generated for each.

[0028] Based on the parameter sequence matrix, a three-dimensional spatial response parameter matrix under a unified time axis is constructed, and each matrix unit is set to correspond to a micro time slice and a physical property group.

[0029] The three-dimensional spatial response parameter matrix is ​​input into a structure inversion neural network based on an attention weight mechanism to identify the evolution path of the sulfur-aluminum coordination state and output its current structural state, future evolution direction and associated hysteresis label.

[0030] Based on the first reaction lag time t1 and the second reaction lag time t2, a lag prediction window is embedded in the coordination evolution output to generate a lag intervention factor set;

[0031] The concentration of ferrous sulfide impurities in the sulfur source raw material is monitored in real time, and the reaction temperature and target alkalinity range are dynamically adjusted according to their changing trends.

[0032] Based on the set of hysteresis intervention factors, temperature, and target alkalinity range, a weight allocation table between sulfur source injection acceleration rate and alkalinity is constructed to regulate sulfur source injection acceleration rate and alkalinity, and liquid polysulfide aluminum chloride is produced based on the regulated sulfur source injection acceleration rate and alkalinity.

[0033] Specifically, the step of inputting the three-dimensional spatial response parameter matrix into a structure inversion neural network based on an attention weight mechanism to identify the evolution path of the sulfur-aluminum coordination state and output its current structural state, future evolution direction, and associated hysteresis labels includes:

[0034] Based on the three-dimensional spatial response parameter matrix, a multi-scale data view set is constructed, and a unique identity index is assigned to each view.

[0035] A multi-scale data view set is input in parallel into an inversion-type deep neural network structure built on an attention weight mechanism to generate a cross-focus map. The network structure scans the weight distribution of time segments of each view in parallel through an attention head.

[0036] Key parameter mutation nodes are identified in the cross-focusing map, and they are strung together in chronological order to form candidate coordination trajectory paths. The relative offset and time overlap between the paths are calculated to form an evolutionary path map cluster.

[0037] The evolutionary path map clusters are subjected to evolutionary consistency verification. The evolutionary trend of each path is marked by drift, stability and mutation three-state labels, and the current coordination structure state and future evolutionary direction are output.

[0038] The index time slice in which the first state mutation occurs in the future evolutionary direction is identified is defined as the associated lag label.

[0039] Specifically, the step of embedding a lag prediction window into the coordination evolution output based on the first response lag time t1 and the second response lag time t2 to generate a lag intervention factor set includes:

[0040] The first reaction lag time t1 and the second reaction lag time t2 are embedded into the coordination evolution trend vector respectively, and the lag interval reference index is constructed with their interval Δt=t2-t1 as the boundary calibration parameter of the lag prediction window.

[0041] Within the lag prediction window, the coordination evolution trend vector is scanned with local attention weights, and a time-weighted sensitivity curve is constructed by weighted integration of time slices. The time-weighted sensitivity curve is used to characterize the strength of the response parameters to structural evolution during the lag period.

[0042] Based on the time-weighted sensitivity curve, the center position of the region with the largest local gradient change rate is extracted and defined as the hysteresis response center index. A symmetrical extended interval is generated starting from the hysteresis response center index to form an intervention envelope cluster.

[0043] The intervention envelope cluster is subjected to lag factor mapping, and a lag intervention factor set is generated based on its distribution pattern. Each factor in the lag intervention factor set contains a label parameter.

[0044] Specifically, the real-time monitoring of the concentration of ferrous sulfide impurities in the sulfur source raw material, and the dynamic adjustment of the reaction temperature and target basicity range based on its changing trend, includes:

[0045] The concentration characteristics of ferrous sulfide in the sulfur source raw material were collected and compared with the preset impurity model to determine its fluctuation trend.

[0046] Based on the fluctuation trend, the reaction deviation region that will occur in the future reaction process is determined, and the temperature control time period is adjusted to construct the corresponding temperature control strategy.

[0047] The alkalinity range is dynamically adjusted based on impurity fluctuation trends and temperature control strategies.

[0048] A composite intelligent production system for liquid polysulfide aluminum chloride, used to implement the aforementioned composite intelligent production method for liquid polysulfide aluminum chloride, includes: a pH trajectory construction module, a first time acquisition module, a second time acquisition module, and a production adjustment module;

[0049] The pH trajectory construction module is used to mix sodium aluminate solution with hydrochloric acid to generate aluminum oxychloride precursor solution and construct an initial pH trajectory model.

[0050] The first time acquisition module is used to add a sulfur source solution to the aluminum hydroxide precursor solution and obtain the first reaction lag time according to the initial pH trajectory model.

[0051] The second time acquisition module is used to identify the sulfur-aluminum coordination ion structure formed in the reaction and extract its dynamic evolution characteristics to obtain the second reaction lag time;

[0052] The production adjustment module adjusts the sulfur source injection acceleration rate and alkalinity based on the first reaction lag time and the second reaction lag time, and produces liquid polysulfide aluminum chloride based on the adjusted sulfur source injection acceleration rate and alkalinity.

[0053] Specifically, the regulating production module includes: a parameter sequence matrix construction unit, an evolution path output unit, an allocation condition adjustment unit, and a regulating production unit;

[0054] The parameter sequence matrix construction unit is used to acquire time-series data streams of Raman spectroscopy, electrochemical impedance spectroscopy, and microwave dielectric changes, and generate independent parameter sequence matrices for each.

[0055] The evolution path output unit is used to identify the evolution path of the sulfur-aluminum coordination state and output its current structural state, future evolution direction and associated hysteresis label.

[0056] The allocation condition adjustment unit is used to generate a set of hysteresis intervention factors and adjust the reaction temperature and target alkalinity range;

[0057] The regulating production unit constructs a weighted allocation table between the sulfur source injection acceleration rate and the alkalinity based on the set of lag intervention factors, temperature, and target alkalinity range, regulates the sulfur source injection acceleration rate and alkalinity, and produces liquid polysulfide aluminum chloride based on the regulated sulfur source injection acceleration rate and alkalinity.

[0058] Compared with the prior art, the beneficial effects of the present invention are:

[0059] This invention proposes a composite intelligent production method and system for liquid polysulfide aluminum chloride. By constructing a pH trajectory model to predict the initial reaction state and guide the addition of the sulfur source, and combining the evolution of the sulfur-aluminum coordination structure, the double hysteresis times t1 and t2 are extracted. A three-dimensional reaction parameter matrix is ​​constructed by integrating Raman spectroscopy, electrochemical impedance spectroscopy, and microwave dielectric data, generating a set of hysteresis intervention factors. Furthermore, by real-time monitoring of raw material impurity trends, the reaction temperature and target alkalinity range are dynamically adjusted to achieve control over the sulfur source addition rate and alkalinity. This method solves the problems of uncertain sulfur-aluminum reaction paths, uncompensated hysteresis behavior, and uncontrollable impurity disturbances in traditional processes, thus improving the production efficiency and quality of liquid polysulfide aluminum chloride. Attached Figure Description

[0060] Figure 1 A flowchart of a composite intelligent production method for liquid polysulfide aluminum chloride provided by the present invention;

[0061] Figure 2 The two-dimensional spectral trajectory diagram of the sulfur-aluminum coordination structure provided by the present invention;

[0062] Figure 3 Raman spectra of the reaction provided by this invention;

[0063] Figure 4 Electrochemical impedance spectroscopy in the reaction provided by this invention;

[0064] Figure 5 The microwave dielectric change diagram provided by this invention;

[0065] Figure 6 This invention provides a composite intelligent production system architecture diagram for liquid polysulfide aluminum chloride. Detailed Implementation

[0066] The present application will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0067] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0068] It should be noted that, unless there is a conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. In addition, the terms "first," "second," and "third" used in this application do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0069] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items.

[0070] Example 1

[0071] Please see Figure 1 The present invention provides an embodiment of a composite intelligent production method for liquid polysulfide aluminum chloride, comprising the following specific steps:

[0072] Step S1: Mix sodium aluminate solution with hydrochloric acid to generate aluminum oxychloride precursor solution and construct an initial pH trajectory model.

[0073] The specific steps of step S1 are as follows:

[0074] Step S101: Hydrochloric acid is injected intermittently with sodium aluminate solution in the reaction tank through the acid injection interface. The pulse interval is adjusted according to the pH change rate of the solution in the previous cycle.

[0075] In this embodiment, during the synthesis of liquid polysulfide aluminum chloride, a 20% sodium aluminate solution is prepared in advance and kept at 45°C in the reaction tank. 31% hydrochloric acid is injected into the reaction tank in a pulse manner using a three-way acid injection valve. The duration of each acid injection is set to 2 seconds, with an initial interval of 15 seconds. The pH change of the reaction solution is monitored in real time at a sampling frequency of 0.5 seconds, and the pH decrease rate within the current cycle is calculated at the end of each acid injection cycle. When the pH change rate is less than 0.03 pH units / second for two consecutive cycles, the next acid injection interval is shortened to 12 seconds; if the change rate is greater than 0.08, it is automatically extended to 18 seconds. The entire acid injection process lasts for 8 minutes, during which the pH curve decreases in a stepwise manner, eventually tending to a slow and stable transition range, forming a homogeneous aluminum oxychloride precursor solution.

[0076] Step S102: Obtain the pH signal of the reaction liquid in the reaction tank by multi-point time series sampling and establish a pH response matrix.

[0077] In this embodiment, during the reaction of sodium aluminate and hydrochloric acid to generate an aluminum oxychloride precursor solution, high-resolution glass pH electrodes are arranged on the walls of the reaction tank at the inlet, the central turbulent zone, near the outlet, at a predetermined distance below the liquid surface, and at the bottom settling zone. Each electrode collects pH values ​​at predetermined intervals and continuously records time-series data. All sampled values ​​are arranged with time as the vertical axis and spatial measurement points as the horizontal axis to generate a two-dimensional pH response matrix. This matrix gradually shows a trend of decreasing vertical gradient and convergence of horizontal distribution as the reaction progresses, reflecting the dynamic homogenization process of the acid-base reaction in the spatial domain.

[0078] Under non-ideal mixing conditions, the pH change of acid-base reactions is a non-uniform diffusion process with obvious differences in local perturbation response. Using this two-dimensional matrix, we can not only determine the strength and phase difference of local pH fluctuations, but also analyze the degree of overlap between the acid diffusion front and the aluminum ion conversion zone in the reaction zone.

[0079] Step S103: Based on the pH response matrix and the evolution law under different initial ratios and temperature conditions, establish multiple trajectory curves, each trajectory curve corresponding to an evolution law.

[0080] In this embodiment, several pH matrices were obtained in the experiment under the initial ratio and temperature conditions. Each matrix was grouped according to its corresponding temperature and molar ratio labels. The average pH value, the location of the maximum gradient interval, and the standard deviation curve at each time point during the reaction were extracted. Based on these characteristics, multiple trajectory curves were summarized.

[0081] Step S104: Match the initial ratio and temperature conditions with the initial ratio and temperature conditions corresponding to each evolution law, and select the trajectory curve corresponding to the current evolution law from the multiple trajectory curves based on the matching results, and use it as the initial pH trajectory model.

[0082] Step S2: Add sulfur source solution dropwise to the aluminum hydroxide precursor solution, and obtain the first reaction lag time according to the initial pH trajectory model.

[0083] The specific steps of step S2 are as follows:

[0084] Step S201: Establish a perturbed flow window, identify the liquid phase shear response boundary, and determine the initial addition point of the sulfur source solution by combining the initial pH trajectory model.

[0085] In this embodiment, a variable frequency driven agitator is used in the reaction tank to apply periodic disturbances during the reaction of sodium aluminate and hydrochloric acid. By deploying shear-sensitive conductivity probes in the reaction liquid, the response time difference and amplitude of the local liquid phase conductivity value under the change of disturbance frequency are recorded. A periodic window is set, and the phase delay value of the liquid response is extracted in different stirring frequency ranges. When the disturbance frequency is greater than the critical threshold, the conductivity response tends to be synchronized, indicating that the system has entered the shear homogenization critical region. At the same time, the pH trajectory collected in this shear window is compared with the previously constructed trajectory model to determine that the pH of the reaction system has entered the non-buffered transition section. It is determined that the initial addition point of the sulfur source should be set at the midpoint of the window when the disturbance frequency is maintained at the critical threshold and the pH decrease rate is higher than the threshold.

[0086] By periodically agitating and stirring, the macroscopic response characteristics of the liquid phase are exposed without changing the chemical composition. The electrical conductivity response delay caused by shear perturbation is consistent with the distribution of reactants at the microscale. When the delay decreases to a critical value, the surface system crosses from the mixing hysteresis zone to the reaction stability zone. Combined with the inflection point segment identified in the pH trajectory model, the physical perturbation behavior and chemical evolution path are cross-located, thereby determining the timing of sulfur source intervention.

[0087] Step S202: Use delayed triggering for dripping, and add the dripping with a timing pulse loaded with random noise disturbance, so that the dripping behavior presents a non-periodic waveform.

[0088] In this embodiment, during the addition of the sulfur source (sodium sulfide), a baseline addition cycle is set, and a random delay perturbation is introduced before each cycle. This delay value is randomly generated by a Gaussian distribution. After each trigger, the pump opening time is 1.5 seconds, and the flow rate is finely adjusted according to the pH fluctuation trend of the previous cycle. The maximum deviation is controlled within ±10%. During the addition process, it was found that the addition flow rate did not have obvious periodic repetition within the preset time window, and the hydrodynamic behavior showed non-steady perturbation characteristics. By tracking the Raman spectrum response of the aluminum-sulfur coordination structure, this non-periodic addition method extends the dynamically adjustable range of the reaction path.

[0089] Step S203: Guide the hydrogen sulfide gas channel at the top of the reaction tank to perform gas chromatography-mass spectrometry dual-mode analysis in a one-step cycle, and establish a gas concentration-time response curve.

[0090] In this embodiment, hydrogen sulfide, as a reaction byproduct during the sulfur source addition process, exhibits a gas-phase release behavior coupled with the reaction rate and coordination competition state of sulfur ions in the liquid phase. Since the solubility of hydrogen sulfide in the liquid phase is significantly affected by pH and aluminum ion complexation, its escape rate shows a clear time lag characteristic. By guiding the gas channel for gas chromatography-mass spectrometry dual-mode analysis, the components of the mixed gas can be separated and qualitatively identified at the molecular level, avoiding interference from other volatile components. A one-step cyclic sampling method can construct a highly time-sensitive concentration response sequence without interrupting the main reaction process; this response curve reflects the rate change of hydrogen sulfide release.

[0091] Step S204: Combine the preset reference response envelope and set the timestamp corresponding to the gas mutation inflection point as the first reaction lag time t1.

[0092] In this embodiment, after the sulfur source is added and hydrogen sulfide gas is collected, the obtained gas concentration-time response curve is compared with the pre-established standard reference envelope. The envelope is the upper and lower limit envelope formed by the hydrogen sulfide response trajectory collected under n different adding rhythms and temperature conditions, which includes the fluctuation range and inflection point interval of the expected reaction window. The envelope matching method is used to extract the change segment in the experimental curve where the instantaneous rise rate is greater than 30% of the reference average rise rate for the first time and lasts for three time points, and the starting time of this segment is marked as the first reaction lag time t1.

[0093] Since the response lag characteristic is condition-dependent, a single peak or rate of change is insufficient as a universal indicator. By mapping the current response curve to the envelope in time series, abnormal behavior intervals are identified without relying on explicit function fitting. Then, the first response lag time t1 is calibrated by matching the rate of change with the critical point of the structural response.

[0094] Step S3: Identify the structure of the sulfur-aluminum coordination ions formed in the reaction and extract their dynamic evolution characteristics to obtain the second reaction lag time.

[0095] The specific steps of step S3 are as follows:

[0096] Step S301: Excite a multi-wavelength scanning sequence in the reaction solution of aluminum oxychloride precursor solution and sulfur source solution to generate a set of instantaneous response spectra of aluminum coordination structure, and establish a snapshot data package after each scan.

[0097] In this embodiment, after each scan is completed, the corresponding scan time, band interval, peak intensity distribution, baseline offset parameters and sampling position information are automatically encapsulated to form a set of structured snapshot data packets. The spectral set generated throughout the reaction process is used to track the configuration evolution of the sulfur-aluminum coordination structure at different time points.

[0098] During the reaction of sulfur source with aluminum oxychloride, the coordination behavior of aluminum ions and sulfide anions forms dynamically changing bridging structures such as Al-S and Al-SO3. Their vibrational characteristics are manifested in Raman spectroscopy as a series of band-specific instantaneous peak clusters. Due to the short lifespan and reversibility of these coordination structures, a multi-wavelength, high-time-density scanning method is required to effectively capture the entire process of their formation, stabilization, and disappearance. The spectral set obtained by scanning is essentially a joint projection of the reaction state in the frequency and time domains. Each set of scanning results is encapsulated as a snapshot data package to preserve the original structural response information. Then, the data is spliced ​​in time sequence, extracted in mode, and classified in spectral form to ultimately reconstruct the dynamic evolution of the coordination structure.

[0099] like Figure 2 As shown, step S302: the snapshot data packet is converted into a two-dimensional spectral trajectory diagram in chronological order, the sulfur-aluminum coordination peak cluster with specific absorption shift behavior is extracted, and its initial appearance time, maximum intensity time and disappearance time are marked to generate a candidate coordination state time node cluster.

[0100] In this embodiment, as Figure 2 As shown, the snapshot dataset is integrated in chronological order along the scanning time axis to form a two-dimensional spectral trajectory diagram, with wavenumber as the horizontal axis, reaction time as the vertical axis, and Raman intensity as the color level. The wavenumber unit is cm. -1 The reaction time is measured in minutes (min), and the Raman intensity is measured in au, used for relative comparison of coordination structure changes during the reaction. Two-dimensional peak extraction was performed on the image, identifying three peak clusters with distinct absorption evolution characteristics, concentrated at wavenumbers of approximately 500 cm⁻¹. -1 650–700cm -1 1100–1130cm -1 For each peak cluster, contour tracking is performed along the time axis to extract its first appearance, the corresponding time of peak intensity, and the time of complete disappearance. Based on this, candidate coordination state time node clusters are constructed and assigned candidate labels according to waveband.

[0101] Presenting snapshot data as two-dimensional images significantly enhances the visual identification of peak position shifts, intensity changes, and overlapping interference. The absorption behavior of specific coordination structures during formation, evolution, and dissociation often exhibits characteristics such as peak shifts, energy bandwidth expansion, and resonance decay. By calibrating these behaviors in the time domain, a mapping relationship between spectral behavior and reaction stages can be established, allowing the extraction of key inflection points in the reaction pathway.

[0102] Step S303: Compare the candidate coordination state time node clusters with the preset sulfur-aluminum coordination model to identify whether there are trans-state mutation signals or reconstruction signals, and deduce the evolution path of the coordination structure.

[0103] In this embodiment, the candidate coordination state time node clusters obtained by spectrum deconstruction are compared one-to-one with the preset sulfur-aluminum coordination model database. The database contains reference Raman profile templates for seven typical aluminum-sulfur complex structures, including Al-S single bridge bond, Al2(μ-S)2 ring structure, Al-SO3 internal ligand, etc. During the comparison, peak position interval, bandwidth expansion, intensity ratio and multi-peak overlap phase are used as feature dimensions, and a weighted scoring algorithm is used to calculate the fitting similarity of each snapshot peak cluster.

[0104] The coordination evolution of the sulfur-aluminum system is highly polymorphic. The same sulfur source forms multiple complex configurations in different reaction windows. The spectrum at a single time point is difficult to reflect the global path. By comparing the spectral shape of the time snapshot with the multivariate reference template, the process of structural reconstruction can be identified from the relative intensity changes, spectral line drift trends and bandwidth broadening behavior. In particular, trans-state abrupt changes often manifest as spectral peak centroid shifts and band splitting, forming identifiable features in the two-dimensional spectral variation diagram.

[0105] Step S304: Define the time point in the evolutionary path where the first spectral switching behavior occurs as the second reaction lag time t2, and mark the coordination state category corresponding to this time point.

[0106] In this embodiment, in the two-dimensional spectral trajectory diagram, the coordination structure evolution path is identified based on spectral shape comparison. The peak drift interval corresponding to each state switch in the path is mapped on the time axis to determine the first behavior and its corresponding time period. By fitting the tangent point at the maximum slope of spectral intensity change in the transition interval, the time point of the switching behavior is determined, and this time point is defined as the second reaction lag time t2.

[0107] The second reaction lag time t2 is essentially the first dominant coordination transition moment after the liquid sulfur source enters the main complexation reaction mechanism, and there is a structural response delay between it and the dropping behavior; the spectral shape switching, as a direct external signal of structural change, has stable characteristic transfer behavior, including phenomena such as sudden changes in the main peak wavenumber, disappearance or new formation of characteristic peaks, and narrowing of bandwidth; by performing spectral intensity derivative analysis on each state node on the evolution path, the significant turning points of the peak cluster can be captured, and the relative starting position of the reaction dynamics can be quantified from the spectral level.

[0108] Step S4: Adjust the sulfur source addition acceleration rate and alkalinity based on the first reaction lag time and the second reaction lag time, and produce liquid polysulfide aluminum chloride based on the adjusted sulfur source addition acceleration rate and alkalinity.

[0109] The specific steps of step S4 are as follows:

[0110] like Figure 3 , Figure 4 and Figure 5As shown, step S401: Collect time-series data streams of Raman spectroscopy, electrochemical impedance spectroscopy and microwave dielectric change, and generate independent parameter sequence matrices for each.

[0111] In this embodiment, three types of sensor acquisition channels are connected during the reaction of liquid polysulfide aluminum chloride: such as Figure 3 As shown, the Raman spectrum covers 400-900 cm⁻¹ -1 Wavenumber intervals are scanned at 30-second intervals to generate characteristic peak intensity arrays. Figure 3 In the diagram, the horizontal axis represents the wave number, with the unit being cm. -1 The vertical axis represents Raman intensity in au; the blue line represents the initial stage of the reaction, the yellow line represents the middle stage, and the red line represents the later stage. Figure 4 As shown, electrochemical impedance spectroscopy was performed using equally spaced Nyquist tests. Impedance nodes with real and imaginary parts were extracted, and an impedance sequence matrix was formed based on the equivalent loop parameters fitted after each test. Figure 4 In the diagram, the horizontal axis represents the real impedance in Ω, and the vertical axis represents the imaginary impedance in Ω. The blue line represents the initial stage of the reaction, where the impedance is high and the interfacial charge transfer is slow, resulting in a large-radius arc. The yellow line represents the middle stage of the reaction, where charge migration accelerates, the impedance decreases, and the arc radius shrinks. The red line represents the later stage of the reaction, where the system tends towards complex coordination stability. Figure 5 As shown, microwave dielectric monitoring samples the dielectric constant and loss angle at a frequency of 2.45 GHz and records their nonlinear decay trend over 5 minutes, generating a complex dielectric parameter matrix indexed by time. Figure 5 In the figure, the horizontal axis represents frequency in GHz, and the vertical axis represents dielectric constant in F / m. The blue line represents the initial stage of the reaction, when the ion concentration is low and the polarizability is moderate. The yellow line represents the middle stage of the reaction, when the polarization increases after the addition of the reagent and the overall dielectric constant increases. The red line represents the later stage of the reaction, when the coordination structure stabilizes and the polarizability decreases.

[0112] Raman spectroscopy, electrochemical impedance spectroscopy, and microwave dielectric parameters respectively map the structural configuration, charge transfer characteristics, and molecular polarization state of the reaction system. Together, they constitute a physical quantification channel for the reaction process at the bond energy-interface-medium level. Due to different acquisition mechanisms, each type of data exhibits asynchronous sampling, different sampling frequencies, and parameter dimensions. By aligning the timestamps of the three and constructing independent parameter sequence matrices, their original variation trends can be preserved to the greatest extent.

[0113] Step S402: Based on the parameter sequence matrix, construct a three-dimensional spatial response parameter matrix under a unified time axis, and set each matrix unit to correspond to a micro time slice and a physical property group.

[0114] Step S403: Input the three-dimensional spatial response parameter matrix into the structure inversion neural network based on the attention weight mechanism to identify the evolution path of the sulfur-aluminum coordination state and output its current structural state, future evolution direction and associated hysteresis label.

[0115] The specific steps of step S403 are as follows:

[0116] Step S4031: Based on the three-dimensional spatial response parameter matrix, construct a multi-scale data view set and assign a unique identity index to each view.

[0117] In this embodiment, based on the constructed three-dimensional spatial reaction parameter matrix, a multi-scale view set is generated according to the time axis sliding window mechanism. Different granularity units and window overlap rates are set, and local sub-matrices with different time spans are extracted from the overall matrix as independent views. Each view retains the complete physical property sequence of the three modes: Raman, electrochemical and dielectric. A series of tensor quantum sets under time scale transformation are generated through matrix slicing operations. After the views are generated, they are assigned an identity index including the window start and end time, mode code and sequence number.

[0118] Step S4032: Input the multi-scale data view set into the inversion-type deep neural network structure built based on the attention weight mechanism in parallel to generate a cross-focus map. The network structure scans the weight distribution of the time segment of each view in parallel through the attention head.

[0119] In this embodiment, the constructed multi-scale data view set is input into the inversion-type deep neural network. The network adopts a multi-head attention mechanism as its core structure. Each attention head scans a sub-view at a specific time scale, focusing on the distribution of physical attribute weights in different time segments within the view. The network performs position embedding on the three types of modal data in the input layer, merges the modal attribute encoding with the time index to form a sequence tensor, and generates query, key, value, and vector groups through linear transformation. Each attention head independently calculates the attention weight and uses it to weight and aggregate the input sequence, finally generating a cross-focusing map in the intermediate layer.

[0120] Step S4033: Identify key parameter mutation nodes in the cross-focusing map, connect them in chronological order to form candidate coordination trajectory paths, and calculate the relative offset and time overlap between paths to form an evolutionary path map cluster.

[0121] In this embodiment, attention value mutation regions within each time scale are extracted from the cross-focusing map output by the inversion neural network. High-intensity weighted superposition nodes of Raman, electrochemical, and dielectric modes appearing at different times are identified as key parameter mutation points. These nodes are concatenated in chronological order to construct multiple candidate coordination trajectory paths. Each path contains a sequence of modal mutation points arranged chronologically, and the physical property type and view source corresponding to each node are labeled. Subsequently, the relative offset index is calculated based on the synchronicity of nodes between paths to measure the response phase difference of different paths in the same structural transition event. Furthermore, the temporal overlap rate between paths is calculated to analyze their consensus region in the evolution process. Finally, a cluster of coordination evolution path maps with internal hierarchical differences and temporal interval intersections is generated as the core graphical representation unit reflecting dynamic structural reconstruction.

[0122] Step S4034: Perform evolutionary consistency verification on the evolutionary path map cluster, label the evolutionary trend of each path with drift, stability and mutation three-state labels, and output the current coordination structure state and future evolutionary direction.

[0123] In this embodiment, for the multiple candidate coordination trajectory paths generated in the previous step, an evolution consistency check is performed in the path map cluster. The check logic is based on the relative offset stability between paths within a continuous time segment, combined with the synchronous weight distribution characteristics of Raman, electrochemical and dielectric modes, to evaluate whether the path exhibits convergence, drift or abrupt change. If adjacent paths have a high overlap rate on the time axis and the difference in physical property response is within the tolerance range, they are marked as stable. If there is a continuous displacement trend between paths and the offset direction is consistent, they are marked as drifting. If there is a sudden view switch and attention focus reconstruction in a short period of time, it is marked as abrupt change. After all path labels are marked, the coordination structure state category of the current reaction and its predicted evolution direction are output according to the label ratio and path aggregation trend in the current time window.

[0124] The essence of evolutionary consistency testing is to identify the phase coordination of these path behaviors. By clustering and trend classifying the local dynamic features of multiple trajectories in a path map cluster within a specific time window, we can determine their evolutionary attributes in physical space.

[0125] Step S4035: Identify the index time slice in the future evolutionary direction where the first state variation occurs, and define it as an associated lag label.

[0126] In this embodiment, after obtaining the coordination trajectory paths labeled with drift-stable-mutation states, the path map clusters are scanned frame by frame along the time axis corresponding to the future evolution trend; the time slice index of the first transition from a stable or drift state to a mutation state is recorded, that is, the starting node of the path set that first shows the tendency of structural reorganization in the predicted evolution is determined, and the time index corresponding to this node is defined as the associated lag label.

[0127] Step S404: Based on the first reaction lag time t1 and the second reaction lag time t2, embed a lag prediction window into the coordination evolution output to generate a lag intervention factor set.

[0128] The specific steps of step S404 are as follows:

[0129] Step S4041: Embed the first reaction lag time t1 and the second reaction lag time t2 into the coordination evolution trend vector respectively, and construct a lag interval reference index with their interval Δt=t2-t1 as the boundary calibration parameter of the lag prediction window.

[0130] In this embodiment, after identifying the first reaction lag time t1 and the second reaction lag time t2, the two time points are embedded into the current coordination evolution trend vector. The trend vector is a multi-dimensional vector group composed of a series of labeled state sequences and characteristic change rates. Its time dimension has been standardized. The window boundary is set with t1 and t2 as anchor points, and the interval Δt=t2-t1 is calculated. This interval is used as the reference index range for constructing the lag interval.

[0131] The lag behavior is essentially a time misalignment between the input disturbance and the structural response. Its boundary is difficult to define clearly by single-point calibration. By calibrating t1 and t2 as the disturbance initiation and response manifestation nodes of the reaction chain, respectively, and using the interval Δt between them as the basis for window quantization, the time domain range of the lag behavior has a clear boundary semantic in the vector space. This lag interval does not depend on the absolute time value, but has structural significance through its association with the evolution trend vector.

[0132] Step S4042: Perform local attention weight scanning on the coordination evolution trend vector within the lag prediction window, and construct a time-weighted sensitivity curve by integrating the weights on each time slice. The time-weighted sensitivity curve is used to characterize the strength of the response parameters to the structural evolution during the lag period.

[0133] In this embodiment, within the lag prediction window Δt defined by the first response lag time t1 and the second response lag time t2, the corresponding segment of the coordination evolution trend vector sequence is extracted, and a local attention weight scan is performed. Using a sliding mechanism based on time slices, the attribute sub-dimensions under different physical modes are weighted and reduced in each time slice. Then, the attention response intensity of each time slice is integrated and superimposed to construct a weighted sensitivity curve that changes with time.

[0134] Step S4043: Based on the time-weighted sensitivity curve, extract the center position of the region with the largest local gradient change rate, define it as the hysteresis response center index, and generate a symmetrical extended interval starting from the hysteresis response center index to form an intervention envelope cluster.

[0135] In this embodiment, the time-weighted sensitivity curve is processed by the first derivative to identify the region with the largest local gradient change rate. The center of curvature of gradient change is calculated for this region, and its index position on the time axis is extracted and defined as the hysteresis response center index. Subsequently, with this center point as the axis of symmetry, a time-symmetric dynamic window is constructed by extending the sensitivity curve forward and backward by a fixed proportion according to the decay trend of the sensitivity curve in the positive and negative time directions, forming an intervention envelope cluster.

[0136] Step S4044: Perform lag factor mapping on the intervention envelope cluster and generate a lag intervention factor set based on its distribution pattern. Each factor in the lag intervention factor set contains a label parameter.

[0137] In this embodiment, the generated intervention envelope cluster is discretized according to the time dimension, and the entire envelope interval is divided into multiple equal-width sub-segments. Each sub-segment is mapped to a lagged intervention factor unit based on multiple parameters such as the gradient value of the sensitivity curve it covers, the distance of the hysteresis response center, and the modal activation intensity. Each factor unit is assigned a set of label parameters, including: the dominant type of the corresponding modal weight, such as electrochemical dominant type; the response position state, such as the post-peak decline segment; the intervention recommendation intensity level, such as medium; and the time index interval. The set of factors of all sub-segments constitutes the final lagged intervention factor set.

[0138] Step S405: Monitor the concentration of ferrous sulfide impurities in the sulfur source raw material in real time, and dynamically adjust the reaction temperature and target alkalinity range according to its changing trend.

[0139] The specific steps of step S405 are as follows:

[0140] Step S4051: Collect the concentration characteristic data of ferrous sulfide in the sulfur source raw material and compare it with the preset impurity model to determine its fluctuation trend.

[0141] In this embodiment, the instantaneous concentration of ferrous sulfide is sampled in real time using spectral absorption or electrochemical detection technology. A sampling period is set, and the collected results are normalized to generate a concentration feature sequence that changes over time. Then, this concentration sequence is fitted and compared with a preset impurity behavior reference model. The reference model is trained by the fluctuation characteristics of FeS (ferrous sulfide) in a stable state in historical batches, including mode features such as short-period oscillations, sudden peaks, and background drift. The similarity between the current sequence's trend and the feature curves in the model is compared to determine whether the current fluctuation trend of ferrous sulfide is abnormal or tends to be unstable.

[0142] Step S4052: Based on the fluctuation trend, determine the reaction offset region that will occur in the future reaction process, adjust the temperature control time period, and construct the corresponding temperature control strategy.

[0143] In this embodiment, based on the ferrous sulfide concentration fluctuation trend identified in the previous step, it is fitted into the established reaction evolution state map of the sulfur-aluminum system to evaluate the coordination shift behavior it will cause on the future time axis. According to the fluctuation amplitude and change direction, multiple potential structural disturbance nodes are set in the reaction chain timeline, and combined with the existing evolution path model, it is determined which time periods are at risk of reaction state deviation. For these predicted reaction shift regions, a thermally sensitive intervention priority window is set, the original temperature control curve is readjusted, and the thermal response action is activated in advance or delayed to form a dynamic temperature control time period.

[0144] Furthermore, the temperature regulation strategy is guided by dynamic feedback control. Driven by the fluctuation trend, the temperature control time node is locally shifted, and the heating or cooling rate is adjusted at the same time to avoid reaction lag or interruption induced by impurities.

[0145] Step S4053: Dynamically adjust the alkalinity range based on the impurity fluctuation trend and temperature adjustment strategy.

[0146] In this embodiment, the influence of reaction parameter drift caused by impurity changes on the final basicity formation mechanism is further analyzed. By using the acid-base titration curve evolution model in the reaction process, the threshold changes of acid or alkali dosage required under different impurity disturbance backgrounds are quantified. The temperature adjustment time period and fluctuation trend are jointly projected onto the acid-base equilibrium evolution space to identify the displacement trend of the basicity inflection point and the degree of reaction window compression. Based on this, the target basicity range is adaptively contracted or expanded.

[0147] Step S406: Based on the set of hysteresis intervention factors, temperature and target alkalinity range, construct a weight allocation table between sulfur source injection acceleration rate and alkalinity, adjust the sulfur source injection acceleration rate and alkalinity, and produce liquid polysulfide aluminum chloride based on the adjusted sulfur source injection acceleration rate and alkalinity.

[0148] In this embodiment, the weight allocation table uses the hysteresis response intensity, mode dominance label, temperature control triggering period, and alkalization window sensitivity as input indicators, and maps the dynamic contribution coefficients of each control unit to the sulfur source rate and alkalization parameters in a two-dimensional matrix structure; each weight node represents the resource priority allocation direction under a specific response state.

[0149] On the one hand, the label of the lag intervention factor is used to set the response slope of the acceleration rate change; on the other hand, the correction range of the target alkalinity and the temperature control strategy form an overlapping discrimination region on the time axis, which is used to determine the adjustment frequency of the alkalinity step size; through the dynamic reference of the weight table, when the coordination reaction enters different evolution stages, the adjustment emphasis of sulfur source and alkalinity can be automatically switched, so that the structure construction remains in a state of continuous and stable evolution.

[0150] Furthermore, this weight table is a real-time adjustment reference mapping table that can be embedded into a neural network or fuzzy control engine to perform priority sorting and adaptive correction. The actual dropping acceleration rate and alkalinity correction function of the sulfur source solution use this table as the core input variable, and form a closed-loop control in the reaction system by sensing the changing trend of the evolution state in real time.

[0151] It should be noted that the final liquid polysulfide aluminum chloride produced achieves a stable configuration based on the above dynamic control, with coordination integrity and hydroxyl distribution reaching a predictable range, thereby improving the structural controllability and stability of the product.

[0152] Example 2

[0153] Please see Figure 6 Another embodiment of the present invention provides: a composite intelligent production system for liquid polysulfide aluminum chloride, comprising: a pH trajectory construction module, a first time acquisition module, a second time acquisition module, and a production adjustment module;

[0154] The pH trajectory construction module is used to mix sodium aluminate solution with hydrochloric acid to generate aluminum oxychloride precursor solution and construct an initial pH trajectory model.

[0155] The first time acquisition module is used to add a sulfur source solution to the aluminum hydroxide precursor solution and obtain the first reaction lag time according to the initial pH trajectory model.

[0156] The second time acquisition module is used to identify the sulfur-aluminum coordination ion structure formed in the reaction and extract its dynamic evolution characteristics to obtain the second reaction lag time;

[0157] The production adjustment module adjusts the sulfur source injection acceleration rate and alkalinity based on the first reaction lag time and the second reaction lag time, and produces liquid polysulfide aluminum chloride based on the adjusted sulfur source injection acceleration rate and alkalinity.

[0158] The production regulation module includes: a parameter sequence matrix construction unit, an evolution path output unit, an allocation condition adjustment unit, and a production regulation unit;

[0159] The parameter sequence matrix construction unit is used to acquire time-series data streams of Raman spectroscopy, electrochemical impedance spectroscopy, and microwave dielectric changes, and generate independent parameter sequence matrices for each.

[0160] The evolution path output unit is used to identify the evolution path of the sulfur-aluminum coordination state and output its current structural state, future evolution direction and associated hysteresis label.

[0161] The allocation condition adjustment unit is used to generate a set of hysteresis intervention factors and adjust the reaction temperature and target alkalinity range;

[0162] The regulating production unit constructs a weighted allocation table between the sulfur source injection acceleration rate and the alkalinity based on the set of lag intervention factors, temperature, and target alkalinity range, regulates the sulfur source injection acceleration rate and alkalinity, and produces liquid polysulfide aluminum chloride based on the regulated sulfur source injection acceleration rate and alkalinity.

[0163] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0164] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A composite intelligent production method for liquid polysulfide aluminum chloride, characterized in that, include: Sodium aluminate solution was mixed with hydrochloric acid to generate aluminum oxychloride precursor solution, and an initial pH trajectory model was constructed. A sulfur source solution is added dropwise to the aluminum hydroxide precursor solution, and the first reaction lag time is obtained according to the initial pH trajectory model. Identify the structure of the sulfur-aluminum coordination ions formed in the reaction and extract their dynamic evolution characteristics to obtain the second reaction lag time; The sulfur source addition acceleration rate and alkalinity are adjusted based on the first reaction lag time and the second reaction lag time, and liquid polysulfide aluminum chloride is produced based on the adjusted sulfur source addition acceleration rate and alkalinity. The process of mixing sodium aluminate solution with hydrochloric acid to generate an aluminum oxychloride precursor solution and constructing an initial pH trajectory model includes: Hydrochloric acid is injected intermittently with sodium aluminate solution in the reaction tank through the acid injection interface. The pulse interval is adjusted according to the pH change rate of the solution in the previous cycle. The pH signal of the reaction liquid in the reaction tank was obtained by multi-point time series sampling, and a pH response matrix was established. Based on the pH response matrix and the evolution patterns under different initial ratios and temperature conditions, multiple trajectory curves are established, each trajectory curve corresponding to an evolution pattern. The initial ratio and temperature conditions are matched with the initial ratio and temperature conditions corresponding to each evolution law, and the trajectory curve corresponding to the current evolution law is selected from the multiple trajectory curves based on the matching results, and used as the initial pH trajectory model.

2. The composite intelligent production method for liquid polysulfide aluminum chloride as described in claim 1, characterized in that, Adding a sulfur source solution dropwise to the aluminum hydroxide precursor solution and obtaining the first reaction lag time includes: Establish a perturbation flow window, identify the liquid phase shear response boundary, and determine the initial addition point of the sulfur source solution by combining the initial pH trajectory model; Using delayed triggering for dripping, the timing pulses loaded with random noise perturbation are used for dripping, so that the dripping behavior presents a non-periodic waveform; Hydrogen sulfide gas was guided through a gas chromatography-mass spectrometry dual-mode analysis in a one-step circulation manner at the top of the reaction tank, and a gas concentration-time response curve was established. Based on the preset reference response envelope, the timestamp corresponding to the gas mutation inflection point is set as the first reaction lag time t1.

3. The composite intelligent production method of liquid polysulfide aluminum chloride as described in claim 2, characterized in that, The process of identifying the structure of the sulfur-aluminum coordination ions formed in the reaction and extracting their dynamic evolution characteristics to obtain the second reaction lag time includes: In the reaction solution of aluminum hydroxide precursor solution and sulfur source solution, a multi-wavelength scanning sequence was excited to generate a set of transient response spectra of aluminum coordination structure, and a snapshot data package was built after each scan. The snapshot data packets are converted into two-dimensional spectral trajectory diagrams in chronological order. Sulfur-aluminum coordination peak clusters with absorption shift behavior are extracted, and their initial appearance time, maximum intensity time, and disappearance time are marked to generate candidate coordination state time node clusters. The candidate coordination state time node clusters are compared with the preset sulfur-aluminum coordination model to identify whether there are trans-state abrupt change signals or reconstruction signals, and the evolution path of the coordination structure is deduced. The time point at which the first spectral switching behavior occurs in the evolutionary path is defined as the second reaction lag time t2, and the coordination state category corresponding to this time point is marked.

4. The composite intelligent production method for liquid polysulfide aluminum chloride as described in claim 3, characterized in that, The adjustment of the sulfur source addition acceleration rate and alkalinity based on the first reaction lag time and the second reaction lag time includes: The time-series data streams of Raman spectroscopy, electrochemical impedance spectroscopy, and microwave dielectric changes were acquired, and independent parameter sequence matrices were generated for each. Based on the parameter sequence matrix, a three-dimensional spatial response parameter matrix under a unified time axis is constructed, and each matrix unit is set to correspond to a micro time slice and a physical property group. The three-dimensional spatial response parameter matrix is ​​input into a structure inversion neural network based on an attention weight mechanism to identify the evolution path of the sulfur-aluminum coordination state and output its current structural state, future evolution direction and associated hysteresis label. Based on the first reaction lag time t1 and the second reaction lag time t2, a lag prediction window is embedded in the coordination evolution output to generate a lag intervention factor set; The concentration of ferrous sulfide impurities in the sulfur source raw material is monitored in real time, and the reaction temperature and target alkalinity range are dynamically adjusted according to their changing trends. Based on the set of hysteresis intervention factors, temperature, and target alkalinity range, a weight allocation table between sulfur source injection acceleration rate and alkalinity is constructed to regulate sulfur source injection acceleration rate and alkalinity, and liquid polysulfide aluminum chloride is produced based on the regulated sulfur source injection acceleration rate and alkalinity.

5. The composite intelligent production method of liquid polysulfide aluminum chloride as described in claim 4, characterized in that, The process involves inputting the three-dimensional spatial response parameter matrix into a structure inversion neural network based on an attention weight mechanism to identify the evolution path of the sulfur-aluminum coordination state and outputting its current structural state, future evolution direction, and associated hysteresis labels, including: Based on the three-dimensional spatial response parameter matrix, a multi-scale data view set is constructed, and a unique identity index is assigned to each view. A multi-scale data view set is input in parallel into an inversion-type deep neural network structure built on an attention weight mechanism to generate a cross-focus map. The network structure scans the weight distribution of time segments of each view in parallel through an attention head. Key parameter mutation nodes are identified in the cross-focusing map, and they are strung together in chronological order to form candidate coordination trajectory paths. The relative offset and time overlap between the paths are calculated to form an evolutionary path map cluster. The evolutionary path map clusters are subjected to evolutionary consistency verification. The evolutionary trend of each path is marked by drift, stability and mutation three-state labels, and the current coordination structure state and future evolutionary direction are output. The index time slice in which the first state mutation occurs in the future evolutionary direction is identified is defined as the associated lag label.

6. The composite intelligent production method of liquid polysulfide aluminum chloride as described in claim 5, characterized in that, The step of embedding a lag prediction window into the coordination evolution output based on the first response lag time t1 and the second response lag time t2 to generate a lag intervention factor set includes: The first reaction lag time t1 and the second reaction lag time t2 are embedded into the coordination evolution trend vector respectively, and the lag interval reference index is constructed with their interval Δt=t2-t1 as the boundary calibration parameter of the lag prediction window. Within the lag prediction window, the coordination evolution trend vector is scanned with local attention weights, and a time-weighted sensitivity curve is constructed by weighted integration of time slices. The time-weighted sensitivity curve is used to characterize the strength of the response parameters to structural evolution during the lag period. Based on the time-weighted sensitivity curve, the center position of the region with the largest local gradient change rate is extracted and defined as the hysteresis response center index. A symmetrical extended interval is generated starting from the hysteresis response center index to form an intervention envelope cluster. The intervention envelope cluster is subjected to lag factor mapping, and a lag intervention factor set is generated based on its distribution pattern. Each factor in the lag intervention factor set contains a label parameter.

7. The composite intelligent production method of liquid polysulfide aluminum chloride as described in claim 6, characterized in that, The real-time monitoring of the concentration of ferrous sulfide impurities in the sulfur source raw material, and the dynamic adjustment of the reaction temperature and target basicity range based on its changing trend, includes: The concentration characteristics of ferrous sulfide in the sulfur source raw material were collected and compared with the preset impurity model to determine its fluctuation trend. Based on the fluctuation trend, the reaction deviation region that will occur in the future reaction process is determined, and the temperature control time period is adjusted to construct the corresponding temperature control strategy. The alkalinity range is dynamically adjusted based on impurity fluctuation trends and temperature control strategies.

8. A composite intelligent production system for liquid polysulfide aluminum chloride, used to implement the composite intelligent production method for liquid polysulfide aluminum chloride according to any one of claims 1-7, characterized in that, include: pH trajectory construction module, first-time acquisition module, second-time acquisition module, and production adjustment module; The pH trajectory construction module is used to mix sodium aluminate solution with hydrochloric acid to generate aluminum oxychloride precursor solution and construct an initial pH trajectory model. The first time acquisition module is used to add a sulfur source solution to the aluminum hydroxide precursor solution and obtain the first reaction lag time according to the initial pH trajectory model. The second time acquisition module is used to identify the sulfur-aluminum coordination ion structure formed in the reaction and extract its dynamic evolution characteristics to obtain the second reaction lag time; The production adjustment module adjusts the sulfur source injection acceleration rate and alkalinity based on the first reaction lag time and the second reaction lag time, and produces liquid polysulfide aluminum chloride based on the adjusted sulfur source injection acceleration rate and alkalinity.

9. The composite intelligent production system for liquid polysulfide aluminum chloride as described in claim 8, characterized in that, The regulation production module includes: a parameter sequence matrix construction unit, an evolution path output unit, an allocation condition adjustment unit, and a regulation production unit; The parameter sequence matrix construction unit is used to acquire time-series data streams of Raman spectroscopy, electrochemical impedance spectroscopy, and microwave dielectric changes, and generate independent parameter sequence matrices for each. The evolution path output unit is used to identify the evolution path of the sulfur-aluminum coordination state and output its current structural state, future evolution direction and associated hysteresis label. The allocation condition adjustment unit is used to generate a set of hysteresis intervention factors and adjust the reaction temperature and target alkalinity range; The regulating production unit constructs a weighted allocation table between the sulfur source injection acceleration rate and the alkalinity based on the set of lag intervention factors, temperature, and target alkalinity range, regulates the sulfur source injection acceleration rate and alkalinity, and produces liquid polysulfide aluminum chloride based on the regulated sulfur source injection acceleration rate and alkalinity.

Images