A method and system for uv online end point determination of an extraction concentration process

Abstract

The application discloses a kind of extraction concentration process ultraviolet online end point determination method and system, it is related to process detection technical field, comprising: multiple modal perception array is laid out, and multiple modal data set is obtained by data acquisition;Adaptive fuzzy inference engine of working condition is constructed;The confidence interval of each mode data is determined, and input to dynamic weight distribution module, and weight coefficient is respectively distributed for ultraviolet full spectrum data, stirring shaft mechanical impedance data and liquid phase ion concentration data;Adaptive fuzzy inference engine of working condition is based on multiple modal data set and corresponding weight coefficient and carries out multistage confidence reasoning, and outputs triad state vector;Based on triad state vector, end point determination signal or cleaning alarm signal is generated, and instruction execution is carried out.The application solves the technical problem that in prior art, it is difficult to accurately determine the concentration end point by relying on ultraviolet signal alone, and achieves the technical effect of improving the accuracy and reliability of online determination of concentration end point.

Description

Technical Field

[0001] This invention relates to the field of process detection technology, specifically to a method and system for determining the endpoint of an extraction and concentration process using ultraviolet light online. Background Technology

[0002] In the extraction and concentration of liquid materials, the concentration endpoint is typically related to the material's moisture content, the enrichment level of the target component, and the quality of the subsequent product. Online ultraviolet (UV) detection can continuously monitor the concentration process by collecting absorption changes of characteristic components in the concentrate, offering advantages such as fast response and no need for frequent sampling. However, as the concentration process progresses, the material viscosity gradually increases, the moisture content decreases, and the component concentration tends to stabilize. This can cause the UV absorption signal to plateau, making it difficult to independently reflect whether the material has truly reached the concentration endpoint. Furthermore, adhesion of the concentrate, changes in turbidity, or contamination of the probe window can also distort the UV signal, increasing the risk of misjudgment in endpoint determination. Summary of the Invention

[0003] This application provides a method and system for determining the ultraviolet online endpoint of the extraction and concentration process, which addresses the technical problem that it is difficult to accurately determine the concentration endpoint by relying solely on ultraviolet signals in the prior art.

[0004] In view of the above problems, this application provides a method and system for determining the ultraviolet online endpoint of the extraction and concentration process.

[0005] The first aspect of this application provides a method for determining the ultraviolet online endpoint of an extraction and concentration process, the method comprising:

[0006] A multimodal sensing array is deployed on the circulation pipeline of the extraction and concentration equipment. This array includes at least an ultraviolet (UV) spectral sensor, a torque sensor, and a conductivity sensor, used to collect real-time UV full-spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data of the concentrate. The UV full-spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data are synchronously aligned on a time axis to obtain a multimodal dataset. A condition-adaptive fuzzy inference engine is constructed, incorporating a complementary verification fusion module and a dynamic weight allocation module. The complementary verification fusion module determines the confidence interval of each modality data based on the current concentration stage and inputs the multimodal dataset into the dynamic weight allocation module. The reassignment module; the dynamic weight allocation module assigns a first weight coefficient, a second weight coefficient, and a third weight coefficient to the UV full-spectrum data, the stirring shaft mechanical impedance data, and the liquid phase ion concentration data respectively based on the current concentration process index; the working condition adaptive fuzzy inference engine, based on the multimodal dataset and the first, second, and third weight coefficients, calls a preset fuzzy inference rule set to perform multi-level confidence inference, outputting a triplet state vector, the triplet state vector containing the first state confidence, the second state confidence, and the third state confidence; based on the triplet state vector, an endpoint determination signal or a cleaning alarm signal is generated and sent to the actuator for instruction execution.

[0007] A second aspect of this application provides an online ultraviolet endpoint determination system for an extraction and concentration process, the system comprising:

[0008] The system includes a sensing deployment unit for deploying a multimodal sensing array on the circulation pipeline of the extraction and concentration equipment. This array includes at least an ultraviolet (UV) spectral sensor, a torque sensor, and a conductivity sensor, used to collect real-time UV full-spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data of the concentrate. A synchronization alignment unit is used to synchronize and align the UV full-spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data on a time axis to obtain a multimodal dataset. An engine construction unit is used to construct a condition-adaptive fuzzy inference engine, which incorporates a complementary verification fusion module and a dynamic weight allocation module. A credibility determination unit is used by the complementary verification fusion module to determine the credibility interval of each modality data based on the current concentration stage and input the multimodal dataset. The dynamic weight allocation module; the weight allocation unit, used by the dynamic weight allocation module to allocate a first weight coefficient, a second weight coefficient, and a third weight coefficient to the ultraviolet full spectrum data, the stirring shaft mechanical impedance data, and the liquid phase ion concentration data respectively according to the current concentration process index; the inference unit, used by the working condition adaptive fuzzy inference engine to call a preset fuzzy inference rule set to perform multi-level confidence inference based on the multimodal dataset and the first weight coefficient, the second weight coefficient, and the third weight coefficient, and output a triplet state vector, the triplet state vector containing a first state confidence, a second state confidence, and a third state confidence; the instruction execution unit, used to generate an endpoint determination signal or a cleaning alarm signal based on the triplet state vector, and send it to the execution mechanism for instruction execution.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] This application deploys a multimodal sensing array on the circulation pipeline of an extraction and concentration device. The multimodal sensing array includes at least an ultraviolet (UV) spectral sensor, a torque sensor, and a conductivity sensor, used to collect real-time UV full-spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data of the concentrate. The UV full-spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data are synchronously aligned on a time axis to obtain a multimodal dataset. A working condition adaptive fuzzy inference engine is constructed, which incorporates a complementary verification fusion module and a dynamic weight allocation module. The complementary verification fusion module determines the confidence interval of each modality data based on the current concentration stage and inputs the multimodal dataset into the dynamic weight allocation module. The system includes a weight allocation module; the dynamic weight allocation module assigns a first weight coefficient, a second weight coefficient, and a third weight coefficient to the UV full-spectrum data, the stirring shaft mechanical impedance data, and the liquid phase ion concentration data respectively based on the current concentration process index; the adaptive fuzzy inference engine, based on the multimodal dataset and the first, second, and third weight coefficients, calls a preset fuzzy inference rule set to perform multi-level confidence inference, outputting a triplet state vector, which includes a first state confidence, a second state confidence, and a third state confidence; based on the triplet state vector, an endpoint determination signal or a cleaning alarm signal is generated and sent to the actuator for instruction execution. This invention solves the technical problem in the prior art where it is difficult to accurately determine the concentration endpoint by relying solely on UV signals. By fusing UV full-spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data and performing adaptive fuzzy inference, it achieves the technical effect of improving the accuracy and reliability of online concentration endpoint determination. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A schematic flowchart of an online ultraviolet endpoint determination method for an extraction and concentration process provided in this application embodiment;

[0013] Figure 2 A broken line diagram illustrating the exponential change of the weight coefficient with the current concentration process in a UV online endpoint determination method for an extraction and concentration process provided in an embodiment of this application;

[0014] Figure 3 This is a schematic diagram of an online ultraviolet endpoint determination system for an extraction and concentration process provided in an embodiment of this application.

[0015] Explanation of reference numerals in the attached figures: Perception deployment unit 11, Synchronization alignment unit 12, Engine construction unit 13, Credibility determination unit 14, Weight allocation unit 15, Inference unit 16, Instruction execution unit 17. Detailed Implementation

[0016] This application provides a method and system for determining the online endpoint of the extraction and concentration process using ultraviolet light. It addresses the technical problem that relying solely on ultraviolet signals makes it difficult to accurately determine the concentration endpoint in existing technologies. By integrating ultraviolet full-spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data and performing adaptive fuzzy inference, the method improves the accuracy and reliability of online endpoint determination.

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0018] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0019] Example 1, as Figure 1 As shown, this application provides a method for determining the ultraviolet online endpoint of an extraction and concentration process, the method comprising:

[0020] Step S100: A multimodal sensing array is installed on the circulation pipeline of the extraction and concentration equipment. The multimodal sensing array includes at least an ultraviolet spectral sensor, a torque sensor, and a conductivity sensor, which are used to collect the ultraviolet full spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data of the concentrate in real time.

[0021] In this embodiment, during the operation of the extraction and concentration equipment, the concentrate is continuously returned to the equipment via a circulation pipeline. The state of the concentrate in the circulation pipeline corresponds to the concentration process within the equipment. By deploying a multimodal sensing array on the circulation pipeline, an online composite detection structure for the concentrate is formed. The multimodal sensing array is a data acquisition array composed of sensors with different detection mechanisms. It includes at least an ultraviolet (UV) spectral sensor, a torque sensor, and a conductivity sensor. The UV spectral sensor performs a UV spectral scan on the concentrate flowing through the circulation pipeline, acquiring the absorption response of the concentrate across the entire UV spectrum in real time to obtain full-spectrum UV data. The torque sensor detects the change in load resistance generated when the stirring shaft drives the concentrate, acquiring the mechanical impedance data of the stirring shaft in real time. The conductivity sensor detects the change in the conductivity of the concentrate phase, acquiring the ion concentration data in real time.

[0022] Step S200: Synchronously align the ultraviolet full spectrum data, the mechanical impedance data of the stirring shaft, and the liquid phase ion concentration data on the time axis to obtain a multimodal dataset.

[0023] In this embodiment, the control clock of the extraction and concentration equipment is used as a unified time reference, a fixed preset sampling period is set, and a time number is generated at the beginning of each preset sampling period. Under the same time number, acquisition trigger signals are sent simultaneously to the ultraviolet spectral sensor, torque sensor, and conductivity sensor, causing the ultraviolet spectral sensor to acquire the ultraviolet full spectrum data of the concentrate, the torque sensor to acquire the mechanical impedance data of the stirring shaft, and the conductivity sensor to acquire the liquid phase ion concentration data. When each sensor completes acquisition, it uploads the acquisition results along with the current time number. When receiving the ultraviolet full spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data, the time number carried by each acquisition result is read first, and then the ultraviolet full spectrum data and stirring shaft mechanical impedance data with the same time number are uploaded together. Impedance data and liquid phase ion concentration data are grouped into the same data record. When all three types of data under the same time number have been received, the data record corresponding to that time number is written into the data sequence. When there is a short delay in the data upload time of any sensor, the sensor data continues to be received and matched within the waiting time window corresponding to that time number. If the corresponding data is not received after the waiting time window, the valid acquisition result of the previous time number of the sensor is used to fill the data field of the current time number. The data records are arranged in ascending order of time number, so that each data record contains UV full spectrum data, stirring shaft mechanical impedance data and liquid phase ion concentration data within the same sampling period, thereby obtaining a multimodal dataset synchronously aligned on the time axis.

[0024] Step S300: Construct a working condition adaptive fuzzy inference engine, wherein the working condition adaptive fuzzy inference engine has a built-in complementary verification fusion module and a dynamic weight allocation module.

[0025] In this embodiment, when constructing the condition-adaptive fuzzy inference engine, the UV full-spectrum data, stirrer shaft mechanical impedance data, and liquid phase ion concentration data in the multimodal dataset are set as input data, and the first state confidence, second state confidence, and third state confidence are set as output data. A complementary verification fusion module is configured in the condition-adaptive fuzzy inference engine. This module pre-stores the confidence intervals corresponding to the UV full-spectrum data, stirrer shaft mechanical impedance data, and liquid phase ion concentration data at different concentration stages. During runtime, it reads the current concentration stage and determines the confidence intervals corresponding to each modal data based on the current concentration stage. The inference engine is configured with a dynamic weight allocation module. This module pre-stores the correspondence between the current concentration process index and the first, second, and third weight coefficients. During runtime, it allocates the first weight coefficient to the UV full spectrum data, the second weight coefficient to the stirring shaft mechanical impedance data, and the third weight coefficient to the liquid phase ion concentration data based on the current concentration process index. The multimodal dataset, the confidence intervals corresponding to each modality, and the first, second, and third weight coefficients are all connected to a preset fuzzy inference rule set, enabling the condition-adaptive fuzzy inference engine to perform multi-level confidence inference according to the preset fuzzy inference rule set.

[0026] Step S400: The complementary verification fusion module determines the confidence interval of each modality data according to the current condensation stage, and inputs the multimodal dataset into the dynamic weight allocation module.

[0027] In this embodiment, the complementary verification fusion module receives the multimodal dataset synchronized and aligned on the time axis and obtains the current concentration stage. Based on the current concentration stage, it determines the confidence intervals for the UV full-spectrum data, the stirring shaft mechanical impedance data, and the liquid phase ion concentration data. When the current concentration stage is the early stage, the confidence interval for the UV full-spectrum data is determined as the first confidence interval, and the confidence interval for the stirring shaft mechanical impedance data is determined as the second confidence interval, with the lower limit of the first confidence interval higher than the upper limit of the second confidence interval. When the current concentration stage is the middle stage, the confidence intervals for the UV full-spectrum data and the stirring shaft mechanical impedance data overlap. When the current concentration stage is the late stage, the confidence interval for the stirring shaft mechanical impedance data is determined as the third confidence interval, and the confidence interval for the UV full-spectrum data is determined as the fourth confidence interval, with the lower limit of the third confidence interval higher than the upper limit of the fourth confidence interval. The confidence interval for the liquid phase ion concentration data remains constant during the early, middle, and late stages of concentration. After the complementary verification fusion module determines the confidence range of each modality data, it inputs the multimodal dataset into the dynamic weight allocation module.

[0028] Furthermore, in the method provided in the application embodiments, the complementary verification fusion module determines the confidence interval of each modality data according to the current condensation stage, and further includes:

[0029] In the early stage of concentration, the reliability range of the UV full spectrum data is the first reliability range, and the reliability range of the stirring shaft mechanical impedance data is the second reliability range. The lower limit of the first reliability range is higher than the upper limit of the second reliability range. In the middle stage of concentration, there is an overlap between the reliability range of the UV full spectrum data and the reliability range of the stirring shaft mechanical impedance data. In the later stage of concentration, the reliability range of the stirring shaft mechanical impedance data is the third reliability range, and the reliability range of the UV full spectrum data is the fourth reliability range. The lower limit of the third reliability range is higher than the upper limit of the fourth reliability range. The reliability range of the liquid phase ion concentration data remains constant throughout the entire concentration process.

[0030] In this embodiment, when the complementary verification fusion module receives the multimodal dataset, it determines the current concentration stage based on the concentration start time of the extraction and concentration device and the preset concentration process duration. Specifically, the concentration start time is denoted as T0, and the preset concentration process duration is denoted as T... max The current sampling time is denoted as T. n When T n -T0 is less than or equal to 1 / 3T max When T is reached, the current concentration stage is determined to be the pre-concentration stage; when T... n -T0 is greater than 1 / 3T maxAnd less than or equal to 2 / 3T max When T is reached, the current concentration stage is determined to be the mid-concentration stage; when T... n -T0 is greater than 2 / 3T max At that time, the current concentration stage is determined to be the late concentration stage. The complementary verification and fusion module determines the corresponding confidence intervals for the UV full spectrum data, stirrer shaft mechanical impedance data, and liquid phase ion concentration data under the same time number, according to the current concentration stage.

[0031] The confidence interval is represented by a normalized numerical range from 0 to 1. In the early stage of concentration, the complementary verification and fusion module sets the confidence interval for the UV full-spectrum data as the first confidence interval and the confidence interval for the stirrer shaft mechanical impedance data as the second confidence interval, ensuring that the lower limit of the first confidence interval is higher than the upper limit of the second confidence interval; for example, the first confidence interval is set to 0.75 to 1.00, and the second confidence interval is set to 0.20 to 0.60. When processing data records from the early stage of concentration, the complementary verification and fusion module maps UV full-spectrum data under the same time number to the first confidence interval and stirrer shaft mechanical impedance data under the same time number to the second confidence interval.

[0032] During the mid-concentration phase, the complementary verification and fusion module sets the confidence intervals of the UV full-spectrum data and the stirrer shaft mechanical impedance data to have overlapping regions. For example, the confidence interval for the UV full-spectrum data is set to 0.50 to 0.85, and the confidence interval for the stirrer shaft mechanical impedance data is set to 0.45 to 0.80, forming an overlapping region between 0.50 and 0.80. When processing data records during the mid-concentration phase, the complementary verification and fusion module maps the UV full-spectrum data and the stirrer shaft mechanical impedance data under the same time number to the aforementioned confidence intervals with overlapping regions.

[0033] During the later stages of concentration, the complementary verification and fusion module sets the confidence interval for the stirrer shaft mechanical impedance data as the third confidence interval and the confidence interval for the UV full-spectrum data as the fourth confidence interval, ensuring that the lower limit of the third confidence interval is higher than the upper limit of the fourth confidence interval; for example, the third confidence interval is set to 0.75 to 1.00, and the fourth confidence interval is set to 0.20 to 0.60. When processing data records from the later stages of concentration, the complementary verification and fusion module maps the stirrer shaft mechanical impedance data under the same time number to the third confidence interval and the UV full-spectrum data under the same time number to the fourth confidence interval.

[0034] The confidence interval for liquid phase ion concentration data remains constant throughout the concentration process; for example, the confidence interval for liquid phase ion concentration data is set to 0.40 to 0.80, and this confidence interval is used in the early, middle, and late stages of concentration. The complementary verification and fusion module determines the confidence intervals for UV full-spectrum data, stirrer shaft mechanical impedance data, and liquid phase ion concentration data according to the current concentration stage, and inputs the corresponding multimodal datasets into the dynamic weight allocation module.

[0035] Step S500: The dynamic weight allocation module assigns a first weight coefficient, a second weight coefficient, and a third weight coefficient to the UV full spectrum data, the stirring shaft mechanical impedance data, and the liquid phase ion concentration data respectively according to the current concentration process index.

[0036] In this embodiment, the dynamic weight allocation module receives the multimodal dataset input by the complementary verification fusion module, and also receives the confidence intervals of each modality data determined by the complementary verification fusion module based on the current concentration stage. Specifically, the UV full-spectrum data corresponds to the confidence interval of the UV full-spectrum data, the stirrer shaft mechanical impedance data corresponds to the confidence interval of the stirrer shaft mechanical impedance data, and the liquid phase ion concentration data corresponds to the confidence interval of the liquid phase ion concentration data.

[0037] The dynamic weight allocation module reads the concentration start time T0 and the current sampling time T from the extraction and concentration equipment. n And the estimated maximum concentration time T max Calculate the current enriched time based on Tn-T0, and then divide the current enriched time by the estimated maximum enriched time T. max The current concentration process index P is obtained, i.e., P = (T n -T0) / T max When P is less than 0, P = 0; when P is greater than 1, P = 1, so that the current concentration process index is kept in the range of 0 to 1.

[0038] The dynamic weight allocation module determines the first, second, and third weight coefficients based on the current concentration progress index P. The first weight coefficient corresponds to the UV full-spectrum data, the second weight coefficient corresponds to the mechanical impedance data of the stirring shaft, and the third weight coefficient corresponds to the liquid phase ion concentration data. The first weight coefficient decreases as the current concentration progress index P increases, the second weight coefficient increases as the current concentration progress index P increases, and the third weight coefficient remains within a preset stable range throughout the entire concentration process.

[0039] In one executable configuration, the initial value of the third weighting coefficient is set to 0.20, the initial value of the first weighting coefficient is set to 0.60 - 0.40P, and the initial value of the second weighting coefficient is set to 0.20 + 0.40P. When the current concentration process index P is 0, the initial values ​​of the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient are all 0.20; when the current concentration process index P is 1, the initial values ​​of the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient are all 0.20.

[0040] like Figure 2 As shown, the horizontal axis represents the current concentration process index P, and the vertical axis represents the weighting coefficients. The initial value of the first weighting coefficient corresponds to the UV full spectrum data. As the current concentration process index P increases from 0 to 1, the initial value of the first weighting coefficient decreases from 0.60 to 0.20. The initial value of the second weighting coefficient corresponds to the mechanical impedance data of the stirring shaft. As the current concentration process index P increases from 0 to 1, the initial value of the second weighting coefficient increases from 0.20 to 0.60. The initial value of the third weighting coefficient corresponds to the liquid phase ion concentration data, which remains at 0.20 during the change of the current concentration process index P. Figure 2 The broken line change corresponds to the process by which the dynamic weight allocation module generates initial values ​​for weight coefficients based on the current concentration process index P.

[0041] The dynamic weight allocation module corrects the initial values ​​of the aforementioned weight coefficients based on the confidence intervals of each modal data. Specifically, the midpoint values ​​of the confidence intervals for the UV full spectrum data, the stirrer shaft mechanical impedance data, and the liquid phase ion concentration data are taken as the confidence participation values ​​for the corresponding modal data. The initial value of the first weight coefficient is multiplied by the confidence participation value of the UV full spectrum data to obtain the first intermediate weight. The initial value of the second weight coefficient is multiplied by the confidence participation value of the stirrer shaft mechanical impedance data to obtain the second intermediate weight. The initial value of the third weight coefficient is multiplied by the confidence participation value of the liquid phase ion concentration data to obtain the third intermediate weight.

[0042] The dynamic weight allocation module normalizes the first, second, and third intermediate weights to obtain the final first, second, and third weight coefficients used in fuzzy inference. Specifically, the first weight coefficient equals the first intermediate weight divided by the sum of the first, second, and third intermediate weights; the second weight coefficient equals the second intermediate weight divided by the sum of the first, second, and third intermediate weights; and the third weight coefficient equals the third intermediate weight divided by the sum of the first, second, and third intermediate weights. After normalization, the sum of the first, second, and third weight coefficients is 1.

[0043] The dynamic weight allocation module assigns the final first weight coefficient, second weight coefficient, and third weight coefficient to the ultraviolet full spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data under each time number in the manner described above. The first weight coefficient, second weight coefficient, and third weight coefficient, along with the multimodal dataset under the corresponding time number, are then input into the fuzzy inference rule set of the working condition adaptive fuzzy inference engine.

[0044] Furthermore, in the method provided in the application embodiments, the calculation of the current concentration process index further includes:

[0045] The ratio obtained by dividing the current concentration time by the estimated maximum concentration time is used as the current concentration process index; the first weighting coefficient decreases as the current concentration process index increases; the second weighting coefficient increases as the current concentration process index increases; and the third weighting coefficient remains within a preset stable range throughout the entire concentration process.

[0046] In this embodiment, the current concentration progress index is determined based on the current concentrated time and the estimated maximum concentration time. The current concentrated time is the cumulative running time of the extraction and concentration equipment from the start of concentration to the current sampling time, and the estimated maximum concentration time is a time parameter preset before the start of the current batch concentration task. The current concentrated time is divided by the estimated maximum concentration time to obtain the current concentration progress index, which is then limited to a range of 0 to 1.

[0047] The UV full-spectrum data, stirrer shaft mechanical impedance data, and liquid phase ion concentration data are weighted according to the current concentration progress index. Specifically, the UV full-spectrum data corresponds to the first weighting coefficient, the stirrer shaft mechanical impedance data to the second weighting coefficient, and the liquid phase ion concentration data to the third weighting coefficient. As the current concentration progress index increases, the first weighting coefficient is adjusted in a decreasing relationship, the second weighting coefficient is adjusted in an increasing relationship, and the third weighting coefficient remains within a preset stable range.

[0048] During the weight allocation process, the dynamic weight allocation module, in conjunction with the confidence interval of each modal data determined by the complementary verification and fusion module, corrects the weight allocation result obtained based on the current concentration process index, so that the ultraviolet full spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data at each sampling time obtain the corresponding first weight coefficient, second weight coefficient, and third weight coefficient, respectively, and the first weight coefficient, second weight coefficient, and third weight coefficient are used to participate in the subsequent adaptive fuzzy inference of the working condition.

[0049] Step S600: The working condition adaptive fuzzy inference engine, based on the multimodal dataset and the first weight coefficient, the second weight coefficient and the third weight coefficient, calls the preset fuzzy inference rule set to perform multi-level confidence inference and outputs a triplet state vector, wherein the triplet state vector includes the first state confidence, the second state confidence and the third state confidence.

[0050] In this embodiment, the working condition adaptive fuzzy inference engine receives a multimodal dataset and a first weight coefficient, a second weight coefficient, and a third weight coefficient. It reads the ultraviolet full spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data at the same sampling time according to the time number. It calculates the ultraviolet spectral stability based on the ultraviolet full spectrum data, the torque change rate based on the stirring shaft mechanical impedance data, and the conductivity fluctuation based on the liquid phase ion concentration data. The ultraviolet spectral stability, torque change rate, and conductivity fluctuation are used as input conditions for a preset fuzzy inference rule set.

[0051] The operating condition adaptive fuzzy inference engine applies a first weighting coefficient to the ultraviolet spectral stability, a second weighting coefficient to the torque change rate, and a third weighting coefficient to the conductivity fluctuation, and then calls a preset fuzzy inference rule set for rule matching. The preset fuzzy inference rule set includes rule conditions corresponding to the first state confidence, second state confidence, and third state confidence; when the ultraviolet spectral stability, torque change rate, and conductivity fluctuation meet the corresponding rule conditions, the corresponding first state confidence, second state confidence, and third state confidence are generated.

[0052] The working condition adaptive fuzzy inference engine combines the first state confidence, second state confidence, and third state confidence obtained under the same time number to form a triplet state vector; the triplet state vector contains the first state confidence, second state confidence, and third state confidence in a preset order.

[0053] Furthermore, in the method provided in the application embodiments, the preset fuzzy inference rule set includes at least:

[0054] First fuzzy inference rule: If the UV spectral stability is high, the torque change rate is positive and medium-high, and the conductivity fluctuation is low, then the confidence level of the first state is high, and the confidence level of the second state is low. Second fuzzy inference rule: If the UV spectral stability is high, the torque change rate is zero and low, and the conductivity fluctuation is low, then the confidence level of the first state is low, and the confidence level of the second state is high. Third fuzzy inference rule: If the UV spectral stability is medium-low, the torque change rate is arbitrary, and the conductivity fluctuation is high, then the confidence level of the third state is high, and the priority of the confidence level of the third state is higher than that of the confidence levels of the first and second states. Fourth fuzzy inference rule: If the UV spectral stability is low, the torque change rate is negative and medium-high, and the conductivity fluctuation is medium, then the confidence level of the third state is medium-high, and a cooling request signal is output simultaneously.

[0055] In this embodiment, before calling the preset fuzzy inference rule set, the condition-adaptive fuzzy inference engine first reads the ultraviolet full-spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data at the same time number from the multimodal dataset, and obtains the ultraviolet spectral stability, torque change rate, and conductivity fluctuation, respectively. Then, the ultraviolet spectral stability, torque change rate, and conductivity fluctuation are multiplied by the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient, respectively, to obtain weighted input variables, and the weighted input variables are used for level determination. The ultraviolet spectral stability is represented by a normalized value from 0 to 1, the torque change rate is represented by a normalized change rate, and the conductivity fluctuation is represented by a normalized value from 0 to 1.

[0056] The adaptive fuzzy inference engine for operating conditions determines the levels of weighted ultraviolet spectral stability, weighted torque change rate, and weighted conductivity fluctuation according to preset level ranges. Specifically, a high ultraviolet spectral stability corresponds to a value range of 0.75 to 1.00; a medium-low ultraviolet spectral stability corresponds to a value range of 0.30 to 0.60; a low ultraviolet spectral stability corresponds to a value range of 0 to 0.35; and a low conductivity fluctuation corresponds to a value range of 0 to 0.25. When the conductivity fluctuation is medium, the corresponding value range is 0.25 to 0.60; when the conductivity fluctuation is high, the corresponding value range is 0.60 to 1.00. When the torque change rate is positive and medium to high, the corresponding value range is 0.30 to 1.00; when the torque change rate is zero and low, the corresponding value range is -0.05 to 0.05; when the torque change rate is negative and medium to high, the corresponding value range is -1.00 to -0.30. When the torque change rate is any value, the value range of the torque change rate is not limited.

[0057] The confidence levels for the first, second, and third states are all represented using normalized values ​​from 0 to 1. A high confidence level is assigned a value of 0.85; a medium-high confidence level is assigned a value of 0.70; and a low confidence level is assigned a value of 0.15. The adaptive fuzzy inference engine then sequentially matches the first, second, third, and fourth fuzzy inference rules based on the aforementioned level determination results.

[0058] If the UV spectral stability falls between 0.75 and 1.00, the torque change rate falls between 0.30 and 1.00, and the conductivity fluctuation falls between 0 and 0.25, then the first fuzzy inference rule is satisfied, i.e., the UV spectral stability is high, the torque change rate is positive and medium-high, and the conductivity fluctuation is low. The confidence level for the first state is assigned 0.85, and the confidence level for the second state is assigned 0.15. If the UV spectral stability falls between 0.75 and 1.00, the torque change rate falls between -0.05 and 0.05, and the conductivity fluctuation falls between 0 and 0.25, then the second fuzzy inference rule is satisfied, i.e., the UV spectral stability is high, the torque change rate is zero and low, and the conductivity fluctuation is low. The confidence level for the first state is assigned 0.15, and the confidence level for the second state is assigned 0.85.

[0059] If the UV spectral stability falls between 0.30 and 0.60, the conductivity fluctuation falls between 0.60 and 1.00, and the torque change rate is arbitrary, then the third fuzzy inference rule is satisfied, i.e., the UV spectral stability is medium to low, the torque change rate is arbitrary, and the conductivity fluctuation is high, and the third state confidence is assigned a value of 0.85. When the output conditions of the first state confidence, the second state confidence, and the third state confidence exist simultaneously under the same time number, the judgment result corresponding to the third state confidence is retained first, according to the priority of the third state confidence being higher than that of the first state confidence and the second state confidence. If the UV spectral stability falls between 0 and 0.35, the torque change rate falls between -1.00 and -0.30, and the conductivity fluctuation falls between 0.25 and 0.60, then the fourth fuzzy inference rule is satisfied, namely, the UV spectral stability is low, the torque change rate is negative and medium-high, and the conductivity fluctuation is medium. The confidence level of the third state is assigned to 0.70, and a cooling request signal is output.

[0060] The adaptive fuzzy inference engine performs rule matching on each data point within a continuous sampling period according to the time number, and uses the first state confidence, second state confidence, and third state confidence obtained under each time number as the inference output for the corresponding sampling time. When the first fuzzy inference rule is matched, the first state confidence is output as 0.85 and the second state confidence as 0.15; when the second fuzzy inference rule is matched, the first state confidence is output as 0.15 and the second state confidence as 0.85; when the third fuzzy inference rule is matched, the third state confidence is output as 0.85 and priority processing is performed; when the fourth fuzzy inference rule is matched, the third state confidence is output as 0.70 and a cooling request signal is generated. Thus, the preset fuzzy inference rule set converts the ultraviolet spectral stability, torque change rate, and conductivity fluctuations into first state confidence, second state confidence, and third state confidence expressed using normalized numerical values.

[0061] Furthermore, the method provided in the application embodiments also includes:

[0062] A first sequence is constructed by selecting the ultraviolet full-spectrum integral values ​​from the most recent first preset number of sampling periods. The coefficient of variation of the first sequence is calculated, and a negative exponential transformation is performed on the coefficient of variation to obtain the ultraviolet spectral stability with a value between zero and one. A weighted linear regression is performed on the torque values ​​from the most recent second preset number of sampling periods, giving higher weight to recent data, and the regression slope is used as the torque change rate. The variance of the conductivity values ​​from the most recent third preset number of sampling periods is calculated and normalized to obtain the conductivity fluctuation. The ultraviolet spectral stability, the torque change rate, and the conductivity fluctuation are multiplied by their corresponding first weighting coefficient, second weighting coefficient, and third weighting coefficient, respectively, to obtain the weighted input variables.

[0063] In this embodiment of the application, the working condition adaptive fuzzy inference engine reads the multimodal dataset according to the time number and calculates the ultraviolet spectral stability, torque change rate and conductivity fluctuation based on the data of continuous sampling periods.

[0064] For ultraviolet full-spectrum data, select the most recent first preset number of sampling periods, such as the most recent 5 sampling periods; within each sampling period, integrate the ultraviolet full-spectrum data collected by the ultraviolet spectrometer according to a preset wavelength range to obtain the ultraviolet full-spectrum integral value of the corresponding sampling period; arrange the ultraviolet full-spectrum integral values ​​of the most recent 5 sampling periods in chronological order to form a first sequence; calculate the mean and standard deviation of the first sequence, and divide the standard deviation by the mean to obtain the coefficient of variation of the first sequence; when the mean is zero, replace the mean with a preset minimum positive number in the calculation; then perform a negative exponential transformation on the coefficient of variation, for example, set the ultraviolet spectral stability to exp(-CV), where CV is the coefficient of variation, so that the value range of the ultraviolet spectral stability is between 0 and 1.

[0065] For the mechanical impedance data of the stirring shaft, select the second most recent preset number of sampling periods, such as the last 5 sampling periods, and read the torque value corresponding to each sampling period. Number the last 5 sampling periods in chronological order as sampling period 1, sampling period 2, sampling period 3, sampling period 4, and sampling period 5, where sampling period 5 is the current sampling period. When performing a weighted linear regression on the torque values ​​of the above 5 sampling periods, set the unnormalized weights corresponding to sampling periods 1 to 5 to 1, 2, 3, 4, and 5 respectively, and divide each unnormalized weight by the sum of the weights, 15, to obtain the corresponding normalized weights 1 / 15, 2 / 15, 3 / 15, 4 / 15, and 5 / 15. Using the sampling period number as the independent variable and the torque value as the dependent variable, perform a weighted linear regression on the torque values ​​of the last 5 sampling periods according to the normalized weights to obtain a regression line. Use the slope of the regression line as the torque change rate. A positive torque change rate indicates that the torque increases as the sampling period progresses; a zero torque change rate indicates that the torque remains stable within the continuous sampling period; and a negative torque change rate indicates that the torque decreases as the sampling period progresses.

[0066] For liquid phase ion concentration data, the third most recent preset number of sampling periods is selected, for example, the last 5 sampling periods, and the conductivity value corresponding to each sampling period is read. The average conductivity value of the last 5 sampling periods is calculated, and then the average of the squares of the differences between each conductivity value and this average value is calculated to obtain the variance of the conductivity value. The variance is divided by the preset maximum variance to obtain the normalized value. Normalized values ​​less than 0 are set to 0, and normalized values ​​greater than 1 are set to 1, resulting in a conductivity fluctuation ranging from 0 to 1. The preset maximum variance is set according to the upper limit of conductivity fluctuation of the extraction and concentration equipment during calibration operation. Thus, the operating condition adaptive fuzzy inference engine obtains the ultraviolet spectral stability through the negative exponential transformation of the coefficient of variation of the ultraviolet full spectrum integral value, obtains the torque change rate through weighted linear regression of the torque value, and obtains the conductivity fluctuation through the normalization of the variance of the conductivity value.

[0067] After obtaining the UV spectral stability, torque change rate, and conductivity fluctuation, the UV spectral stability is multiplied by a first weighting coefficient to obtain the weighted UV spectral stability; the torque change rate is multiplied by a second weighting coefficient to obtain the weighted torque change rate; and the conductivity fluctuation is multiplied by a third weighting coefficient to obtain the weighted conductivity fluctuation. The operating condition adaptive fuzzy inference engine uses the weighted UV spectral stability, weighted torque change rate, and weighted conductivity fluctuation as input variables to a preset fuzzy inference rule set.

[0068] Step S700: Based on the triplet state vector, generate an endpoint determination signal or a cleaning alarm signal, and send it to the actuator for instruction execution.

[0069] In this embodiment, after the adaptive fuzzy inference engine outputs a triplet state vector, signal determination is performed based on the first state confidence, second state confidence, and third state confidence in the triplet state vector. The second state confidence is used to participate in the generation of the endpoint determination signal, and the third state confidence is used to participate in the generation of the cleaning alarm signal. When the second state confidence exceeds a first dynamic threshold and the third state confidence is lower than the second dynamic threshold, the concentration process is determined to meet the endpoint output conditions, and an endpoint determination signal is generated. When the third state confidence exceeds the third dynamic threshold, the cleaning alarm signal is generated first, regardless of the value of the second state confidence, in accordance with the output order of the cleaning alarm signal taking precedence over the endpoint determination signal. After receiving the endpoint determination signal, the actuator controls the concentration equipment to stop heating or stop concentration. After receiving the cleaning alarm signal, the actuator triggers the self-cleaning device of the ultraviolet spectral sensor to perform in-situ cleaning, and automatically performs baseline correction and zero-point reset after cleaning, restoring the ultraviolet spectral sensor to online detection status.

[0070] Furthermore, in the method provided in the application embodiment, based on the triplet state vector, an endpoint determination signal or a cleaning alarm signal is generated and sent to the actuator for instruction execution, which further includes:

[0071] When the confidence level of the second state exceeds the first dynamic threshold and the confidence level of the third state is lower than the second dynamic threshold, an endpoint determination signal is output; when the confidence level of the third state exceeds the third dynamic threshold, regardless of the value of the confidence level of the second state, a cleaning alarm signal is output first; the actuator controls the concentration equipment to stop heating or stop concentration according to the endpoint determination signal, or triggers the self-cleaning device of the ultraviolet spectral sensor to perform in-situ cleaning according to the cleaning alarm signal, and automatically performs baseline correction and zero-point reset after cleaning to restore online detection.

[0072] In this embodiment, after the working condition adaptive fuzzy inference engine outputs a triplet state vector, it reads the second and third state confidence scores from the triplet state vector and performs signal output judgments according to a preset judgment order. The preset judgment order first performs the cleaning alarm signal judgment, and then performs the endpoint judgment signal judgment.

[0073] In the cleaning alarm signal judgment, the third state confidence level is compared with the third dynamic threshold. When the third state confidence level exceeds the third dynamic threshold, a cleaning alarm signal is directly generated, and the second state confidence level is no longer used as the endpoint judgment condition. After the cleaning alarm signal is generated, it is sent to the actuator, which sends a cleaning start command to the self-cleaning device of the ultraviolet spectral sensor. The self-cleaning device opens the cleaning fluid passage or cleaning gas passage according to the cleaning start command, so that the cleaning fluid or cleaning gas acts on the detection window of the ultraviolet spectral sensor, and performs in-situ cleaning according to the preset cleaning time. After the in-situ cleaning is completed, the cleaning fluid passage or cleaning gas passage is closed, and the baseline correction program of the ultraviolet spectral sensor is started. The baseline correction program collects the reference spectral data when the detection window is in a clean state after cleaning, and uses the reference spectral data as the subtraction reference for subsequent ultraviolet full spectrum data. Then, the zero-point reset program is executed to adjust the detection output in the absence of sample or reference state to the zero-point range. After completion, the online detection of the ultraviolet spectral sensor is restored.

[0074] If the confidence level of the third state does not exceed the third dynamic threshold, an endpoint determination signal is executed. The confidence level of the second state is compared with the first dynamic threshold, and the confidence level of the third state is compared with the second dynamic threshold. When the confidence level of the second state exceeds the first dynamic threshold and the confidence level of the third state is lower than the second dynamic threshold, an endpoint determination signal is generated. After the endpoint determination signal is generated, it is sent to the actuator. The actuator sends a stop heating command to the heating control circuit of the concentration equipment according to the endpoint determination signal, cutting off or reducing the output of the heating source; or sends a stop concentration command to the concentration control circuit of the concentration equipment, stopping the evaporation, circulation, or vacuum concentration operation, so that the concentration equipment stops heating or stops concentration.

[0075] Furthermore, in the method provided in the application embodiment, the first dynamic threshold, the second dynamic threshold, and the third dynamic threshold are all adaptively adjusted according to the current concentration process index: the first dynamic threshold decreases as the current concentration process index increases; the second dynamic threshold increases as the current concentration process index increases; and the third dynamic threshold remains within a preset constant range throughout the entire concentration process.

[0076] In this embodiment, the first dynamic threshold, the second dynamic threshold, and the third dynamic threshold are set based on the current concentration process index. Specifically, the current concentration process index P corresponding to the current sampling time is read, and the value range of the current concentration process index P is 0 to 1; the current concentration process index P is input into the threshold calculation program, and the first dynamic threshold, the second dynamic threshold, and the third dynamic threshold are calculated respectively.

[0077] The first dynamic threshold is calculated using a linearly decreasing method. Let the first dynamic threshold be A1, then A1 = 0.85 - 0.25P; when P = 0, A1 = 0.85, and when P = 1, A1 = 0.60, so that the first dynamic threshold decreases as the current concentration process exponentially increases. The first dynamic threshold is used to compare with the second state confidence level. As the concentration process progresses, the judgment criterion for the second state confidence level is updated according to the above linear decreasing relationship.

[0078] The second dynamic threshold is calculated using a linear increment method. Let the second dynamic threshold be A2, then A2 = 0.35 + 0.25P; when P = 0, A2 = 0.35, and when P = 1, A2 = 0.60, so that the second dynamic threshold increases with the current concentration process exponent. The second dynamic threshold is used to compare with the confidence level of the third state. It is recalculated at each sampling time based on the current concentration process exponent P and used as the threshold to determine whether the confidence level of the third state is lower than the corresponding constraint condition.

[0079] The third dynamic threshold is maintained within a preset constant range throughout the concentration process. Specifically, the preset constant range of the third dynamic threshold can be set to 0.70 to 0.80, and the third dynamic threshold A3 can be set to 0.75; when P changes from 0 to 1, A3 remains at 0.75. The third dynamic threshold is used to compare with the third state confidence level. When the third state confidence level exceeds the third dynamic threshold, the cleaning alarm signal output process is initiated.

[0080] After generating the triplet state vector corresponding to each time number, the current concentration process index P is substituted into the above threshold calculation method to obtain the first dynamic threshold A1, the second dynamic threshold A2, and the third dynamic threshold A3 under the current time number. Then, the second state confidence is compared with the first dynamic threshold A1, and the third state confidence is compared with the second dynamic threshold A2 and the third dynamic threshold A3, respectively. Thus, the first dynamic threshold, the second dynamic threshold, and the third dynamic threshold are updated periodically by sampling using the current concentration process index P, so that the endpoint determination signal and the cleaning alarm signal are generated according to the threshold conditions corresponding to the current concentration process.

[0081] Furthermore, the method provided in the application embodiments also includes endpoint confidence evolution verification:

[0082] When the adaptive fuzzy inference engine outputs a second state confidence level exceeding the first dynamic threshold for the first time, it does not immediately trigger a shutdown, but instead starts a verification timing window of a fourth preset duration; it continuously monitors the changing trend of the second state confidence level within the fourth preset duration; if the second state confidence level remains monotonically increasing or remains above the first dynamic threshold within the fourth preset duration, it outputs a final endpoint determination signal at the end of the fourth preset duration; if the second state confidence level decreases and falls below the first dynamic threshold within the fourth preset duration, it cancels the current endpoint determination and returns to the normal condensation state.

[0083] In this embodiment, during the endpoint confidence evolution verification, the working condition adaptive fuzzy inference engine continuously outputs the second state confidence according to a preset sampling period, and reads the first dynamic threshold corresponding to the current sampling time in each sampling period. The second state confidence output at the current sampling time is compared with the first dynamic threshold corresponding to the current sampling time; when the second state confidence is detected to exceed the first dynamic threshold for the first time, the sampling time is recorded as the verification start time, and a verification identifier is generated. The verification identifier indicates that the current endpoint determination has entered the endpoint confidence evolution verification process; after generating the verification identifier, no stop command is sent to the execution mechanism.

[0084] The fourth preset duration is the duration of the verification timing window, set according to the preset number of sampling periods. For example, if the preset sampling period is 10 seconds and the fourth preset duration is set to 60 seconds, the verification timing window includes 6 consecutive sampling periods starting from the verification start time. After the verification timing window starts, the second state confidence level continues to be acquired in each sampling period, and the second state confidence level corresponding to each sampling period is written into the verification sequence in chronological order; at the same time, the first dynamic threshold corresponding to each sampling period is written into the threshold sequence, so that each second state confidence level in the verification sequence corresponds to a first dynamic threshold at the same sampling time.

[0085] Within a fourth preset time period, the verification sequence is continuously compared. Specifically, the confidence level of the second state in the current sampling period is compared with the confidence level of the second state in the previous sampling period. If two adjacent confidence levels of the second state in the verification sequence both satisfy the condition that the confidence level of the second state in the subsequent sampling period is greater than or equal to the confidence level of the second state in the previous sampling period, then it is determined that the confidence level of the second state remains monotonically increasing within the fourth preset time period. Alternatively, each confidence level of the second state in the verification sequence is compared with the first dynamic threshold at the same sampling time in the threshold sequence. If each confidence level of the second state is greater than or equal to the corresponding first dynamic threshold, then it is determined that the confidence level of the second state remains above the first dynamic threshold within the fourth preset time period.

[0086] When the fourth preset time ends, if the confidence level of the second state remains monotonically increasing within the fourth preset time, or if the confidence level of the second state remains above the first dynamic threshold within the fourth preset time, then a final endpoint determination signal is generated and sent to the actuator. After receiving the final endpoint determination signal, the actuator sends a stop heating command to the heating control circuit of the concentration equipment, or sends a stop concentration command to the concentration control circuit of the concentration equipment.

[0087] Within the fourth preset time period, if the confidence level of the second state in any sampling period is less than the confidence level of the second state in the previous sampling period, and the confidence level of the second state in this sampling period is less than the first dynamic threshold corresponding to the same sampling time, then it is determined that the confidence level of the second state has decreased and fallen below the first dynamic threshold within the fourth preset time period. At this time, the current endpoint determination is cancelled, the verification identifier is deleted, the verification sequence and threshold sequence are cleared, the final endpoint determination signal is not output, and the extraction and concentration equipment returns to the normal concentration state. After returning to the normal concentration state, the working condition adaptive fuzzy inference engine continues to output the confidence level of the second state according to the subsequent sampling periods, and restarts the endpoint confidence level evolution verification when the confidence level of the second state exceeds the first dynamic threshold for the first time in the next sampling period.

[0088] Furthermore, the method provided in the application embodiments also includes:

[0089] The fourth preset duration is dynamically adjusted according to the current concentration process index: the fourth preset duration is equal to the product of the basic verification duration and the first coefficient, and the first coefficient increases as the current concentration process index increases.

[0090] In this embodiment, the fourth preset duration is dynamically adjusted based on the current concentration process index. Specifically, when the endpoint confidence evolution verification is initiated, the current concentration process index P corresponding to the current sampling time is read, and the preset basic verification duration T is read. base The basic verification duration is the verification baseline time set by the extraction and concentration equipment before the start of the current batch concentration task. It is set as an integer multiple of the preset sampling period. For example, when the preset sampling period is 10 seconds, the basic verification duration T is... base Set to 60 seconds.

[0091] The first coefficient K1 is calculated based on the current concentration process index P. The first coefficient K1 increases as the current concentration process index P increases. In one specific implementation, K1 = 1 + 0.5P; when P = 0, K1 = 1; when P = 0.5, K1 = 1.25; and when P = 1, K1 = 1.5. Subsequently, the basic verification duration T is... base Multiplying by the first coefficient K1, we get the fourth preset duration T4, that is, T4 = T base ×K1.

[0092] After calculating the fourth preset duration T4, the number of sampling periods included in the verification timing window is determined according to the preset sampling period. For example, the preset sampling period is 10 seconds, and the basic verification duration T... base When the current concentration process index P = 0, the fourth preset duration T4 = 60 seconds, corresponding to 6 sampling periods; if the current concentration process index P = 0.5, the fourth preset duration T4 = 75 seconds, corresponding to 8 sampling periods; if the current concentration process index P = 1, the fourth preset duration T4 = 90 seconds, corresponding to 9 sampling periods. If the fourth preset duration T4 is not divisible by the preset sampling period, the number of sampling periods is determined by rounding up.

[0093] After the endpoint confidence evolution verification is initiated, the second state confidence is continuously collected within the verification timing window corresponding to the fourth preset duration, and a verification sequence is formed in chronological order. When the second state confidence remains monotonically increasing or remains above the first dynamic threshold within the fourth preset duration, the final endpoint determination signal is output at the end of the fourth preset duration. When the second state confidence decreases and falls below the first dynamic threshold within the fourth preset duration, the endpoint determination is cancelled and the system returns to the normal concentration state.

[0094] In summary, the embodiments of this application have at least the following technical effects:

[0095] This application deploys a multimodal sensing array on the circulation pipeline of an extraction and concentration device. The multimodal sensing array includes at least an ultraviolet (UV) spectral sensor, a torque sensor, and a conductivity sensor, used to collect real-time UV full-spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data of the concentrate. The UV full-spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data are synchronously aligned on a time axis to obtain a multimodal dataset. A working condition adaptive fuzzy inference engine is constructed, which incorporates a complementary verification fusion module and a dynamic weight allocation module. The complementary verification fusion module determines the confidence interval of each modality data based on the current concentration stage and inputs the multimodal dataset into the dynamic weight allocation module. The system includes a weight allocation module; the dynamic weight allocation module assigns a first weight coefficient, a second weight coefficient, and a third weight coefficient to the UV full-spectrum data, the stirring shaft mechanical impedance data, and the liquid phase ion concentration data respectively based on the current concentration process index; the adaptive fuzzy inference engine, based on the multimodal dataset and the first, second, and third weight coefficients, calls a preset fuzzy inference rule set to perform multi-level confidence inference, outputting a triplet state vector, which includes a first state confidence, a second state confidence, and a third state confidence; based on the triplet state vector, an endpoint determination signal or a cleaning alarm signal is generated and sent to the actuator for instruction execution. This invention solves the technical problem in the prior art where it is difficult to accurately determine the concentration endpoint by relying solely on UV signals. By fusing UV full-spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data and performing adaptive fuzzy inference, it achieves the technical effect of improving the accuracy and reliability of online concentration endpoint determination.

[0096] Example 2, based on the same inventive concept as the ultraviolet online endpoint determination method for the extraction and concentration process in the foregoing examples, such as... Figure 3 As shown, this application provides an online ultraviolet endpoint determination system for an extraction and concentration process. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0097] The sensing deployment unit 11 is used to deploy a multimodal sensing array on the circulation pipeline of the extraction and concentration equipment. The multimodal sensing array includes at least an ultraviolet spectral sensor, a torque sensor, and a conductivity sensor, which are used to collect ultraviolet full-spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data of the concentrate in real time, respectively. The synchronization alignment unit 12 is used to synchronize and align the ultraviolet full-spectrum data, the stirring shaft mechanical impedance data, and the liquid phase ion concentration data on the time axis to obtain a multimodal dataset. The engine construction unit 13 is used to construct a working condition adaptive fuzzy inference engine, which has a built-in complementary verification fusion module and a dynamic weight allocation module. The credibility determination unit 14 is used by the complementary verification fusion module to determine the credibility interval of each modality data according to the current concentration stage and output the multimodal dataset. The data is fed into the dynamic weight allocation module; the weight allocation unit 15 is used by the dynamic weight allocation module to allocate a first weight coefficient, a second weight coefficient, and a third weight coefficient to the ultraviolet full spectrum data, the stirring shaft mechanical impedance data, and the liquid phase ion concentration data respectively according to the current concentration process index; the inference unit 16 is used by the working condition adaptive fuzzy inference engine to call a preset fuzzy inference rule set to perform multi-level confidence inference based on the multimodal dataset and the first weight coefficient, the second weight coefficient, and the third weight coefficient, and output a triplet state vector, the triplet state vector containing a first state confidence, a second state confidence, and a third state confidence; the instruction execution unit 17 is used to generate an endpoint determination signal or a cleaning alarm signal based on the triplet state vector, and send it to the execution mechanism for instruction execution.

[0098] Furthermore, the system is also used to implement the following functions:

[0099] In the early stage of concentration, the reliability range of the UV full spectrum data is the first reliability range, and the reliability range of the stirring shaft mechanical impedance data is the second reliability range. The lower limit of the first reliability range is higher than the upper limit of the second reliability range. In the middle stage of concentration, there is an overlap between the reliability range of the UV full spectrum data and the reliability range of the stirring shaft mechanical impedance data. In the later stage of concentration, the reliability range of the stirring shaft mechanical impedance data is the third reliability range, and the reliability range of the UV full spectrum data is the fourth reliability range. The lower limit of the third reliability range is higher than the upper limit of the fourth reliability range. The reliability range of the liquid phase ion concentration data remains constant throughout the entire concentration process.

[0100] Furthermore, the system is also used to implement the following functions:

[0101] The ratio obtained by dividing the current concentration time by the estimated maximum concentration time is used as the current concentration process index; the first weighting coefficient decreases as the current concentration process index increases; the second weighting coefficient increases as the current concentration process index increases; and the third weighting coefficient remains within a preset stable range throughout the entire concentration process.

[0102] Furthermore, the system is also used to implement the following functions:

[0103] First fuzzy inference rule: If the UV spectral stability is high, the torque change rate is positive and medium-high, and the conductivity fluctuation is low, then the confidence level of the first state is high, and the confidence level of the second state is low. Second fuzzy inference rule: If the UV spectral stability is high, the torque change rate is zero and low, and the conductivity fluctuation is low, then the confidence level of the first state is low, and the confidence level of the second state is high. Third fuzzy inference rule: If the UV spectral stability is medium-low, the torque change rate is arbitrary, and the conductivity fluctuation is high, then the confidence level of the third state is high, and the priority of the confidence level of the third state is higher than that of the confidence levels of the first and second states. Fourth fuzzy inference rule: If the UV spectral stability is low, the torque change rate is negative and medium-high, and the conductivity fluctuation is medium, then the confidence level of the third state is medium-high, and a cooling request signal is output simultaneously.

[0104] Furthermore, the system is also used to implement the following functions:

[0105] A first sequence is constructed by selecting the ultraviolet full-spectrum integral values ​​from the most recent first preset number of sampling periods. The coefficient of variation of the first sequence is calculated, and a negative exponential transformation is performed on the coefficient of variation to obtain the ultraviolet spectral stability with a value between zero and one. A weighted linear regression is performed on the torque values ​​from the most recent second preset number of sampling periods, giving higher weight to recent data, and the regression slope is used as the torque change rate. The variance of the conductivity values ​​from the most recent third preset number of sampling periods is calculated and normalized to obtain the conductivity fluctuation. The ultraviolet spectral stability, the torque change rate, and the conductivity fluctuation are multiplied by their corresponding first weighting coefficient, second weighting coefficient, and third weighting coefficient, respectively, to obtain the weighted input variables.

[0106] Furthermore, the system is also used to implement the following functions:

[0107] When the confidence level of the second state exceeds the first dynamic threshold and the confidence level of the third state is lower than the second dynamic threshold, an endpoint determination signal is output; when the confidence level of the third state exceeds the third dynamic threshold, regardless of the value of the confidence level of the second state, a cleaning alarm signal is output first; the actuator controls the concentration equipment to stop heating or stop concentration according to the endpoint determination signal, or triggers the self-cleaning device of the ultraviolet spectral sensor to perform in-situ cleaning according to the cleaning alarm signal, and automatically performs baseline correction and zero-point reset after cleaning to restore online detection.

[0108] Furthermore, the system is also used to implement the following functions:

[0109] The first dynamic threshold, the second dynamic threshold, and the third dynamic threshold are all adaptively adjusted according to the current concentration process index: the first dynamic threshold decreases as the current concentration process index increases; the second dynamic threshold increases as the current concentration process index increases; and the third dynamic threshold remains within a preset constant range throughout the entire concentration process.

[0110] Furthermore, the system is also used to implement the following functions:

[0111] Endpoint confidence evolution verification: When the working condition adaptive fuzzy inference engine outputs the second state confidence exceeding the first dynamic threshold for the first time, it does not immediately trigger a shutdown, but instead starts a verification timing window of a fourth preset duration; within the fourth preset duration, it continuously monitors the changing trend of the second state confidence; if the second state confidence remains monotonically increasing or remains above the first dynamic threshold within the fourth preset duration, it outputs the final endpoint determination signal at the end of the fourth preset duration; if the second state confidence decreases and falls below the first dynamic threshold within the fourth preset duration, it cancels the current endpoint determination and returns to the normal condensation state.

[0112] Furthermore, the system is also used to implement the following functions:

[0113] The fourth preset duration is dynamically adjusted according to the current concentration process index: the fourth preset duration is equal to the product of the basic verification duration and the first coefficient, and the first coefficient increases as the current concentration process index increases.

[0114] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0115] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for determining the endpoint of an extraction and concentration process using ultraviolet light, characterized in that, The method includes: A multimodal sensing array is deployed on the circulation pipeline of the extraction and concentration equipment. The multimodal sensing array includes at least an ultraviolet spectral sensor, a torque sensor, and a conductivity sensor, which are used to collect ultraviolet full spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data of the concentrate in real time, respectively. The ultraviolet full spectrum data, the mechanical impedance data of the stirring shaft, and the liquid phase ion concentration data are synchronized and aligned on the time axis to obtain a multimodal dataset. A working condition adaptive fuzzy inference engine is constructed, which has a built-in complementary verification and fusion module and a dynamic weight allocation module. The complementary verification fusion module determines the confidence interval of each modality data according to the current condensation stage, and inputs the multimodal dataset into the dynamic weight allocation module; The dynamic weight allocation module assigns a first weight coefficient, a second weight coefficient, and a third weight coefficient to the ultraviolet full spectrum data, the mechanical impedance data of the stirring shaft, and the liquid phase ion concentration data respectively based on the current concentration process index. The working condition adaptive fuzzy inference engine, based on the multimodal dataset and the first weight coefficient, the second weight coefficient, and the third weight coefficient, calls a preset fuzzy inference rule set to perform multi-level confidence inference and outputs a triplet state vector, wherein the triplet state vector includes the first state confidence, the second state confidence, and the third state confidence. Based on the triplet state vector, an endpoint determination signal or a cleaning alarm signal is generated and sent to the actuator for instruction execution.

2. The method for determining the endpoint of an extraction and concentration process using ultraviolet light as described in claim 1, characterized in that, The complementary verification fusion module determines the confidence interval of each modality data based on the current condensation stage, including: In the early stage of concentration, the confidence interval of the ultraviolet full spectrum data is the first confidence interval, and the confidence interval of the mechanical impedance data of the stirring shaft is the second confidence interval. The lower limit of the first confidence interval is higher than the upper limit of the second confidence interval. During the middle stage of concentration, there is an overlap between the reliability range of the ultraviolet full spectrum data and the reliability range of the stirring shaft mechanical impedance data; In the later stage of concentration, the confidence interval of the mechanical impedance data of the stirring shaft is the third confidence interval, and the confidence interval of the UV full spectrum data is the fourth confidence interval. The lower limit of the third confidence interval is higher than the upper limit of the fourth confidence interval. The reliability range of the liquid phase ion concentration data remained constant throughout the concentration process.

3. The method for determining the endpoint of an extraction and concentration process using ultraviolet light as described in claim 1, characterized in that, The calculation of the current concentration process index includes: The ratio of the current enriched time to the estimated maximum enriched time is used as the current enrichment progress index. The first weighting coefficient decreases as the current concentration process index increases; The second weighting coefficient increases as the current concentration process index increases; The third weighting coefficient remains within a preset stable range throughout the entire concentration process.

4. The method for determining the endpoint of an extraction and concentration process using ultraviolet light as described in claim 1, characterized in that, The preset fuzzy reasoning rule set includes at least: First fuzzy inference rule: If the ultraviolet spectral stability is high, the torque change rate is positive and medium to high, and the conductivity fluctuation is low, then the confidence level of the first state is high, and the confidence level of the second state is low. The second fuzzy inference rule is: if the ultraviolet spectral stability is high, the torque change rate is zero and low, and the conductivity fluctuation is low, then the confidence level of the first state is low, and the confidence level of the second state is high. The third fuzzy inference rule is: if the ultraviolet spectral stability is medium to low, the torque change rate is arbitrary, and the conductivity fluctuation is high, then the confidence level of the third state is high, and the priority of the confidence level of the third state is higher than that of the confidence level of the first state and the confidence level of the second state. The fourth fuzzy inference rule is: if the ultraviolet spectral stability is low, the torque change rate is negative and medium-high, and the conductivity fluctuation is medium, then the confidence level of the third state is medium-high, and a cooling request signal is output at the same time.

5. The method for determining the endpoint of an extraction and concentration process using ultraviolet light as described in claim 4, characterized in that, The first sequence is constructed by selecting the ultraviolet full spectrum integral values ​​of the most recent first preset number of sampling periods, the coefficient of variation of the first sequence is calculated, and the coefficient of variation is subjected to negative exponential transformation to obtain the ultraviolet spectral stability with a value between zero and one. A weighted linear regression is performed on the torque values ​​of the most recent second preset number of sampling periods, giving higher weight to recent data, and the regression slope is used as the torque change rate. Calculate the variance of the conductivity values ​​for the most recent third preset number of sampling periods, and perform normalization processing to obtain the conductivity fluctuation; The ultraviolet spectral stability, the torque change rate, and the conductivity fluctuation are multiplied by their respective first weighting coefficient, second weighting coefficient, and third weighting coefficient to obtain the weighted input variables.

6. The method for determining the endpoint of an extraction and concentration process using ultraviolet light as described in claim 1, characterized in that, Based on the triplet state vector, an endpoint determination signal or a cleaning alarm signal is generated and sent to the actuator for instruction execution, including: When the confidence level of the second state exceeds the first dynamic threshold and the confidence level of the third state is lower than the second dynamic threshold, an endpoint determination signal is output. When the confidence level of the third state exceeds the third dynamic threshold, regardless of the confidence level of the second state, a cleaning alarm signal is output first. The actuator controls the concentration equipment to stop heating or concentration based on the endpoint determination signal, or triggers the self-cleaning device of the ultraviolet spectral sensor to perform in-situ cleaning based on the cleaning alarm signal, and automatically performs baseline correction and zero-point reset after cleaning to restore online detection.

7. The method for determining the endpoint of an extraction and concentration process using ultraviolet light as described in claim 6, characterized in that, The first dynamic threshold, the second dynamic threshold, and the third dynamic threshold are all adaptively adjusted according to the current concentration process index. The first dynamic threshold decreases as the current concentration process exponentially increases; The second dynamic threshold increases exponentially with the current concentration process. The third dynamic threshold is maintained within a preset constant range throughout the entire concentration process.

8. The method for determining the endpoint of an extraction and concentration process using ultraviolet light as described in claim 7, characterized in that, The method also includes endpoint confidence evolution verification: When the working condition adaptive fuzzy inference engine outputs the second state confidence level for the first time, it does not immediately trigger a shutdown, but instead starts a verification timing window of a fourth preset duration. The trend of change in the confidence level of the second state is continuously monitored within the fourth preset time period; If the confidence level of the second state remains monotonically increasing or remains above the first dynamic threshold within the fourth preset time period, then the final endpoint determination signal is output at the end of the fourth preset time period. If the confidence level of the second state decreases and falls below the first dynamic threshold within the fourth preset time period, the current endpoint determination is cancelled and the state returns to normal concentration.

9. The method for determining the endpoint of an extraction and concentration process using ultraviolet light as described in claim 8, characterized in that, The fourth preset duration is dynamically adjusted according to the current concentration process index: the fourth preset duration is equal to the product of the basic verification duration and the first coefficient, and the first coefficient increases as the current concentration process index increases.

10. A UV online endpoint determination system for an extraction and concentration process, characterized in that, The system is used to execute the ultraviolet online endpoint determination method for an extraction and concentration process as described in any one of claims 1-9, and the system comprises: A sensing deployment unit is used to deploy a multimodal sensing array on the circulation pipeline of the extraction and concentration equipment. The multimodal sensing array includes at least an ultraviolet spectral sensor, a torque sensor, and a conductivity sensor, which are used to collect ultraviolet full spectrum data, stirring shaft mechanical impedance data, and liquid phase ion concentration data of the concentrate in real time, respectively. The synchronization alignment unit is used to synchronize and align the ultraviolet full spectrum data, the mechanical impedance data of the stirring shaft, and the liquid phase ion concentration data on the time axis to obtain a multimodal dataset; An engine building unit is used to build a working condition adaptive fuzzy inference engine, which has a built-in complementary verification and fusion module and a dynamic weight allocation module. A credibility determination unit is used by the complementary verification fusion module to determine the credibility range of each modality data according to the current condensation stage, and input the multimodal dataset into the dynamic weight allocation module; The weighting allocation unit is used by the dynamic weighting allocation module to allocate a first weighting coefficient, a second weighting coefficient, and a third weighting coefficient to the ultraviolet full spectrum data, the mechanical impedance data of the stirring shaft, and the liquid phase ion concentration data according to the current concentration process index. The inference unit is used by the working condition adaptive fuzzy inference engine to call a preset fuzzy inference rule set to perform multi-level confidence inference based on the multimodal dataset and the first weight coefficient, the second weight coefficient and the third weight coefficient, and output a triplet state vector, wherein the triplet state vector includes the first state confidence, the second state confidence and the third state confidence. The instruction execution unit is used to generate an endpoint determination signal or a cleaning alarm signal based on the triplet state vector, and send it to the execution mechanism for instruction execution.

Images