Micro-jet mixing efficiency dynamic calibration method and system based on multi-sensor data fusion

Abstract

The invention provides a micro-jet mixing efficiency dynamic calibration method and system based on multi-sensor data fusion, and the method comprises the steps: obtaining formamide flow data and hydrazine hydrate flow data; acquiring spectral data of the mixed material; calculating the flow ratio of formamide to hydrazine hydrate, and detecting the mixing efficiency according to the flow ratio and the spectral data of the mixed material to obtain a detection result; if the detection result is that mixing efficiency calibration needs to be carried out, triggering data fusion analysis, and obtaining a data fusion analysis result; and positioning a deviation source according to a data fusion analysis result, and performing dynamic calibration. According to the micro-jet mixing efficiency dynamic calibration method and system based on multi-sensor data fusion, through nanoscale micro-channel mixing and spectral flow two-dimensional monitoring, the mixing uniformity and reaction consistency in the triazole production process are improved, in addition, the calibration response time is shortened through a calibration mechanism of data fusion analysis, and the calibration accuracy is improved. The generation of byproducts is reduced, and the production stability is greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of micro-fluid mixing, in particular to a micro-fluid mixing efficiency dynamic calibration method and system based on multi-sensor data fusion. BACKGROUND

[0002] 1,2,4-triazole (triazole) as an important pharmaceutical intermediate, the optimization of its synthesis process is of great significance to the pharmaceutical industry. The traditional triazole production process uses a batch reaction kettle, which generates formamide and hydrazine hydrate by direct reaction at high temperature. The material mixing is uneven, and as the reaction proceeds, the concentration of formamide gradually decreases, resulting in a significant increase in the generation of by-product 4-amino triazole (4-AT), which affects the quality of the product.

[0003] Continuous flow micro-reaction technology is gradually replacing traditional batch reaction process. However, although micro-fluid mixing process can improve mixing efficiency, there are problems such as flow fluctuation and micro-channel blockage during operation. Micro-fluid mixer realizes nanoscale mixing of formamide and hydrazine hydrate through high-speed shearing and cavitation effect, but this process is highly dependent on the stability of fluid dynamics conditions, and existing monitoring means is too single to accurately locate the cause of the decrease in mixing efficiency.

[0004] Therefore, there is an urgent need for a micro-fluid mixing efficiency dynamic calibration method and system based on multi-sensor data fusion to at least solve the above problems. SUMMARY

[0005] One of the purposes of the present application is to provide a micro-fluid mixing efficiency dynamic calibration method and system based on multi-sensor data fusion, which performs mixing efficiency verification according to the obtained material flow data and mixed material spectrum data. If any index is out of standard, the dynamic calibration process is triggered, data fusion analysis is performed, deviation sources are located according to the data fusion analysis results and dynamic calibration is performed. Through nanoscale micro-channel mixing and spectrum flow double-dimensional monitoring, the mixing uniformity and reaction consistency in the production process of triazole are improved. In addition, the calibration mechanism of data fusion analysis shortens the calibration response time, reduces the generation of by-products, and greatly improves the production stability.

[0006] The micro-fluid mixing efficiency dynamic calibration method based on multi-sensor data fusion provided by the embodiments of the present application comprises: Based on the electromagnetic flowmeter of the formamide inlet and the hydrazine hydrate inlet of the micro-fluid mixer, the formamide flow data and the hydrazine hydrate flow data are obtained; Based on the infrared spectrum sensor of the outlet of the micro-fluid mixer, the mixed material spectrum data are obtained; The flow ratio of formamide and hydrazine hydrate is calculated, the mixing efficiency is detected according to the flow ratio and the mixed material spectrum data, and the detection result is obtained; If the detection result is that the mixing efficiency needs to be calibrated, triggering data fusion analysis, obtaining data fusion analysis result; According to the data fusion analysis result, locating the deviation source and performing dynamic calibration.

[0007] Preferably, the flow ratio of formamide and hydrazine hydrate is calculated, and the mixing efficiency is detected according to the flow ratio and the mixture material spectrum data, including: According to the formamide flow data and the hydrazine hydrate flow data, the flow ratio of formamide and hydrazine hydrate is calculated; According to the flow ratio and the preset standard flow ratio, the flow ratio deviation is calculated; The spectral similarity of the mixture material spectrum data and the standard spectrum data is calculated; If the flow ratio deviation is greater than the preset deviation threshold or the spectral similarity is less than the preset spectral similarity threshold, the detection result is that the mixing efficiency needs to be calibrated; Otherwise, the detection result is that the mixing efficiency does not need to be calibrated.

[0008] Preferably, data fusion analysis is triggered, and data fusion analysis result is obtained, including: The temperature data and pressure data of the inlet and outlet of the tubular plug flow reactor are obtained; According to the weighted algorithm, the formamide flow data, the hydrazine hydrate flow data, the mixture material spectrum data, the temperature data and the pressure data are fused and calculated to obtain the data fusion analysis result.

[0009] Preferably, according to the weighted algorithm, the formamide flow data, the hydrazine hydrate flow data, the mixture material spectrum data, the temperature data and the pressure data are fused and calculated to obtain the data fusion analysis result, including: According to the formamide flow data, the hydrazine hydrate flow data, the mixture material spectrum data, the temperature data and the pressure data, a deviation data set is outputted; According to the deviation data set, the preset basic weight set is adjusted to output a dynamic weight set; According to the deviation data set and the dynamic weight set, a deviation contribution set and a comprehensive deviation index are calculated; The comprehensive deviation index and the deviation contribution set are taken as the data fusion analysis result.

[0010] Preferably, according to the data fusion analysis result, the deviation source is located and dynamic calibration is performed, including: According to the data fusion analysis result and the preset deviation mode recognition rule library, the deviation source is determined; According to the calibration rule library corresponding to the deviation source, a calibration rule is matched; According to the calibration rule, corresponding dynamic calibration is performed.

[0011] Preferably, the deviation source is located according to the data fusion analysis result, and dynamic calibration is carried out, and the method further comprises the following steps of: According to the deviation mode identification rule base, the deviation source positioning model is trained. According to the calibration rule base corresponding to the deviation source, the dynamic calibration model is trained. Based on the deviation source positioning model and the dynamic calibration model, according to the data fusion analysis result, corresponding dynamic calibration is carried out.

[0012] Preferably, the preset basic weight set is adjusted according to the deviation data set, and a dynamic weight set is output, comprising the following steps of: According to the deviation data set, the deviation significance index set is determined. According to the basic weight set and the deviation significance index set, the preliminary adjustment weight set is determined. The preliminary adjustment weight set is normalized and the abnormal mode weight compensation is compensated to determine the dynamic weight set.

[0013] Preferably, according to the deviation data set, the deviation significance index set is determined, comprising the following steps of: Based on the triazole synthesis reaction kinetics and the internal flow field distribution characteristics of the microfluidic mixer, a coupling relationship model of byproduct generation amount and multi-parameter deviation is established; Based on the coupling relationship model, the influence level of each deviation type on product quality and safety is identified, and a three-level threshold system is established. According to the boundary value of the warning threshold and the attention threshold, the response surface design is used for optimization. According to the threshold interval of each deviation data, the deviation significance index set is determined.

[0014] Preferably, according to the basic weight set and the deviation significance index set, the preliminary adjustment weight set is determined, comprising the following steps of: According to the microfluidic mixing mechanism, a parameter cross-influence model is established. According to historical operation experience data, the adjustment situation is determined. Based on the parameter cross-influence model, according to the adjustment situation and the historical quality abnormal event analysis result, the weight adjustment rule is determined. Based on the weight adjustment rule, according to the basic weight set and the deviation significance index set, the preliminary adjustment weight set is determined.

[0015] The microfluidic mixing efficiency dynamic calibration system based on multi-sensor data fusion provided by the embodiment of the application comprises: The flow data acquisition module is used for acquiring the formamide flow data and the hydrazine hydrate flow data based on the electromagnetic flow meters of the formamide inlet and the hydrazine hydrate inlet of the microfluidic mixer. The spectrum data acquisition module is used for acquiring the mixed material spectrum data based on the infrared spectrum sensor of the outlet of the microfluidic mixer. The mixing efficiency detection module is configured to calculate a flow ratio of the formamide and the hydrazine hydrate, detect mixing efficiency according to the flow ratio and spectrum data of the mixed material, and obtain a detection result. The fusion analysis module is configured to trigger data fusion analysis and obtain a data fusion analysis result if the detection result indicates that the mixing efficiency calibration is needed. The dynamic calibration module is configured to locate a deviation source according to the data fusion analysis result and perform dynamic calibration.

[0016] The present application has the following advantages: The present application performs mixing efficiency calibration according to the obtained material flow data and spectrum data of the mixed material, triggers a dynamic calibration process if any index exceeds a standard, performs data fusion analysis, locates a deviation source according to the data fusion analysis result, and performs dynamic calibration.

[0017] Other features and advantages of the present application will be further described in the following description, and will become apparent from the description, or will be learned through practice of the present application. The objects and other advantages of the present application will be realized and achieved by particularly pointed out in the specification.

[0018] The technical solutions of the present application will be further described in detail below with reference to the drawings and embodiments. DETAILED DESCRIPTION

[0019] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, which together with the embodiments serve to explain the present application, and do not constitute a limitation to the present application. In the drawings: Figure 1 FIG. 1 is a schematic diagram of a microjet mixing efficiency dynamic calibration method based on multi-sensor data fusion in an embodiment of the present application; Figure 2 FIG. 2 is a schematic diagram of a microjet mixing efficiency dynamic calibration system based on multi-sensor data fusion in an embodiment of the present application. DETAILED DESCRIPTION

[0020] The preferred embodiments of the present application will be described below with reference to the drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and do not limit the present application.

[0021] The present application provides a microjet mixing efficiency dynamic calibration method based on multi-sensor data fusion, as shown in FIG. 1, which comprises the following steps: Figure 1 ​Step 1: Obtain formamide flow data and hydrazine hydrate flow data based on electromagnetic flow meters of the formamide inlet and the hydrazine hydrate inlet of the microfluidic mixer.

[0022] In this embodiment, the microfluidic mixer is designed with a microchannel structure, which can realize nanoscale mixing of formamide preheated to 150-160°C and hydrazine hydrate at room temperature through high-speed jet flow. The electromagnetic flow meters are installed on the material inlet (formamide inlet and hydrazine hydrate inlet) pipelines of the microfluidic mixer. When the formamide and hydrazine hydrate flow through the magnetic field area respectively, the sensor detects the induced electromotive force and converts it into digital signal output, and the analysis of the digital signal output obtains the formamide flow data and hydrazine hydrate flow data.

[0023] Step 2: Obtain mixed material spectrum data based on the infrared spectrum sensor of the outlet of the microfluidic mixer.

[0024] In this embodiment, the mixed material spectrum data is a spectrum signal reflecting the uniformity of the mixing of formamide and hydrazine hydrate. When mixed uniformly, the characteristic peak distribution is stable, and when mixed unevenly, the characteristic peak intensity and position will change.

[0025] Step 3: Calculate the flow ratio of formamide and hydrazine hydrate, conduct mixing efficiency detection according to the flow ratio and the mixed material spectrum data, and obtain the detection result. Step 3 specifically includes: According to the formamide flow data and the hydrazine hydrate flow data, the flow ratio of formamide and hydrazine hydrate is calculated.

[0026] According to the flow ratio and the preset standard flow ratio, the flow ratio deviation is calculated.

[0027] In this embodiment, the preset standard flow ratio is set by the process engineer in advance, for example, 5:1. The flow ratio deviation is the relative deviation of the actual flow ratio and the standard flow ratio. Specifically, the absolute value of the difference between the actual flow ratio and the standard flow ratio is calculated, and the ratio of this absolute value to the standard flow ratio is the flow ratio deviation.

[0028] Calculate the spectral similarity of the mixed material spectrum data and the standard spectrum data.

[0029] In this embodiment, the spectral similarity is the similarity degree of the mixed material spectrum data and the standard uniform mixing spectrum calculated by principal component analysis or cosine similarity, which reflects the mixing uniformity.

[0030] If the flow ratio deviation is greater than the preset deviation threshold or the spectral similarity is less than the preset spectral similarity threshold, the detection result is that the mixing efficiency needs to be calibrated.

[0031] In this embodiment, the preset deviation threshold is 0.5%, and the preset spectral similarity threshold is 98%.

[0032] Otherwise, the detection result is that the mixing efficiency calibration is not needed.

[0033] Step 4: If the detection result is that the mixing efficiency calibration is needed, triggering the data fusion analysis to obtain the data fusion analysis result. Specifically, step 4 includes: Obtaining the temperature data and pressure data of the inlet and outlet of the tubular plug flow reactor.

[0034] In this embodiment, the tubular plug flow reactor is a reaction device after micro-jet mixing, and the temperature data and pressure data are obtained through temperature / pressure sensors installed at the inlet and outlet of the tubular plug flow reactor.

[0035] According to the formamide flow data, hydrazine hydrate flow data, mixed material spectrum data, temperature data and pressure data, output the deviation data set.

[0036] In this embodiment, the deviation data set is a standardized data set containing flow ratio deviation, spectrum deviation, temperature deviation and pressure deviation.

[0037] According to the deviation data set, adjusting the preset basic weight set to output the dynamic weight set.

[0038] In this embodiment, the basic weight set is a set containing the preset basic weight of each deviation data type, and the preset basic weight of the deviation data type reflects the basic importance of the parameter in the mixing efficiency evaluation, which is set by the process engineer in advance. The dynamic weight set is a set of real-time weights obtained by dynamically adjusting the weight of the deviation data type according to the actual deviation situation, reflecting the real-time importance of the parameter in the mixing efficiency evaluation.

[0039] According to the deviation data set and the dynamic weight set, calculating the deviation contribution set and the comprehensive deviation index.

[0040] In this embodiment, the deviation contribution set is a set of results (deviation contributions) obtained by multiplying the deviation data and the corresponding dynamic weight. The comprehensive deviation index is the sum of the deviation contributions.

[0041] Taking the comprehensive deviation index and the deviation contribution set as the data fusion analysis result.

[0042] Step 5: According to the data fusion analysis result, locating the deviation source and performing dynamic calibration.

[0043] Specifically, step 5 includes: According to the data fusion analysis result and the preset deviation mode recognition rule library, determining the deviation source.

[0044] In this embodiment, the deviation mode recognition rule base is a database constructed based on the correspondence between multi-parameter deviation modes and problem sources, for example, if the flow deviation contribution is > 60% and the spectrum similarity is good, the corresponding problem source is the flow control system problem; if the spectrum deviation contribution is > 70% and the pressure difference is abnormally increased, the corresponding problem source is the mixer channel blockage. The deviation mode recognition rule base is constructed: the abnormal events are collected through historical data, and the deviation source types are labeled by process experts and summarized into the database.

[0045] The calibration rule corresponding to the deviation source is matched according to the calibration rule base.

[0046] In this embodiment, the calibration rule base is a database of preset calibration strategies corresponding to different problems, which is constructed by manually screening historical calibration strategies according to historical calibration data and corresponding calibration effects.

[0047] According to the calibration rule, corresponding dynamic calibration is performed.

[0048] In this embodiment, according to the calibration rule, corresponding dynamic calibration is performed: the equipment parameters are automatically adjusted according to the matched calibration rule (calibration strategy) to restore the mixing efficiency, for example, the metering pump speed is adjusted when the flow deviation occurs, and the mixer channel collision angle is fine-tuned when the mixing is uneven.

[0049] Step 5 further comprises: According to the deviation mode recognition rule base, a deviation source positioning model is trained.

[0050] In this embodiment, the deviation source positioning model adopts a random forest algorithm, the input is 4-dimensional deviation data, the output is 4 types of deviation sources, and the training data is the data in the deviation mode recognition rule base.

[0051] According to the calibration rule corresponding to the deviation source, a dynamic calibration model is trained.

[0052] In this embodiment, the dynamic calibration model adopts a Q-learning reinforcement learning algorithm, the state is the deviation source and the severity, the action is the adjustment strategy, and the reward function is the degree of decline of the historical comprehensive deviation index after adjustment.

[0053] Based on the deviation source positioning model and the dynamic calibration model, corresponding dynamic calibration is performed according to the data fusion analysis result.

[0054] In this embodiment, the data fusion analysis result is first input into the deviation source positioning model, the input port of the dynamic calibration model is determined according to the output of the deviation source positioning model, then the data fusion analysis result and the deviation source positioning label are input into the input port of the corresponding dynamic calibration model, and the dynamic calibration model adaptively executes the adjustment strategy.

[0055] The working principle and beneficial effects of the above technical solutions are: The application carries out mixing efficiency verification according to the obtained material flow data and mixture spectrum data, triggers a dynamic calibration process if any index exceeds the standard, carries out data fusion analysis, locates the deviation source according to the data fusion analysis result and carries out dynamic calibration, improves the mixing uniformity and reaction consistency in the production process of triazole through nanoscale micro-channel mixing and spectrum flow double-dimensional monitoring, in addition, the calibration mechanism of data fusion analysis shortens the calibration response time, reduces the generation of by-products, and greatly improves the production stability.

[0056] In one embodiment, the preset basic weight set is adjusted according to the deviation data set, and a dynamic weight set is output, including: According to the deviation data set , a deviation significance index set is determined ; Wherein, is the flow deviation data, is the spectrum deviation data, is the temperature deviation data, is the pressure deviation data; is the significance index of flow deviation, is the significance index of spectrum deviation, is the significance index of temperature deviation, is the significance index of pressure deviation.

[0057] In this embodiment, the deviation significance index is the discrete significance score converted from each parameter deviation, which is used to judge the severity of the deviation and determine the adjustment range of the subsequent weight. Specifically, according to the deviation data set, a deviation significance index set is determined, including: Step A: Based on the kinetics of triazole synthesis reaction and the internal flow field distribution characteristics of micro-fluidic mixer, a coupling relationship model of by-product generation amount and multi-parameter deviation is established: ; Wherein, is the mass percentage content of by-products in triazole product, is the basic by-product content under ideal working conditions (experimental measured value); When , it reflects the contribution of flow ratio deviation to by-products; , it reflects the contribution of spectrum deviation (mixing unevenness) to by-products; , it reflects the influence of temperature deviation on reaction rate and selectivity; , it reflects the influence of pressure deviation (usually indicating blockage) on residence time distribution; , reflecting the cross-coupling effect between parameters (for example: flow deviation and temperature deviation will exacerbate local overheating; temperature deviation and pressure deviation indicate serious plugging); Model coefficients , , , and Determined by: using orthogonal experimental design, carrying out multi-level parameter perturbation experiment on pilot plant, collecting different working condition data (covering normal, slight deviation, serious deviation three states), using multivariate nonlinear regression fitting model parameters.

[0058] Step B: Based on the coupling relationship model, identify the influence level of each deviation type on product quality and safety, and establish a three-level threshold system; In this embodiment, the coupling relationship model is combined to carry out hazard and operability analysis, and a three-level threshold system is established: Safety threshold: flow deviation (corresponding to ); spectral similarity , temperature deviation (corresponding to tower kettle temperature 115℃ or 105℃, triggering automatic cut-off); pressure deviation 0.5MPa (corresponding to serious plugging risk); safety threshold cannot be broken; Warning threshold: flow deviation (corresponding to ); spectral similarity ; temperature deviation ; 0.3MPa pressure deviation 0.5Mpa; parameters reaching the warning threshold need to be calibrated immediately; Attention threshold: flow deviation (corresponding to ); spectral similarity ; temperature deviation ; 0.1MPa pressure deviation 0.3Mpa; parameters reaching the attention threshold need to be continuously detected.

[0059] Step C: According to the boundary values of the warning threshold and the attention threshold, response surface design is used for optimization; In this embodiment, for the boundary values of the warning threshold and the attention threshold, Box-Behnken response surface design is used for optimization. The optimization goal is to minimize the false positive rate under the premise of ensuring ; the constraint condition is that the false negative rate 2%; the optimization variable is the lower limit of the alert threshold of the four parameters; the response index is the calibration trigger frequency, the quality compliance rate after calibration, and the continuous production duration. The optimal threshold combination is determined through response surface analysis, for example: the flow deviation alert lower limit is optimized from 0.5 to 0.48, and the spectral similarity alert lower limit is optimized from 0.95 to 0.96.

[0060] Step D: Determine the deviation significance index set according to the threshold interval in which each deviation data is located.

[0061] In this embodiment, when the deviation data is in the normal interval (deviation lower limit of the attention threshold), the deviation significance index is 0; when the deviation data is in the attention interval (attention threshold lower limit deviation alert threshold lower limit), the deviation significance index is 0.3; when the deviation data is in the alert interval (alert threshold lower limit deviation safety threshold lower limit), the deviation significance index is 0.6; when the deviation data is in the danger interval (deviation safety threshold lower limit), the deviation significance index is 1.

[0062] When the system accumulates more than a preset batch (for example, 1000 batches) of historical data, a machine learning method can be used instead of steps A-D, which includes: Construct a multi-modal feature data set: In addition to the four deviation parameters, extract time series features (mean, variance, and change rate in a 1-minute sliding window), frequency domain features (wavelet transform coefficients of spectral data), and operating condition features (raw material batch, ambient temperature, and equipment running time), and construct a feature vector; Train a gradient boosting decision tree model: includes two related tasks, task 1 is a regression task to predict 4-AT content, and task 2 is a classification task to identify the dominant deviation source, sharing the underlying feature representation through multi-task learning; Use SHAP values to dynamically extract significance indexes: for each real-time sample, calculate the SHAP value of each parameter feature (representing the marginal contribution of the feature to the prediction result), and obtain the significance index after normalization; Online adaptive update: after completing each calibration cycle, update the model parameters through reinforcement learning according to the calibration effect (quality improvement degree), and dynamically adjust the threshold boundary according to the recent false positive rate and false negative rate.

[0063] According to the basic weight set and the deviation significance index set, determine the preliminary adjustment weight set ; wherein, is the basic weight of the flow deviation, is the basic weight of the spectral deviation, is the base weight for temperature deviation, is the base weight for pressure deviation; is the preliminary adjustment weight for flow deviation, is the preliminary adjustment weight for spectrum deviation, is the preliminary adjustment weight for temperature deviation, is the preliminary adjustment weight for pressure deviation.

[0064] In this embodiment, the base weight is adaptively adjusted according to the deviation significance index, the decision weight of the abnormal parameter is enhanced, and the preliminary adjustment weight is obtained. The adjustment rule is, for example: Base enhancement: When the flow deviation is significant, increase its weight; Cross-inhibition: When the spectrum is also abnormal, reduce the flow weight; Joint enhancement: when temperature and pressure are simultaneously abnormal ( And ), increase the synergistic weight of the two, reflecting the working condition characteristics that "temperature and pressure are simultaneously abnormal indicating system blockage", and the specific adjustment is: ; ; When determining the adjustment rule, according to the microfluidic mixing mechanism, a parameter cross-influence model is established: ; wherein, is the mass percentage content of by-products in the triazole product in the triazole synthesis process. According to historical operation experience data, the adjustment situation is determined, for example, when the inlet and outlet pressure difference is greater than 0.3 Mpa, the temperature deviation is greater than 0.3, indicating that the system may be blocked, the weights of temperature and pressure need to be increased.

[0065] The determination method of each adjustment coefficient is: The base enhancement coefficient 0.18 is determined by statistical analysis of historical batch single-parameter abnormal event; the cross-inhibition coefficient 0.05 is determined according to the proportion of coupling terms; the synergistic coefficient 0.12 in joint enhancement is determined by analyzing historical blockage events; The decision tree algorithm is used to extract the weight adjustment rule from the historical data, which is stored in the control system rule library. The input features of the decision tree are 8-dimensional vectors , the output label is the optimal weight adjustment amount (determined by post-analysis: the adjustment amount after which the comprehensive deviation index has the highest correlation with the actual quality deviation), the training data is 500 batches of historical production data, and the decision tree parameter setting is: maximum depth 10, minimum leaf node sample number 20.

[0066] The preliminary adjustment weight set is normalized and abnormal mode weight compensation is performed to determine the dynamic weight set.

[0067] In this embodiment, the abnormal mode weight compensation: an abnormal mode rule base is constructed, and a compensation vector is predefined, the Rete algorithm is used to match the current working condition with the predefined mode, and the weight distribution strategy is adjusted for specific risk working conditions.

[0068] The working principle and beneficial effects of the above technical solution are: The quantitative relationship model is established based on historical experimental data and process mechanism, the flow, spectrum, temperature and pressure deviation data collected by each sensor are converted into discrete significance indicators, and these indicators accurately reflect the influence degree of the deviation on the product quality (especially the content of 4-AT byproduct). According to the internal correlation characteristics of the chemical process, the adaptive weight adjustment rule is designed, when a parameter is significantly abnormal, the weight is enhanced, and when multiple parameters are associated with abnormalities, the synergistic enhancement strategy is implemented, thereby improving the diagnostic accuracy. Finally, through normalization processing and abnormal mode matching based on the Rete algorithm, the weight compensation is performed, so that the final weight set can reflect the characteristics of the current working condition and meet the safety operation specification. The present application can automatically optimize the parameter weight distribution according to different working conditions, improve the accuracy and response speed of the mixing efficiency diagnosis. Through accurate control of the mixing process, the product batch difference and safety risk caused by uneven mixing in the traditional process are avoided, and the production stability of the triazole is improved.

[0069] The embodiment of the present application provides a micro-fluidic mixing efficiency dynamic calibration system based on multi-sensor data fusion, as shown in the figure, which comprises: Figure 2 The flow data acquisition module 1 is used for acquiring the methanamide flow data and the hydrazine hydrate flow data based on the electromagnetic flow meters of the methanamide inlet and the hydrazine hydrate inlet of the micro-fluidic mixer; The spectrum data acquisition module 2 is used for acquiring the mixed material spectrum data based on the infrared spectrum sensor of the outlet of the micro-fluidic mixer; The mixing efficiency detection module 3 is used for calculating the flow ratio of the methanamide and the hydrazine hydrate, performing mixing efficiency detection according to the flow ratio and the mixed material spectrum data, and acquiring a detection result; The fusion analysis module 4 is used for triggering data fusion analysis and acquiring a data fusion analysis result if the detection result needs to be calibrated for the mixing efficiency; The dynamic calibration module 5 is used for positioning a deviation source according to the data fusion analysis result and performing dynamic calibration; The mixing efficiency detection module 3 performs the following operations: Temperature data and pressure data of the inlet and outlet of the tubular plug flow reactor are acquired; ​Output a deviation data set according to the formamide flow data, the hydrazine hydrate flow data, the mixture material spectrum data, the temperature data and the pressure data; Adjust a preset basic weight set according to the deviation data set, and output a dynamic weight set; Calculate a deviation contribution set and a comprehensive deviation index according to the deviation data set and the dynamic weight set; Take the comprehensive deviation index and the deviation contribution set as a data fusion analysis result; The method further comprises: Determine a deviation significance index set according to the deviation data set; Determine a preliminary adjustment weight set according to the basic weight set and the deviation significance index set; Carry out normalization processing and abnormal mode weight compensation on the preliminary adjustment weight set, and determine the dynamic weight set; The method further comprises: Based on the triazole synthesis reaction kinetics and the internal flow field distribution characteristics of the microfluidic mixer, a coupling relationship model of byproduct generation amount and multi-parameter deviation is established; Based on the coupling relationship model, the influence level of each deviation type on product quality and safety is identified, and a three-level threshold system is established; According to the boundary values of the warning threshold and the attention threshold, response surface design is adopted for optimization; According to the threshold interval in which each deviation data is located, the deviation significance index set is determined; The method further comprises: According to the microfluidic mixing mechanism, a parameter cross-influence model is established; According to historical operation experience data, an adjustment situation is determined; Based on the parameter cross-influence model, according to the adjustment situation and historical quality abnormal event analysis results, a weight adjustment rule is determined; Based on the weight adjustment rule, the preliminary adjustment weight set is determined according to the basic weight set and the deviation significance index set.

[0070] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and variations.

Claims

1. A dynamic calibration method for microjets mixing efficiency based on multi-sensor data fusion, characterized in that, include: Electromagnetic flow meters based on the formamide inlet and hydrazine hydrate inlet of a microjet mixer are used to acquire formamide flow data and hydrazine hydrate flow data. Based on the infrared spectral sensor at the outlet of the microjet mixer, the spectral data of the mixed materials are acquired; Calculate the flow ratio of formamide and hydrazine hydrate, and perform mixing efficiency testing based on the flow ratio and spectral data of the mixture to obtain the test results; If the test result indicates that mixing efficiency calibration is required, trigger data fusion analysis and obtain the data fusion analysis results; Based on the data fusion analysis results, the source of deviation is located and dynamic calibration is performed.

2. The method for dynamic calibration of microjets mixing efficiency based on multi-sensor data fusion as described in claim 1, characterized in that, Calculate the flow ratio of formamide and hydrazine hydrate, and perform mixing efficiency testing based on the flow ratio and spectral data of the mixture, including: Calculate the flow ratio of formamide and hydrazine hydrate based on the flow data of formamide and hydrazine hydrate. Calculate the flow ratio deviation based on the flow ratio and the preset standard flow ratio; Calculate the spectral similarity between the spectral data of the mixture and the standard spectral data; If the flow ratio deviation is greater than the preset deviation threshold or the spectral similarity is less than the preset spectral similarity threshold, the detection result indicates that mixing efficiency calibration is required. Otherwise, the test result indicates that no mixing efficiency calibration is required.

3. The method for dynamic calibration of microjets mixing efficiency based on multi-sensor data fusion as described in claim 2, characterized in that, Trigger data fusion analysis and obtain the results, including: Acquire temperature and pressure data at the inlet and outlet of the tubular plug flow reactor; The formamide flow rate data, hydrazine hydrate flow rate data, mixture spectral data, temperature data, and pressure data are fused and calculated using a weighted algorithm to obtain the data fusion analysis results.

4. The method for dynamic calibration of microjets mixing efficiency based on multi-sensor data fusion as described in claim 3, characterized in that, The formamide flow rate data, hydrazine hydrate flow rate data, mixture spectral data, temperature data, and pressure data are fused using a weighted algorithm to obtain the data fusion analysis results, including: Based on formamide flow data, hydrazine hydrate flow data, mixture spectral data, temperature data, and pressure data, output a deviation dataset; Adjust the preset base weight set based on the deviation dataset, and output a dynamic weight set; Based on the deviation dataset and dynamic weight set, calculate the deviation contribution set and the comprehensive deviation index; The comprehensive deviation index and deviation contribution set are used as the results of data fusion analysis.

5. The method for dynamic calibration of microjets mixing efficiency based on multi-sensor data fusion as described in claim 1, characterized in that, Based on the data fusion analysis results, the source of deviation is located and dynamic calibration is performed, including: Based on the data fusion analysis results and the pre-set deviation pattern recognition rule base, the source of deviation is determined; Match calibration rules according to the calibration rule library corresponding to the deviation source; Perform appropriate dynamic calibration according to the calibration rules.

6. The method for dynamic calibration of microjets mixing efficiency based on multi-sensor data fusion as described in claim 1, characterized in that, Based on the data fusion analysis results, the source of deviation is located and dynamic calibration is performed, which also includes: Based on the deviation pattern recognition rule base, train the deviation source localization model; A dynamic calibration model is trained based on the calibration rule library corresponding to the deviation sources. Based on the deviation source localization model and the dynamic calibration model, corresponding dynamic calibration is performed according to the data fusion analysis results.

7. The method for dynamic calibration of microjets mixing efficiency based on multi-sensor data fusion as described in claim 4, characterized in that, Adjust the preset base weight set based on the deviation dataset, and output a dynamic weight set, including: Based on the deviation dataset, determine the set of deviation significance indicators; Based on the basic weight set and the deviation significance index set, determine the initial adjustment weight set; The initial weight set is normalized and abnormal mode weights are compensated to determine the dynamic weight set.

8. The method for dynamic calibration of microjets mixing efficiency based on multi-sensor data fusion as described in claim 7, characterized in that, Based on the deviation dataset, determine the set of deviation significance indicators, including: Based on the kinetics of triazole synthesis reaction and the internal flow field distribution characteristics of the microjet mixer, a coupling relationship model between by-product generation and multi-parameter deviation is established. Based on the coupling relationship model, the impact level of each deviation type on product quality and safety is identified, and a three-level threshold system is established. Based on the boundary value between the warning threshold and the attention threshold, response surface methodology is used for optimization. Based on the threshold range in which each deviation data falls, a set of deviation significance indicators is determined.

9. The method for dynamic calibration of microjets mixing efficiency based on multi-sensor data fusion as described in claim 7, characterized in that, Based on the basic weight set and the set of deviation significance indicators, a preliminary adjustment weight set is determined, including: Based on the microjets mixing mechanism, a parameter cross-influence model is established; Based on historical operational experience data, determine the adjustment scenarios; Based on the parameter cross-influence model, the weight adjustment rules are determined according to the adjustment scenarios and the analysis results of historical quality anomaly events; Based on the weight adjustment rules, a preliminary adjustment weight set is determined according to the basic weight set and the deviation significance index set.

10. A dynamic calibration system for microjets mixing efficiency based on multi-sensor data fusion, characterized in that, include: The flow data acquisition module is used to acquire formamide flow data and hydrazine flow data based on the electromagnetic flow meter at the formamide inlet and hydrazine hydrate inlet of the microjet mixer. The spectral data acquisition module is used to acquire spectral data of the mixed materials based on the infrared spectral sensor at the outlet of the microjet mixer; The mixing efficiency detection module is used to calculate the flow ratio of formamide and hydrazine hydrate, and to detect the mixing efficiency based on the flow ratio and the spectral data of the mixture to obtain the detection results; The fusion analysis module is used to trigger data fusion analysis and obtain the data fusion analysis results if the detection result indicates that mixing efficiency calibration is required. The dynamic calibration module is used to locate the source of deviation based on the data fusion analysis results and perform dynamic calibration.

Images