Variable diameter propeller blade type reaction kettle self-cleaning stirring and uniform mixing system

Abstract

The present application relates to the technical field of reaction kettle stirring, and discloses a variable-diameter propeller blade type reaction kettle self-cleaning stirring and uniform mixing system.The system comprises a variable-diameter blade parameter acquisition module, a historical stirring data classification module, a stirring performance sequence construction module, a mixing trend analysis module, a self-cleaning parameter correction module and a mixing strategy decision module.The system acquires the current working condition information of the target blade, combines the historical stirring data, classifies the historical working conditions and analyzes the stirring performance, constructs a historical stirring performance sequence to obtain the mixing efficiency and the cleaning efficiency.The efficiency parameters are corrected by matching the deviation of the target working condition and the historical working condition, and then the mixing strategy scheme is decided.This system can dynamically adapt to different working conditions, improve the uniformity of the reaction kettle stirring and the self-cleaning effect, and is suitable for the reaction kettle stirring scene in the fields of chemical industry, pharmaceuticals and the like.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of reaction kettle stirring, in particular to a variable-diameter propeller blade type reaction kettle self-cleaning stirring and uniform mixing system. BACKGROUND

[0002] In the industrial production fields of chemical industry, medicine, food, etc., the reaction kettle, as the core equipment for material mixing and reaction, its stirring effect directly affects the product quality and production efficiency. The traditional reaction kettle stirring system mostly adopts fixed size blade structure, which is difficult to balance the stirring uniformity and self-cleaning ability when facing materials with different viscosity, density and solid-liquid ratio. When dealing with high viscosity materials, fixed blades are prone to problems such as material adhesion to the kettle wall and stirring dead angle, which leads to insufficient mixing of local materials and affects the reaction process. At the same time, after long-term use, the residual materials on the kettle wall will form scale, which not only reduces the heat transfer efficiency, but also may cause cross contamination, increasing the equipment cleaning cost and downtime.

[0003] The control mode of the existing stirring system mostly depends on preset parameters, lacking dynamic adjustment ability to real-time working conditions. For example, when the material properties change with reaction time, fixed stirring speed, blade angle and other parameters cannot be adapted in time, leading to decreased mixing efficiency. Although some systems introduce simple feedback regulation mechanism, they can only adjust based on single working condition parameter, and are difficult to comprehensively consider the mixing rules in historical operation data, and cannot form accurate stirring strategy. In addition, the traditional self-cleaning design mostly realizes through increasing mechanical structures such as scrapers, which not only increases the complexity and energy consumption of the equipment, but also may reduce the stability of the system due to mechanical wear, making it difficult to meet the needs of continuous production.

[0004] With the increasing requirements of industrial production for automation and intelligentization, the balance problem between mixing uniformity, self-cleaning effect and energy consumption control of the traditional stirring system is increasingly prominent. The working condition difference of different batches of materials and the dynamic change of material state in the reaction process make it difficult to maintain optimal performance by relying on experience setting of stirring parameters, and a stirring and self-cleaning integrated system that can combine historical data and real-time working conditions to realize self-adaptive adjustment is needed. SUMMARY

[0005] The purpose of the present application is to provide a variable-diameter propeller blade type reaction kettle self-cleaning stirring and uniform mixing system to solve the problems raised in the background.

[0006] To achieve the above purpose, the present application provides a variable-diameter propeller blade type reaction kettle self-cleaning stirring and uniform mixing system, which comprises:

[0007] A variable-diameter blade parameter acquisition module is configured to acquire a target blade currently operated by the reactor and collect target working condition information of the target blade currently operated by the reactor, and index historical stirring data of the target blade in a historical time of the reactor according to the target blade;

[0008] A historical stirring data classification module is configured to classify historical working condition information in the historical stirring data, obtain a plurality of classified historical working conditions, extract a historical mixing information set and a plurality of historical stirring data sets under the plurality of classified historical working conditions, and process to obtain a plurality of historical stirring performance sets;

[0009] A stirring performance sequence construction module is configured to respectively optimize and reduce dimensions of the plurality of historical stirring performance sets, obtain a plurality of reduced-dimension historical stirring performance sets, and arrange in time sequence to obtain a plurality of historical stirring performance sequences;

[0010] A mixing trend analysis module is configured to perform mixing trend analysis according to the plurality of historical stirring performance sequences to obtain mixing efficiency and cleaning efficiency;

[0011] A self-cleaning parameter correction module is configured to match the target working condition information with the plurality of classified historical working conditions to obtain a matching historical working condition, and correct the mixing efficiency and the cleaning efficiency according to a deviation of the target working condition information from standard matching working condition information of the matching historical working condition to obtain corrected mixing efficiency and corrected cleaning efficiency;

[0012] A mixing strategy decision module is configured to perform mixing strategy decision according to the corrected mixing efficiency and the corrected cleaning efficiency to obtain a mixing strategy scheme, and perform mixing strategy execution.

[0013] Preferably, the target blade currently operated by the reactor is acquired, and the target working condition information of the target blade currently operated by the reactor is collected, and the historical stirring data of the target blade in a historical time of the reactor is indexed according to the target blade, including:

[0014] The target blade currently operated by the reactor is acquired, and the upper working condition parameter and the lower working condition parameter of the target blade are collected as the target working condition information;

[0015] The historical stirring data of the target blade is obtained by indexing in the historical operation record of the reactor according to the target blade.

[0016] Preferably, the historical working condition information in the historical stirring data is classified to obtain a plurality of classified historical working conditions, the historical mixing information set and the plurality of historical stirring data sets under the plurality of classified historical working conditions are extracted, and the plurality of historical stirring performance sets are processed, including:

[0017] Obtaining a plurality of historical working condition information in the historical stirring data, performing classification processing, and obtaining a plurality of classified historical working conditions;

[0018] Extracting a standard mixing state of a target blade in the historical stirring data under the plurality of classified historical working conditions, obtaining a historical mixing information set, and extracting a plurality of historical mixing state information sets and a plurality of historical stirring execution time sets generated by stirring of the reaction kettle under the plurality of classified historical working conditions;

[0019] According to the deviation amplitude of the plurality of historical mixing state information sets and the historical mixing information set, a plurality of historical basic performance sets are classified and obtained;

[0020] According to the preset execution time threshold and the ratio of the plurality of historical stirring execution time sets, the plurality of historical basic performance sets are corrected and calculated to obtain a plurality of historical stirring performance sets.

[0021] Preferably, the plurality of historical stirring performance sets are respectively optimized and reduced in dimension to obtain a plurality of reduced dimension historical stirring performance sets, and are arranged in time sequence to obtain a plurality of historical stirring performance sequences, including:

[0022] In the first historical stirring performance set in the plurality of historical stirring performance sets, a first reference performance value is selected and obtained;

[0023] According to the difference between the first reference performance value and other historical stirring performance in the first historical stirring performance set, a distribution probability is assigned to obtain a first basic probability distribution, wherein the size of the difference and the size of the distribution probability are negatively correlated;

[0024] According to the first basic probability distribution, the first historical stirring performance set is optimized and reduced in dimension to obtain a first reduced dimension performance set;

[0025] According to the time stamp information of the plurality of historical stirring performance in the first reduced dimension historical stirring performance set, the first historical stirring performance sequence is sorted and obtained;

[0026] The plurality of historical stirring performance sets are optimized and reduced in dimension and arranged in time sequence to obtain a plurality of historical stirring performance sequences.

[0027] Preferably, according to the first basic probability distribution, the first historical stirring performance set is optimized and reduced in dimension to obtain a first reduced dimension performance set, including:

[0028] A preset number of historical stirring performances are randomly extracted in the first historical stirring performance set to obtain a first reduced dimension historical performance set;

[0029] According to the difference between the historical stirring performance in the first reduced dimension historical performance set and the first benchmark performance value, a distribution probability is assigned to obtain a first reduced dimension probability distribution;

[0030] The similarity between the first reduced dimension probability distribution and the first basic probability distribution is calculated as a first reduced dimension fitness;

[0031] The first historical stirring performance set is randomly extracted again to obtain a second reduced dimension historical performance set, and a second reduced dimension fitness is obtained by processing;

[0032] The optimization of the reduced dimension continues until convergence, and the reduced dimension historical performance set with the maximum reduced dimension fitness is output as the first reduced dimension performance set.

[0033] Preferably, a mixing trend analysis is performed on the plurality of historical stirring performance sequences to obtain mixing efficiency and cleaning efficiency, including:

[0034] According to the sample operation data of the plurality of reaction kettles, a sample stirring performance sequence set is collected, and a sample mixing efficiency set and a sample cleaning efficiency set are obtained according to the performance change identification in each sample stirring performance sequence;

[0035] The sample stirring performance sequence set is used as the classification input, and the sample mixing efficiency set and the sample cleaning efficiency set are used as the classification output to construct a mixing trend analyzer;

[0036] Based on the mixing trend analyzer, the mixing trend of the plurality of historical stirring performance sequences is classified to obtain a plurality of working condition mixing efficiencies and a plurality of working condition cleaning efficiencies;

[0037] The similarity between the target working condition information and the plurality of classified historical working conditions is analyzed, and the plurality of working condition mixing efficiencies and the plurality of working condition cleaning efficiencies are weighted calculated according to the size of the plurality of working condition similarities to obtain the mixing efficiency and the cleaning efficiency.

[0038] Preferably, the target working condition information is matched with the plurality of classified historical working conditions to obtain a matching historical working condition, and the mixing efficiency and the cleaning efficiency are corrected according to the deviation of the target working condition information and the standard matching working condition information of the matching historical working condition to obtain a corrected mixing efficiency and a corrected cleaning efficiency, including:

[0039] The classified historical working condition with the largest similarity is selected as the matching historical working condition, and the standard matching working condition information of the matching historical working condition is obtained;

[0040] According to the deviation of the target working condition information and the standard matching working condition information of the matching historical working condition, an efficiency correction coefficient is set;

[0041] The efficiency correction coefficient is used to correct and calculate the mixing efficiency and the cleaning efficiency, so as to obtain a corrected mixing efficiency and a corrected cleaning efficiency.

[0042] Preferably, according to the corrected mixing efficiency and the corrected cleaning efficiency, a mixing strategy decision is made to obtain a mixing strategy scheme, and a mixing strategy execution is performed, including:

[0043] A sample corrected mixing efficiency set and a sample corrected cleaning efficiency set are collected, and according to the size of each sample corrected mixing efficiency and sample corrected cleaning efficiency, a sample mixing strategy scheme is set to obtain a sample mixing strategy scheme set, wherein each sample mixing strategy scheme includes a strategy parameter, and the size of the sample corrected mixing efficiency and the sample corrected cleaning efficiency is negatively correlated with the size of the strategy parameter.

[0044] The sample corrected mixing efficiency set and the sample corrected cleaning efficiency set are used as decision inputs, and the sample mixing strategy scheme set is used as decision output, so as to construct a mixing strategy decision maker.

[0045] The mixing strategy decision maker is used to make a mixing strategy decision on the corrected mixing efficiency and the corrected cleaning efficiency, so as to obtain a mixing strategy scheme.

[0046] Preferably, a plurality of historical working condition information in the historical stirring data is obtained, and a classification processing is performed to obtain a plurality of classified historical working conditions, including:

[0047] The working condition characteristic parameters in the plurality of historical working condition information are identified, including a paddle speed, a material viscosity and a reaction temperature.

[0048] According to a similarity threshold of the working condition characteristic parameters, the plurality of historical working condition information is clustered to obtain a plurality of classified historical working conditions, wherein each classified historical working condition contains at least three similar working condition characteristic parameter combinations.

[0049] Preferably, a preset dimension reduction number of historical stirring performances is randomly extracted from the first historical stirring performance set to obtain a first dimension reduction historical performance set, including:

[0050] The preset dimension reduction number is set to one third of the original set number.

[0051] A random number generator is used to generate a corresponding number of random numbers in the index range of the first historical stirring performance set, and the historical stirring performances corresponding to the indexes are extracted to form the first dimension reduction historical performance set.

[0052] Compared with the prior art, the present application has the following advantages:

[0053] By integrating historical stirring data and real-time operating information, a dynamic adaptive stirring control mechanism is constructed, effectively solving the deficiencies of traditional reaction kettle stirring systems in mixing uniformity and self-cleaning ability. The core lies in obtaining the real-time operating conditions of the target paddle using the variable-diameter paddle parameter acquisition module, and combining the historical data classification module to deeply mine the past operating data. Through the construction of stirring performance sequence and mixing trend analysis, the mixing law and cleaning effect under different operating conditions can be accurately captured.

[0054] Compared with the traditional fixed parameter control stirring system, this system breaks through the limitations of single operating parameter adjustment through classification and analysis of historical stirring data, and realizes dynamic evaluation of mixing efficiency and cleaning efficiency. The self-cleaning parameter correction module corrects the parameters based on the deviation between the target operating condition and the historical operating condition, so that the stirring strategy can be adjusted in real time with the changes of material properties and reaction stages, avoiding the problems of insufficient mixing or incomplete cleaning caused by operating condition fluctuations. When dealing with high-viscosity and easily-attached materials, the system can reduce the residual material on the kettle wall and reduce the frequency of manual cleaning, while ensuring that the material can maintain good mixing state in different reaction stages.

[0055] This system does not need to rely on additional mechanical cleaning structure, and realizes self-cleaning function through intelligent adjustment of paddle parameters, simplifies the equipment structure, reduces the risk of mechanical wear and tear, and prolongs the service life of the equipment. In multi-batch continuous production, the system can continuously optimize the stirring strategy based on historical data, adapt to the characteristic differences of different batches of materials, reduce production fluctuations caused by improper parameter adjustment, and improve the stability and continuity of the production process. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 The working principle diagram of the variable-diameter propeller blade type reaction kettle self-cleaning stirring and uniform mixing system described in the present application;

[0057] Figure 2 The flowchart for historical stirring data classification and performance set generation;

[0058] Figure 3 The flowchart for historical stirring performance sequence construction;

[0059] Figure 4 The flowchart for optimizing and dimensionality reducing historical stirring performance set;

[0060] Figure 5 The flowchart for mixing trend analyzer construction and efficiency calculation. DETAILED DESCRIPTION

[0061] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described, obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0062] Please refer to Figures 1-5 The present application provides a self-cleaning stirring and uniform mixing system for a variable-diameter propeller blade type reaction kettle, which comprises.

[0063] A variable-diameter blade parameter acquisition module is configured to acquire a target blade currently operated by the reaction kettle, acquire target working condition information of the target blade currently operated, and index historical stirring data of the target blade in a historical time of the reaction kettle according to the target blade.

[0064] A historical stirring data classification module is configured to classify historical working condition information in the historical stirring data, obtain a plurality of classified historical working conditions, extract a historical mixing information set and a plurality of historical stirring data sets under the plurality of classified historical working conditions, and process to obtain a plurality of historical stirring performance sets.

[0065] A stirring performance sequence construction module is configured to respectively optimize and reduce dimensions of the plurality of historical stirring performance sets, obtain a plurality of reduced-dimension historical stirring performance sets, and arrange in time sequence to obtain a plurality of historical stirring performance sequences.

[0066] A mixing trend analysis module is configured to perform mixing trend analysis according to the plurality of historical stirring performance sequences, and obtain mixing efficiency and cleaning efficiency.

[0067] A self-cleaning parameter correction module is configured to match the target working condition information with the plurality of classified historical working conditions, obtain a matching historical working condition, and correct the mixing efficiency and the cleaning efficiency according to a deviation of the target working condition information and standard matching working condition information of the matching historical working condition, and obtain a corrected mixing efficiency and a corrected cleaning efficiency.

[0068] A mixing strategy decision module is configured to perform mixing strategy decision according to the corrected mixing efficiency and the corrected cleaning efficiency, obtain a mixing strategy scheme, and perform mixing strategy execution.

[0069] In the operation process of the self-cleaning stirring and uniform mixing system for the variable-diameter propeller blade type reaction kettle, the variable-diameter blade parameter acquisition module performs the operation of acquiring a target blade currently operated by the reaction kettle, acquiring target working condition information of the target blade currently operated, and indexing historical stirring data of the target blade in a historical time of the reaction kettle according to the target blade.

[0070] The variable-diameter paddle parameter acquisition module first determines the target paddle that the reactor is currently running. This process is achieved through the operation monitoring system of the reactor, which can identify the model and specifications of the paddle currently participating in the stirring work in real time, thereby determining the specific information of the target paddle.

[0071] After determining the target paddle, the variable-diameter paddle parameter acquisition module begins to collect the upper working condition parameters and the lower working condition parameters of the target paddle, and integrates these parameters as the target working condition information. Among them, the upper working condition parameters cover the relevant operating data of the upper half of the paddle during the stirring process, such as the speed fluctuation of the upper half of the paddle, the distance change with the inner wall of the upper half of the reactor, the instantaneous pressure of the material contacted by the upper half, etc.; the lower working condition parameters include the operating data of the lower half of the paddle, such as the torque change of the lower half of the paddle, the flow speed of the material in the lower half, the distance to the bottom of the reactor, etc. These upper and lower working condition parameters together constitute the target working condition information that can fully reflect the current running state of the target paddle.

[0072] After the target paddle and target working condition information are obtained, the variable-diameter paddle parameter acquisition module indexes the historical running records of the reaction kettle according to the determined target paddle to obtain the historical stirring data of the target paddle. The working condition characteristic parameters specifically include the paddle speed (unit: r / min), the material viscosity (unit: cP), and the reaction temperature (unit: ℃), which directly determine the load characteristics, the material flow state, and the reaction rate of the stirring process, and are the core indexes for distinguishing the differences between working conditions. The similarity threshold is determined based on the actual running fluctuation range of the three characteristic parameters and the stirring effect sensitivity, and a dual standard of a parameter threshold and a comprehensive similarity threshold is adopted: the parameter threshold sets the allowable deviation range of a single parameter according to the design running range of the reaction kettle, that is, the paddle speed similarity threshold is ±15 r / min, and if the speed of a certain historical working condition is 300 r / min, the speed of other working conditions is considered to be similar to this parameter within the range of 285-315 r / min; the material viscosity similarity threshold is ±20 cP, and if the viscosity of a certain historical working condition is 100 cP, the viscosity of other working conditions is considered to be similar to this parameter within the range of 80-120 cP; the reaction temperature similarity threshold is ±5 ℃, and if the temperature of a certain historical working condition is 80 ℃, the temperature of other working conditions is considered to be similar to this parameter within the range of 75-85 ℃. When at least two of the three parameters meet the parameter threshold requirement, and the comprehensive deviation rate of the three parameters is ≤10%, it is determined that the working conditions are similar. The calculation method of the comprehensive deviation rate is: (|current speed-base speed| / base speed+|current viscosity-base viscosity| / base viscosity+|current temperature-base temperature| / base temperature) x 100% / 3, wherein the base speed, the base viscosity, and the base temperature are the characteristic parameter values of a certain reference working condition. If a historical working condition with a speed of 300 r / min, a viscosity of cP, and a temperature of 80 ℃ is taken as a reference base, and the parameters of a certain to-be-classified historical working condition are a speed of 310 r / min, a viscosity of 110 cP, and a temperature of 83 ℃, then: the speed deviation is 10 r / min (≤±15 r / min), the viscosity deviation is 10 cP (≤±20 cP), and the temperature deviation is 3 ℃ (≤±5 ℃), all of which meet the parameter threshold; the comprehensive deviation rate is (10 / 300+10 / 100+3 / 80) x 100% / 3≈(3.33%+10%+3.75%) / 3≈5.69% (≤10%), so the to-be-classified working condition and the reference base working condition are classified into the same category. The historical running records of the reaction kettle store in detail the running conditions of different paddles in the past time periods, including but not limited to the working condition parameters, the stirring time, the material mixing effect records, the wear condition of the paddle, etc. Through the specific identification information of the target paddle, accurate retrieval is performed in the historical running records, so that all the historical stirring data related to the target paddle can be quickly located, thereby providing complete basic data for subsequent historical stirring data processing and analysis.

[0073] Through the above process, the variable diameter paddle parameter acquisition module can comprehensively and accurately obtain the relevant information of the target paddle and the corresponding historical data, providing solid data support for the subsequent work of the system's historical stirring data classification, stirring performance sequence construction, and mixing trend analysis modules, ensuring that the entire system can effectively self-clean the stirring and uniform mixing control based on sufficient information.

[0074] In embodiment 2, the variable diameter paddle parameter acquisition module acquires the target paddle currently running in the reaction kettle, collects the target working condition information of the target paddle currently running, and acquires the historical stirring data of the target paddle in the reaction kettle history time according to the target paddle index. The variable diameter paddle parameter acquisition module determines the target paddle currently in working state through the operation log of the reaction kettle, which records the start time, running state and switching record of each paddle in real time. After determining the target paddle, the variable diameter paddle parameter acquisition module starts the sensor group installed at different positions of the paddle, in which the pressure sensor installed at the upper half of the paddle records the contact pressure value with the inner wall of the reaction kettle every 0.5 seconds, the vibration sensor installed at the lower half of the paddle continuously captures the amplitude change of the paddle in the stirring process, the temperature sensor collects the real-time temperature of the paddle surface and the material contact area, and the humidity sensor records the humidity value of the material near the paddle. These sets composed of pressure, vibration, temperature and humidity data are integrated into target working condition information. The variable diameter paddle parameter acquisition module matches in the historical database of the reaction kettle through the unique code of the target paddle, which stores the stirring time, material type, stirring speed change curve and paddle surface state image at each running time of the paddle in the past 12 months in chronological order, and packs these data as historical stirring data after indexing.

[0075] In the process of classifying and processing the historical stirring data by the historical stirring data classification module to obtain a plurality of historical stirring performance sets, the historical basic performance set is obtained and corrected in the following manner:

[0076] The acquisition logic of the historical basic performance set extracts the historical mixing information set under multiple classified historical working conditions, i.e., the standard mixing state of the target blade under the working condition, including the material mixing uniformity reference value, the blade surface residual-free reference state, and the material flow dead angle-free reference range. Then, the deviation of the historical mixing state information set under each classified historical working condition from the standard mixing state is compared one by one. For the material mixing uniformity, the consistency of the composition of the material in different regions in the historical mixing state is detected, and the consistency of the composition of the standard mixing state is compared with the reference value to determine the deviation amplitude. For example, if the standard reference value is that the composition difference in each region is ≤5%, and there are three regions with composition differences of 6%, 8%, and 7% in a certain historical mixing state, the deviation amplitude in this dimension is calculated according to the average difference value. For the blade surface residual, the surface state data of the blade after shutdown in the historical record is compared with the standard reference value of no visible residual to determine the deviation amplitude of the residual area. For the material flow dead angle, the material flow velocity distribution data in the historical mixing process is compared with the standard reference value of no flow velocity ≤0.1 m / s in the region to determine the area deviation amplitude of the dead angle region. The deviation amplitudes in the above three dimensions are comprehensively calculated according to the uniformity weight of 40%, the residual weight of 30%, and the dead angle weight of 30%. According to the interval classification of the comprehensive deviation amplitude: the comprehensive deviation amplitude ≤10% is A level, 10%-20% is B level, 20%-30% is C level, and >30% is D level. The performance parameter set corresponding to each level is a historical basic performance set. For example, the A-level historical basic performance set contains the mixing performance data under the working condition with a comprehensive deviation amplitude ≤10%. The correction logic of the historical basic performance set presets the execution time threshold as the standard mixing completion time corresponding to the classified historical working condition, and extracts the historical mixing execution time set under each classified historical working condition. The correction calculation method is to multiply the original performance parameters in the historical basic performance set by the correction coefficient of the preset execution time threshold / actual mixing execution time. For example, the preset execution time threshold of a certain classified historical working condition is 60 minutes, and the actual mixing execution time is 75 minutes. Therefore, the correction coefficient is 60 / 75=0.8. If the initial value of the mixing efficiency in the historical basic performance set corresponding to this mixing is 12 units / hour, the corrected mixing efficiency is 12×0.8=9.6 units / hour. The corrected performance parameters are re-integrated to form a historical mixing performance set.

[0077] The historical stirring data processing module classifies the historical stirring data to obtain classified historical stirring data, and constructs a historical stirring performance sequence of the target paddle based on the classified historical stirring data. The historical stirring data processing module first extracts material type information from the historical stirring data, and divides stirring data corresponding to the same material type into the same data group. For records containing a mixture of multiple materials, secondary classification is performed according to the main material type. After classification, the average value of the stirring speed, the cumulative value of the stirring time length, and the definition parameter of the paddle surface state image are extracted from each group of data, and these parameters are arranged in chronological order to form a historical stirring performance sequence of the target paddle under different material types. Among them, the average value of the speed is calculated by the instantaneous speed recorded every 5 minutes, the cumulative value of the time length is determined by the sum of the continuous running time period, and the image definition parameter is converted from the pixel ratio of the paddle surface stain coverage area by the image analysis algorithm.

[0078] The mixing trend analysis module analyzes the mixing trend of the target paddle under the current target working condition based on the target working condition information and the historical stirring performance sequence, and generates a mixing trend analysis result. The mixing trend analysis module compares the real-time pressure value in the target working condition information with the pressure average value under the same material type in the historical stirring performance sequence. When the real-time pressure value is outside the ±15% range of the historical average value, it is marked as a pressure abnormal point. At the same time, the real-time vibration amplitude is fitted with the amplitude change curve in the historical sequence to calculate the fitting degree value. The area with a fitting degree less than 80% is marked as a vibration abnormal section. The real-time data of temperature and humidity are respectively associated with the corresponding parameters in the historical data to establish a correlation model, and the deviation coefficient of the current parameter and the historical parameter is calculated through the model output. The mixing trend analysis result includes the occurrence time of the pressure abnormal point, the duration of the vibration abnormal section, the temperature deviation coefficient and the humidity deviation coefficient. These data together constitute a quantitative description of the current mixing trend.

[0079] The self-cleaning regulation module generates a diameter regulation instruction based on the mixed trend analysis result and the variable diameter parameter of the target paddle, and sends the diameter regulation instruction to the variable diameter driving device of the reaction kettle. The self-cleaning regulation module extracts the diameter adjustment range of the upper half and the lower half of the paddle from the design parameters of the target paddle. The diameter of the upper half can be adjusted between 300 mm and 500 mm, and the diameter of the lower half can be adjusted between 200 mm and 400 mm. The adjustment accuracy is 10 mm. When an abnormal point appears in the mixed trend analysis result, the self-cleaning regulation module calculates the paddle position corresponding to the abnormal point. If the abnormal point is located in the upper half, the diameter of the upper half is adjusted in the decreasing direction, and the adjustment amplitude is 10 mm each time until the real-time pressure value returns to the historical average value range. If the abnormal point is located in the lower half, the diameter of the lower half is adjusted, and the adjustment direction is determined according to the positive and negative of the pressure value exceeding the range. For the vibration abnormal segment, the self-cleaning regulation module adjusts the paddle diameter at a rate of 5 mm / min in the corresponding time period according to the duration of the abnormal segment, and records the vibration amplitude change in the adjustment process. The deviation coefficient of temperature and humidity is converted into a compensation value for diameter adjustment. When there is a positive deviation, the diameter of the corresponding paddle segment is increased, and when there is a negative deviation, the diameter is decreased. The size of the compensation value is proportional to the absolute value of the deviation coefficient. The diameter regulation instruction includes specific parameters of the adjustment position, the adjustment direction, the adjustment amplitude and the adjustment rate. These parameters are sent to the execution unit of the variable diameter driving device through the internal control bus of the reaction kettle.

[0080] The variable diameter driving device executes the variable diameter regulation instruction to drive the target paddle to perform variable diameter operation, so as to realize self-cleaning stirring and uniform mixing of the reaction kettle. After receiving the variable diameter regulation instruction, the servo motor of the variable diameter driving device drives the telescopic shaft inside the paddle to move through the gear transmission mechanism. The displacement of the telescopic shaft is determined by the adjustment amplitude in the instruction, and the displacement accuracy is controlled within ±2 mm. When adjusting the diameter of the upper half, the blades of the upper half paddle expand or contract along the radial direction through the hinge structure. During the expansion process, the contact area between the anti-slip lines on the blade surface and the material gradually increases, and during the contraction process, the relative movement between the scraping strips on the blade edge and the inner wall of the reaction kettle occurs. When adjusting the diameter of the lower half, the pitch of the spiral structure at the bottom of the paddle changes. When the pitch increases, the material forms an upward vortex below the paddle, and when the pitch decreases, the vortex direction changes to downward. The displacement sensor of the driving device monitors the actual change value of the paddle diameter in real time, and feeds back the value to the internal control system. When the deviation between the actual value and the instruction value exceeds 5 mm, the servo motor automatically performs secondary fine adjustment. During the entire variable diameter operation process, the rotating speed of the paddle remains stable. By adjusting the diameter proportion of different segments, the material forms multi-directional convection in the reaction kettle. The self-cleaning coating on the paddle surface rubs with the material when the diameter changes, reducing the adhesion of stains.

[0081] The stirring effect monitoring module monitors the stirring effect of the target blade in real time during the diameter changing operation of the target blade and generates stirring effect monitoring data. The camera of the stirring effect monitoring module is installed at the observation window position on the top of the reaction kettle. The camera takes a picture of the material mixing state inside the reaction kettle every 2 minutes. The color uniformity of the material in the picture is converted into the standard deviation of the RGB value by the color analysis algorithm. The smaller the standard deviation, the more uniform the mixing. The sampling device installed at the bottom of the reaction kettle extracts a material sample every 10 minutes. The viscosity value of the sample is measured by a viscometer. The fluctuation range of the viscosity value is recorded as the viscosity stability parameter. The infrared sensor on the surface of the blade continuously scans the temperature distribution on the surface of the blade. The uniformity of the temperature distribution is represented by the maximum temperature difference. The area with a temperature difference exceeding 5°C is marked as a temperature abnormal area. The monitoring data also includes the adhesion area of the material on the inner wall of the reaction kettle, which is calculated by the pixels of the reflective area on the inner wall in the picture. All monitoring data are stored in chronological order to form a continuous stirring effect monitoring data stream.

[0082] The parameter feedback adjustment module adjusts the diameter changing control instruction based on the stirring effect monitoring data and sends the adjusted diameter changing control instruction to the diameter changing driving device. When the RGB value standard deviation in the stirring effect monitoring data is greater than the historical average value for three consecutive measurements, the parameter feedback adjustment module calculates the diameter adjustment amplitude that needs to be increased. The amplitude value is proportional to the excess proportion of the standard deviation. If the viscosity stability parameter exceeds the historical fluctuation range, the parameter feedback adjustment module sends an instruction to adjust the pitch of the lower half of the blade to the diameter changing driving device according to the viscosity value. When the viscosity is high, the pitch is increased. When the viscosity is low, the pitch is decreased. When the area ratio of the temperature abnormal area exceeds 20%, the diameter of the blade segment in the corresponding area is adjusted to regulate the temperature distribution by changing the contact area with the material. When the pixel ratio of the inner wall adhesion area increases by 5%, the parameter feedback adjustment module increases the diameter adjustment frequency of the upper half of the blade. The adjustment amplitude is kept within the range of 3-5 mm. The adjusted diameter changing control instruction contains new adjustment parameters, which are sent to the diameter changing driving device through the same transmission path as the original instruction. The driving device continues to perform the diameter changing operation according to the new instruction until each parameter in the stirring effect monitoring data returns to the historical normal range.

[0083] In embodiment 3, the stirring performance sequence construction module optimizes and reduces the dimension of each historical stirring performance set to obtain multiple reduced historical stirring performance sets, and arranges the multiple reduced historical stirring performance sets in chronological order to obtain the operation of the multiple historical stirring performance sequences.

[0084] In the first historical stirring performance set, the stirring performance sequence construction module first selects a first benchmark performance value. The set contains 300 groups of performance data of a certain variable pitch propeller blade stirring different viscosity materials (viscosity range 50-500 cP) in the past 6 months, each group of data covering three dimensions of stirring uniformity (represented by the composition deviation value detected by sampling the material), blade cleanliness (weight percentage of residual material on the blade surface), and energy consumption coefficient (ratio of power consumption to stirring volume per unit time). In the screening process, the stirring uniformity 95%, the cleanliness 90%, and the energy consumption coefficient 0.8 kW·h / kg are used as standard thresholds, and the performance data that meet all three thresholds are selected as the first benchmark performance value. If there are multiple groups of data that meet the conditions, the group with the shortest stirring time is selected.

[0085] After determining the first benchmark performance value, the stirring performance sequence construction module calculates the difference between the first benchmark performance value and other historical stirring performance in the first historical stirring performance set. The calculation method is as follows: for each dimension of each group of performance data, subtract the corresponding dimension value of the first benchmark performance value from the value of the group of data to obtain the uniformity difference, the cleanliness difference, and the energy consumption difference. For example, the uniformity of a certain group of data is 92%, the cleanliness is 85%, and the energy consumption coefficient is 0.9 kW·h / kg, then the corresponding differences are -3%, -5%, and 0.1 kW·h / kg, respectively. According to the size of the difference, the distribution probability is allocated, and the smaller the absolute value of the difference, the larger the distribution probability. Specifically, the probability allocation formula is set as:

[0086]

[0087] where P is the distribution probability, and D is the difference of a certain dimension (the uniformity difference is substituted with the percentage value, the cleanliness difference is substituted with the percentage value, and the energy consumption difference is substituted with the kW·h / kg value). For each group of performance data, the average of the distribution probabilities of the three dimensions is taken as the comprehensive distribution probability of the group of data, and thus the first basic probability distribution is obtained.

[0088] Based on the first basic probability distribution, the stirring performance sequence construction module optimizes and reduces the dimension of the first historical stirring performance set. The preset dimension reduction number is set to one third of the original set number, i.e., 100 groups are selected from the 300 groups of data. A random number generator is used to generate 100 non-repeating random numbers within the index range (1-300) of the first historical stirring performance set, and the historical stirring performance corresponding to the index is extracted to form the first reduced dimension historical performance set. Then, the difference between each group of data in the reduced dimension set and the first benchmark performance value is calculated, and the first reduced dimension probability distribution is obtained according to the same formula as above. By calculating the cosine similarity between the first reduced dimension probability distribution and the first basic probability distribution, the first reduced dimension fitness is obtained, and the higher the similarity, the larger the fitness value.

[0089] Then, the stirring performance sequence construction module randomly extracts 100 groups of data from the first historical stirring performance set again to form a second reduced dimension historical performance set, repeats the above calculation process to obtain a second reduced dimension fitness. This is repeated 20 times, each time generating a new reduced dimension set and calculating the fitness, and finally the reduced dimension historical performance set with the largest reduced dimension fitness is selected as the first reduced dimension performance set. If the difference between the reduced dimension set generated this time and the fitness of the previous time is less than 0.01, the iteration is stopped in advance, and the current reduced dimension set is taken as the first reduced dimension performance set.

[0090] After obtaining the first reduced dimension performance set, the stirring performance sequence construction module extracts the time stamp information corresponding to each historical stirring performance in the set. The time stamp is accurate to the minute and records the stirring operation start time corresponding to each group of performance data. According to the chronological order of the time stamp, the 100 groups of data in the first reduced dimension performance set are arranged to form the first historical stirring performance sequence. In the sequence, each group of data is arranged in chronological order, and the time interval between adjacent two groups of data may be 1 hour, 2 hours or longer, depending on the actual execution time of the historical stirring operation.

[0091] For other multiple historical stirring performance sets (such as a second historical stirring performance set for different reaction temperature ranges, a third historical stirring performance set for different paddle speed ranges, etc.), the stirring performance sequence construction module uses the same operation process. Taking the second historical stirring performance set as an example, the set contains 240 groups of data, and 80 groups are retained after dimension reduction. The reduced dimension performance set is obtained in the same way as random extraction, probability distribution calculation, fitness comparison, and then sorted according to the time stamp to form the second historical stirring performance sequence. Through the above process, multiple historical stirring performance sequences are finally obtained, each corresponding to the historical stirring performance change under a certain classification standard.

[0092] Example 4: The mixing trend analysis module performs mixing trend analysis according to multiple historical stirring performance sequences to obtain mixing efficiency and cleaning efficiency.

[0093] The mixed trend analysis module first collects sample operation data of multiple reaction kettles. These sample operation data cover stirring records of different types of variable-diameter propeller blades in the past three years, including blade speed change curves over time, material flow velocity field distribution in the reaction kettle, blade surface stain adhesion amount per hour, and material mixing uniformity detection results. From these sample operation data, a sample stirring performance sequence set is extracted, and each sample stirring performance sequence is composed of continuous stirring performance data, for example, a certain sample sequence includes stirring uniformity (expressed in percentage of composition deviation), blade cleanliness (expressed in percentage of stain coverage area), and energy consumption data recorded every 10 minutes at a speed of 300 r / min, a material viscosity of 200 cP, and a reaction temperature of 80°C for a blade, lasting for 5 hours. At the same time, a sample mixing efficiency set and a sample cleaning efficiency set are obtained according to the performance change indicators in each sample stirring performance sequence. Performance change indicators include stirring uniformity improvement rate, stain adhesion amount reduction rate, etc. Mixing efficiency is calculated by the amount of stirring uniformity improvement per unit time, and cleaning efficiency is calculated by the amount of stain coverage area reduction per unit time, for example, in a certain sample sequence, the stirring uniformity is improved from 60% to 90% in 2 hours, and the mixing efficiency is 15% / hour; the stain coverage area is reduced from 30% to 10% in 2 hours, and the cleaning efficiency is 10% / hour.

[0094] The sample stirring performance sequence set is used as the classification input, and the sample mixing efficiency set and the sample cleaning efficiency set are used as the classification output to construct a mixed trend analyzer. The analyzer is trained by a deep learning algorithm, and the algorithm's hidden layer contains 128 neurons, which uses ReLU activation function to process the input sequence data, and the output layer corresponds to the predicted values of mixing efficiency and cleaning efficiency. During training, the sample data is divided into training set and validation set in the ratio of 7:3, the iteration number is set to 500, and the connection weight between neurons is adjusted each time until the prediction error of the model on the validation set stabilizes within the preset range.

[0095] Based on the hybrid trend analyzer, the multiple historical stirring performance sequences are classified by hybrid trend to obtain multiple working condition hybrid efficiencies and multiple working condition cleaning efficiencies. The historical stirring performance sequence includes performance data of a certain target blade under different classified historical working conditions, for example, under classified historical working condition A (blade speed 250 r / min, material viscosity 150 cP, reaction temperature 60℃), the historical stirring performance sequence records the change process of stirring uniformity from 55% to 92% and the corresponding change of stain coverage area in 10 consecutive stirring processes; under classified historical working condition B (blade speed 350 r / min, material viscosity 250 cP, reaction temperature 90℃), the sequence records the related performance changes of 8 stirring processes. After inputting these sequences into the hybrid trend analyzer, the analyzer outputs the working condition hybrid efficiency and the working condition cleaning efficiency corresponding to each classified historical working condition, for example, the hybrid efficiency of working condition A is 12% / hour, and the cleaning efficiency is 8% / hour, the hybrid efficiency of working condition B is 18% / hour, and the cleaning efficiency is 12% / hour.

[0096] The similarity of the target working condition information and the multiple classified historical working conditions is analyzed, and the multiple working condition hybrid efficiencies and multiple working condition cleaning efficiencies are weighted calculated according to the size of the multiple working condition similarities, to obtain the hybrid efficiency and the cleaning efficiency. The target working condition information includes the current target blade speed 280 r / min, material viscosity 220 cP, and reaction temperature 75℃. When calculating the similarity, the Euclidean distance algorithm is used to calculate the distance value of the target working condition and each classified historical working condition after standardizing the working condition characteristic parameters (speed, viscosity, temperature), and the smaller the distance value is, the greater the similarity is. For example, the distance value of the target working condition and working condition A is 15, the distance value of the target working condition and working condition B is 20, and the distance value of the target working condition and working condition C (speed 260 r / min, material viscosity 210 cP, reaction temperature 70℃) is 10. According to the size of the similarity, the weights are assigned, and the total weight is 1. The weight of working condition C with the smallest distance value is 0.5, the weight of working condition A is 0.3, and the weight of working condition B is 0.2. After multiplying the hybrid efficiency and the cleaning efficiency of each working condition by the corresponding weight and summing, the hybrid efficiency is (12% / hour x 0.3 + 18% / hour x 0.2 + 14% / hour x 0.5) = 13.6% / hour, and the cleaning efficiency is (8% / hour x 0.3 + 12% / hour x 0.2 + 10% / hour x 0.5) = 9.4% / hour.

[0097] The self-cleaning parameter correction module matches the target working condition information with the multiple classified historical working conditions to obtain the matching historical working conditions, and corrects the hybrid efficiency and the cleaning efficiency according to the deviation of the target working condition information and the standard matching working condition information of the matching historical working conditions, to obtain the corrected hybrid efficiency and the corrected cleaning efficiency.

[0098] The self-cleaning parameter correction module selects the most similar classified historical working condition as the matching historical working condition, i.e., the above-mentioned working condition C, and obtains the standard matching working condition information of the matching historical working condition. The standard matching working condition information is the typical parameters of working condition C recorded in multiple runs, including the standard speed 260 r / min, the standard material viscosity 210 cP, the standard reaction temperature 70℃, and the corresponding standard mixing efficiency 14% / hour and the standard cleaning efficiency 10% / hour under the standard working condition.

[0099] According to the deviation of the target working condition information and the standard matching working condition information of the matching historical working condition, the efficiency correction coefficient is set. The deviation values of each parameter are calculated, the speed deviation is 280 r / min-260 r / min=20 r / min, the viscosity deviation is 220 cP-210 cP=10 cP, and the temperature deviation is 75℃-70℃=5℃. The deviation influence weight is set for each parameter, and the weights of speed, viscosity and temperature are 0.4, 0.3 and 0.3 respectively. The relative deviation rate is calculated, the speed relative deviation rate is 20 / 260≈7.69%, the viscosity relative deviation rate is 10 / 210≈4.76%, and the temperature relative deviation rate is 5 / 70≈7.14%. The efficiency correction coefficient is calculated according to the comprehensive relative deviation rate, the comprehensive relative deviation rate is (7.69%×0.4+4.76%×0.3+7.14%×0.3)≈6.57%, and the correction coefficient is set to (1-comprehensive relative deviation rate), i.e., 0.9343.

[0100] The efficiency correction coefficient is used to correct and calculate the mixing efficiency and the cleaning efficiency to obtain the corrected mixing efficiency and the corrected cleaning efficiency. The corrected mixing efficiency=mixing efficiency×correction coefficient=13.6% / hour×0.9343≈12.71% / hour, and the corrected cleaning efficiency=cleaning efficiency×correction coefficient=9.4% / hour×0.9343≈8.78% / hour.

[0101] The mixing strategy decision module makes a mixing strategy decision according to the corrected mixing efficiency and the corrected cleaning efficiency, obtains a mixing strategy scheme, and performs the mixing strategy.

[0102] The mixed strategy decision module collects sample correction mixing efficiency set and sample correction cleaning efficiency set, which are from historical decision data of different reactors under various working conditions. For example, the sample correction mixing efficiency covers different values from 5% / hour to 25% / hour, and the sample correction cleaning efficiency covers different values from 3% / hour to 18% / hour. According to the size of each sample correction mixing efficiency and sample correction cleaning efficiency, a sample mixing strategy scheme is set, and a sample mixing strategy scheme set is obtained. Each sample mixing strategy scheme includes strategy parameters such as paddle speed adjustment amplitude, stirring direction switching frequency, paddle diameter change rate, etc. The size of the sample correction mixing efficiency and the sample correction cleaning efficiency is negatively correlated with the size of the strategy parameter, i.e. when the sample correction mixing efficiency is 20% / hour and the sample correction cleaning efficiency is 15% / hour, the strategy parameter can be speed adjustment amplitude ±10r / min, stirring direction switching every 30 minutes, and paddle diameter change rate 5mm / min; while when the sample correction mixing efficiency is 8% / hour and the sample correction cleaning efficiency is 5% / hour, the strategy parameter can be speed adjustment amplitude ±30r / min, stirring direction switching every 10 minutes, and paddle diameter change rate 15mm / min.

[0103] The sample correction mixing efficiency set and the sample correction cleaning efficiency set are used as decision input, and the sample mixing strategy scheme set is used as decision output to construct a mixed strategy decision maker. The decision maker is constructed based on decision tree algorithm, and is trained by sample data so that the decision maker can output the corresponding mixing strategy scheme according to the input correction mixing efficiency and correction cleaning efficiency. For example, when the input correction mixing efficiency is 12% / hour and the correction cleaning efficiency is 8% / hour, the decision maker will match the most similar sample data and output the corresponding strategy parameters: speed adjustment amplitude ±20r / min, stirring direction switching every 20 minutes, and paddle diameter change rate 10mm / min.

[0104] The mixed strategy decision maker is used to make mixed strategy decisions on the correction mixing efficiency and the correction cleaning efficiency to obtain a mixed strategy scheme. When the correction mixing efficiency 12.71% / hour and the correction cleaning efficiency 8.78% / hour are input into the mixed strategy decision maker, the mixed strategy scheme output by the decision maker includes: the paddle speed can be adjusted within the range of 260r / min to 300r / min based on the current 280r / min; the stirring direction is switched every 20 minutes from clockwise to counterclockwise or vice versa; the paddle diameter changes at a rate of 10mm / min, and the diameter size is adjusted within the allowed range according to the real-time state of the material in the reactor. The mixed strategy scheme is sent to the actuator of the reactor, and the actuator controls the operation of the variable-diameter propeller according to the parameters in the scheme to realize self-cleaning stirring and uniform mixing.

[0105] Embodiment 5: The mixed strategy decision module makes mixed strategy decision according to the revised mixing efficiency and the revised cleaning efficiency, obtains a mixed strategy scheme, and performs mixed strategy execution.

[0106] The mixed strategy decision module collects a sample revised mixing efficiency set and a sample revised cleaning efficiency set. These sample data are derived from the operation records of the variable-diameter propeller blade type reaction kettle in the past two years, covering the revised efficiency data of different material types (such as liquid material, semi-solid material, and material containing particle suspension), different reaction stages (such as initial mixing stage, mid-reaction stage, and end-reaction stage). The sample revised mixing efficiency set contains continuous values from 4% / h to 22% / h, and the sample revised cleaning efficiency set contains continuous values from 2% / h to 16% / h, each efficiency value corresponding to a complete set of stirring process parameter records.

[0107] According to the size of each sample revised mixing efficiency and sample revised cleaning efficiency, a sample mixed strategy scheme is set, and a sample mixed strategy scheme set is obtained. Each sample mixed strategy scheme includes multiple strategy parameters, specifically paddle speed adjustment range (unit: r / min), stirring direction switching period (unit: min), paddle diameter variation amplitude (unit: mm), stirring intensity level (divided into 1-5 levels). The size of the sample revised mixing efficiency and the sample revised cleaning efficiency is negatively related to the size of the strategy parameters, i.e. when the sample revised mixing efficiency is 20% / h and the sample revised cleaning efficiency is 15% / h, the corresponding strategy parameters can be: paddle speed adjustment range ±10 r / min, stirring direction switching period 40 min, paddle diameter variation amplitude ±5 mm, stirring intensity level 2 levels; when the sample revised mixing efficiency is 6% / h and the sample revised cleaning efficiency is 3% / h, the corresponding strategy parameters can be: paddle speed adjustment range ±30 r / min, stirring direction switching period 15 min, paddle diameter variation amplitude ±15 mm, stirring intensity level 4 levels.

[0108] The sample corrected mixing efficiency set and the sample corrected cleaning efficiency set are used as the decision input, and the sample mixing strategy scheme set is used as the decision output to construct the mixing strategy decision maker. The construction of the mixing strategy decision maker is based on the support vector machine algorithm, and the mapping of the input and output is realized by training the sample data. During the training process, the sample data is divided into a training set and a test set in a ratio of 8:2, wherein the training set is used to adjust the hyperparameters (such as the penalty coefficient and the kernel function parameter) of the algorithm, and the test set is used to verify the output accuracy of the decision maker. The kernel function of the algorithm is selected as the radial basis function, the penalty coefficient is determined as 10 by the grid search method, and the kernel function parameter is determined as 0.1 to ensure that the decision maker can maintain stable output performance on the training set and the test set. The input layer of the decision maker receives two-dimensional parameters (corrected mixing efficiency and corrected cleaning efficiency), and the output layer corresponds to four-dimensional strategy parameters (rotation speed adjustment range, switching period, diameter change amplitude, and force level).

[0109] The mixing strategy decision maker is used to make a mixing strategy decision on the corrected mixing efficiency and the corrected cleaning efficiency, and a mixing strategy scheme is obtained. When the current corrected mixing efficiency of the reaction kettle is 12% / h and the corrected cleaning efficiency is 9% / h, the two parameters are input into the mixing strategy decision maker. The decision maker first performs standardization processing on the input parameters, converting them to values between 0 and 1 (the corrected mixing efficiency of 12% / h corresponds to 0.4, and the corrected cleaning efficiency of 9% / h corresponds to 0.5), and then calculates the corresponding strategy parameters through the internal mapping relationship: the paddle rotation speed adjustment range is ±20 r / min, the stirring direction switching period is 25 min, the paddle diameter change amplitude is ±10 mm, and the stirring force level is 3 levels. These parameters jointly constitute the mixing strategy scheme, which is transmitted to the control system of the reaction kettle. The control system adjusts the operating state of the variable-diameter propeller blade in real time according to the parameters in the scheme, for example, based on the current rotation speed of 280 r / min, it allows fluctuation between 260-300 r / min, switches the stirring direction from clockwise to counterclockwise every 25 min, simultaneously controls the paddle diameter to change at a rate of 5 mm / min within a range of ±10 mm, and maintains the stirring force level at 3 levels.

[0110] During the mixed strategy execution, the sensors of the reactor collect the running parameters of the paddle (such as actual speed, direction switching time, current diameter) and the state parameters of the material (such as mixing uniformity, paddle surface cleanliness) in real time, and feed these parameters back to the mixed strategy decision maker. The decision maker judges the execution effect of the current strategy scheme according to the feedback data. If the mixing uniformity change rate collected for three consecutive times is lower than the preset value, the fine tuning mechanism of the strategy parameters is automatically triggered, for example, the speed adjustment range is expanded by ±5r / min, or the direction switching period is shortened by 5min, to adapt to the change of the material state. This dynamic adjustment mechanism ensures that the mixed strategy scheme can continuously adapt to the actual operation of the reactor until the stirring process is completed.

[0111] It should be noted that, in this document, the terms such as first and second are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between such entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such a process, method, article or equipment.

[0112] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and variations can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A self-cleaning stirring and uniform mixing system for a variable-diameter propeller blade type reaction vessel, characterized by, The system comprises: A variable-pitch blade parameter acquisition module is configured to acquire a target blade currently operated by the reactor and collect target working condition information of the target blade currently operated by the reactor, and index historical stirring data of the target blade in a historical time of the reactor according to the target blade. A historical stirring data classification module is configured to classify historical working condition information in the historical stirring data, obtain a plurality of classified historical working conditions, extract a historical mixing information set and a plurality of historical stirring data sets under the plurality of classified historical working conditions, and process to obtain a plurality of historical stirring performance sets. A stirring performance sequence construction module is configured to respectively optimize and reduce dimensions of the plurality of historical stirring performance sets to obtain a plurality of reduced historical stirring performance sets, and arrange in time sequence to obtain a plurality of historical stirring performance sequences. A mixing trend analysis module is configured to perform mixing trend analysis according to the plurality of historical stirring performance sequences to obtain mixing efficiency and cleaning efficiency. A self-cleaning parameter correction module is configured to match the target working condition information with the plurality of classified historical working conditions to obtain a matching historical working condition, and correct the mixing efficiency and the cleaning efficiency according to a deviation of the target working condition information and standard matching working condition information of the matching historical working condition to obtain a corrected mixing efficiency and a corrected cleaning efficiency. A mixing strategy decision module is configured to perform mixing strategy decision according to the corrected mixing efficiency and the corrected cleaning efficiency to obtain a mixing strategy scheme, and perform mixing strategy execution.

2. The self-cleaning agitating and homogenizing system for a variable pitch propeller blade type reaction vessel according to claim 1, wherein The target blade currently operated by the reactor is acquired, and the target working condition information of the target blade currently operated by the reactor is collected, and the historical stirring data of the target blade in a historical time of the reactor is indexed according to the target blade, comprising: The target blade currently operated by the reactor is acquired, and the upper working condition parameter and the lower working condition parameter of the target blade are collected as the target working condition information. According to the target blade, the historical stirring data of the target blade is obtained by indexing in the historical operation record of the reactor.

3. The self-cleaning agitating and homogenizing system for a variable pitch propeller blade type reaction vessel according to claim 1, wherein The historical working condition information in the historical stirring data is classified to obtain a plurality of classified historical working conditions, the historical mixing information set and the plurality of historical stirring data sets under the plurality of classified historical working conditions are extracted, and the plurality of historical stirring performance sets are processed and obtained, comprising: A plurality of historical working condition information in the historical stirring data is acquired, classified and processed to obtain a plurality of classified historical working conditions. The standard mixing state of the target blade in the historical stirring data under the plurality of classified historical working conditions is extracted to obtain a historical mixing information set, and a plurality of historical mixing state information sets and a plurality of historical stirring execution time sets generated by stirring of the reactor under the plurality of classified historical working conditions are extracted. According to the deviation amplitude of the plurality of historical mixing state information sets and the historical mixing information set, a plurality of historical basic performance sets are classified and obtained. According to a preset execution time threshold and a ratio of the plurality of historical stirring execution time sets, the plurality of historical basic performance sets are corrected and calculated to obtain a plurality of historical stirring performance sets.

4. The self-cleaning agitating and homogenizing system for variable pitch propeller blade type reaction vessel according to claim 1, wherein, respectively, and in time sequence, to obtain a plurality of historical stirring performance sequences, comprising: In a first historical stirring performance set in the plurality of historical stirring performance sets, a first reference performance value is obtained by selection; According to the difference between the first reference performance value and other historical stirring performances in the first historical stirring performance set, a distribution probability is assigned to obtain a first basic probability distribution, wherein the size of the difference and the size of the distribution probability are negatively correlated; According to the first basic probability distribution, the first historical stirring performance set is optimized and dimensionally reduced to obtain a first reduced performance set; According to the time stamp information of the plurality of historical stirring performances in the first reduced historical stirring performance set, the first historical stirring performance sequence is sorted to obtain a first historical stirring performance sequence; The plurality of historical stirring performance sets are optimized and dimensionally reduced and arranged in time sequence to obtain a plurality of historical stirring performance sequences.

5. The self-cleaning agitating and homogenizing system for variable pitch propeller blade type reaction vessel according to claim 4, characterized in that, According to the first basic probability distribution, the first historical stirring performance set is optimized and dimensionally reduced to obtain a first reduced performance set, comprising: In the first historical stirring performance set, a first reduced historical performance set is obtained by randomly extracting a preset number of historical stirring performances; According to the difference between the first reference performance value and other historical stirring performances in the first historical stirring performance set, a distribution probability is assigned to obtain a first basic probability distribution, wherein the size of the difference and the size of the distribution probability are negatively correlated; The similarity between the first reduced probability distribution and the first basic probability distribution is calculated as the first reduced fitness; Again, in the first historical stirring performance set, a second reduced historical performance set is obtained by randomly extracting a preset number of historical stirring performances, and a second reduced fitness is obtained by processing; Continue to optimize and reduce the dimension, until convergence, output the reduced historical performance set with the largest reduced fitness as the first reduced performance set.

6. The self-cleaning agitating and uniform mixing system of variable pitch propeller blade type reaction vessel according to claim 1, characterized in that, According to the plurality of historical stirring performance sequences, a mixing trend analysis is performed to obtain a mixing efficiency and a cleaning efficiency, comprising: According to the sample operation data of the plurality of reaction kettles, a sample stirring performance sequence set is collected, and a sample mixing efficiency set and a sample cleaning efficiency set are obtained according to the performance change identification in each sample stirring performance sequence; The sample stirring performance sequence set is used as the classification input, and the sample mixing efficiency set and the sample cleaning efficiency set are used as the classification output to construct a mixing trend analyzer; Based on the mixing trend analyzer, the plurality of historical stirring performance sequences are classified according to the mixing trend to obtain a plurality of working condition mixing efficiencies and a plurality of working condition cleaning efficiencies; The similarity between the target working condition information and the plurality of classified historical working conditions is analyzed, and the plurality of working condition mixing efficiencies and the plurality of working condition cleaning efficiencies are weighted calculated according to the size of the plurality of working condition similarities to obtain a mixing efficiency and a cleaning efficiency.

7. The self-cleaning agitating and uniform mixing system of variable pitch propeller blade type reaction vessel according to claim 6, characterized in that, The target working condition information is matched with the plurality of classified historical working conditions to obtain a matching historical working condition, and the mixing efficiency and the cleaning efficiency are corrected according to the deviation of the standard matching working condition information between the target working condition information and the matching historical working condition to obtain a corrected mixing efficiency and a corrected cleaning efficiency, comprising: selecting a classified historical working condition with the largest similarity as a matching historical working condition, and obtaining standard matching working condition information of the matching historical working condition; setting an efficiency correction coefficient according to a deviation between the target working condition information and the standard matching working condition information of the matching historical working condition; correcting and calculating the mixing efficiency and the cleaning efficiency by using the efficiency correction coefficient to obtain a corrected mixing efficiency and a corrected cleaning efficiency.

8. The self-cleaning agitating and uniform mixing system of variable pitch propeller blade type reaction vessel according to claim 1, characterized in that, performing mixing strategy decision according to the corrected mixing efficiency and the corrected cleaning efficiency to obtain a mixing strategy scheme, and performing mixing strategy execution, including: collecting a sample corrected mixing efficiency set and a sample corrected cleaning efficiency set, and setting a sample mixing strategy scheme according to a size of each sample corrected mixing efficiency and sample corrected cleaning efficiency to obtain a sample mixing strategy scheme set, wherein each sample mixing strategy scheme includes a strategy parameter, and the size of the sample corrected mixing efficiency and the sample corrected cleaning efficiency is negatively correlated with the size of the strategy parameter; using the sample corrected mixing efficiency set and the sample corrected cleaning efficiency set as decision input, and using the sample mixing strategy scheme set as decision output to construct a mixing strategy decision maker; using the mixing strategy decision maker to perform mixing strategy decision on the corrected mixing efficiency and the corrected cleaning efficiency to obtain a mixing strategy scheme.

9. The self-cleaning agitating and uniform mixing system of variable pitch propeller blade type reaction vessel according to claim 3, characterized in that, obtaining a plurality of historical working condition information in the historical stirring data, performing classification processing to obtain a plurality of classified historical working conditions, including: identifying working condition characteristic parameters in the plurality of historical working condition information, including paddle speed, material viscosity, and reaction temperature; performing clustering on the plurality of historical working condition information according to a similarity threshold of the working condition characteristic parameters to obtain a plurality of classified historical working conditions, wherein each classified historical working condition contains at least three similar working condition characteristic parameter combinations.

10. The self-cleaning agitating and uniform mixing system of variable pitch propeller blade type reaction vessel according to claim 5, characterized in that, randomly extracting a preset dimension reduction number of historical stirring performances in the first historical stirring performance set to obtain a first dimension reduction historical performance set, including: setting the preset dimension reduction number as one third of the original set number; using a random number generator to generate a corresponding number of random numbers in the index range of the first historical stirring performance set, extracting historical stirring performances corresponding to the indexes to form the first dimension reduction historical performance set.

Images