Verification system and verification method for viscometer

Abstract

The invention relates to the technical field of viscosity measurement, and discloses a verification system for a viscometer, comprising: a dynamic temperature compensation module for collecting temperature data inside and outside the viscometer in real time; the embedded controller is responsible for receiving the temperature data, executing a temperature compensation process and outputting a temperature compensation coefficient and an initial viscosity value; the multi-modal data fusion calibration module is used for receiving a temperature compensation coefficient provided by the multi-modal data fusion calibration module and data from a pressure sensor, a shear rate sensor and a vibration sensor, and carrying out joint modeling on the data by utilizing a weighted average method; the invention also discloses a verification method for the viscometer, which comprises the following steps: acquiring temperature data inside and outside the viscometer through the temperature sensor, and performing temperature compensation through the embedded controller. According to the invention, the multi-modal data fusion calibration module is introduced, and data from a plurality of sensors are integrated through a weighted average method, so that the optimized viscosity value is more comprehensive and reliable.

Description

Technical Field

[0001] The present invention relates to the technical field of viscosity measurement, in particular to a calibration system and a calibration method for a viscometer. Background Art

[0002] Accurately measuring liquid viscosity is crucial in many areas of industrial production and scientific research. As one of the most important physical properties of fluids, viscosity measurement accuracy directly impacts product quality, production process stability, and the reliability of scientific research results. Viscometers, key instruments for measuring viscosity, have become increasingly diverse with the continuous advancement of technology, including capillary viscometers, rotational viscometers, and falling-ball viscometers. To ensure the accuracy and reliability of viscometer measurements, regular calibration is essential.

[0003] Traditional technologies mostly rely on static temperature sensors. While this approach can provide basic temperature monitoring, it often faces various challenges in practical use. Temperature fluctuations directly affect the viscosity of the liquid, and a single temperature sensor often cannot fully respond to these changes, resulting in unstable measurement results. Furthermore, existing technologies often rely on a single or limited data source for data fusion, which results in a lack of breadth and depth in viscosity measurements. This is especially true for complex fluid properties, where comprehensive references are difficult to obtain, thus affecting the accuracy of the final results. In some older viscometers, the calibration process relies on manual adjustment and fixed parameters. This practice is not only time-consuming but also can lead to inconsistent calibration data due to differences in the properties of different liquid batches, reducing the efficiency of the device. Furthermore, traditional viscometers often lack the necessary self-learning capabilities to cope with changes in operating conditions. If the set parameters cannot adapt to the new fluid or environmental changes, the device's measurement accuracy will also decrease. This requires many users to constantly intervene manually during operation, increasing the complexity and inconvenience of operation. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a calibration system for viscometers, which solves the problem that traditional viscometers rely on static temperature sensors and cannot respond to temperature changes in real time, resulting in unstable measurement results. In addition, the fusion method of a single data source lacks comprehensiveness, affecting the measurement accuracy of complex fluids. The manual calibration process of old equipment is time-consuming and prone to inconsistencies. Traditional technologies that lack self-learning capabilities make the equipment unable to adapt to new fluids or environmental changes, increasing the need for manual intervention.

[0005] To achieve the above object, the present invention is implemented through the following technical solution: A calibration system for a viscometer, characterized by comprising the following steps:

[0006] Dynamic temperature compensation module collects temperature data inside and outside the viscometer in real time;

[0007] An embedded controller is responsible for receiving temperature data, performing a temperature compensation process, and outputting a temperature compensation coefficient and a preliminary viscosity value. The embedded controller is also connected to a pressure sensor, a shear rate sensor, and a vibration sensor to acquire and monitor data from each sensor in real time.

[0008] The multimodal data fusion calibration module receives the temperature compensation coefficients provided by the module and data from the pressure sensor, shear rate sensor, and vibration sensor. It uses the weighted average method to jointly model these data to optimize the viscometer's measurement parameters and feeds the calibration results back to the embedded controller.

[0009] The self-learning calibration process module is connected to the multimodal data fusion calibration module. Based on the calibration results fed back by the module, a physical simulation model of the viscosity response is constructed. The PPO algorithm is used to automatically adjust and optimize the calibration strategy of the viscometer, thereby generating a dynamic characteristic model.

[0010] 3. Preferably, the dynamic temperature compensation module includes:

[0011] By installing multiple temperature sensors inside and outside the viscometer, the temperature data inside and outside the viscometer are continuously monitored in real time;

[0012] A long short-term memory network (LSTM) model is used to dynamically model the temperature data of the viscometer;

[0013] The output formula of the LSTM model is:

[0014]

[0015] Among them, h t is the hidden state vector at the current time t, which represents the model's memory and state of the current input information; is the hidden state vector at the previous time t-1, which represents the model's memory state of the previous input information; t is the input vector at the current time t, representing the data W input to the LSTM unit h is the hidden state weight matrix, used to connect the previous hidden state h t-1 With the current calculation; W x is the input weight matrix, used to connect the current input x t With the current calculation; b h is the bias term of the hidden state, which adjusts the current calculation; f is the activation function, which is used to introduce nonlinear characteristics and calculate the current hidden state; is an intermediary variable, which represents the output result after the activation function is processed and is usually used in the following calculations; Wy is the output weight matrix, which is used to map the current hidden state to the output vector; b y is the output bias term, which adjusts the output result; t is the output vector at the current time t, which represents the result generated by the LSTM model based on the current state.

[0016] Preferably, the embedded controller includes:

[0017] a central processing unit for processing received temperature data and sensor data;

[0018] Memory, used to store temperature compensation coefficient, preliminary viscosity value and other related parameters;

[0019] Communication interface, used for data exchange with dynamic temperature compensation module, sensors and other modules;

[0020] The pressure sensor is installed between the liquid inlet and outlet of the viscometer to monitor the pressure changes of the fluid in real time;

[0021] Shear rate sensors are usually installed in the fluid flow area to provide the necessary shear rate information for viscosity calculation;

[0022] The vibration sensor is installed on the housing of the viscometer or at a location connected to the viscometer to monitor the vibration information generated by the viscometer during operation.

[0023] Preferably, the output temperature compensation coefficient and preliminary viscosity value include:

[0024] Sending the temperature compensation coefficient to the display via a formatted output signal;

[0025] Preliminary viscosity values are generated after data processing and output in standard units in accordance with relevant measurement standards.

[0026] Preferably, the multimodal data fusion calibration module includes:

[0027] The data from each sensor is fused through the weighted average algorithm. The specific algorithm is:

[0028]

[0029] Among them, V 最终 is the optimized viscosity value; V P 、V S 、V R are the viscosity values corresponding to pressure, shear rate and vibration data respectively; W1, W2 and W3 are the corresponding data weight coefficients.

[0030] Preferably, the self-learning calibration process module includes:

[0031] Generate viscosity response data through simulation models to reflect system behavior under different input conditions of temperature, pressure and shear rate;

[0032] Based on real-time feedback and historical data, the reinforcement learning method is used to optimize the calibration strategy, and the optimization formula for updating the calibration process is:

[0033]

[0034] Among them, θ t+1 is the parameter vector of the next iteration, indicating the new value of the model parameters after the update; θ t is the parameter vector of the current iteration, which represents the parameter value of the model at the current time t; α is the learning rate, which represents the step size or rate of parameter update. It controls the size of the parameter movement at each update. A larger value will increase the update step, resulting in unstable convergence; a too small value will lead to slow convergence. is the current parameter θ t The gradient of the loss function L represents the rate of change of the loss function with respect to the parameters.

[0035] Preferably, the building of the physical simulation model includes:

[0036] The system dynamics is described as the relationship between inputs and outputs using transfer functions or state-space models, specifically:

[0037]

[0038] Where G(s) is the transfer function of the system, which represents the relationship between input and output. s is a variable in the complex frequency domain, which is usually used to describe the dynamic behavior of a linear time-invariant system. Y(s) is the Laplace transform of the system output, which represents the performance of the system output signal in the complex frequency domain. X(s) is the Laplace transform of the system input, which represents the performance of the system input signal in the complex frequency domain. 4. b0, b1, ..., b n are the coefficients of the output equation, which are the coefficients corresponding to the various terms in the output signal Y(s), where b0 is the coefficient of the highest-order term, b1 is the coefficient of the second-highest-order term, and so on. n are the coefficients of the lowest-order terms; a0, a1, ..., a m are the coefficients of the input equation, which are the coefficients corresponding to the terms in the input signal X(s), where a0 is the coefficient of the highest-order term, a1 is the coefficient of the second-highest-order term, and so on. m is the coefficient of the lowest-order term; s n , s n-1 ,...,s 0represents the different orders in the complex frequency domain and is used to construct the polynomial of the output signal Y(s), with the highest order being n; s m , s m-1 ,...,s 0 Represents different orders in the complex frequency domain and is used to construct a polynomial of the input signal X(s), with the highest order being m.

[0039] Preferably, the PPO algorithm includes:

[0040] Optimize the strategy through the policy gradient method so that the action selected in a given state can maximize the cumulative reward;

[0041] The PPO algorithm can be defined as: using the constraints of proximal policy optimization during the update period so that the change between the new policy and the old policy is limited to a threshold, specifically expressed as:

[0042]

[0043] Among them, L CLIP (θ) represents the clipping loss function in the PPO algorithm; represents the expected value (i.e., average value) at time t, and represents the expected value calculation for the entire data sample; r t (θ) represents the ratio calculated using the current policy θ at time t; this ratio can be used to measure the ratio of the probability of the current policy selecting an action to the probability of the old policy; is the odds; π θ (a t |s t ) means in state s t Under the current strategy θ, action a is selected t probability; In state s t Next, the old strategy selects action a t probability; represents the advantage function at time t; the advantage function is used to measure the goodness of the current action relative to the average action, and is calculated as the expected return of the current action minus the benchmark return; clip(r t (θ), 1-ε, 1+ε) represents the contrast ratio r t (θ) is clipped; the ε hyperparameter is used to control the limit of the policy update; is the advantage function.

[0044] Preferably, the dynamic characteristic model includes:

[0045] The response characteristics of the viscometer under different working conditions, using a nonlinear regression model to establish the relationship between input temperature, pressure and shear rate and output viscosity;

[0046] Model equation description:

[0047] η=β0+β1T+β2P+β3S+∈;

[0048] Where η is the viscosity value; T, P, and S represent temperature, pressure, and shear rate, respectively; β0, β1, β2, and β3 are the coefficients of the model; and ∈ is a random error term used to capture the variation that the model cannot explain.

[0049] The present invention provides a calibration system for a viscometer, which has the following beneficial effects:

[0050] 1. This invention utilizes a dynamic temperature compensation module to collect real-time internal and external temperature data from the viscometer, achieving higher measurement accuracy. Compared to existing solutions using a single temperature sensor, this solves the problem of inaccurate viscosity measurements caused by temperature fluctuations.

[0051] 2. The present invention introduces a multimodal data fusion calibration module, which integrates data from multiple sensors through a weighted average method, making the optimized viscosity value more comprehensive and reliable. Compared with the limitation of traditional methods that rely on a single data source, it significantly improves the effectiveness of data fusion.

[0052] 3. This invention utilizes a self-learning calibration process module to automatically adjust the calibration strategy based on real-time feedback. This approach continuously improves viscometer performance through learning. Compared to existing technologies that rely on static models, this approach introduces dynamic optimization features, making measurement results more adaptable.

[0053] 4. This invention uses a physical simulation model combined with a PPO algorithm for optimization, enabling the viscometer to maintain excellent response characteristics under different operating conditions. Compared with the shortcomings of previous methods that have poor adaptability to environmental variables, this provides a more intelligent solution and improves the overall reliability of the device. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 Schematic diagram of the system architecture of the present invention;

[0055] Figure 2 Schematic diagram of the method of the present invention. DETAILED DESCRIPTION

[0056] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0057] Please see the attached Figure 1 , an embodiment of the present invention provides a calibration system for a viscometer, including.

[0058] Dynamic temperature compensation module collects temperature data inside and outside the viscometer in real time;

[0059] The main purpose of the dynamic temperature compensation module is to collect temperature data inside and outside the viscometer in real time to achieve dynamic compensation and improve measurement accuracy;

[0060] The module includes multiple temperature sensors, which are arranged inside and outside the viscometer to monitor temperature changes in real time. The internal sensor is responsible for monitoring the temperature of the fluid, while the external sensor monitors the ambient temperature.

[0061] First, a real-time temperature monitoring model is established through multiple temperature sensors. The temperature data of the model includes the internal temperature T in and external temperature T out ,The temperature data will be input to the embedded controller for processing.,Secondly, after receiving the temperature data, the embedded controller will perform dynamic modeling,through the long short-term memory network LSTM model to improve the accuracy of,temperature compensation;

[0062] The calculation formula of the model is:

[0063]

[0064] Among them, h t is the hidden state vector at the current time t, which represents the model's memory and state of the current input information; h t-1 is the hidden state vector at the previous moment t-1, which represents the model's memory state of the previous input information; t is the input vector at the current time t, representing the data W input to the LSTM unit h is the hidden state weight matrix, used to connect the previous hidden state h t-1 With the current calculation; W x is the input weight matrix, used to connect the current input x t With the current calculation; b h is the bias term of the hidden state, which adjusts the current calculation; f is the activation function, which is used to introduce nonlinear characteristics and calculate the current hidden state; is an intermediary variable, which represents the output result after the activation function is processed and is usually used in the following calculations; W y is the output weight matrix, which is used to map the current hidden state to the output vector; b y is the output bias term, which adjusts the output result; tis the output vector at the current time t, which represents the result generated by the LSTM model based on the current state.

[0065] An embedded controller is responsible for receiving temperature data, performing a temperature compensation process, and outputting a temperature compensation coefficient and a preliminary viscosity value. The embedded controller is also connected to a pressure sensor, a shear rate sensor, and a vibration sensor to acquire and monitor data from each sensor in real time.

[0066] The embedded controller's hardware structure includes a central processing unit (CPU), memory, communication interfaces, and sensor interfaces. The CPU is responsible for processing received temperature and sensor data, monitoring the status of each sensor in real time, and executing the temperature compensation algorithm. The memory is used to store temperature compensation coefficients, preliminary viscosity values, and other related parameters, providing persistent storage and fast access to data.

[0067] The pressure sensor is installed between the liquid inlet and outlet of the viscometer. It measures the force exerted by the fluid on the pressure sensor and outputs a voltage signal proportional to the fluid pressure. The voltage signal is converted into a digital signal by an analog-to-digital converter and input into the embedded controller. The embedded controller reads the signal and calculates the real-time pressure value for subsequent viscosity calculation.

[0068] Shear rate sensors are usually installed in the area where the fluid flows. They calculate the current shear rate by measuring the shear strain generated by the fluid flow. The sensor converts the measured shear rate into an electrical signal, which is then processed by the ADC and transmitted to the embedded controller. The controller obtains the signal and processes it into a shear rate value represented by a digital signal, providing the necessary data for viscosity calculation.

[0069] The vibration sensor is installed on the housing of the viscometer or at the position connected to the viscometer. The sensor detects the mechanical vibration generated by the viscometer during operation and outputs an electrical signal related to the vibration amplitude. After ADC conversion, the signal is transmitted to the embedded controller, which monitors the vibration signal V in real time. ib to evaluate the working stability of the equipment;

[0070] When receiving temperature data, the system first converts the temperature value T collected by the temperature sensor into in and ambient temperature T out For comparison, dynamic temperature compensation algorithm is used to calculate the temperature compensation coefficient K T , specifically:

[0071] K T =K base ·(T in -T out );

[0072] Among them, Kbase is the basic compensation coefficient, which represents the influence coefficient of temperature change on viscosity. This factor is obtained from the standard data initially set by the system and is continuously optimized based on historical data. In the data processing after temperature compensation, the embedded controller uses the current temperature compensation coefficient to correct the viscosity value calculation and generate the preliminary viscosity value V initial The specific formula is:

[0073] V initial =V raw +K T ;

[0074] Among them, V raw The temperature compensation coefficient and preliminary viscosity value obtained in the calculation process are sent to the display through the communication interface for users to view.

[0075] The multimodal data fusion calibration module receives the temperature compensation coefficients provided by the module and data from the pressure sensor, shear rate sensor, and vibration sensor. It uses the weighted average method to jointly model these data to optimize the viscometer's measurement parameters and feeds the calibration results back to the embedded controller.

[0076] First, the temperature compensation coefficient T is received from the embedded controller. comp , pressure data P, shear rate data And vibration data V, these data are connected to the embedded controller in real time through the communication interface and perform data interaction. The input data of multiple sensors can be expressed as follows: T comp is the temperature after compensation; P is the pressure of the fluid; is the shear rate of flow; V: vibration signal monitored by the vibration sensor;

[0077] Then, the collected data are jointly modeled using a weighted average algorithm to optimize the viscometer's measurement parameters;

[0078] The expression of the weighted average algorithm is:

[0079]

[0080] Among them, V 最终 is the optimized viscosity value; V P 、V S 、V R are the viscosity values corresponding to pressure, shear rate and vibration data respectively; W1, W2 and W3 are the corresponding data weight coefficients;

[0081] After processing this information, the embedded controller can output formatted temperature compensation coefficients and preliminary viscosity values, which will be presented to the user through a display for easy observation and recording;

[0082] Once the optimized viscosity value is obtained, it will be fed back to the embedded controller through the communication interface to update the device's viscosity measurement parameters. The embedded controller will use the received calibration results to improve the next viscosity measurement process, ensuring high-precision measurements of the viscometer under different conditions.

[0083] The design of this multimodal data fusion calibration module effectively reduces the impact of environmental changes on the viscometer, thereby improving the accuracy and reliability of viscometer measurements. This technical solution emphasizes the integrity of the dynamic fusion and adaptive calibration process, ensuring comprehensive disclosure of technical features to facilitate practical application by those skilled in the art. The clear physical and logical connections between each technical feature facilitate optimized real-time data processing and analysis.

[0084] The self-learning calibration process module is connected to the multimodal data fusion calibration module. Based on the calibration results fed back by the module, a physical simulation model of the viscosity response is constructed. The PPO algorithm is used to automatically adjust and optimize the calibration strategy of the viscometer, thereby generating a dynamic characteristic model.

[0085] The main purpose of the self-learning calibration process module is to use the calibration results fed back from the multimodal data fusion calibration module to build a physical simulation model of the viscosity response, thereby automatically adjusting and optimizing the viscometer calibration strategy to generate a dynamic characteristic model. This module includes the following steps:

[0086] First, the calibration results from the multimodal data fusion calibration module are received. The results include the optimized viscosity value and its related parameters. By analyzing this feedback information, the necessary data support can be provided for subsequent model generation. Based on the feedback results, a physical simulation model needs to be constructed. The physical simulation model is constructed by using a transfer function or a state space model to describe the system dynamics by the relationship between input and output. Specifically,

[0087]

[0088] Where G(s) is the transfer function of the system, which represents the relationship between input and output. s is a variable in the complex frequency domain, which is usually used to describe the dynamic behavior of a linear time-invariant system. Y(s) is the Laplace transform of the system output, which represents the performance of the system output signal in the complex frequency domain. X(s) is the Laplace transform of the system input, which represents the performance of the system input signal in the complex frequency domain. b0, b1, ..., b n are the coefficients of the output equation, which are the coefficients corresponding to the various terms in the output signal Y(s), where b0 is the coefficient of the highest-order term, b1 is the coefficient of the second-highest-order term, and so on. n are the coefficients of the lowest-order terms; a0, a1, ..., a mare the coefficients of the input equation, which are the coefficients corresponding to the terms in the input signal X(s), where a0 is the coefficient of the highest-order term, a1 is the coefficient of the second-highest-order term, and so on. m is the coefficient of the lowest-order term; s n , s n-1 ,...,s 0 represents the different orders in the complex frequency domain and is used to construct the polynomial of the output signal Y(s), with the highest order being n; s m , s m-1 , ..., s represents different orders in the complex frequency domain and is used to construct the polynomial of the input signal X(s), with the highest order being m;

[0089] Then, the proximal policy optimization (PPO) algorithm is used for model training and optimization. The algorithm automatically adjusts the calibration strategy through reinforcement learning to further improve the measurement accuracy. The core update formula of the PPO algorithm is expressed as:

[0090]

[0091] Among them, L CLIP (θ) represents the clipping loss function in the PPO algorithm; Represents the expected value at time t, representing the expected value calculation of the entire data sample; r t (θ) represents the ratio calculated using the current policy θ at time t; this ratio can be used to measure the ratio of the probability of the current policy selecting an action to the probability of the old policy; is the odds; π θ (a t |s t ) means in state s t Under the current strategy θ, action a is selected t probability; In state s t Next, the old strategy selects action a t probability; represents the advantage function at time t; the advantage function is used to measure the goodness of the current action relative to the average action, and is calculated as the expected return of the current action minus the benchmark return; clip(r t (θ), 1-ε, 1+ε) represents the contrast ratio r t (θ) is clipped; the ε hyperparameter is used to control the limit of the policy update; is the advantage function;

[0092] After each update, the current policy parameter θ is adjusted based on real-time feedback and historical data. The specific update formula is:

[0093]

[0094] Among them, θ t is the current strategy parameter; α is the learning rate; is the loss function, which indicates the effect of the strategy;

[0095] Ultimately, the generated dynamic characteristic model will include analytical characteristics under different operating conditions to achieve the viscometer's adaptive calibration capability. The establishment of this model ensures that the system has real-time calibration capabilities and can quickly adjust its measurement strategy according to changes in the external environment, thereby maintaining high accuracy and high stability.

[0096] The following description of a calibration method for a viscometer and the above description of a calibration system for a viscometer may refer to each other.

[0097] Please see the attached Figure 2 The present invention also provides a calibration system for a viscometer, comprising:

[0098] The temperature sensor collects the internal and external temperature data of the viscometer, and the embedded controller performs temperature compensation, calculates the temperature compensation coefficient and preliminarily estimates the viscosity value;

[0099] The embedded controller is connected to multiple sensors, including a pressure sensor, a shear rate sensor, and a vibration sensor, to acquire and monitor data from the multiple sensors in real time, providing multi-dimensional information required for viscosity measurement;

[0100] The temperature compensation coefficient and data from multiple sensors are input into the multimodal data fusion calibration module, where a weighted average method is used for joint modeling to optimize the viscometer's measurement parameters. The calibration results are then fed back to the embedded controller for updating.

[0101] According to the feedback of calibration results, the PPO algorithm is used to build a physical simulation model, and the calibration strategy of the viscometer is automatically optimized to generate a dynamic characteristic model.

[0102] The device of this embodiment can be used to execute the above method embodiment, and its principles and technical effects are similar, so they will not be repeated here.

[0103] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A calibration system for a viscometer, characterized in that: include: Dynamic temperature compensation module collects temperature data inside and outside the viscometer in real time; An embedded controller is responsible for receiving temperature data, performing a temperature compensation process, and outputting a temperature compensation coefficient and a preliminary viscosity value. The embedded controller is also connected to a pressure sensor, a shear rate sensor, and a vibration sensor to acquire and monitor data from each sensor in real time. The multimodal data fusion calibration module receives the temperature compensation coefficients and data from the pressure sensor, shear rate sensor, and vibration sensor, and uses the weighted average method to jointly model these data to optimize the viscometer's measurement parameters. The calibration results are then fed back to the embedded controller. The self-learning calibration process module is connected to the multimodal data fusion calibration module. Based on the calibration results fed back by the module, a physical simulation model of the viscosity response is constructed. The PPO algorithm is used to automatically adjust and optimize the calibration strategy of the viscometer, thereby generating a dynamic characteristic model.

2. A viscometer calibration system according to claim 1, characterized in that: The dynamic temperature compensation module includes: By installing multiple temperature sensors inside and outside the viscometer, the temperature data inside and outside the viscometer are continuously monitored in real time; A long short-term memory network (LSTM) model is used to dynamically model the temperature data of the viscometer; The output formula of the LSTM model is: Among them, h t is the hidden state vector at the current time t, which represents the model's memory and state of the current input information; is the hidden state vector at the previous time t-1, which represents the model's memory state of the previous input information; t is the input vector at the current time t, representing the data W input to the LSTM unit h is the hidden state weight matrix, used to connect the previous hidden state h t-1 With the current calculation; W x is the input weight matrix, used to connect the current input x t With the current calculation; b h is the bias term of the hidden state, which adjusts the current calculation; f is the activation function, which is used to introduce nonlinear characteristics and calculate the current hidden state; is an intermediary variable, which represents the output result after the activation function is processed and is usually used in the following calculations; W y is the output weight matrix, which is used to map the current hidden state to the output vector; b y is the output bias term, which adjusts the output result; t is the output vector at the current time t, which represents the result generated by the LSTM model based on the current state.

3. The viscometer calibration system according to claim 1, characterized in that: The embedded controller includes: a central processing unit for processing received temperature data and sensor data; Memory, used to store temperature compensation coefficient, preliminary viscosity value and other related parameters; Communication interface, used for data exchange with dynamic temperature compensation module, sensors and other modules; The pressure sensor is installed between the liquid inlet and outlet of the viscometer to monitor the pressure changes of the fluid in real time; Shear rate sensors are usually installed in the fluid flow area to provide the necessary shear rate information for viscosity calculation; The vibration sensor is installed on the housing of the viscometer or at a location connected to the viscometer to monitor the vibration information generated by the viscometer during operation.

4. The viscometer calibration system according to claim 1, characterized in that: The output temperature compensation coefficient and preliminary viscosity value include: Sending the temperature compensation coefficient to the display via a formatted output signal; Preliminary viscosity values are generated after data processing and output in standard units in accordance with relevant measurement standards.

5. The viscometer calibration system according to claim 1, characterized in that: The multimodal data fusion calibration module includes: The data from each sensor is fused through the weighted average algorithm. The specific algorithm is: Among them, V 最终 is the viscosity value after optimization; V P 、V S 、V R are the viscosity values corresponding to pressure, shear rate and vibration data respectively; W1, W2 and W3 are the corresponding data weight coefficients.

6. The viscometer calibration system according to claim 1, characterized in that: The self-learning calibration process module includes: Generate viscosity response data through simulation models to reflect system behavior under different input conditions of temperature, pressure and shear rate; Based on real-time feedback and historical data, the reinforcement learning method is used to optimize the calibration strategy, and the optimization formula for updating the calibration process is: Among them, θ t+1 is the parameter vector of the next iteration, indicating the new value of the model parameters after the update; θ t is the parameter vector of the current iteration, which represents the parameter value of the model at the current time t; α is the learning rate, which represents the step size or rate of parameter update. It controls the size of the parameter movement at each update. A larger value will increase the update step, resulting in unstable convergence; a too small value will lead to slow convergence. is the current parameter θ t The gradient of the loss function L represents the rate of change of the loss function with respect to the parameters.

7. The viscometer calibration system according to claim 1, characterized in that: The construction of the physical simulation model comprises: The system dynamics is described as the relationship between inputs and outputs using transfer functions or state-space models, specifically: Where G(s) is the transfer function of the system, which represents the relationship between input and output. s is a variable in the complex frequency domain, which is usually used to describe the dynamic behavior of a linear time-invariant system. Y(s) is the Laplace transform of the system output, which represents the performance of the system output signal in the complex frequency domain. X(s) is the Laplace transform of the system input, which represents the performance of the system input signal in the complex frequency domain.

4. b0, b1, ..., b n are the coefficients of the output equation, which are the coefficients corresponding to the various terms in the output signal Y(s), where b0 is the coefficient of the highest-order term, b1 is the coefficient of the second-highest-order term, and so on. n are the coefficients of the lowest-order terms; a0, a1, ..., a m are the coefficients of the input equation, which are the coefficients corresponding to the terms in the input signal X(s), where a0 is the coefficient of the highest-order term, a1 is the coefficient of the second-highest-order term, and so on. m is the coefficient of the lowest-order term; s n , s n-1 ,...,s 0 represents the different orders in the complex frequency domain and is used to construct the polynomial of the output signal Y(s), with the highest order being n; s m , s m-1 , ..., s represent different orders in the complex frequency domain and are used to construct polynomials of the input signal X(s), with the highest order being m.

8. The viscometer calibration system according to claim 1, characterized in that: The PPO algorithm includes: Optimize the strategy through the policy gradient method so that the action selected in a given state can maximize the cumulative reward; The PPO algorithm can be defined as: using the constraints of proximal policy optimization during the update period so that the change between the new policy and the old policy is limited to a threshold, specifically expressed as: Among them, L CLIP (θ) represents the clipping loss function in the PPO algorithm; represents the expected value (i.e., average value) at time t, and represents the expected value calculation for the entire data sample; r t (θ) represents the ratio calculated using the current policy θ at time t; this ratio can be used to measure the ratio of the probability of the current policy selecting an action to the probability of the old policy; is the odds; π θ (a t |s t ) means in state s t Under the current strategy θ, action a is selected t probability; In state s t Next, the old strategy selects action a t probability; represents the advantage function at time t; the advantage function is used to measure the goodness of the current action relative to the average action, and is calculated as the expected return of the current action minus the benchmark return; clip(r t (θ), 1-∈, 1+∈) represents the contrast ratio r t (θ) is clipped; ∈ hyperparameter, used to control the limit range of policy update; is the advantage function.

9. The viscometer calibration system according to claim 1, characterized in that: The dynamic characteristic model includes: The response characteristics of the viscometer under different working conditions, using a nonlinear regression model to establish the relationship between input temperature, pressure and shear rate and output viscosity; Model equation description: η=β0+β1T+β2P+β3S+∈; Where η is the viscosity value; T, P, and S represent temperature, pressure, and shear rate, respectively; β0, β1, β2, and β3 are the coefficients of the model; and ∈ is a random error term used to capture the variation that the model cannot explain.

10. A method for calibrating a viscometer, applied to a calibration system for a viscometer according to any one of claims 1 to 9, characterized in that: The following steps are involved: The temperature sensor collects the internal and external temperature data of the viscometer, and the embedded controller performs temperature compensation, calculates the temperature compensation coefficient and preliminarily estimates the viscosity value; The embedded controller is connected to multiple sensors, including a pressure sensor, a shear rate sensor, and a vibration sensor, to acquire and monitor data from the multiple sensors in real time, providing multi-dimensional information required for viscosity measurement; The temperature compensation coefficient and data from multiple sensors are input into the multimodal data fusion calibration module, where a weighted average method is used for joint modeling to optimize the viscometer's measurement parameters. The calibration results are then fed back to the embedded controller for updating. Based on the feedback of calibration results, the PPO algorithm is used to build a physical simulation model, and the calibration strategy of the viscometer is automatically optimized to generate a dynamic characteristic model.

Images