Self-adaptive real-time calibration method based on impulse method injection rule test force sensor

Abstract

The invention relates to the field of fuel injector testing, in particular to a self-adaptive real-time calibration method for testing a force sensor based on an impulse method injection rule, which comprises the following steps: firstly, carrying out basic dynamic calibration by using a volumetric method or a weighing method or a fuel injection quantity separator, and further utilizing an advanced sensor fusion algorithm, a real-time feedback loop and a machine learning technology; sensor reading drift or error due to fluctuations in environmental or operating conditions is detected and compensated. In the self-adaptive calibration process, calibration parameters of the sensor are regularly adjusted based on real-time data, and it is ensured that the force sensor always keeps accurate calibration regardless of external influences or test condition changes. The dynamic calibration method provided by the invention not only improves the accuracy and reliability of force measurement, but also reduces the demand of manual intervention, so that the test process is more simplified and automated.

Description

Technical Field

[0001] This invention relates to the field of fuel injector testing, and in particular to an adaptive real-time calibration method for a force sensor based on the impulse method for testing injection law. Background Technology

[0002] In traditional systems, force sensors typically require manual calibration before the start of a test procedure. Calibration is usually done by comparing sensor readings to known reference values, which can require specialized equipment and significant time. While effective in well-controlled test environments, manual calibration becomes extremely tedious when testing dynamic or complex systems, especially in real-time testing or long-term experimental runs. Furthermore, force sensor calibration tends to degrade over time, particularly under fluctuating conditions such as temperature variations, fluid viscosity changes, pressure fluctuations, or sensor drift. These factors can lead to measurement errors, resulting in inaccurate sensor data and consequently affecting the reliability of test results. In many cases, engineers must periodically recalibrate sensors during testing, which can cause interruptions, delays, and potential errors in data collection.

[0003] Sensor drift is a particularly challenging problem in force measurement systems. As sensors are exposed to changing environmental factors, such as temperature variations or changes in fluid composition, their accuracy can degrade. Drift in sensor readings can lead to significant inaccuracies that accumulate over time, and without frequent recalibration, this drift can distort test results, resulting in misleading interpretations of system performance. Traditional systems rely on periodic manual adjustments to compensate for drift, which are typically labor-intensive and susceptible to human error. Furthermore, dynamic test conditions, such as fluctuations in fluid properties (density, viscosity, pressure, and temperature), pose additional challenges to maintaining the accuracy of force sensors. Force sensors may exhibit nonlinear behavior in response to these changing conditions, further increasing the complexity of their calibration. In such cases, simple static calibration techniques are insufficient to ensure accurate measurements because they fail to account for the real-time variability of the test environment. Summary of the Invention

[0004] The purpose of this invention is to address the problems existing in the prior art. The adaptive real-time calibration system of this invention provides an innovative solution. This system continuously monitors sensor data, detects calibration drift or measurement errors caused by environmental factors, and dynamically adjusts calibration parameters during testing. By combining sensor fusion algorithms and machine learning techniques, the system can analyze sensor data in real time, identify discrepancies, and automatically correct them. This closed-loop feedback mechanism ensures that calibration parameters remain optimal throughout the testing process, thereby maintaining the accuracy of sensor readings without manual intervention. The adaptive real-time calibration system offers significant improvements over traditional systems, eliminating the need for manual recalibration and minimizing errors caused by environmental fluctuations. It also enables continuous, uninterrupted testing, thereby improving data collection efficiency, reducing testing time, and increasing the reliability of test results. This is particularly valuable in high-performance testing environments, where time is critical, and accurate sensor data is essential to ensuring the optimal performance of fluid jet systems. This invention also addresses the problem of sensor drift, which can accumulate over time and lead to performance degradation in traditional systems. By continuously adjusting calibration parameters based on real-time data, the system ensures that sensors maintain accuracy even under fluctuating testing conditions. Furthermore, the system is designed to be modular, allowing it to adapt to new sensors or changes in test configurations without requiring major redesign.

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

[0006] An adaptive real-time calibration method for a force sensor based on the impulse method for testing injection laws, as disclosed in this invention, includes the following steps:

[0007] Step 1: Establish the experimental scenario by placing the fuel injector in the constant volume projectile. A force sensor is placed outside the nozzle of the fuel injector, and sensors for measuring environmental parameters are also placed inside the constant volume projectile.

[0008] Step 2: Using a force sensor, measure the static calibration coefficient of the force sensor using a static calibration method, and input it into the database for storage;

[0009] Step 3: Based on the static calibration coefficient, using a force sensor, the dynamic calibration coefficient of the fuel injector is calculated through the impulse method and other cyclic injection quantity test methods, and then entered into the database for storage. In addition, the environmental parameters of the fluid injection system during the test of the impulse method and other cyclic injection quantity test methods are also entered into the database for storage.

[0010] Step 4: Repeat the tests in Step 2 and Step 3 to continuously obtain new static calibration coefficients, new dynamic calibration coefficients, and new environmental parameters. Input these into the database and save them. Together with the historical static calibration coefficients, historical dynamic calibration coefficients, and historical environmental parameters obtained before Step 2 and Step 3, they will form a reference calibration value in the database.

[0011] Step 5: Perform real-time adaptive dynamic calibration;

[0012] First, the real-time data collected by each sensor, including static calibration coefficients and dynamic calibration coefficients, are input into the data preprocessing module for preprocessing.

[0013] Then, some of the preprocessed data (i.e., the cleaned data) is directly input into the real-time adaptive calibration module, while some of the preprocessed data and reference calibration values ​​(i.e., historical static calibration coefficients, historical dynamic calibration coefficients, historical environmental parameters, and purely theoretical calculation calibration data, etc.) are checked for initial calibration by the initial calibration module to obtain initial calibration parameters. The reference calibration values ​​are also input into the real-time adaptive calibration module as reference calibration data. At the same time, environmental parameters (such as temperature, pressure, etc.), fuel characteristic parameters (such as density, viscosity, etc.), and test conditions (injection pressure, injection pulse width, etc.) are also input into the real-time adaptive calibration module.

[0014] The purely theoretical calibration data is obtained by applying Bernoulli's equation upstream and downstream of the injector orifice. Compared with the orifice exit velocity, the velocity upstream of the orifice inlet is considered negligible, thus obtaining the theoretical fuel flow velocity at the orifice exit. In the formula, u th p is the theoretical fuel flow rate at the nozzle outlet. i p is the fuel injection pressure at the nozzle inlet. b The ambient pressure at the nozzle exit;

[0015] Among them, fuel characteristic parameters specifically refer to the physical characteristics of the fuel itself, such as density, viscosity, modulus, etc. Environmental parameters refer to the environment in which the fuel is injected into the injector, such as the pressure, temperature, etc. provided by the container when it is injected into a constant volume bomb. Test conditions refer to the injection pressure, injection pulse width, etc. provided to the injector by the equipment that generates injection behavior on a real engine, such as the high-pressure common rail injection system. The relationship between the three is parallel and they do not have an inclusive relationship.

[0016] Finally, through the real-time adaptive calibration module and the real-time sensor calibration application module (the data obtained from the adaptive module enters the real-time sensor calibration application module for the actual sensor calibration process), high-quality, calibrated data is output through internal processing, including comparison with historical data, empirical parameter correction, and AI training model correction, for use by the analysis software.

[0017] As a further preferred option, the sensors used to measure environmental parameters in step one include an ambient temperature sensor, an ambient humidity sensor, and an ambient pressure sensor.

[0018] As a further preferred option, in step two, the static calibration method uses the weight method. During the calibration process, five different calibration points were selected to test the sensor performance within different ranges. Ten repeated trials were performed at each calibration point, and the average value was taken. The actual curve used is the fitted curve obtained by the least squares method, and the fitted curve is as follows:

[0019] (1)

[0020] In the formula, x is the applied force value, and y is the output electrical signal.

[0021] As a further preferred approach, the reference value of the static calibration method is measured by calculating the Pearson correlation coefficient, which is defined as follows:

[0022] (2)

[0023] in and The sample average value of force and output voltage is used to calculate r. If r is close to 1, there is a strong linear relationship between the output voltage signal of the force sensor and the applied force, and the static calibration of the force sensor response can be used as a reference value. If r is close to 0, there may be a nonlinear relationship between the output of the force sensor and the applied force, which requires further correction or analysis.

[0024] As a further preferred option, in step three, the impulse method assumes the fluid inside the nozzle is a liquid and simplifies the geometric flow area of ​​the nozzle based on the law of conservation of mass, thereby obtaining the relationship between the effective flow area, transient jet rate, and spray momentum flow (impulse) and the average velocity at the nozzle exit:

[0025] (3)

[0026] (4)

[0027] In the formula: The transient jetting rate at the nozzle exit; For spray momentum; u m To simplify the average velocity at the nozzle exit; ρ f For fuel density, A geo The geometric flow cross-sectional area of ​​the nozzle;

[0028] The relationship between the fuel injection rate by volume and the simplified average exit velocity is as follows:

[0029] (5)

[0030] Combining equations (4) and (5), the relationship between the fuel injection rate and the spray impulse at the nozzle outlet is as follows:

[0031] (6)

[0032] Equation (6) leads to the conclusion that the fuel injection rate of each injector hole can be calculated by measuring the instantaneous spray impulse of each hole.

[0033] In actual measurement, a simplified result was obtained:

[0034] (7)

[0035] Therefore, by modifying equation (6), the injection rate is:

[0036] (8)

[0037] By measuring the spray impact force of a single nozzle using formula (8), the injection rate of that nozzle can be calculated. The obtained fuel injection rate formula (8) is then integrated according to the actual injection duration, and the result is the cyclic injection quantity of each nozzle:

[0038] (9)

[0039] In the formula, q f This refers to the amount of oil injected during the nozzle cycle.

[0040] As a further preferred option, the other method for testing the cyclic injection quantity in step three is the volumetric method, which measures the cyclic injection quantity q. v To allow for cyclic injection volumes applicable to multiple operating conditions, the dynamic calibration factor for a specific operating condition is:

[0041] .

[0042] As a further preferred option, the data preprocessing module cleans the raw data, including denoising, filling out outliers, and standardizing the data to make it suitable for use in subsequent calibration processes.

[0043] As a further preferred solution, the system also features a continuous feedback loop to monitor sensor data, test parameters, and environmental parameters in real time. When data drift is detected (e.g., force signal deviation > 0.5%FS / min) or sudden environmental changes (temperature change rate > 1℃ / s, pressure change > 5%), a dynamic calibration cycle is automatically triggered. The calibration parameters are updated through a real-time adaptive calibration module to ensure the timeliness and accuracy of the calibration output.

[0044] As a further preferred option, the training method of the real-time adaptive calibration module is based on an AI-based prediction model (gradient boosting regressor / LSTM / feedforward neural network) trained according to historical calibration data. The input is real-time force signal, temperature, pressure, fuel density, viscosity and operating parameters, and the output is dynamic calibration coefficient or correction factor. Through supervised learning, a mapping relationship between multiple parameters and calibration deviation is established to achieve adaptive prediction of calibration parameters.

[0045] As a further preferred approach, fuzzy logic-based calibration employs a rule base of predefined empirical correction parameters (example: IF Temperature = High AND Viscosity = Low THEN Correction Factor = 0.97). Input variables (temperature, pressure, density, viscosity, original force value) are processed by the Mamdani inference system and defuzzified using the centroid method to generate calibration coefficient correction factors. This factor acts multiplicatively on the basic calibration coefficients, achieving real-time compensation for environmental and fuel characteristics.

[0046] This invention provides an adaptive real-time calibration method for a force sensor based on the impulse method for measuring jet propulsion. It offers the following advantages:

[0047] (1) Multi-stage error compensation mechanism: By integrating a three-layer technical architecture of static calibration, benchmark dynamic calibration and real-time adaptive dynamic calibration, the accuracy and environmental adaptability of fuel injection testing are significantly improved. In the static calibration stage, multi-point repeated loading (at least 5 calibration points, 10 tests at each point) and least squares fitting are used to reduce the nonlinear error and zero drift of the sensor from the source; the benchmark dynamic calibration corrects the momentum calculation deviation caused by uneven fuel atomization through cross-validation of the impulse method and the volume method; and the real-time adaptive calibration further introduces the LSTM or gradient boosting regressor model, laying the technical foundation for high-precision testing.

[0048] (2) Full-parameter environmental adaptability Facing complex environmental interference, the system exhibits strong robustness through multi-parameter fusion and intelligent compensation mechanisms. For example, the integrated temperature sensor data dynamically adjusts the calibration coefficient, effectively suppressing thermal drift under extreme temperatures; the environmental pressure adaptive module corrects the fuel density and flow rate model based on real-time air pressure, avoiding momentum calculation deviations under high-altitude low-pressure or boosted conditions; at the same time, the isolated signal processing circuit and software filtering algorithm work together to eliminate electrical noise interference in high humidity environments.

[0049] (3) Improved applicability across multiple fuels and operating conditions: The system's design compatibility allows it to seamlessly adapt to various fuels and complex operating conditions. For traditional diesel and biodiesel, the momentum model is automatically updated by inputting real-time density and viscosity parameters. For the low viscosity of hydrogen fuel or the high volatility of methanol, the gradient boosting regressor can effectively handle non-Newtonian fluid characteristics, avoiding calibration failures caused by differences in physical properties. In terms of operating condition adaptability, the system supports 20~300 MPa ultra-high pressure injection and multi-stage injection strategies. It uses an LSTM model to analyze the superposition effect of transient impact forces, thereby accurately capturing flow changes in complex modes such as pre-injection, main injection, and post-injection. This flexibility makes it not only suitable for diesel engine high-pressure common rail systems, but also extend to high-value scenarios such as aero-engine fuel nozzle testing and hydrogen fuel cell injector verification. Attached Figure Description

[0050] Figure 1 Hardware layout diagram;

[0051] Figure 2 Data flow diagram;

[0052] Figure 3 Real-time calibration flowchart. Detailed Implementation

[0053] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0054] In the field of fluid dynamics testing, particularly in the testing and optimization of fluid injection systems such as fuel injectors, aerospace propulsion systems, and industrial fluid distribution equipment, accurate measurement of the forces exerted by the fluid is crucial for understanding system performance. In flow-momentum injection testing, force sensors are widely used to measure the forces generated during injection events. These measurements help engineers evaluate various aspects of the system, such as fuel efficiency, emissions, flow control, and system reliability. However, accurately calibrating these force sensors under different operating conditions is a significant challenge.

[0055] This invention introduces an adaptive real-time calibration method for force sensors based on the impulse method for testing injection patterns. This process is crucial in the evaluation and optimization of injection systems in fluid dynamics and fuel injection technology. The system addresses the problem of accurate dynamic calibration of force sensors operating under complex time-varying conditions during testing. Force sensors are essential for accurately measuring the spray impact force during fuel injection, which is vital for understanding and optimizing the performance of automotive engines and industrial injection systems in various applications. In traditional fuel injection pattern testing, force sensors typically require manual calibration (e.g., using weights), which is time-consuming and error-prone. Furthermore, test conditions, including variations in temperature and pressure, as well as the properties of the injected fuel, can affect sensor accuracy over time, leading to sensor reading drift.

[0056] This invention overcomes these limitations by introducing an adaptive calibration mechanism that continuously monitors and adjusts the force sensor calibration in real time, ensuring data accuracy and consistency throughout the testing process. The system first performs basic dynamic calibration using a volumetric method, weighing method, or fuel injection separator. It then utilizes advanced sensor fusion algorithms, real-time feedback loops, and machine learning techniques to detect and compensate for sensor reading drift or errors caused by fluctuations in environmental or operating conditions. The adaptive calibration process periodically adjusts the sensor's calibration parameters based on real-time data, ensuring that the force sensor remains accurately calibrated regardless of external influences or changes in test conditions. This dynamic calibration method not only improves the accuracy and reliability of force measurements but also reduces the need for manual intervention, thus simplifying and automating the testing process. Furthermore, the system includes an intelligent compensation algorithm that considers changes in fluid properties (such as viscosity and density) during testing. The term "adaptive real-time calibration framework" refers to a series of techniques specifically designed to handle the impact of changes in fuel physical properties (density and viscosity) on measurement accuracy. Its core purpose is to ensure that the physical property parameters in the formula are real-time and accurate, or to directly correct errors caused by changes in physical properties through intelligent models / rules. For example, parameter updates based on physical models: In the core formula of the impulse method, fuel density is a key parameter. The compensation algorithm acquires or estimates the current fuel density in real time (possibly through table lookup-temperature correspondence, online densitometer, or calculation based on a model of fuel type and temperature), and dynamically substitutes it into the formula to calculate the injection rate and injection quantity, compensating for errors caused by density changes. Viscosity may be compensated for in a similar way for its impact on the model (especially in AI models) or the sensor itself. Implicit compensation based on AI models: The input of a trained AI model includes parameters such as fuel density and viscosity (and possibly type encoding). When predicting real-time calibration coefficients or correction force values, the model inherently learns and compensates for the impact of changes in these physical property parameters on the relationship between sensor readings and actual physical quantities (force, injection quantity). Rule-based compensation based on fuzzy logic: The rule base for empirical correction parameters of fuzzy logic explicitly includes rules such as "IF viscosity IS high THEN…" and "IF density IS low THEN…". These rules output corresponding calibration coefficient correction factors based on the real-time input fuzzy values ​​such as viscosity and density, thus compensating for changes in these physical properties. By adapting to these changes, the system ensures that the force sensor maintains accurate readings even when the properties of the injected fluid change. This adaptability is particularly suitable for applications with different types of fluids or variable test conditions, ensuring that the system can handle a wide range of test scenarios. In addition to improving sensor accuracy, this adaptive real-time calibration system also improves the overall efficiency of flow momentum injection testing.It allows for continuous, uninterrupted data collection without recalibration, significantly reducing test time and increasing productivity. Integration with modern data acquisition systems further facilitates seamless data integration, storage, and analysis, providing real-time performance insights into the tested injection system. By implementing this adaptive calibration system, industries involved in fluid dynamics testing, fuel injection systems, and precision instrumentation benefit from more accurate, reliable, and efficient testing processes capable of handling diverse operating conditions. This innovation enhances the ability to optimize injection system performance, improve fuel efficiency, and ensure the accuracy and reliability of force measurements in critical testing environments, automatically adjusting to changes in environment, fuel properties, and operating conditions during the testing process.

[0057] An adaptive real-time calibration system for force sensors used in flow momentum injection testing;

[0058] It includes multiple sensors for measuring the physical parameters of the injection system, including injection pressure, spray impact force, ambient temperature, and ambient pressure, providing real-time data related to the characteristics of the injected fuel and test conditions; it also includes a data preprocessing module that applies filtering technology to remove noise, smooth data, and process outliers and irrelevant data, ensuring that the sensor data has been cleaned before calibration, and is used to clean and filter the raw sensor data in preparation for calibration.

[0059] It includes an initial calibration module, which compares the cleaned sensor data with known reference values, i.e., inputting static calibration and basic dynamic calibration data to generate initial calibration parameters;

[0060] It includes a Real-Time Adaptive Calibration (RTCE) module (Real-Time Adaptive Calibration, Closed-Loop Feedback) for processing data from post-cleaning, initial calibration parameters, and continuously updated calibration parameters in the feedback loop. It adjusts the calibration parameters, taking into account drift or errors in the data, and operates in closed-loop feedback mode. In this mode, the calibration parameters are continuously adjusted based on real-time sensor data without manual intervention. The RTCE module utilizes sensor fusion algorithms and machine learning techniques to continuously analyze the received sensor data in conjunction with environmental parameters (such as temperature and pressure), fuel characteristic parameters (such as density and viscosity), and test condition parameters (injection pressure, injection pulse width, etc.) to determine the optimal calibration adjustment to minimize measurement errors.

[0061] It includes a real-time sensor calibration application module, which receives updated calibration parameters from the real-time adaptive calibration module, performs dynamic calibration tasks, combines raw sensor data with real-time updated calibration parameters, outputs accurate and usable calibration data, and feeds it back to the adaptive calibration module to continuously improve system performance.

[0062] It includes a feedback loop for continuously monitoring and updating calibration parameters. When additional sensor data is received during testing, it performs corrections and compensations to apply the adjusted calibration parameters to the sensor data in real time. The calibration parameters are continuously and dynamically updated according to the feedback loop to ensure that the force sensor remains accurate throughout the testing process, regardless of changes in external environment or operating conditions. The feedback loop uses real-time sensor data to detect any measurement discrepancies or drift, provides automatic compensation, and ensures that the system remains accurate throughout long-term testing.

[0063] It includes a final data output module for outputting calibrated sensor data for performance analysis, reporting, or further testing. The final calibrated sensor data is provided in a standardized format for easy interpretation by downstream analysis systems, thereby improving data interpretation efficiency.

[0064] An adaptive real-time calibration method of the present invention is used for a force sensor in the injection pattern test of injectors based on the impulse method, comprising the following steps:

[0065] 1. Collect raw data from multiple force sensors configured to measure parameters such as force, pressure, temperature, and fluid properties, as well as operating condition parameters transmitted from the test bench;

[0066] 2. Preprocess the collected data to remove noise and filter out irrelevant or erroneous readings;

[0067] 3. Perform preliminary calibration based on known reference values ​​or preset calibration settings;

[0068] 4. Upon receiving new sensor data, adjust calibration parameters in real time and use adaptive algorithms to adjust for data drift, measurement errors, and changing conditions;

[0069] 5. Apply the adjusted calibration parameters to the sensor data in real time to ensure accurate force measurement;

[0070] 6. Output calibrated sensor data for performance analysis, reporting, or further testing.

[0071] The adaptive algorithm of this invention uses machine learning or sensor fusion technology to evaluate and adjust calibration parameters based on continuous feedback from sensor data during the testing process.

[0072] The following description, in conjunction with the accompanying drawings of the embodiments of the present invention, addresses the issue that existing calibration techniques for spray impact force sensors used in spray tests are susceptible to the influence of operating environments such as temperature, humidity, and ambient pressure. Furthermore, they are prone to zero drift due to changes in actual operating conditions and fuel composition. Existing static calibration techniques often employ weight methods or other load-applying methods, inevitably leading to inaccurate calibration results and failing to provide high-precision measurement and real-time correction of the transient impact force of fluid spray. This application primarily provides an adaptive real-time calibration method for force sensors based on the impulse method for testing spray patterns.

[0073] To achieve the above objectives, this application employs the following method: First, a static calibration method is used to obtain the original calibration coefficients of the force sensor as the basic calibration parameters for measuring injection characteristics using the impulse method. Second, a benchmark dynamic calibration method is used. First, the original calibration coefficients are substituted into the sensor to calculate the cyclic injection quantity of each injector orifice using the impulse method-based injection characteristic calculation method. Simultaneously, using the same injector, different cyclic injection quantity test methods are used to obtain the cyclic injection quantity. This is used as a basis and compared with the cyclic injection quantity calculated by the impulse method-based injection characteristic calculation method to obtain the benchmark dynamic calibration coefficients. Third, a real-time adaptive dynamic calibration system is used. Environmental parameters such as temperature and pressure, fuel characteristic parameters such as fuel density and viscosity, and test conditions such as injection pressure and injection pulse width are input into the adaptive dynamic calibration system. Real-time adaptive calibration is performed based on methods such as gradient boosting regressors, long short-term memory networks, and feedforward neural networks to maintain transient accuracy of the force sensor, ultimately improving the testing accuracy of the injection characteristics of each injector orifice.

[0074] The static calibration method uses a weight of known mass to apply a standard force and measures the output signal of the force sensor. The accuracy of the sensor is determined by comparing the standard force with the sensor's output.

[0075] The weight method is a commonly used static calibration method for measuring the accuracy of force sensors. It involves applying a standard force using weights of known mass and measuring the force sensor's output signal U. The sensor's accuracy can be determined by comparing the output relationship between the standard force and the sensor's output. During calibration, five different calibration points were selected to test the sensor's performance within different ranges. Ten repeated trials were performed at each calibration point, and the average value was taken to reduce measurement error. Due to various factors, the sensor characteristics may exhibit some nonlinearity and bias. Therefore, the actual curve used is the fitted curve obtained by the least squares method, and the fitted curve is as follows:

[0076] (1)

[0077] In the formula, x is the applied force and y is the output electrical signal.

[0078] The Pearson correlation coefficient is a statistical method used to measure the strength and direction of the linear relationship between two variables. It is typically used to measure the linear relationship between two continuous variables, and its definition is as follows:

[0079] (2)

[0080] in and The value of r is calculated from the sample average of the force and the output voltage. If r is close to 1, there is a strong linear relationship between the sensor's output voltage signal and the applied force, meaning that the sensor's static response calibration can be used as a reference value. If r is close to 0, there may be a non-linear relationship between the sensor's output and the applied force, requiring further correction or analysis.

[0081] The benchmark dynamic calibration method, taking diesel injector testing as an example, involves using a fuel pump test bench to drive the fuel injection pump. The injector is connected to the fuel injection pump via a high-pressure fuel line. A force sensor to be tested is installed on the injector at a certain distance from the orifice. The spray momentum of each orifice is proportional to the flow rate of that orifice. When the fuel pump test bench is started, fuel is sprayed out through the nozzle on the injector. Using a statically calibrated force sensor, the fuel jet impact force of the same orifice is used as the dynamic load. Based on the law of conservation of momentum and relevant injector parameters, the spray impact force received by the sensor is used to calculate and convert the cyclic injection quantity. Using the same injector, the cyclic injection quantity is measured using methods such as the volumetric method as a basic reference value. The reciprocal of the ratio of the cyclic injection quantity calculated by the impulse method to the cyclic injection quantity calculated by the volumetric method is the basic dynamic calibration parameter. Calibration is performed under different operating conditions to obtain dynamic calibration parameters under different operating conditions.

[0082] Existing research has described the impulse method in detail; this section summarizes that research. The fluid inside the nozzle is assumed to be liquid, and the geometric flow area of ​​the nozzle is simplified based on the law of conservation of mass, thus obtaining the effective flow area. The measurement principle is illustrated in the diagram. The relationship between transient jet rate and spray momentum flow (impulse) and the average velocity at the nozzle exit is as follows:

[0083] (3)

[0084] (4)

[0085] In the formula: The transient jetting rate at the nozzle exit; For spray momentum; u mTo simplify the average velocity at the nozzle exit; ρ f For fuel density, A geo Let be the geometric flow cross-sectional area of ​​the nozzle.

[0086] The relationship between the fuel injection rate by volume and the simplified average exit velocity is as follows:

[0087] (5)

[0088] Combining equations (4) and (5), the relationship between the fuel injection rate and the spray impulse at the nozzle outlet is as follows:

[0089] (6)

[0090] Equation (6) shows that the fuel injection rate of each orifice of the injector can be calculated by measuring the instantaneous spray impulse of each orifice.

[0091] In actual measurement, a simplified result was obtained:

[0092] (7)

[0093] Therefore, by modifying equation (6), the injection rate is:

[0094] (8)

[0095] By measuring the spray impact force of a single nozzle using formula (8), the injection rate of that nozzle can be calculated. The obtained fuel injection rate formula (8) is then integrated according to the actual injection duration, and the result is the cyclic injection quantity of each nozzle:

[0096] (9)

[0097] In the formula, q f This refers to the amount of oil injected during the nozzle cycle.

[0098] The cyclic fuel injection quantity q was measured using the volumetric method. v To allow for cyclic injection volumes applicable to multiple operating conditions, the dynamic calibration coefficient for a specific operating condition is:

[0099]

[0100] Based on the above method, two basic types of calibration parameters were obtained and input into the database as reference calibration parameters. Among them, the static calibration parameters are greater than or equal to 1, and the number of basic dynamic calibration parameters is greater than or equal to 5 (i.e., 5 different working conditions) to have reference value. In addition, each test of the same sensor will be stored in the database as basic parameters.

[0101] The real-time adaptive dynamic calibration process can be structurally represented by a data flow diagram, divided into a top-level layer (Level 0) and a core process layer (Level 1), covering the entire process from data acquisition to feedback calibration and output results. In the top-level layer (Level 0), the external entities of the system mainly include various sensors, testing equipment, databases, and analysis software.

[0102] The sensor arrangement and related parameter signal output directions are shown in the figure. It includes at least one temperature sensor and one ambient pressure sensor for environmental testing. The high-pressure common rail test bench directly transmits injection signals, including injection pressure signals and injection pulse width signals (i.e., solenoid valve energizing signals). For the injector, at least one force sensor is installed to provide a raw signal proportional to the force acting on it by the fuel flow. An injection pressure sensor is used to compare the measured injection pressure with the injection pressure output signal of the high-pressure common rail test bench.

[0103] The sensor continuously inputs raw data (such as pressure, force, temperature, etc.), and the testing equipment triggers the sensor's operating state and coordinates the data acquisition process. The database provides the system with reference calibration values ​​(i.e., static calibration parameters and basic dynamic calibration parameters), historical environmental parameters, and historical calibration parameters. It should also include relevant parameters of the test fluid, including but not limited to density, viscosity, and bulk modulus at different temperatures. The analysis software ultimately receives the calibrated data output by the adaptive dynamic calibration system.

[0104] Adaptive dynamic calibration, as the core system, receives sensor data and database information, and outputs high-quality, calibrated data for analysis software through internal processing. Further refined to Level 1, the core data processing flow of adaptive dynamic calibration, this system comprises multiple consecutive processing modules.

[0105] First, the system collects raw data from the sensors. This data is input directly into the data preprocessing module without processing. In the preprocessing stage, the system cleans the raw data, including denoising, imputing outliers, and standardizing the data to make it suitable for use in subsequent calibration processes.

[0106] The cleaned data will be temporarily stored and used as the basis for subsequent module calls. Next, the system performs an initial calibration check, which involves combining the reference calibration values ​​(i.e., static calibration parameters and basic dynamic calibration parameters) from the database with the preprocessed real-time data to extract the initial calibration parameters and store them in the "calibration parameter library". This process is the starting point for achieving adaptive calibration, ensuring that subsequent adjustments have a solid foundation.

[0107] At its core, the real-time adaptive calibration module uses cleaned data and initial parameters to dynamically assess real-time changes and generate updated calibration parameters. This module also receives feedback from the "calibration application" module, forming an internal closed-loop mechanism for continuous parameter optimization.

[0108] The real-time calibration sensor application module actually performs the dynamic calibration task, combining the sensor's raw data with the real-time updated calibration parameters to output accurate and usable calibration data, and feeds it back to the adaptive calibration module to continuously improve system performance.

[0109] In addition, the system has a continuous feedback loop. When data drift or sudden environmental changes are detected, the system automatically triggers an update mechanism, dynamically adjusting parameters through a new calibration cycle to ensure the timeliness and accuracy of the calibration output. Finally, the system outputs high-quality calibration data, after multiple rounds of adaptive optimization, to analysis software for further processing and analysis by upper-level systems or engineers.

[0110] The training method for the real-time adaptive calibration module is based on an AI model trained on historical data to predict calibration data. Model types include, but are not limited to, supervised learning models such as Gradient Boosting Regressors, Long Short-Term Memory (LSTM) networks, or Feedforward Neural Networks. The dataset sources include historical injector test datasets, force sensor transmission data, known injected fuel quantities (bench fuel consumption, fuel consumption tested by volumetric methods, etc.), environmental conditions (temperature, pressure), injector type, and fuel type operating parameters. Specific engineering features include: ambient temperature, ambient pressure, fuel viscosity, fuel density, initial force reading, injector ID, and bench injection pressure. Model training: Training uses 70% of the data for learning, 15% for validation, and 15% for testing. The model predicts calibration coefficients, which are then applied to the initial force readings.

[0111] # AI Model Inference Example

[0112] def realtime_calibration(raw_force, temp, pressure):

[0113] lstm_model.load_weights('calibration_model.h5')

[0114] return lstm_model.predict([raw_force, temp, pressure])

[0115] The fuzzy logic-based calibration uses predefined rules to adjust the calibration coefficients based on environmental factors and fuel type. Input variables include, but are not limited to, temperature, ambient pressure, fuel density, fuel viscosity, and force sensor signals. The inference engine assumes the use of a Mamdani-style fuzzy inference system, employing centroid defuzzification to generate real-time calibration coefficient factors.

[0116] Example definition rules

[0117] Temperature: Low (≤15°C), Medium (15–40°C), High (>40°C)

[0118] Environmental pressure: Low (<1MPa), Normal (2-4MPa), High (>4MPa)

[0119] Viscosity: Low, Medium, High

[0120] Example rule base (experience-corrected parameters):

[0121] Rule number Conditional parameters Calibration coefficient correction R1 Low temperature and high viscosity Increase by 5% R2 High temperature and high environmental pressure Reduced by 3% R3 The viscosity is medium and the original force signal is low, less than the threshold. No correction R4 Medium temperature, low pressure, and low density Increase by 8% R5 All parameters are within the normal range. Using benchmark calibration coefficients

[0122] The force sensor adaptive real-time calibration system based on the impulse method developed in this invention has core application value in the field of internal combustion engine fuel system research and development. For high-pressure common rail diesel injection systems (200-300MPa ultra-high pressure environment), it accurately captures the superimposed effects of multiple injection stages, including pre-injection, main injection, and post-injection. Simultaneously, the temperature adaptive module can suppress thermal deformation drift across the entire temperature range, significantly improving the reliability of injector consistency testing and common rail system fault diagnosis for China VI / Euro VII diesel engines. In alternative fuel compatibility, it addresses issues such as the wide viscosity fluctuations of biodiesel, density instability caused by the high volatility of methanol, and the non-Newtonian fluid effects resulting from the ultra-low viscosity of hydrogen fuel.

[0123] In the field of aerospace propulsion systems, this technology provides a solution for the rapid switching between high-altitude, low-pressure conditions and wide temperature ranges of aero-engine fuel nozzles. An adaptive environmental pressure module dynamically updates air density parameters and, combined with multi-sensor fusion technology, corrects the impulse formula in real time, addressing the problem of spray momentum calculation deviations caused by sudden pressure changes in traditional calibration on high-altitude simulation benches. Simultaneously, it ensures measurement stability under extreme conditions in rocket engine propellant injection verification scenarios.

[0124] Precision fluid control in industrial applications is another important application of this patent. In the printing industry, the impact-flow mapping model enables ink jet positioning; in minimally invasive medical surgical robots, an adaptive viscosity compensation mechanism can maintain real-time drug dosage accuracy when switching between blood and contrast agents; in the semiconductor photoresist spraying process, a multi-nozzle collaborative calibration matrix is ​​used to eliminate individual differences and ensure consistent coating thickness; and in agricultural variable spraying systems, the environmental drift self-correction function maintains reliable pesticide flow control under field dust and humidity interference.

[0125] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An adaptive real-time calibration method for a force sensor based on the impulse method for measuring jet propulsion, characterized in that, Includes the following steps: S1: Establish the experimental scenario, place the fuel injector in the constant volume projectile, place a force sensor outside the nozzle of the fuel injector, and place sensors for measuring environmental parameters in the constant volume projectile. S2: Using a force sensor, the static calibration coefficient of the force sensor is measured using a static calibration method and then input into the database; S3: Based on the static calibration coefficient, using a force sensor, the dynamic calibration coefficient of the fuel injector is calculated through the impulse method and other cyclic injection quantity test methods, and then entered into the database for storage. In addition, the environmental parameters of the fluid injection system during the test of the impulse method and other cyclic injection quantity test methods are also entered into the database. S4: Repeat the tests of S2 and S3 to continuously obtain new static calibration coefficients, new dynamic calibration coefficients, and new environmental parameters. Input these into the database and, together with the historical static calibration coefficients, historical dynamic calibration coefficients, and historical environmental parameters obtained before S2 and S3, form a reference calibration value in the database. S5: Perform real-time adaptive dynamic calibration The real-time data collected by each sensor is input into the data preprocessing module for preprocessing. Some of the preprocessed data is directly input into the real-time adaptive calibration module, while some of the preprocessed data and the reference calibration value are checked by the initial calibration module to obtain the initial calibration parameters. The reference calibration value is also input into the real-time adaptive calibration module as reference calibration data. At the same time, environmental parameters, fuel characteristic parameters, and test conditions are input into the real-time adaptive calibration module. Through the real-time adaptive calibration module and the real-time sensor calibration application module, the internal processing flow includes comparison with historical data, empirical parameter correction, and AI training model correction to output high-quality, calibrated data for use by the analysis software.

2. The adaptive real-time calibration method for a force sensor based on the impulse method for testing jet laws, as described in claim 1, is characterized in that: The sensors used to measure environmental parameters in step one include an ambient temperature sensor, an ambient humidity sensor, and an ambient pressure sensor.

3. The adaptive real-time calibration method for a force sensor based on the impulse method for testing jet laws, as described in claim 1, is characterized in that: In step two, the static calibration method uses the weight method. During the calibration process, five different calibration points were selected to test the sensor performance within different ranges. Each calibration point underwent 10 repeated trials, and the average value was taken. The actual curve used is the fitted curve obtained by the least squares method, and the fitted curve is as follows: (1) In the formula, x is the applied force value, and y is the output electrical signal.

4. An adaptive real-time calibration method for a force sensor based on the impulse method for testing jet laws, as described in claim 2 or 3, characterized in that: The reference value of static calibration methods is measured by calculating the Pearson correlation coefficient, which is defined as follows: (2) in and The sample average value of force and output voltage is used to calculate r. If r is close to 1, there is a strong linear relationship between the output voltage signal of the force sensor and the applied force, and the static calibration of the force sensor response can be used as a reference value. If r is close to 0, there may be a nonlinear relationship between the output of the force sensor and the applied force, which requires further correction or analysis.

5. The adaptive real-time calibration method for a force sensor based on the impulse method for testing jet laws according to claim 1, characterized in that: In step three, the impulse method assumes the fluid inside the nozzle is liquid and simplifies the geometric flow area of ​​the nozzle based on the law of conservation of mass, thus obtaining the effective flow area, the relationship between the transient jet rate and the spray momentum flow (impulse) and the average velocity at the nozzle exit: (3) (4) In the formula: The transient jetting rate at the nozzle exit; For spray momentum; u m To simplify the average velocity at the nozzle exit; ρ f For fuel density, A geo The geometric flow cross-sectional area of ​​the nozzle; The relationship between the fuel injection rate by volume and the simplified average exit velocity is as follows: (5) Combining equations (4) and (5), the relationship between the fuel injection rate and the spray impulse at the nozzle outlet is as follows: (6) Equation (6) leads to the conclusion that the fuel injection rate of each injector hole can be calculated by measuring the instantaneous spray impulse of each hole. In actual measurement, a simplified result was obtained: (7) Therefore, by modifying equation (6), the injection rate is: (8) By measuring the spray impact force of a single nozzle using formula (8), the injection rate of that nozzle can be calculated. The obtained fuel injection rate formula (8) for each nozzle is integrated according to the actual injection duration, and the result is the cyclic injection quantity of each nozzle: (9) In the formula, q f This refers to the amount of oil injected during the nozzle cycle.

6. The adaptive real-time calibration method for a force sensor based on the impulse method for testing jet laws, as described in claim 5, is characterized in that: The other method for testing the cyclic injection quantity in step three is the volumetric method, which measures the cyclic injection quantity q. v To allow for cyclic injection volumes applicable to multiple operating conditions, the dynamic calibration factor for a specific operating condition is: 。 7. The adaptive real-time calibration method for a force sensor based on the impulse method for testing jet laws according to claim 1, characterized in that: The data preprocessing module cleans the raw data, including noise reduction, outlier completion, and standardization, to make the data suitable for subsequent calibration processes.

8. The adaptive real-time calibration method for a force sensor based on the impulse method for testing jet laws according to claim 1, characterized in that: The system also features a continuous feedback loop that monitors sensor data, test parameters, and environmental parameters in real time. When data drift or environmental changes are detected, a dynamic calibration cycle is automatically triggered, returning to the real-time adaptive calibration module. The calibration parameters are then updated through the real-time adaptive calibration module to ensure the timeliness and accuracy of the calibration output.

9. The adaptive real-time calibration method for a force sensor based on the impulse method for testing injection patterns according to claim 1, characterized in that: The AI-based prediction model is trained based on historical calibration data. It takes real-time force signals, temperature, pressure, fuel density, viscosity and operating parameters as input, and outputs dynamic calibration coefficients or correction factors. Through supervised learning, it establishes a mapping relationship between multiple parameters and calibration deviations to achieve adaptive prediction of calibration parameters.

10. The adaptive real-time calibration method for a force sensor based on the impulse method for testing injection laws according to claim 1, characterized in that: The fuzzy logic-based calibration uses a rule base of predefined empirical correction parameters. Input variables, including temperature, pressure, density, viscosity, and original force values, are processed by the Mamdani inference system and defuzzified using the centroid method to generate calibration coefficient correction factors. These factors act on the basic calibration coefficients in a multiplicative manner to achieve real-time compensation for environmental and fuel characteristics.

Images