Intelligent calibration method and system for sensor signals controlled by plc

Abstract

This application discloses a PLC-controlled intelligent calibration method and system for sensor signals. It acquires raw readings from a target sensor and multiple related auxiliary sensors, preprocesses them to form a context data vector containing rich process information, and then inputs this vector into an association prediction model to generate a predicted value for the target sensor. Based on this, a multi-dimensional consistency score is performed on the target sensor's processed value, digital state, and predicted value, and intelligent adjustments are made in conjunction with dynamic operating condition information to determine the sensor's health status. This allows the calibration logic to adapt to transient and steady-state changes in industrial processes, avoiding false alarms and missed alarms. This approach significantly improves the intelligence level of sensor signal calibration, effectively avoids the real-time performance degradation caused by forcibly placing complex calculations on the PLC, and ensures the deterministic control of the PLC.

Description

Technical Field

[0001] This application relates to the field of intelligent calibration, specifically to a PLC-controlled intelligent calibration method and system for sensor signals. Background Technology

[0002] In modern industrial automation systems, sensors, as fundamental components for sensing the state of the production process, are crucial for the accuracy and reliability of their measurement data. Any anomaly in sensor signals, such as drift, deviation, or malfunction, can directly affect the control decisions of programmable logic controllers (PLCs), thereby posing a serious threat to process stability, product quality, and even production safety. Therefore, how to intelligently, in real-time, and efficiently calibrate sensor signals under PLC control to ensure continuous data accuracy is a universal challenge facing the industrial sector today.

[0003] Currently, traditional sensor calibration methods mainly rely on periodic manual checks or preset fixed threshold alarms. These methods are not only inefficient and labor-intensive, but also struggle to capture subtle degradations in sensor performance in real time, often leading to delayed fault detection and failing to meet the demands of modern industry for high precision, high reliability, and low maintenance costs. Although some intelligent sensor signal calibration schemes have been proposed, their widespread application in actual industrial scenarios still faces significant technical bottlenecks. A core challenge lies in the inherent characteristics of PLCs: the core task of a PLC is to ensure the real-time and deterministic nature of control logic, typically requiring cyclic scanning to be completed within milliseconds. However, intelligent calibration often involves complex algorithms such as pattern recognition and predictive modeling, which are highly computationally intensive. Integrating these complex algorithms directly into a PLC would severely consume its limited computing power, leading to prolonged PLC scan cycles, affecting its inherent real-time control performance, and even causing system instability or safety risks. This fundamental contradiction between the limited computing power of PLCs and complex calibration algorithms—the extreme requirement of time determinism in real-time control systems versus the strong dependence of complex intelligent algorithms on computationally intensive algorithms—is a major obstacle currently hindering the large-scale application of intelligent calibration technology in industrial settings.

[0004] Therefore, an optimized intelligent calibration method for sensor signals controlled by PLC is needed. Summary of the Invention

[0005] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a PLC-controlled intelligent calibration method and system for sensor signals. It acquires raw readings from a target sensor and multiple related auxiliary sensors, preprocesses them to form a context data vector containing rich process information, and then inputs this vector into an association prediction model to generate a predicted value for the target sensor. Based on this, a multi-dimensional consistency score is performed on the target sensor's processed value, digital state, and predicted value, and intelligent adjustments are made in conjunction with dynamic operating condition information to determine the sensor's health status. This allows the calibration logic to adapt to transient and steady-state changes in industrial processes, avoiding false alarms and missed alarms. In this way, the intelligence level of sensor signal calibration can be significantly improved, effectively avoiding the real-time performance degradation caused by forcibly placing complex calculations on the PLC, and ensuring the deterministic control of the PLC.

[0006] According to one aspect of this application, a PLC-controlled intelligent calibration method for sensor signals is provided, comprising: Acquire the raw reading of P1, the digital status of P2, the raw reading of T1, the raw reading of F1, and the raw reading of S1; Preprocess the raw readings of P1, P2 digital status, T1, F1, and S1 to obtain the processed value of P1, the processed status of P2, and the context data vector. Input the context data vector into the P1 association prediction model to obtain the P1 prediction value; Multi-dimensional consistency scores were performed on the P1 processed value, P2 processed status, and P1 predicted value to obtain consistency scores and difference indicators. Sensor health status is determined based on consistency scores and difference indicators to obtain P1 health status and system alarm codes.

[0007] According to another aspect of this application, a PLC-controlled intelligent calibration system for sensor signals is provided, comprising: The raw reading acquisition module is used to acquire the raw reading of P1, the digital status of P2, the raw reading of T1, the raw reading of F1, and the raw reading of S1. The preprocessing module is used to preprocess the raw readings of P1, P2 digital status, T1, F1, and S1 to obtain the P1 processed value, P2 processed status, and context data vector. The P1 association prediction module is used to input the context data vector into the P1 association prediction model to obtain the P1 prediction value. The multi-dimensional consistency scoring module is used to perform multi-dimensional consistency scoring on the P1 processing value, P2 processing status, and P1 prediction value to obtain a consistency score and a difference indicator. The sensor health status determination module is used to determine the sensor health status based on the consistency score and difference flag to obtain the P1 health status and system alarm code.

[0008] Compared with existing technologies, this application provides a PLC-controlled intelligent calibration method and system for sensor signals. It acquires raw readings from a target sensor and multiple related auxiliary sensors, preprocesses them to form a contextual data vector containing rich process information, and then inputs this vector into an association prediction model to generate a predicted value for the target sensor. Based on this, it determines the sensor's health status by performing a multi-dimensional consistency score on the target sensor's processed value, digital state, and predicted value, and by combining dynamic operating condition information for intelligent adjustments. This allows the calibration logic to adapt to transient and steady-state changes in industrial processes, avoiding false alarms and missed alarms. In this way, the intelligence level of sensor signal calibration can be significantly improved, effectively avoiding the real-time performance degradation caused by forcibly placing complex calculations on the PLC, and ensuring the deterministic control of the PLC. Attached Figure Description

[0009] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 This is a flowchart of a PLC-controlled intelligent calibration method for sensor signals according to an embodiment of this application; Figure 2 This is a schematic diagram of the data flow in the PLC-controlled intelligent calibration method for sensor signals according to an embodiment of this application; Figure 3 This is a block diagram of a PLC-controlled intelligent calibration system for sensor signals according to an embodiment of this application. Detailed Implementation

[0011] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0012] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0013] While this application makes various references to certain modules of the systems according to embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.

[0014] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0015] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0016] The present application proposes a PLC-controlled intelligent calibration method for sensor signals. Figure 1 This is a flowchart of a PLC-controlled intelligent calibration method for sensor signals according to an embodiment of this application. Figure 2 This is a schematic diagram of the data flow in a PLC-controlled intelligent calibration method for sensor signals according to an embodiment of this application. Figure 1 and Figure 2 As shown, the PLC-controlled intelligent calibration method for sensor signals according to an embodiment of this application includes the following steps: S1, acquiring the original readings of P1, P2 digital status, T1, F1, and S1; S2, preprocessing the original readings of P1, P2 digital status, T1, F1, and S1 to obtain the P1 processed value, P2 processed status, and context data vector; S3, inputting the context data vector into the P1 correlation prediction model to obtain the P1 predicted value; S4, performing multi-dimensional consistency scoring on the P1 processed value, P2 processed status, and P1 predicted value to obtain a consistency score and a difference flag; S5, determining the sensor health status based on the consistency score and the difference flag to obtain the P1 health status and system alarm code.

[0017] Specifically, S1 involves acquiring the raw readings of P1, P2 digital status, T1, F1, and S1. It should be understood that sensor data is the cornerstone of industrial process control, and its accuracy and completeness directly affect the effectiveness of PLC control decisions and the stability of the process flow. In the technical solution of this application, acquiring diverse raw readings includes not only the data from the main sensor to be calibrated (P1), but also a series of auxiliary sensors (T1, F1, S1) and digital status (P2) reflecting the system's operating status. Specifically, the P1 raw reading is the initial measurement value of a key process parameter that is the primary focus for intelligent calibration, such as the raw reading of the main pressure sensor; the P2 digital status represents a discrete equipment state or operating mode, such as the open / closed state of a valve or the running / stopping state of a pump; the T1 raw reading refers to the raw data from temperature sensors such as ambient temperature and reactor temperature; the F1 raw reading refers to the raw data from flow sensors such as liquid flow and gas flow; and the S1 raw reading refers to the raw data from other key process parameters such as equipment speed and liquid level. These multi-dimensional raw readings together construct the environmental context of the system at a specific moment. By comprehensively capturing this raw data, it is possible to ensure that the system has enough information to accurately assess the health status of the sensors and ultimately generate reliable calibrations or alarms.

[0018] In practice, raw readings are acquired directly from physical sensors deployed in the industrial field. These sensors are typically integrated with the PLC system via analog input modules (e.g., 4mA-20mA current signals, 0V-10V voltage signals) or digital input modules (e.g., switch signals, industrial bus data). The PLC, as the control core, acquires data from the connected input modules at an extremely high frequency (typically on the millisecond level) during each scan cycle, thereby reading and updating the raw analog readings from P1, T1, F1, and S1, as well as the digital status of P2, in real time.

[0019] Specifically, step S2 preprocesses the raw readings of P1, P2 digital status, T1, F1, and S1 to obtain the processed values ​​of P1, P2 processing status, and a context data vector. It should be understood that the raw readings are inevitably affected by environmental noise, electromagnetic interference, and other factors, resulting in random fluctuations or spikes in the data, i.e., noise. Without processing, this noise will severely interfere with subsequent model training and prediction accuracy. Furthermore, different types of sensors have different ranges, units, and output characteristics (e.g., pressure sensors output 4mA-20mA current, temperature sensors output analog voltage, and digital status sensors output Boolean values). These heterogeneous data need to be unified to the same scale and format for effective comparison, correlation, and modeling. Therefore, in the technical solution of this application, the raw readings of P1, P2 digital status, T1, F1, and S1 are preprocessed to purify the data, unify the dimensions, and construct a context data vector that comprehensively reflects the system's operating conditions, providing high-quality input for the subsequent P1 correlation prediction model.

[0020] In practice, firstly, the raw readings of P1, T1, F1, and S1 are filtered, denoised, and normalized to obtain the processed values ​​of P1, T1, F1, and S1, respectively. Specifically, for the raw readings of analog sensors (P1, T1, F1, S1), digital signal processing techniques are used to remove measurement noise. For example, algorithms such as moving average filtering, exponential smoothing filtering, Kalman filtering, or Butterworth filtering can be used to eliminate random fluctuations and instantaneous spikes in the data, making the data trend smoother and more realistic, improving the signal-to-noise ratio, and thus avoiding misjudgment of noise in subsequent models. Secondly, the denoised data is normalized to eliminate differences in the dimensions, ranges, and numerical ranges of data from different sensors. Commonly used normalization methods include min-max normalization, which maps data to the [0,1] region; Z-score normalization can also be used to give the data zero mean and unit variance. Normalization makes all input data numerically comparable, preventing a single variable from dominating the model due to its excessively large numerical range, which helps stabilize the model's training and improve its predictive performance. Next, the digital state of P2 is assigned to the processing state of P2. Since P2 is a digital state, it typically represents a discrete quantity, such as a switching quantity or mode selection, and does not contain analog noise and does not require dimensionless standardization. Therefore, in the technical solution of this application, the original digital state of P2 is directly retained and assigned to the processing state of P2 to ensure its semantics and information integrity; Furthermore, the processed values ​​T1, F1, and S1 are vectorized to obtain the context data vector. That is, the auxiliary sensor data (T1, F1, and S1 values), which have already undergone filtering, noise reduction, and normalization, are combined in a predefined order into a multi-dimensional vector, the context data vector. This vector is a digital representation of the environment and operating conditions of the main sensor P1, containing information about other physical quantities closely related to P1 and capable of influencing its measurements, such as temperature, flow rate, and velocity. In this way, multi-dimensional environmental information can be input into the subsequent P1 correlation prediction model in a compact and structured manner.

[0021] Specifically, in step S3, the context data vector is input into the P1 correlation prediction model to obtain the predicted P1 value. It should be understood that ensuring accurate evaluation of the P1 sensor reading requires a dynamic and intelligent benchmark. In complex industrial production environments, normal sensor readings are not static but fluctuate with changes in process parameters (such as temperature, flow rate, and pressure). Traditional calibration methods often rely on static thresholds or human experience, which are difficult to effectively handle this dynamism and can easily lead to misjudgments or even delays in identifying genuine faults. Therefore, in the technical solution of this application, the context data vector is input into the P1 correlation prediction model to generate the predicted P1 value by establishing an external prediction model. This establishes a dynamic reference standard for subsequent multi-dimensional consistency scoring, thereby achieving true intelligent calibration, rather than relying solely on static thresholds for judgment.

[0022] In practice, after receiving the context data vector, the P1 correlation prediction model performs rapid calculations based on its inherent algorithmic logic. During the training phase, the model has learned and solidified the complex relationships between the P1 sensor and these context variables, namely, the output patterns of the P1 sensor when it is healthy under various operating conditions. These algorithms can be based on statistical regression analysis, machine learning-based neural network models, or function mappings learned offline from historical data. The model utilizes these learned relationships to simulate the normal response of P1 under the current input context conditions. Finally, after internal inference calculations, the model outputs a single numerical value, the P1 predicted value, which represents the theoretical or expected value that the P1 sensor reading should be at when it is in a healthy state under the current system operating conditions defined by the context data vector.

[0023] Specifically, S4 involves performing multi-dimensional consistency scoring on the processed value of P1, the processed state of P2, and the predicted value of P1 to obtain a consistency score and a difference indicator. It should be understood that traditional deviation calculation mechanisms, which judge consistency by calculating the absolute deviation between the sensor's measured value and the model's predicted value, have an inherent technical flaw: state blindness. This mechanism assesses the absolute value of the deviation in isolation, completely ignoring the close correlation between system dynamics and error tolerance in complex industrial scenarios such as chemical reactors. However, an industrial process is not static; it inevitably includes transient phases such as heating, feeding, or adjusting process parameters. During these phases, drastic and normal fluctuations in system parameters are expected. Correspondingly, there are also steady-state phases with stable process parameters and continuous production. The actual meaning of the deviation changes fundamentally depending on the system's state. In transient processes, due to small lags in the model response or complex dynamic coupling between multiple variables, a large deviation between the measured and predicted values ​​is normal. Existing mechanisms, therefore, generate numerous false alarms, misjudging normal process adjustments as sensor malfunctions. Conversely, when the system is in a highly stable steady state, a small, persistent deviation is likely a sign of early sensor failure, but existing mechanisms may ignore it due to its small absolute value, leading to missed detection. Therefore, this static measurement method cannot adapt to the dynamic characteristics of industrial processes and lacks the ability to intelligently adjust error tolerance under different operating conditions, thus seriously affecting the accuracy and robustness of fault diagnosis.

[0024] To address the aforementioned technical shortcomings, a dynamic situation-adaptive deviation penalty mechanism is introduced. This mechanism quantifies the current dynamic state of the system and dynamically amplifies or suppresses the basic error, thereby generating a state-adaptive deviation score that can adapt to both transient and steady-state conditions in real time.

[0025] In practical implementation, firstly, system fluctuation quantification is performed on the current context data vector and the previous context data vector to obtain the system fluctuation index. It should be understood that the dynamic nature of industrial processes determines that the actual meaning of deviations varies depending on the system's state. For example, in transient phases such as heating and feeding, drastic parameter fluctuations are normal, and large deviations between measured and predicted values ​​should be tolerated; however, in the steady-state phase of stable continuous production, small deviations may be precursors to early failures and require highly sensitive identification. To overcome this deficiency, the technical solution of this application firstly creates an index that can objectively and quantitatively describe the overall dynamic intensity of the system, namely, the system fluctuation index. Specifically, firstly, the instantaneous changes of various sensor variables (such as temperature, flow rate, and rotational speed) in the context are calculated, and then, through weighted aggregation, these multi-dimensional change information are fused into a single, dimensionless system fluctuation index; this process is expressed by the formula:

[0026] in, The system volatility index is denoted by N; N is the dimension of the context data vector. and These are the current time t and the previous time t of the i-th context variable, respectively. The value of 1; It is the weight coefficient assigned to the i-th variable, used to characterize the importance of the fluctuation of the variable on the overall state of the system; This represents the time interval between two sampling moments. This formula shows that when process parameters are stable and the system is in a steady state, The value approaches zero; however, when the process is drastically adjusted and the system enters a transient state... The value will increase significantly. This provides a basis for decision-making in subsequent adaptive bias assessments. Next, the basic relative error is quantified by analyzing the processed value and the predicted value of P1 to obtain the basic relative error. It should be understood that directly using absolute deviation is affected by the scale and magnitude of the measured quantity, while relative error more fairly reflects the severity of the deviation. By calculating the basic relative error, the influence of numerical magnitude on error judgment can be eliminated, producing a standardized basic error metric that better reflects engineering practice. Specifically, the absolute difference between the measured value and the model predicted value of the main pressure sensor is calculated, and this difference is normalized using the predicted value itself to obtain a dimensionless basic relative error; this process is expressed by the formula:

[0027] in, Represents the fundamental relative error; This is the P1 processing value (the current measured value of the main pressure sensor). It is the P1 predicted value (the stress value predicted by the correlation model based on the context); This is a tiny positive constant set to prevent the denominator from being zero. This step frees error assessment from the constraints of absolute numerical magnitude. For example, a deviation of 1 Bar at 100 Bar (relative error 1%) is obviously far less serious than a deviation of 1 Bar at 10 Bar (relative error 10%), and this difference is well captured by the formula. In this way, a standardized, more practically meaningful basic error metric can be output, providing a fair starting point for the final adaptive scoring.

[0028] Furthermore, based on the system volatility index (SVI), a state-adaptive deviation adjustment is performed on the fundamental relative error to obtain a state-adaptive deviation score as a consistency score. That is, the dynamic state of the system is directly correlated with the penalty intensity of the error. Specifically, a nonlinear penalty function is constructed, using the system volatility index (SVI) calculated in the first step as input. This function applies a high penalty when the system is stable and a low penalty when the system is volatile. This dynamic penalty coefficient is then multiplied by the fundamental relative error calculated in the second step to obtain the final state-adaptive deviation score. This process is expressed by the following formula:

[0029] in, Represents the adaptive deviation score of the state; It is the basic relative error; It is a system volatility index; It is the steady-state penalty gain, a constant greater than zero, used to control the maximum amplification factor of the basis error in absolute steady state; This is the fluctuation attenuation coefficient, a constant greater than zero, used to control the rate at which the deviation penalty decays as system volatility increases. In other words, an exponential decay function links the abstract system dynamics with the specific error penalty: when the system is in steady state (SVI approaches 0), the exponential term approaches 1, significantly amplifying the basic error and making the system extremely sensitive to small anomalies in steady state; when the system is in transient state (SVI is large), the exponential term approaches 0, the penalty term disappears, and the system has a very high tolerance for normal, larger deviations during transient processes. This results in a final, intelligent deviation assessment score that dynamically adjusts its sensitivity, replacing the original static deviation value as the core basis for subsequent fault diagnosis.

[0030] Through the aforementioned mechanism, the state-blindness defect of the original mechanism is completely resolved, enabling the fault diagnosis mechanism to evolve from a static and rigid evaluation model into an intelligent and robust system capable of sensing and adapting to the dynamic characteristics of industrial processes. Specifically, this mechanism can effectively suppress deviations caused by normal fluctuations during transient periods of drastic process parameter adjustments, thereby significantly reducing false alarms and ensuring the continuity of production. Simultaneously, during the steady-state period of stable equipment operation, it can amplify the penalty coefficient, exhibiting extremely high sensitivity to extremely small deviation signals that foreshadow early faults, thus achieving early warning and detection of faults and avoiding missed alarms. Finally, by generating state-adaptive deviation scores, it provides unprecedentedly high-precision and high-reliability decision inputs for subsequent intelligent calibration gating logic, ensuring that calibration is only performed when the sensor is confirmed to be truly in a drifting state rather than a faulty state, fundamentally improving the safety, accuracy, and intelligence level of the entire intelligent calibration system.

[0031] Specifically, in step S5, the sensor health status is determined based on the consistency score and the difference flag to obtain the P1 health status and system alarm code. It should be understood that traditional fault diagnosis methods, due to their inability to accurately distinguish between the inherent dynamics of industrial processes and actual sensor faults, often misjudge normal parameter fluctuations as sensor faults during system transient phases, leading to numerous false alarms; while during stable system operation, they may fail to report minute deviations indicating early sensor degradation due to insufficient sensitivity. This situation severely affects the accuracy and robustness of fault diagnosis. In the technical solution of this application, by integrating the previously generated consistency score (which already considers the impact of system dynamics on deviation tolerance) and the difference flag, a mechanism is provided for accurate, robust, and intelligent determination of sensor health status. This enables early warning and detection of faults, reduces false alarms, and ensures that calibration or maintenance actions are triggered only when the sensor is confirmed to have a genuine anomaly rather than normal fluctuations, thereby improving the safety, accuracy, and intelligence of the entire intelligent calibration system.

[0032] In practice, the first step is to prioritize arbitration of hardware faults based on discrepancy flags to obtain intermediate health status and intermediate alarm codes. This step reflects the principle of prioritizing severe anomalies in fault diagnosis. During sensor data acquisition and preprocessing, if extreme anomalies are detected in the raw P1 reading or the digital status of P2, such as complete loss of the P1 signal, exceeding the physical range (e.g., displaying negative pressure), or sudden and repeated switching of the P2 digital status when it should not occur, these are identified as discrepancy flags. These flags typically indicate hard faults in the sensor itself or its connecting lines (e.g., open wires, short circuits, sensor damage). Once a discrepancy flag is triggered, the system immediately arbitrates and derives a preliminary intermediate health status (e.g., "P1 sensor hardware fault") and the corresponding intermediate alarm code. This mechanism ensures a rapid response and reporting of physical faults that may pose a system safety risk or cause serious production disruptions, with a higher priority than consistency scoring analysis of minor deviations. Next, based on the consistency score, the intermediate health status and intermediate alarm codes are assessed using a consistency score-based health status classification to obtain the P1 health status and system alarm code. If the difference flag is not triggered (i.e., no hard fault is detected), or after priority arbitration of hard faults is completed, the system will further utilize the consistency score to perform a refined assessment of the sensor's health status. During this classification assessment, the system compares the consistency score with several preset thresholds to determine the specific health status level of the P1 sensor. For example, the following threshold ranges can be set: When the consistency score is lower than the preset first threshold, the sensor is determined to be "healthy and normal" and a system alarm code indicating normal operation is generated (e.g., "ALERT_P1_OK"). When the consistency score is between the first and second thresholds, the sensor may show slight deviations or early signs of degradation, which will be judged as a "warning" and generate a corresponding alarm code (e.g., "ALERT_P1_DRIFT_LOW"), indicating that close attention is needed. When the consistency score is between the second and third thresholds, the deviation has reached the level that requires intervention and calibration. The sensor is determined to be "needing calibration" and a calibration request alarm code (e.g., "ALERT_P1_CAL_REQUIRED") is generated. When the consistency score exceeds the third threshold, the sensor deviation is too large and is judged as a "fault", generating the highest level fault alarm code (e.g., "ALERT_P1_CRITICAL_FAULT"), which may trigger a safe shutdown of the system or switch to redundant equipment.

[0033] This hierarchical assessment ultimately yields the P1 health status and a system alarm code with semantic information, ensuring the accuracy and usability of the assessment results.

[0034] In summary, the intelligent calibration method for sensor signals controlled by a PLC according to the embodiments of this application is explained. It acquires the raw readings of the target sensor and multiple related auxiliary sensors, preprocesses them to form a context data vector containing rich process information, and then inputs this vector into an association prediction model to generate a predicted value for the target sensor. Based on this, a multi-dimensional consistency score is performed on the target sensor's processed value, digital state, and predicted value, and intelligent adjustments are made in conjunction with dynamic operating condition information to determine the sensor's health status. This allows the calibration logic to adapt to transient and steady-state changes in industrial processes, avoiding false alarms and missed alarms. In this way, the intelligence level of sensor signal calibration can be significantly improved, effectively avoiding the real-time performance degradation caused by forcibly placing complex calculations on the PLC, and ensuring the deterministic control of the PLC.

[0035] Furthermore, a PLC-controlled intelligent calibration system for sensor signals is also provided.

[0036] Figure 3 This is a block diagram of a PLC-controlled intelligent calibration system for sensor signals according to an embodiment of this application. Figure 3As shown, the PLC-controlled intelligent calibration system 300 for sensor signals according to an embodiment of this application includes: a raw reading acquisition module 310, used to acquire raw readings of P1, P2 digital status, T1, F1, and S1; a preprocessing module 320, used to preprocess the raw readings of P1, P2 digital status, T1, F1, and S1 to obtain a P1 processing value, a P2 processing status, and a context data vector; a P1 association prediction module 330, used to input the context data vector into a P1 association prediction model to obtain a P1 predicted value; a multi-dimensional consistency scoring module 340, used to perform multi-dimensional consistency scoring on the P1 processing value, P2 processing status, and P1 predicted value to obtain a consistency score and a difference flag; and a sensor health status determination module 350, used to determine the sensor health status based on the consistency score and the difference flag to obtain the P1 health status and a system alarm code.

[0037] As described above, the PLC-controlled intelligent sensor signal calibration system 300 according to embodiments of this application can be implemented in various wireless terminals, such as servers with PLC-controlled intelligent sensor signal calibration algorithms. In one possible implementation, the PLC-controlled intelligent sensor signal calibration system 300 according to embodiments of this application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the PLC-controlled intelligent sensor signal calibration system 300 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the PLC-controlled intelligent sensor signal calibration system 300 can also be one of many hardware modules of the wireless terminal.

[0038] Alternatively, in another example, the PLC-controlled intelligent calibration system 300 for sensor signals and the wireless terminal can also be separate devices, and the PLC-controlled intelligent calibration system 300 for sensor signals can be connected to the wireless terminal via wired and / or wireless networks, and transmit interactive information in accordance with an agreed data format.

[0039] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A PLC-controlled intelligent calibration method for sensor signals, characterized in that, include: Acquire the raw reading of P1, the digital status of P2, the raw reading of T1, the raw reading of F1, and the raw reading of S1; Preprocess the raw readings of P1, P2 digital status, T1, F1, and S1 to obtain the processed value of P1, the processed status of P2, and the context data vector. Input the context data vector into the P1 association prediction model to obtain the P1 prediction value; Multi-dimensional consistency scores were performed on the P1 processed value, P2 processed status, and P1 predicted value to obtain consistency scores and difference indicators. Sensor health status is determined based on consistency scores and difference indicators to obtain P1 health status and system alarm codes.

2. The intelligent calibration method for sensor signals controlled by PLC according to claim 1, characterized in that, The raw readings of P1, P2 digital status, T1, F1, and S1 are preprocessed to obtain the processed value of P1, the processed status of P2, and the context data vector, including: The original readings of P1, T1, F1, and S1 are filtered, denoised, and normalized to obtain the processed values ​​of P1, T1, F1, and S1. Assign the digital state of P2 to the processing state of P2; The T1, F1, and S1 processing values ​​are vectorized to obtain the context data vector.

3. The intelligent calibration method for sensor signals controlled by PLC according to claim 1, characterized in that, A multi-dimensional consistency score was performed on the P1 treatment value, P2 treatment status, and P1 predicted value to obtain a consistency score and discrepancy indicators, including: The system fluctuation index is obtained by performing system fluctuation quantization on the current context data vector and the previous context data vector; The basic relative error is obtained by quantizing the basic relative error of the processed value of P1 and the predicted value of P1. Based on the system fluctuation index, the state adaptive deviation adjustment is performed on the basic relative error to obtain the state adaptive deviation score as the consistency score.

4. The intelligent calibration method for sensor signals controlled by PLC according to claim 3, characterized in that, To obtain the system fluctuation index, system fluctuation quantization is performed on the current context data vector and the previous context data vector, including: performing system fluctuation quantization on the current context data vector and the previous context data vector using the following formula: in, The system volatility index is denoted by N; N is the dimension of the context data vector. and These are the current time t and the previous time t of the i-th context variable, respectively. The value of 1; These are the weight coefficients assigned to the i-th variable; This represents the time interval between two sampling moments.

5. The intelligent calibration method for sensor signals controlled by PLC according to claim 3, characterized in that, The basic relative error is obtained by quantizing the processed value of P1 and the predicted value of P1, including: quantizing the basic relative error of the processed value of P1 and the predicted value of P1 using the following formula: in, Represents the fundamental relative error; It is the value processed by P1; This is the predicted value for P1; It is a very small positive number set to prevent the denominator from being zero.

6. The intelligent calibration method for sensor signals controlled by PLC according to claim 3, characterized in that, Based on the system fluctuation index, state adaptive deviation adjustment is performed on the basic relative error to obtain a state adaptive deviation score as a consistency score. This includes: adjusting the basic relative error using the following formula, where the formula is: in, Represents the adaptive deviation score of the state; It is the basic relative error; It is a system volatility index; It is the steady-state penalty gain; It is the fluctuation attenuation coefficient.

7. The intelligent calibration method for sensor signals controlled by PLC according to claim 1, characterized in that, Sensor health status is determined based on consistency scores and difference indicators to obtain P1 health status and system alarm codes, including: Priority arbitration of hardware faults is performed based on differential flags to obtain intermediate health status and intermediate alarm codes; Based on the consistency score, the intermediate health status and intermediate alarm code are evaluated by a health status classification based on the consistency score to obtain the P1 health status and system alarm code.

8. A PLC-controlled intelligent calibration system for sensor signals, characterized in that, include: The raw reading acquisition module is used to acquire the raw reading of P1, the digital status of P2, the raw reading of T1, the raw reading of F1, and the raw reading of S1. The preprocessing module is used to preprocess the raw readings of P1, P2 digital status, T1, F1, and S1 to obtain the P1 processed value, P2 processed status, and context data vector. The P1 association prediction module is used to input the context data vector into the P1 association prediction model to obtain the P1 prediction value. The multi-dimensional consistency scoring module is used to perform multi-dimensional consistency scoring on the P1 processing value, P2 processing status, and P1 prediction value to obtain a consistency score and a difference indicator. The sensor health status determination module is used to determine the sensor health status based on the consistency score and difference flag to obtain the P1 health status and system alarm code.

Images