An anti-interference temperature transmitter measurement and control system and method

Abstract

This invention discloses an anti-interference temperature transmitter measurement and control system and method, relating to the field of signal processing technology. The invention collects field temperature data in the working environment and transmits the electrical signal to an amplifier for amplification; constructs a labeled dataset as a training set to build a one-dimensional hybrid model; inputs the amplified electrical signal into the model to identify the noise type and main interference frequency in the electrical signal; dynamically adjusts the ADC sampling rate; performs time-domain analysis on the electrical signal, extracts the signal baseline using the moving average method, and eliminates low-frequency drift; then analyzes the frequency domain of the electrical signal, selects a filter, and constructs an adaptive notch filter to suppress frequency domain interference noise; weightedly fuses the two signals after time-domain and frequency-domain analysis to output the noise-reduced electrical signal; and uses the output electrical signal to calculate the anti-interference performance of the temperature transmitter and verify its anti-interference performance.

Description

Technical Field

[0001] This invention relates to the field of signal processing technology, specifically to an anti-interference temperature transmitter measurement and control system and method. Background Technology

[0002] Precise temperature measurement and control are crucial in numerous fields, including industrial production, scientific research, and environmental monitoring. Temperature transmitters, as key devices connecting temperature sensors and control systems, convert the temperature signals acquired by the sensors into standard electrical signals for subsequent display, recording, and control. However, in practical applications, temperature transmitters face various interference sources, which severely affect the accuracy and stability of temperature measurements.

[0003] Early temperature transmitters had relatively simple structures and functions, resulting in weak anti-interference capabilities. They primarily employed basic hardware protection measures, such as adding simple filter capacitors at the power input to reduce the impact of power fluctuations on the circuit, and using shielded cables on the signal transmission lines to reduce the influence of external electromagnetic interference. However, these simple anti-interference measures had limited effectiveness in complex interference environments and could not meet the demands of high-precision temperature measurement.

[0004] In recent years, with the rapid development of technologies such as artificial intelligence and the Internet of Things, temperature transmitter measurement and control systems have gradually evolved towards intelligence. Intelligent anti-interference technologies mainly include adaptive interference suppression, neural network algorithms, and machine learning algorithms. Adaptive interference suppression technology can automatically adjust anti-interference strategies based on real-time changes in interference signals, improving anti-interference effectiveness. Neural network algorithms and machine learning algorithms can learn from large amounts of historical data to establish interference models, achieving intelligent identification and elimination of interference signals. The application of these intelligent anti-interference technologies enables temperature transmitter measurement and control systems to operate more accurately and stably in complex interference environments, providing more reliable temperature measurement and control solutions for industrial production and scientific research.

[0005] However, when using algorithms to process signal interference in temperature transmitters, most current methods employ a single algorithm or model to eliminate a single type of interference. Furthermore, the elimination of interference across different dimensions is incomplete, and the sampling frequency of the temperature transmitter cannot be intelligently adjusted based on the amount of interference in the signal. This results in a significant reduction in the efficiency of the temperature transmitter and an increase in the workload of signal processing. Different interference frequencies require different sampling rates to ensure the output quality of the temperature transmitter. Summary of the Invention

[0006] The purpose of this invention is to provide an anti-interference temperature transmitter measurement and control system and method to solve the problems raised in the prior art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] An anti-interference temperature transmitter measurement and control method, the method comprising the following steps:

[0009] S100: The ambient temperature is collected in the working environment. The collected temperature signal is calculated and converted into an electrical signal in the temperature transmitter. The electrical signal is then sent to the amplifier for amplification.

[0010] Furthermore, the specific steps for transmitting the electrical signal to the amplifier for amplification are as follows:

[0011] S101. Use a temperature sensor to collect the real-time temperature of the work site. The temperature sensor transmits the collected temperature signal to the temperature transmitter. The temperature transmitter supports temperature signals transmitted by different types of temperature sensors, including thermocouple temperature sensors and resistance temperature sensors. The temperature transmitter calculates the received temperature signal and converts it into a weak electrical signal, and sends the generated weak electrical signal to the amplifier circuit in the temperature transmitter.

[0012] S102. When the temperature sensor is a thermocouple type, the temperature transmitter amplifies the received weak electrical signal, calculates the temperature value corresponding to the temperature signal according to the thermocouple calibration table, and finally converts it into a standard current or voltage signal for output.

[0013] When the temperature sensor is a resistance temperature detector (RTD) type, the temperature transmitter amplifies the received weak electrical signal and converts it into a standard current or voltage signal output through the internal circuitry of the transmitter.

[0014] S200: Collect temperature signals from the working environment, inject interference noise signals, construct a labeled dataset as a training set, and construct a one-dimensional hybrid model; input the amplified electrical signal into the model to identify the noise type and main interference frequency in the electrical signal;

[0015] Furthermore, the specific steps for inputting the amplified electrical signal into the model to identify the noise type and main interference frequency in the electrical signal are as follows:

[0016] S201. Acquire a pure temperature signal under interference-free conditions using a constant temperature device, and use the pure temperature signal as a reference signal x. ref Different types of interference noise signals are artificially designed and added to the reference signal. The signal after the interference noise is injected is used as the original signal xs(t); xs(t) = x ref (t)+n(t), where n(t) represents the interference noise signal;

[0017] Different types of artificially designed interference noise are labeled to obtain classification tags. The interference noise is then analyzed using Fourier transform to obtain the frequency f of the main interference noise. noise Using the main interference noise frequency as the regression label, a labeled dataset is constructed. The data format in the labeled dataset is {xs(t), y cls y reg}, where y cls Indicates category label, y reg Indicates the regression label;

[0018] S202. Design the input layer shape as (N, 1), where N represents the time series length; construct three convolutional layers with kernel sizes of 3, 5, and 7 respectively; the number of filters are 16, 32, and 64 respectively; add the ReLU activation function to each layer; apply max pooling after each layer to compress the feature dimension; design an LSTM neural network layer with 64 units; design the output layer to include a classification branch and a regression branch, with the classification branch outputting the probability P of the interference noise type. cls The regression branch outputs the estimated main interference frequency fg. noise Generate a one-dimensional CNN-LTSM mixture model; train the one-dimensional CNN-LTSM mixture model using the labeled dataset; calculate the classification loss and regression loss using the probability of interference noise types and the estimated frequency of the main interference, respectively, using the following formulas:

[0019]

[0020] In the formula, This represents the true label of the i-th class. The model predicts the probability of the i-th class; C represents the number of noise classes; M represents the batch size; f j noise This represents the j-th dominant interference frequency predicted by the model; This represents the estimated frequency of the j-th main interference.

[0021] The total loss function is constructed as follows:

[0022] L=α×L cls +(1-α)×L reg ;

[0023] In the formula, L represents the total loss, L cls L represents the classification loss. reg Represents the regression loss, where α represents the classification loss weight;

[0024] S203. The generated one-dimensional CNN-LTSM hybrid model is lightweighted using the TensorFlow Lite framework. A sliding window mechanism is designed, where the length of the time series in the model input layer is set as the window length, and one window contains N milliseconds of data. The amplified electrical signal is used as the real signal x(t) and input into the one-dimensional CNN-LTSM hybrid model. The probability of the interference noise type and the estimated value of the main interference frequency of the real signal are output. The results are averaged after three repetitions and the final result is output to determine whether there is interference noise in the real signal. When interference noise exists, the type of interference noise is determined and the main interference frequency is obtained.

[0025] In temperature transmitters, deep learning technology is used to determine the type of interference noise and the dominant interference frequency in the amplified electrical signal. A one-dimensional CNN-LTSM hybrid model is constructed, combining the local feature extraction capability of convolutional neural networks and the temporal modeling capability of long short-term memory networks to achieve accurate classification and dominant frequency estimation of complex noise. Multi-task learning, such as classification + regression, reduces model redundancy and improves inference efficiency. It operates stably in the complex and variable electromagnetic environment of industrial sites, avoiding overfitting to a single noise type.

[0026] S300: Dynamically adjust the ADC sampling rate based on the main interference frequency characteristics output by the one-dimensional hybrid model;

[0027] Furthermore, the specific steps for dynamically adjusting the ADC sampling rate are as follows:

[0028] S301. Collect interference frequencies from unavailable signals in the historical working environment. Calculate the average and standard deviation of the interference frequencies of unavailable signals. Subtract the standard deviation from the average to obtain the high-frequency interference threshold f. thre ;

[0029] S302. Using the high-frequency interference threshold to determine the main interference frequency f output by the one-dimensional CNN-LTSM mixture model. noise Make a judgment when f noise >f thre When f is detected, it is determined to be a high-frequency interference scenario; when f noise ≤f thre At that time, it was determined to be a low-frequency interference scenario;

[0030] S303. Collect the available signal sampling rate in the historical working environment when the signal is in a high-frequency interference scenario and a low-frequency interference scenario, respectively. Construct a linear regression model to analyze the relationship between the available signal sampling rate and the interference frequency in the high-frequency interference scenario and the relationship between the available signal sampling rate and the interference frequency in the low-frequency interference scenario. Obtain the linear influence function of the two relationships respectively, and extract the coefficient of the available signal sampling rate in the function as the dynamic adjustment weight.

[0031] When it is determined that the real signal is in a high-frequency interference scenario, f s =δ1×f noise In the formula, f s The sampling rate is dynamically adjusted, and δ1 represents the dynamic adjustment weight of high-frequency interference; when the real signal is determined to be in a low-frequency interference scenario, f s =δ2×f noise δ2 represents the dynamic adjustment weight of low-frequency interference.

[0032] The sampling rate is dynamically adjusted for high-frequency and low-frequency interference scenarios respectively, improving digital filtering resolution, avoiding spectral overlap caused by aliasing, and ensuring signal integrity; extending the battery life of battery-powered devices; achieving more precise temperature control, improving system stability and control accuracy, and reducing temperature overshoot or undershoot.

[0033] S400 performs time-domain analysis on the electrical signal, extracts the signal baseline using the moving average method, and eliminates low-frequency drift; then it analyzes the frequency domain of the electrical signal, selects a filter, and constructs an adaptive notch filter to suppress frequency domain interference noise.

[0034] Furthermore, the specific steps for selecting a filter to eliminate interference and noise are as follows:

[0035] S401. Calculate the signal mean within the sliding window, and then eliminate low-frequency drift. The formula is:

[0036]

[0037] y corrected (t)=x(t)-y base (t);

[0038] In the formula, y base (t) represents the mean of the signal within the sliding window, characterizing the low-frequency drift component; W represents the length of the sliding window, and x(t) represents the actual input signal. corrected (t) represents the signal after eliminating low-frequency drift; y corrected (t) serves as the time-domain correction signal;

[0039] The possible values ​​of W are:

[0040] If f noise For frequencies ≤10Hz, select a larger window to fully smooth out low-frequency noise;

[0041] If f noise For frequencies >10Hz, select a smaller window to avoid high-frequency signal distortion;

[0042] S402. Select the corresponding frequency filter based on the type of interference noise in the real signal output by the one-dimensional CNN-LTSM hybrid model, and then construct an adaptive notch filter; the formula is:

[0043]

[0044] In the formula, H(z) represents an adaptive notch filter, r represents the pole radius, and z -1 Represents the unit delay operator;

[0045] The digital angular frequency is calculated based on the main interference frequency of the real signal, using the following formula: In the formula, θ represents the digital angular frequency, f s Represents the actual signal sampling rate; updates the adaptive notch filter coefficients, updating the cosine coefficients of the numerator and denominator to 2×cos(θ); and adjusts the signal y after eliminating low-frequency drift. corrected (t) is input to an adaptive notch filter to suppress frequency domain interference in the real signal; the output suppressed signal is y. notch (t), y notch (t) serves as the frequency-filtered signal.

[0046] S500: After performing time-domain analysis and frequency-domain analysis on the electrical signal, the two signals are weighted and fused to output the noise-reduced electrical signal.

[0047] Furthermore, the specific steps for weighted fusion of the two signals after time-domain and frequency-domain analysis to output the noise-reduced electrical signal are as follows:

[0048] S501. The time-domain corrected signal and the frequency-domain filtered signal are fused according to weights, using the following formula:

[0049] y final (t)=β×y corrected (t)+(1-β)×y notch (t);

[0050] In the formula, y final (t) represents the fused signal, and β represents the time-domain signal weight, which is dynamically adjusted according to the noise intensity.

[0051] If high-frequency impulse noise is detected: reduce β and focus on frequency domain wavelet denoising;

[0052] If low-frequency power frequency noise is detected: increase β to preserve time-domain details.

[0053] S502. The fused signal is finally optimized using the Kalman filter model, and the final output signal is y(t).

[0054] In temperature transmitters, interference can lead to signal distortion, increased bit error rate, and other problems, affecting signal quality. Time-domain and frequency-domain dual-domain joint interference cancellation technology can significantly enhance the system's anti-interference capability, effectively reduce the impact of interference on the signal, and improve signal integrity and accuracy. For signals affected by multiple interference sources, dual-domain joint interference cancellation can address interference from different angles, making the signal more stable during transmission, reducing signal fluctuations and distortion, and improving signal quality and availability.

[0055] By employing appropriate processing methods in both the time and frequency domains, various types of interference can be effectively suppressed and eliminated, demonstrating strong flexibility and adaptability.

[0056] S600 uses the output electrical signal to calculate the anti-interference performance of the temperature transmitter and verifies its anti-interference performance.

[0057] Further, the specific steps for verifying the anti-interference performance are as follows:

[0058] S601. Simulate different types of interference combinations in the temperature control chamber, record the real signal, the processed signal, and the reference signal, and use the three signals to calculate the signal-to-noise ratio (SNR) improvement. When the SNR improvement is negative, verify that the anti-interference performance is reduced, and re-acquire data to rebuild the model; when the SNR improvement is positive, verify that the anti-interference performance is improved.

[0059] An anti-interference temperature transmitter measurement and control system, comprising a data acquisition module, an interference noise identification module, a sampling rate adjustment module, a dual-domain noise reduction module, and a performance verification module;

[0060] The data acquisition module is used to collect the frequency and sampling rate of signal interference noise in historical work.

[0061] The interference noise identification module is used to construct a one-dimensional hybrid model using deep learning; the amplified electrical signal is input into the model to identify the noise type and main interference frequency in the electrical signal;

[0062] The sampling rate adjustment module is used to dynamically adjust the ADC sampling rate based on the main interference frequency characteristics output by the one-dimensional hybrid model.

[0063] The dual-domain noise reduction module is used to perform time-domain and frequency-domain analysis on the signal to eliminate interference noise in the signal;

[0064] The performance verification module is used to calculate the signal-to-noise ratio of the processed signal and verify its anti-interference performance.

[0065] The interference and noise identification module includes a training set construction unit, a model construction unit, and a real-time inference unit;

[0066] The training set construction unit is used to collect and set the baseline signal, the original signal, and the type of interference noise to generate a label dataset.

[0067] The model building unit is used to design the input layer, convolutional layer, neural network layer and output layer respectively, and to build a one-dimensional CNN-LTSM hybrid model;

[0068] The real-time inference unit is used to design a sliding window mechanism, which uses a sliding window to input real data into the model for judgment.

[0069] The dual-domain noise reduction module includes a time-domain noise reduction unit, a frequency-domain suppression unit, and a dual-domain fusion unit;

[0070] The time-domain noise reduction unit is used to calculate the signal mean within the sliding window and then eliminate low-frequency drift.

[0071] The frequency domain suppression unit is used to design an adaptive notch filter to suppress frequency domain interference in the real signal.

[0072] The dual-domain fusion unit is used to fuse the time-domain corrected signal and the frequency-domain filtered signal according to weights.

[0073] Compared with the prior art, the beneficial effects of the present invention are:

[0074] 1. This invention utilizes deep learning technology in temperature transmitters to determine the type of interference noise and the main interference frequency in the amplified electrical signal. It constructs a one-dimensional CNN-LTSM hybrid model that combines the local feature extraction capability of convolutional neural networks with the time-series modeling capability of long short-term memory networks to achieve accurate classification and main frequency estimation of complex noise.

[0075] 2. This invention dynamically adjusts the sampling rate in high-frequency interference scenarios and low-frequency interference scenarios respectively, improves the resolution of digital filtering, avoids spectral overlap caused by aliasing, ensures signal integrity, and extends the battery life of battery-powered devices. Attached Figure Description

[0076] Figure 1 This is a module distribution diagram of an anti-interference temperature transmitter measurement and control system according to the present invention;

[0077] Figure 2 This is a schematic diagram illustrating the steps of an anti-interference temperature transmitter measurement and control method according to the present invention. Detailed Implementation

[0078] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] Example: Figures 1-2 As shown, the present invention provides a technical solution.

[0080] An anti-interference temperature transmitter measurement and control method, the method comprising the following steps:

[0081] S100: The ambient temperature is collected in the working environment. The collected temperature signal is calculated and converted into an electrical signal in the temperature transmitter. The electrical signal is then sent to the amplifier for amplification.

[0082] The specific steps for transmitting an electrical signal to an amplifier for amplification are as follows:

[0083] S101. Use a temperature sensor to collect the real-time temperature of the work site. The temperature sensor transmits the collected temperature signal to the temperature transmitter. The temperature transmitter supports temperature signals transmitted by different types of temperature sensors, including thermocouple temperature sensors and resistance temperature sensors. The temperature transmitter calculates the received temperature signal and converts it into a weak electrical signal, and sends the generated weak electrical signal to the amplifier circuit in the temperature transmitter.

[0084] S102. When the temperature sensor is a thermocouple type, the temperature transmitter amplifies the received weak electrical signal, calculates the temperature value corresponding to the temperature signal according to the thermocouple calibration table, and finally converts it into a standard current or voltage signal for output.

[0085] When the temperature sensor is a resistance temperature detector (RTD) type, the temperature transmitter amplifies the received weak electrical signal and converts it into a standard current or voltage signal output through the internal circuitry of the transmitter.

[0086] S200: Collect temperature signals from the working environment, inject interference noise signals, construct a labeled dataset as a training set, and construct a one-dimensional hybrid model; input the amplified electrical signal into the model to identify the noise type and main interference frequency in the electrical signal;

[0087] The specific steps for inputting the amplified electrical signal into the model to identify the noise type and main interference frequency in the electrical signal are as follows:

[0088] S201. Acquire a pure temperature signal under interference-free conditions using a constant temperature device, and use the pure temperature signal as a reference signal x. refDifferent types of interference noise signals are artificially designed and added to the reference signal. The signal after the interference noise is injected is used as the original signal xs(t); xs(t) = x ref (t)+n(t), where n(t) represents the interference noise signal;

[0089] Different types of artificially designed interference noise are labeled to obtain classification tags. The interference noise is then analyzed using Fourier transform to obtain the frequency f of the main interference noise. noise Using the main interference noise frequency as the regression label, a labeled dataset is constructed. The data format in the labeled dataset is {xs(t), y cls y reg}, where y cls Indicates category label, y reg Indicates the regression label;

[0090] S202. Design the input layer shape as (N, 1), where N represents the time series length; construct three convolutional layers with kernel sizes of 3, 5, and 7 respectively; the number of filters are 16, 32, and 64 respectively; add the ReLU activation function to each layer; apply max pooling after each layer to compress the feature dimension; design an LSTM neural network layer with 64 units; design the output layer to include a classification branch and a regression branch, with the classification branch outputting the probability P of the interference noise type. cls The regression branch outputs the estimated main interference frequency fg. noise Generate a one-dimensional CNN-LTSM mixture model; train the one-dimensional CNN-LTSM mixture model using the labeled dataset; calculate the classification loss and regression loss using the probability of interference noise types and the estimated frequency of the main interference, respectively, using the following formulas:

[0091]

[0092]

[0093] In the formula, This represents the true label of the i-th class. The model predicts the probability of the i-th class; C represents the number of noise classes; M represents the batch size; f j noise This represents the j-th dominant interference frequency predicted by the model; This represents the estimated frequency of the j-th main interference.

[0094] The total loss function is constructed as follows:

[0095] L=α×L cls +(1-α)×L reg ;

[0096] In the formula, L represents the total loss, Lcls L represents the classification loss. reg Represents the regression loss, where α represents the classification loss weight;

[0097] S203. The generated one-dimensional CNN-LTSM hybrid model is lightweighted using the TensorFlow Lite framework. A sliding window mechanism is designed, where the length of the time series in the model input layer is set as the window length, and one window contains N milliseconds of data. The amplified electrical signal is used as the real signal x(t) and input into the one-dimensional CNN-LTSM hybrid model. The probability of the interference noise type and the estimated value of the main interference frequency of the real signal are output. The results are averaged after three repetitions and the final result is output to determine whether there is interference noise in the real signal. When interference noise exists, the type of interference noise is determined and the main interference frequency is obtained.

[0098] In temperature transmitters, deep learning technology is used to determine the type of interference noise and the dominant interference frequency in the amplified electrical signal. A one-dimensional CNN-LTSM hybrid model is constructed, combining the local feature extraction capability of convolutional neural networks and the temporal modeling capability of long short-term memory networks to achieve accurate classification and dominant frequency estimation of complex noise. Multi-task learning, such as classification + regression, reduces model redundancy and improves inference efficiency. It operates stably in the complex and variable electromagnetic environment of industrial sites, avoiding overfitting to a single noise type.

[0099] S300: Dynamically adjust the ADC sampling rate based on the main interference frequency characteristics output by the one-dimensional hybrid model;

[0100] The specific steps for dynamically adjusting the ADC sampling rate are as follows:

[0101] S301. Collect interference frequencies from unavailable signals in the historical working environment. Calculate the average and standard deviation of the interference frequencies of unavailable signals. Subtract the standard deviation from the average to obtain the high-frequency interference threshold f. thre ;

[0102] S302. Using the high-frequency interference threshold to determine the main interference frequency f output by the one-dimensional CNN-LTSM mixture model. noise Make a judgment when f noise >f thre When f is detected, it is determined to be a high-frequency interference scenario; when f noise ≤f thre At that time, it was determined to be a low-frequency interference scenario;

[0103] S303. Collect the available signal sampling rate in the historical working environment when the signal is in a high-frequency interference scenario and a low-frequency interference scenario, respectively. Construct a linear regression model to analyze the relationship between the available signal sampling rate and the interference frequency in the high-frequency interference scenario and the relationship between the available signal sampling rate and the interference frequency in the low-frequency interference scenario. Obtain the linear influence function of the two relationships respectively, and extract the coefficient of the available signal sampling rate in the function as the dynamic adjustment weight.

[0104] When it is determined that the real signal is in a high-frequency interference scenario, f s =δ1×f noise In the formula, f s The sampling rate is dynamically adjusted, and δ1 represents the dynamic adjustment weight of high-frequency interference; when the real signal is determined to be in a low-frequency interference scenario, f s =δ2×f noise δ2 represents the dynamic adjustment weight of low-frequency interference.

[0105] The sampling rate is dynamically adjusted for high-frequency and low-frequency interference scenarios respectively, improving digital filtering resolution, avoiding spectral overlap caused by aliasing, and ensuring signal integrity; extending the battery life of battery-powered devices; achieving more precise temperature control, improving system stability and control accuracy, and reducing temperature overshoot or undershoot.

[0106] S400 performs time-domain analysis on the electrical signal, extracts the signal baseline using the moving average method, and eliminates low-frequency drift; then it analyzes the frequency domain of the electrical signal, selects a filter, and constructs an adaptive notch filter to suppress frequency domain interference noise.

[0107] The specific steps for selecting a filter to eliminate interference and noise are as follows:

[0108] S401. Calculate the signal mean within the sliding window, and then eliminate low-frequency drift. The formula is:

[0109]

[0110] y corrected (t)=x(t)-y base (t);

[0111] In the formula, y base (t) represents the mean of the signal within the sliding window, characterizing the low-frequency drift component; W represents the length of the sliding window, and x(t) represents the actual input signal. corrected (t) represents the signal after eliminating low-frequency drift; y corrected (t) serves as the time-domain correction signal;

[0112] The possible values ​​of W are:

[0113] If f noiseFor frequencies ≤10Hz, select a larger window to fully smooth out low-frequency noise;

[0114] If f noise For frequencies >10Hz, select a smaller window to avoid high-frequency signal distortion;

[0115] S402. Select the corresponding frequency filter based on the type of interference noise in the real signal output by the one-dimensional CNN-LTSM hybrid model, and then construct an adaptive notch filter; the formula is:

[0116]

[0117] In the formula, H(z) represents an adaptive notch filter, r represents the pole radius, and z -1 Represents the unit delay operator;

[0118] The digital angular frequency is calculated based on the main interference frequency of the real signal, using the following formula: In the formula, θ represents the digital angular frequency, f s Represents the actual signal sampling rate; updates the adaptive notch filter coefficients, updating the cosine coefficients of the numerator and denominator to 2×cos(θ); and adjusts the signal y after eliminating low-frequency drift. corrected (t) is input to an adaptive notch filter to suppress frequency domain interference in the real signal; the output suppressed signal is y. notch (t), y notch (t) serves as the frequency-filtered signal.

[0119] S500: After performing time-domain analysis and frequency-domain analysis on the electrical signal, the two signals are weighted and fused to output the noise-reduced electrical signal.

[0120] The specific steps for weighted fusing the two signals (after time-domain analysis and frequency-domain analysis) to output the noise-reduced electrical signal are as follows:

[0121] S501. The time-domain corrected signal and the frequency-domain filtered signal are fused according to weights, using the following formula:

[0122] y final (t)=β×y corrected (t)+(1-β)×y notch (t);

[0123] In the formula, y final (t) represents the fused signal, and β represents the time-domain signal weight, which is dynamically adjusted according to the noise intensity.

[0124] If high-frequency impulse noise is detected: reduce β and focus on frequency domain wavelet denoising;

[0125] If low-frequency power frequency noise is detected: increase β to preserve time-domain details.

[0126] S502. The fused signal is finally optimized using the Kalman filter model, and the final output signal is y(t).

[0127] In temperature transmitters, interference can lead to signal distortion, increased bit error rate, and other problems, affecting signal quality. Time-domain and frequency-domain dual-domain joint interference cancellation technology can significantly enhance the system's anti-interference capability, effectively reduce the impact of interference on the signal, and improve signal integrity and accuracy. For signals affected by multiple interference sources, dual-domain joint interference cancellation can address interference from different angles, making the signal more stable during transmission, reducing signal fluctuations and distortion, and improving signal quality and availability.

[0128] By employing appropriate processing methods in both the time and frequency domains, various types of interference can be effectively suppressed and eliminated, demonstrating strong flexibility and adaptability.

[0129] S600 uses the output electrical signal to calculate the anti-interference performance of the temperature transmitter and verifies its anti-interference performance.

[0130] The specific steps for verifying anti-interference performance are as follows:

[0131] S601. Simulate different types of interference combinations in the temperature control chamber, record the real signal, the processed signal, and the reference signal, and use the three signals to calculate the signal-to-noise ratio (SNR) improvement. When the SNR improvement is negative, verify that the anti-interference performance is reduced, and re-acquire data to rebuild the model; when the SNR improvement is positive, verify that the anti-interference performance is improved.

[0132] An anti-interference temperature transmitter measurement and control system, comprising a data acquisition module, an interference noise identification module, a sampling rate adjustment module, a dual-domain noise reduction module, and a performance verification module;

[0133] The data acquisition module is used to collect the frequency and sampling rate of signal interference noise in historical work.

[0134] The interference noise identification module is used to construct a one-dimensional hybrid model using deep learning; the amplified electrical signal is input into the model to identify the noise type and main interference frequency in the electrical signal;

[0135] The sampling rate adjustment module is used to dynamically adjust the ADC sampling rate based on the main interference frequency characteristics output by the one-dimensional hybrid model.

[0136] The dual-domain noise reduction module is used to perform time-domain and frequency-domain analysis on the signal to eliminate interference noise in the signal;

[0137] The performance verification module is used to calculate the signal-to-noise ratio of the processed signal and verify its anti-interference performance.

[0138] The interference and noise identification module includes a training set construction unit, a model construction unit, and a real-time inference unit;

[0139] The training set construction unit is used to collect and set the baseline signal, the original signal, and the type of interference noise to generate a label dataset.

[0140] The model building unit is used to design the input layer, convolutional layer, neural network layer and output layer respectively, and to build a one-dimensional CNN-LTSM hybrid model;

[0141] The real-time inference unit is used to design a sliding window mechanism, which uses a sliding window to input real data into the model for judgment.

[0142] The dual-domain noise reduction module includes a time-domain noise reduction unit, a frequency-domain suppression unit, and a dual-domain fusion unit;

[0143] The time-domain noise reduction unit is used to calculate the signal mean within the sliding window and then eliminate low-frequency drift.

[0144] The frequency domain suppression unit is used to design an adaptive notch filter to suppress frequency domain interference in the real signal.

[0145] The dual-domain fusion unit is used to fuse the time-domain corrected signal and the frequency-domain filtered signal according to weights.

[0146] Example:

[0147] The temperature signal is superimposed with 50Hz power frequency interference (amplitude 5mV) and random pulses (amplitude 10mV).

[0148] Processing result:

[0149] Time-domain correction: Eliminates baseline drift below 0.1 Hz (e.g., slow changes in ambient temperature);

[0150] Frequency domain notch filtering: 50Hz noise attenuation ≥30dB;

[0151] Wavelet denoising: The pulse amplitude was reduced from 10mV to 0.5mV.

[0152] Fusion output: Signal-to-noise ratio (SNR) improved from 20dB to 45dB, response time <50ms.

[0153] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for measuring and controlling an anti-interference temperature transmitter, characterized in that: The method includes the following steps: S100: The ambient temperature is collected in the working environment. The collected temperature signal is calculated and converted into an electrical signal in the temperature transmitter. The electrical signal is then sent to the amplifier for amplification. S200. A pure temperature signal is collected under interference-free conditions using a constant temperature device, and an interference noise signal is injected to construct a labeled dataset as a training set, and a one-dimensional hybrid model is constructed. The amplified electrical signal is input into the model to identify the noise type and main interference frequency in the electrical signal. S300: Dynamically adjust the ADC sampling rate based on the main interference frequency characteristics output by the one-dimensional hybrid model; The specific steps for dynamically adjusting the ADC sampling rate are as follows: S301. Collect interference frequencies from unavailable signals in the historical working environment. Calculate the average and standard deviation of the interference frequencies of unavailable signals. Subtract the standard deviation from the average to obtain the high-frequency interference threshold f. thre ; S302. Using the high-frequency interference threshold to determine the main interference frequency f output by the one-dimensional CNN-LTSM mixture model. noise Make a judgment when f noise >f thre When f is detected, it is determined to be a high-frequency interference scenario; when f noise ≤f thre At that time, it was determined to be a low-frequency interference scenario; S303. Collect the available signal sampling rate in the historical working environment when the signal is in a high-frequency interference scenario and a low-frequency interference scenario, respectively. Construct a linear regression model to analyze the relationship between the available signal sampling rate and the interference frequency in the high-frequency interference scenario and the relationship between the available signal sampling rate and the interference frequency in the low-frequency interference scenario. Obtain the linear influence function of the two relationships respectively, and extract the coefficient of the available signal sampling rate in the function as the dynamic adjustment weight. When it is determined that the real signal is in a high-frequency interference scenario In the formula, f s This represents the dynamically adjusted sampling rate, and δ1 represents the dynamically adjusted weight of high-frequency interference; when the real signal is determined to be in a low-frequency interference scenario, δ2 represents the dynamic adjustment weight of low-frequency interference; S400 performs time-domain analysis on the electrical signal, extracts the signal baseline using the moving average method, and eliminates low-frequency drift; then it analyzes the frequency domain of the electrical signal, selects a filter, and constructs an adaptive notch filter to suppress frequency domain interference noise. S500: After performing time-domain analysis and frequency-domain analysis on the electrical signal, the two signals are weighted and fused to output the noise-reduced electrical signal. S600 uses the output electrical signal to calculate the anti-interference performance of the temperature transmitter and verifies its anti-interference performance.

2. The anti-interference temperature transmitter measurement and control method according to claim 1, characterized in that: The specific steps in S100 for transmitting the electrical signal to the amplifier for amplification are as follows: S101. Use a temperature sensor to collect the real-time temperature of the work site. The temperature sensor transmits the collected temperature signal to the temperature transmitter. The temperature transmitter supports temperature signals transmitted by different types of temperature sensors, including thermocouple temperature sensors and resistance temperature sensors. The temperature transmitter calculates the received temperature signal and converts it into a weak electrical signal, and sends the generated weak electrical signal to the amplifier circuit in the temperature transmitter. S102. When the temperature sensor is a thermocouple type, the temperature transmitter amplifies the received weak electrical signal, calculates the temperature value corresponding to the temperature signal according to the thermocouple calibration table, and finally converts it into a standard current or voltage signal for output. When the temperature sensor is a resistance temperature detector (RTD) type, the temperature transmitter amplifies the received weak electrical signal and converts it into a standard current or voltage signal output through the internal circuitry of the transmitter.

3. The anti-interference temperature transmitter measurement and control method according to claim 1, characterized in that: The specific steps in S200 for inputting the amplified electrical signal into the model to identify the noise type and main interference frequency in the electrical signal are as follows: S201. Acquire a pure temperature signal under interference-free conditions using a constant temperature device, and use the pure temperature signal as a reference signal x. ref Different types of artificially designed interference noise signals are added to the reference signal (t), and the signal after the interference noise is injected is used as the original signal xs(t). The different types of artificially designed interference noise are labeled to obtain classification labels, and the interference noise is analyzed using Fourier transform to obtain the main interference noise frequency f. noise Using the main interference noise frequency as the regression label, a labeled dataset is constructed. The data format in the labeled dataset is {xs(t), y...} cls y reg }, where y cls Indicates category label, y reg Indicates the regression label; S202. Design the input layer shape as (N, 1), where N represents the time series length; construct three convolutional layers with kernel sizes of 3, 5, and 7 respectively; the number of filters are 16, 32, and 64 respectively; add the ReLU activation function to each layer; and apply max pooling after each layer to compress the feature dimension. Design an LSTM neural network with 64 units; design the output layer to include a classification branch and a regression branch, with the classification branch outputting the probability P of the type of interference noise. cls The regression branch outputs the estimated main interference frequency fg. noise Generate a one-dimensional CNN-LTSM mixture model; train and learn the one-dimensional CNN-LTSM mixture model using a labeled dataset; The classification loss and regression loss are calculated using the probability of the interference noise type and the estimated frequency of the main interference, respectively, and the total loss function is constructed as follows: ； In the formula, L represents the total loss, L cls L represents the classification loss. reg Represents the regression loss, where α represents the classification loss weight; S203. The generated one-dimensional CNN-LTSM hybrid model is lightweighted using the TensorFlow Lite framework. A sliding window mechanism is designed, where the length of the time series in the model input layer is set as the window length, and one window contains N milliseconds of data. The amplified electrical signal is used as the real signal x(t) and input into the one-dimensional CNN-LTSM hybrid model. The probability of the interference noise type and the estimated value of the main interference frequency of the real signal are output. The results are averaged after three repetitions and the final result is output to determine whether there is interference noise in the real signal. When interference noise exists, the type of interference noise is determined and the main interference frequency is obtained.

4. The anti-interference temperature transmitter measurement and control method according to claim 3, characterized in that: The specific steps for selecting a filter to eliminate interference and noise in S400 are as follows: S401. Calculate the signal mean within the sliding window, and then eliminate low-frequency drift. The formula is: ； In the formula, y base (t) represents the mean of the signal within the sliding window, characterizing the low-frequency drift component; W represents the length of the sliding window, x(t) represents the actual input signal, and y corrected (t) represents the signal after eliminating low-frequency drift; y corrected (t) serves as the time-domain correction signal; S402. Select the corresponding frequency filter based on the type of interference noise in the real signal output by the one-dimensional CNN-LTSM hybrid model, and then construct an adaptive notch filter. The digital angular frequency is calculated based on the main interference frequency of the real signal, using the following formula: In the formula, θ represents the digital angular frequency, f s Represents the actual signal sampling rate; updates the adaptive notch filter coefficients, updating the cosine coefficients of the numerator and denominator to 2×cos(θ); and adjusts the signal y after eliminating low-frequency drift. corrected (t) is input into an adaptive notch filter to suppress frequency domain interference in the real signal; the output suppressed signal is y. notch (t), y notch (t) serves as the frequency-filtered signal.

5. The anti-interference temperature transmitter measurement and control method according to claim 4, characterized in that: The specific steps in S500 for weighted fusing the two signals (after time-domain analysis and frequency-domain analysis) and outputting the noise-reduced electrical signal are as follows: S501. The time-domain corrected signal and the frequency-domain filtered signal are fused according to weights, using the following formula: ； In the formula, y final (t) represents the fused signal, and β represents the time-domain signal weight, which is dynamically adjusted according to the noise intensity; S502. The fused signal is finally optimized using the Kalman filter model, and the final output signal is y(t).

6. The anti-interference temperature transmitter measurement and control method according to claim 5, characterized in that: The specific steps for verifying the anti-interference performance of the S600 are as follows: S601. Simulate different types of interference combinations in the temperature control chamber, record the real signal, the processed signal, and the reference signal, and use the three signals to calculate the signal-to-noise ratio (SNR) improvement. When the SNR improvement is negative, verify that the anti-interference performance is reduced, and re-acquire data to rebuild the model; when the SNR improvement is positive, verify that the anti-interference performance is improved.

7. An anti-interference temperature transmitter measurement and control system applying the anti-interference temperature transmitter measurement and control method according to any one of claims 1-6, characterized in that: The temperature transmitter measurement and control system includes a data acquisition module, an interference and noise identification module, a sampling rate adjustment module, a dual-domain noise reduction module, and a performance verification module. The data acquisition module is used to collect the frequency and sampling rate of signal interference noise in historical work. The interference noise identification module is used to construct a one-dimensional hybrid model using deep learning; the amplified electrical signal is input into the model to identify the noise type and main interference frequency in the electrical signal; The sampling rate adjustment module is used to dynamically adjust the ADC sampling rate based on the main interference frequency characteristics output by the one-dimensional hybrid model. The dual-domain noise reduction module is used to perform time-domain and frequency-domain analysis on the signal to eliminate interference noise in the signal; The performance verification module is used to calculate the signal-to-noise ratio of the processed signal and verify its anti-interference performance.

8. The anti-interference temperature transmitter measurement and control system according to claim 7, characterized in that: The interference noise identification module includes a training set construction unit, a model construction unit, and a real-time inference unit. The training set construction unit is used to collect and set the baseline signal, the original signal, and the type of interference noise to generate a label dataset. The model building unit is used to design the input layer, convolutional layer, neural network layer and output layer respectively, and to build a one-dimensional CNN-LTSM hybrid model; The real-time inference unit is used to design a sliding window mechanism, which uses a sliding window to input real data into the model for judgment.

9. The anti-interference temperature transmitter measurement and control system according to claim 7, characterized in that: The dual-domain noise reduction module includes a time-domain noise reduction unit, a frequency-domain suppression unit, and a dual-domain fusion unit; The time-domain noise reduction unit is used to calculate the signal mean within the sliding window and then eliminate low-frequency drift. The frequency domain suppression unit is used to design an adaptive notch filter to suppress frequency domain interference in the real signal. The dual-domain fusion unit is used to fuse the time-domain corrected signal and the frequency-domain filtered signal according to weights.

Images