A signal denoising method and system, and a glass electric melting furnace signal transmission system

Abstract

The application provides a signal denoising method and system and a glass electric melting furnace signal transmission system, and belongs to the technical field of industrial signal transmission. The method comprises the following steps: synchronously collecting a first signal and a second signal of a homologous interference; respectively performing ensemble empirical mode decomposition to obtain a plurality of first intrinsic mode function components and a first residual function corresponding to the first signal and a plurality of second intrinsic mode function components corresponding to the second signal; calculating permutation entropy values of the plurality of first intrinsic mode function components; calculating similarity measurement values of the first and second intrinsic mode function components; screening effective first intrinsic mode function components; and reconstructing the first signal after denoising. The beneficial effects are as follows: the parallel ensemble empirical mode decomposition is adopted, the second signal of the homologous interference is taken as a reference, each component after the decomposition of the first signal is screened, the noise interference is removed, the real signal characteristics are retained, the pure signal with a high signal-to-noise ratio is reconstructed, and the precision and reliability of the signal denoising are improved.

Description

Technical Field

[0001] This invention relates to the field of industrial signal transmission technology, and in particular to a signal noise reduction method and system, and a signal transmission system for glass electric melting furnaces. Background Technology

[0002] In traditional glass melting furnaces, temperature measurement typically relies on type K compensated thermocouples. Type K compensated thermocouples use compensating wires whose thermoelectric properties are similar to those of the thermocouple itself, thus ensuring the accuracy of signal transmission in glass electric melting furnaces.

[0003] However, in practical applications of electric melting furnaces, measurement signals often exhibit errors. The main reasons for these errors are as follows: First, fluctuations in ambient temperature directly affect the thermocouple measurement results; second, the compensating wires and their connection points have a certain resistance, which introduces thermal noise, distorting the measurement signal; third, external electromagnetic fields have a particularly significant impact on thermocouple signals. In the complex electromagnetic environment of an electric melting furnace, electromagnetic fields easily interact with thermocouple signals, leading to measurement errors.

[0004] Furthermore, during the glass melting process, some equipment that directly contacts the molten glass, such as agitators, is also subject to interference from complex electromagnetic fields, making it impossible to accurately acquire equipment motor signals, such as current and frequency signals. Although shielding and grounding can mitigate electromagnetic interference to some extent, the nonlinear effects they may cause can still generate thermal noise, leading to signal errors, reduced melting furnace control efficiency, and hindering the glass factory's goal of energy conservation and emission reduction, while increasing production costs and energy consumption. Summary of the Invention

[0005] To address the above technical problems, the present invention provides a signal noise reduction method; furthermore, it provides a signal noise reduction system; and still further, it provides a signal transmission system for a glass electric melting furnace.

[0006] The technical problem solved by this invention can be achieved by the following technical solutions: A signal denoising method, comprising: Step S1: Synchronously acquire the first signal and the second signal that is interfering with the first signal from the same source; Step S2: Perform ensemble empirical mode decomposition on the first signal and the second signal respectively to obtain multiple first intrinsic mode function components and a first residual function corresponding to the first signal, and multiple second intrinsic mode function components corresponding to the second signal. Step S3: Calculate the permutation entropy value of multiple first intrinsic mode function components; Step S4: Calculate the similarity metric between the first intrinsic mode function component and the second intrinsic mode function component; Step S5: Filter the valid first intrinsic mode function components based on the permutation entropy value and similarity metric value; Step S6: Reconstruct the signal corresponding to the denoised first signal by using the selected valid first intrinsic mode function components and the first residual function.

[0007] Preferably, the first signal is a temperature signal from a thermocouple; The second signal is the motor current signal.

[0008] Preferably, in step S2, the first signal and the second signal are used as signals to be decomposed for ensemble empirical mode decomposition, wherein the ensemble empirical mode decomposition includes: Step S21: Add a zero-mean Gaussian white noise sequence to the signal to be decomposed to obtain a new sequence; Step S22: Perform empirical mode decomposition on the new sequence to obtain multiple intrinsic mode function components and residual functions; Step S23: Repeat steps S21-S22 a preset number of times, with each time a different white noise sequence is added; Step S24: The mean value of each intrinsic mode function component obtained after empirical mode decomposition for a preset number of times is determined as the final intrinsic mode function component and residual function.

[0009] Preferably, step S3 includes: Step S31: Reconstruct the phase space for each of the first intrinsic mode function components to obtain a set of reconstruction vectors; Step S32: Rearrange the vector elements in each reconstructed vector in ascending order of their values, and record the symbol sequence formed by the index positions; Step S33: Calculate the relative frequency of each symbol sequence relative to the total number of reconstructed vectors; Step S34: Calculate the permutation entropy value corresponding to each first intrinsic mode function component based on the relative frequency; Step S35: Normalize the permutation entropy value to obtain the normalized permutation entropy value.

[0010] Preferably, the method for calculating the permutation entropy value is as follows:

[0011] in, Indicates relative frequency; k Indicates the symbol sequence index; m Indicates the embedding dimension; Represents the permutation entropy value; The normalized permutation entropy value is:

[0012] in, This represents the normalized permutation entropy value.

[0013] Preferably, the similarity metric is the Pearson cross-correlation coefficient, and step S4 includes: Step S41: Select one first intrinsic mode function component from multiple first intrinsic mode function components, and simultaneously select one second intrinsic mode function component from multiple second intrinsic mode function components; Step S42: Calculate the covariance of the selected first intrinsic mode function component and the second intrinsic mode function component, and calculate the first standard deviation corresponding to the selected first intrinsic mode function component and the second standard deviation corresponding to the selected second intrinsic mode function component respectively. Step S43: Calculate the Pearson cross-correlation coefficients of the selected first intrinsic mode function component and the second intrinsic mode function component based on the covariance, the first standard deviation, and the second standard deviation, and determine the calculated Pearson cross-correlation coefficients as the similarity measure.

[0014] Preferably, the Pearson cross-correlation coefficient is calculated as follows:

[0015] in, Represents the first intrinsic mode function component Second intrinsic mode function component covariance; This represents the first standard deviation corresponding to the selected first intrinsic mode function component; This represents the second standard deviation corresponding to the selected second intrinsic mode function component; This represents the Pearson cross-correlation coefficient.

[0016] Preferably, step S5 includes: Step S51: For each first intrinsic mode function component, detect whether a preset condition is met; wherein, the preset condition is that the permutation entropy value is less than a first preset threshold and the similarity metric value with the second intrinsic mode function component is less than a second preset threshold. If the preset conditions are met, the first intrinsic mode function component is retained; If the preset conditions are not met, the first intrinsic mode function component is removed; Step S52: Iterate through each of the first intrinsic mode function components in all the first intrinsic mode function components until all the first intrinsic mode function components have been traversed, so as to filter out all valid first intrinsic mode function components.

[0017] On the other hand, a signal denoising system is provided for implementing the signal denoising method as described above, including: The signal acquisition module is used to simultaneously acquire the first signal and the second signal that interferes with the first signal from the same source. The mode decomposition module, connected to the signal acquisition module, is used to perform ensemble empirical mode decomposition on the first signal and the second signal respectively, to obtain multiple first intrinsic mode function components and a first residual function corresponding to the first signal, and multiple second intrinsic mode function components corresponding to the second signal; The permutation entropy calculation module, connected to the mode decomposition module, is used to calculate the permutation entropy values ​​of multiple first intrinsic mode function components; The similarity measurement module, connected to the permutation entropy calculation module, is used to calculate the similarity measurement value between the first intrinsic mode function component and the second intrinsic mode function component; The filtering module, connected to the similarity measurement module, is used to filter valid first intrinsic mode function components based on permutation entropy values ​​and similarity measurement values. The signal reconstruction module, connected to the filtering module, is used to reconstruct the signal based on the filtered valid first intrinsic mode function components and the first residual function, thereby obtaining the signal corresponding to the denoised first signal.

[0018] On another front, a signal transmission system for a glass electric melting furnace is provided, comprising a transmitter and a receiver, wherein the transmitter is used to perform the signal noise reduction method described above on a first signal.

[0019] The advantages or beneficial effects of the technical solution of this invention are as follows: This invention synchronously acquires a second signal that interferes with the first signal from the same source, and uses parallel ensemble empirical mode decomposition technology. Taking advantage of the characteristics of noise originating from the same source and signals being heterogeneous, the second signal that interferes from the same source is used as a reference to assist in screening the components after the first signal is decomposed, eliminating noise interference and retaining the true signal characteristics, and finally reconstructing a clean signal with a high signal-to-noise ratio, thereby improving the accuracy and reliability of signal denoising. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a preferred embodiment of the signal noise reduction method of the present invention. Figure 2 This is a schematic diagram of the process of set empirical mode decomposition in a preferred embodiment of the present invention; Figure 3 This is a schematic diagram of the permutation entropy calculation process in a preferred embodiment of the present invention; Figure 4 This is a schematic diagram of the similarity measurement process in a preferred embodiment of the present invention; Figure 5This is a schematic diagram of the process for filtering effective intrinsic mode function components in a preferred embodiment of the present invention; Figure 6 This is a structural block diagram of the signal noise reduction system in a preferred embodiment of the present invention; Figure 7 This is a structural block diagram of the signal transmission system for a glass electric melting furnace, as shown in a preferred embodiment of the present invention. Detailed Implementation

[0021] 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.

[0022] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0023] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0024] Empirical Mode Decomposition (EMD) is a method for decomposing signals based on the inherent time-scale characteristics of data. It can progressively filter out local feature signals at different time scales in the original signal to obtain intrinsic mode function (IMF) components with the same characteristics. IMF components should satisfy the following two conditions: 1. The mean of the envelope formed by the local maxima and minima of the signal is zero; 2. The difference between the number of poles and the number of zeros is no greater than 1.

[0025] For a given original sequence The empirical mode decomposition steps are as follows: Step 1.1: Obtain the original sequence All extreme points (maximum and minimum values) are identified, and then the envelope is fitted using cubic spline interpolation. and lower envelope .

[0026] Step 1.2: Calculate the mean of the two envelopes. Then, the values ​​of the sequence elements are obtained. and mean The difference ,in:

[0027]

[0028] Step 1.3: Determine the difference Does it meet the conditions for IMF component? If satisfied, then the difference For the first valid IMF component, let be... And calculate the difference between the original signal and the IMF component. ; If not satisfied, then let the difference... Given the original sequence, repeat the first two steps until the condition is met.

[0029] Step 1.4: Difference As the signal to be decomposed, repeat steps 1.1-1.3 until the remaining signal is obtained. When the function is monotonic, the decomposition stops. The original sequence can then be represented as:

[0030] in, Indicates the first i One valid IMF component; n This represents the total number of valid IMF components; Original signal and the first n The difference between the IMF components.

[0031] While traditional EMD reduces the influence of human factors, it suffers from mode aliasing when dealing with intermittent interference (such as sudden spike pulse interference generated by motor start-up and shutdown or electrode discharge in electric melting furnaces). This means that signal components from different time scales appear in the same IMF, making it impossible to effectively separate noise from useful signals. Ensemble Empirical Mode Decomposition (EEMD) addresses the shortcomings of Empirical Mode Decomposition by proposing a noise-assisted data analysis method. Traditional EEMD effectively suppresses mode aliasing by adding a white noise sequence to the original sequence and performing multiple ensemble averagings with the assistance of white noise. This results in exceptionally clear and stable decomposition results, reliably extracting pure signals from strong, sudden electromagnetic interference. The specific steps are as follows: Step 2.1: In the original sequence A new sequence is obtained by adding zero-mean Gaussian white noise to the sequence. Its noise variance is Where p is the white noise intensity parameter, for The standard deviation.

[0032] Step 2.2: For EMD decomposition was performed to obtain several IMF components.

[0033] Step 2.3: Repeat N Steps 2.1-2.2 involve adding different white noise sequences each time.

[0034] Step 2.4: N The mean of each component obtained from the second decomposition is used as the original sequence. The final IMF component.

[0035] Based on this, the method and system of this invention introduce ensemble empirical mode decomposition (EEMD) into the field of signal transmission to analyze and filter the generated intrinsic mode function (IMF) components, thereby reducing signal errors. Ensemble empirical mode decomposition (EEMD) has advantages such as adaptive decomposition, multi-scale analysis, and anti-modal aliasing, and is particularly suitable for processing nonlinear and non-stationary signals containing thermal noise. By applying this technology, the accuracy and stability of signal processing can be effectively improved, the level of intelligence in production processes can be enhanced, control efficiency can be optimized, energy conservation and emission reduction can be contributed, and a solid technical foundation can be laid for the sustainable development of industrial production.

[0036] In a preferred embodiment of the present invention, based on the above-mentioned problems existing in the prior art, a signal noise reduction method is provided, such as... Figure 1 As shown, it includes: Step S1: Synchronously acquire the first signal and the second signal that is interfering with the first signal from the same source; Specifically, a temperature signal channel and a motor current signal channel are set up. The data source for the temperature signal channel is a type-K compensated thermocouple, while the data for the motor current signal channel comes from a current sensor, such as a Hall current sensor, connected to equipment that directly contacts the molten glass, such as a stirrer or blower. By synchronously acquiring the signals from these two channels, strict time alignment is ensured.

[0037] In this embodiment, the first signal is the temperature signal from a thermocouple, and the second signal is the motor current signal. These two signals are fundamentally different. The useful real signal in the thermocouple temperature signal is the low-frequency, slowly changing temperature trend. The useful real signal in the motor current signal is the power frequency (50 / 60Hz) and its harmonics, reflecting the motor's operating power and status.

[0038] The key point is that the temperature signal and the motor current signal are both affected by the same electromagnetic interference (EMI) source. EMI, especially sudden spikes, exhibits highly similar time-domain and frequency-domain characteristics in both signals, i.e., similar IMF components. However, their respective useful signals are quite different. This embodiment utilizes this characteristic of "common noise source, heterogeneous signal," using the motor current signal as a reference to analyze its noise mode characteristics, and assists in identifying and eliminating noise components in the temperature signal to be denoised. Through multi-signal correlation analysis, the ability to suppress common-mode noise is improved, further enhancing the accuracy and reliability of temperature measurement.

[0039] Alternatively, the temperature signal can be used as a reference to assist in denoising the motor current signal.

[0040] Step S2: Perform ensemble empirical mode decomposition on the first signal and the second signal respectively to obtain multiple first intrinsic mode function components and a first residual function corresponding to the first signal, and multiple second intrinsic mode function components corresponding to the second signal. In step S2, the first signal and the second signal are used as the signals to be decomposed for ensemble empirical mode decomposition, as follows: Figure 2 As shown, set empirical mode decomposition includes: Step S21: Add a zero-mean Gaussian white noise sequence to the signal to be decomposed to obtain a new sequence; Step S22: Perform empirical mode decomposition on the new sequence to obtain multiple intrinsic mode function components and residual functions; Step S23: Repeat steps S21-S22 a preset number of times, with each time a different white noise sequence is added; Step S24: The mean value of each intrinsic mode function component obtained after empirical mode decomposition for a preset number of times is determined as the final intrinsic mode function component and residual function.

[0041] Specifically, for the synchronously acquired temperature signal and motor current signal, ensemble empirical mode decomposition (EEMD) is performed respectively, resulting in two sets of intrinsic mode functions (IMF): Temperature signal IMF group: IMFT 1 IMFT 2 ,...,IMFTn ; Motor current signal IMF group: IMFC 1 IMFC 2 ,...,IMFCn . Step S3: Calculate the permutation entropy values ​​of multiple first intrinsic mode function components; After the EEMD decomposes the IMF components, in order to automatically and accurately determine which are noise components and which are signal components, in a preferred embodiment of the present invention, the permutation entropy (PE) is used as an index to quantify the randomness of each intrinsic mode function (IMF) to achieve adaptive identification of noise components.

[0042] like Figure 3 As shown, step S3 includes: Step S31: Reconstruct the phase space for each first intrinsic mode function component to obtain a set of reconstruction vectors; For any intrinsic mode function component to be analyzed, its time series is denoted as... X ={ x ( i )}, i =1,2,..., N , N Where is the sequence length. Phase space reconstruction is performed according to the following formula to obtain a set of reconstructed vectors:

[0043] in, m This represents the embedding dimension. m The value ranges from 3 to 7, with 6 being the preferred value; L Indicates the delay time. L It is a positive integer, and the delay time is usually... L The preferred value is 1, but other positive integers can also be selected. j Represents a vector index. j =1,2,..., K , K Indicates the number of reconstructed vectors. ; x ( j ) represents the first element in the reconstructed vector. j 1 vector element; X ( j ) represents the reconstructed vector.

[0044] Step S32: Rearrange the vector elements in each reconstructed vector in ascending order of their values, and record the symbol sequence formed by the index positions; For each reconstructed vector X ( j ), and among them mThe elements are rearranged in ascending order of value, and the symbol sequence formed by their index positions is recorded, which is called the "Ordinal Pattern".

[0045] For example, for an embedding dimension m =3 reconstructed vector[ x j+2 , x j , x j+1 [19, 9, 15]. First, sorting by numerical ascending order gives: 9, 15, 19; the corresponding original index order is: j , j +1, j +2; Therefore, the permutation pattern corresponding to this reconstructed vector is ( j , j +1, j +2).

[0046] for m A dimensional vector, the total number of possible permutations is m ! represents the factorial symbol. m ! represents the factorial of m, i.e. m !=1×2×3×…× m .

[0047] Step S33: Calculate the relative frequency of each symbol sequence relative to the total number of reconstructed vectors; For any symbol sequence, count the number of times the symbol sequence appears in the entire time series, and divide the number of times by the total number of reconstructed vectors to obtain the relative frequency of the symbol sequence.

[0048] For example, for the first k a sequence of symbols, k =1,2,..., m !, The statistics show that the first k The number of occurrences of a symbol sequence, divided by the total number of reconstructed vectors. K The relative frequency of this arrangement pattern is obtained. P k : P k = C k / K .

[0049] in, C k Indicates the first k The number of times a symbol sequence appears.

[0050] Step S34: Calculate the permutation entropy value corresponding to each first intrinsic mode function component based on the relative frequency; Based on the information entropy formula, the relative frequency obtained in step S33 is used. P k Calculate the permutation entropy of the IMF component. H PE :

[0051] in, Indicates relative frequency; k Indicates the symbol sequence index; m Indicates the embedding dimension; Represents the permutation entropy value; Step S35: Normalize the permutation entropy value to obtain the normalized permutation entropy value.

[0052] For easier comparison, the above permutation entropy calculation result is divided by its theoretical maximum value ln( m !), thus obtaining the normalized permutation entropy value:

[0053] in, This represents the normalized permutation entropy value.

[0054] This invention introduces permutation entropy (PE), a nonlinear dynamic index, as an adaptive selection criterion for IMF components. Permutation entropy can quantify the randomness and complexity of a time series. The higher the complexity of the noise component, the larger the permutation entropy value; the stronger the regularity of the real temperature signal, the smaller the permutation entropy value. Normalized permutation entropy. H NPE The value range is [0,1]. The closer its value is to 1, the stronger the randomness and complexity of the time series; the closer its value is to 0, the stronger the regularity and certainty of the time series.

[0055] This invention sets a permutation entropy threshold and automatically identifies and discards noise components by comparing the permutation entropy value of the IMF with the preset threshold. This allows for fully automatic and adaptive screening and reconstruction of IMF components without manual intervention, greatly improving the intelligence level and universality of the method under different working conditions.

[0056] Step S4: Calculate the similarity metric between the first intrinsic mode function component and the second intrinsic mode function component; The similarity metric is the Pearson cross-correlation coefficient, such as... Figure 4 As shown, step S4 includes: Step S41: Select one first intrinsic mode function component from multiple first intrinsic mode function components, and simultaneously select one second intrinsic mode function component from multiple second intrinsic mode function components; Specifically, one first intrinsic mode function component is arbitrarily selected from the multiple first intrinsic mode function components obtained by EEMD decomposition of the first signal, denoted as . X ={ x 1, x 2,..., x N}

[0057] Simultaneously, from the multiple second intrinsic mode function components obtained by EEMD decomposition of the second signal that co-interferes with the first signal, one second intrinsic mode function component is arbitrarily selected and denoted as . Y= { y 1 ,y 2 ,...,y N}

[0058] Two intrinsic mode function (IMF) components were selected as the two components to be compared.

[0059] To ensure comparability, the sequence lengths of the first intrinsic mode function component X and the second intrinsic mode function component Y are equal, both being... N Furthermore, strict alignment of time points is ensured during the initial synchronous data collection phase.

[0060] Step S42: Calculate the covariance of the selected first intrinsic mode function component and the second intrinsic mode function component, and calculate the first standard deviation corresponding to the selected first intrinsic mode function component and the second standard deviation corresponding to the selected second intrinsic mode function component respectively. Specifically, first, the arithmetic mean of the sequences of the two intrinsic mode function components is calculated separately:

[0061] in, Represents the first intrinsic mode function component The i One element; N Indicates the total number of elements in the component; Represents the first intrinsic mode function component The arithmetic mean;

[0062] in, Represents the second intrinsic mode function component The i One element; Represents the second intrinsic mode function component The arithmetic mean.

[0063] Then, the components of the first intrinsic mode function are calculated. Second intrinsic mode function component The covariance. Covariance measures the correlation between the population errors of two variables, and it is calculated as follows:

[0064] in, Represents the first intrinsic mode function component Second intrinsic mode function component The covariance.

[0065] Next, the standard deviations of the selected first and second intrinsic mode function components are calculated respectively.

[0066] The first standard deviation corresponding to the selected first intrinsic mode function component is:

[0067] in, This represents the first standard deviation corresponding to the selected first intrinsic mode function component.

[0068] The second standard deviation corresponding to the selected second intrinsic mode function component is:

[0069] in, This represents the second standard deviation corresponding to the selected second intrinsic mode function component.

[0070] Step S43, based on covariance First standard deviation Second standard deviation Calculate the Pearson cross-correlation coefficients of the selected first intrinsic mode function component and the second intrinsic mode function component, and determine the calculated Pearson cross-correlation coefficients as the similarity measure.

[0071] The normalized Pearson cross-correlation coefficient is calculated as follows:

[0072] in, Represents the first intrinsic mode function component Second intrinsic mode function component covariance; This represents the first standard deviation corresponding to the selected first intrinsic mode function component; This represents the second standard deviation corresponding to the selected second intrinsic mode function component; This represents the Pearson cross-correlation coefficient. The range of values ​​is .when When it indicates a perfect positive correlation; when When it indicates a perfect negative correlation; when Time indicates wireless correlation.

[0073] According to Pearson's cross-correlation coefficient Filter out valid IMF components. If the absolute value of the Pearson cross-correlation coefficient... If the correlation between the two is weak, it is necessary to combine other indicators such as their permutation entropy value to comprehensively judge whether the IMF component is effective.

[0074] If the absolute value of Pearson's cross-correlation coefficient If the two IMF components are highly correlated, the noise in the IMF component corresponding to the temperature signal is mainly common-mode electromagnetic interference, which should be discarded during reconstruction.

[0075] It should be noted that although the operations of the method of the present invention are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order. The order of steps listed in the embodiments is merely one possible order of execution among many steps, and does not represent the only possible order. For example, steps S3 and S4 are not in any particular order. In practical applications, step S4 can be executed first, followed by step S3, or they can be executed simultaneously.

[0076] Step S5: Filter the valid first intrinsic mode function components based on the permutation entropy value and similarity metric value; Among them, such as Figure 5 As shown, step S5 includes: Step S51: For each first intrinsic mode function component, detect whether it meets the preset conditions; wherein, the preset conditions are that the permutation entropy value is less than the first preset threshold and the similarity metric value with the second intrinsic mode function component is less than the second preset threshold. If the preset conditions are met, the first intrinsic mode function component is retained; If the preset conditions are not met, the first intrinsic mode function component is removed. Step S52: Iterate through each of the first intrinsic mode function components in all the first intrinsic mode function components until all the first intrinsic mode function components have been traversed, so as to filter out all valid first intrinsic mode function components.

[0077] Specifically, in this embodiment, adaptive selection of IMF components is achieved by combining permutation entropy and similarity measurement. First, a first preset threshold corresponding to permutation entropy is set. H th And a second preset threshold corresponding to a similarity metric. .

[0078] For any IMF component, if its permutation entropy value H NPE < H th And the absolute value of Pearson's cross-correlation coefficient If the signal is positive, the component is considered a valid signal component and is retained during signal reconstruction.

[0079] Conversely, for other situations, such as H NPE ≥ H th or If the signal is not clear, the component is determined to be noise or interference and is discarded during signal reconstruction.

[0080] In this embodiment, the first preset threshold Hth The value can range from 0.4 to 0.6, preferably 0.55. Second preset threshold. The preferred value is 0.7.

[0081] Step S6: Reconstruct the signal corresponding to the denoised first signal by using the selected effective first intrinsic mode function components and the first residual function.

[0082] Finally, all the retained effective IMF components are added to the residual to reconstruct a high-precision temperature signal after noise removal.

[0083] Furthermore, the reconstructed temperature signal after noise removal is used to achieve closed-loop control of the melting furnace, thereby achieving energy saving and emission reduction.

[0084] This invention provides a signal noise reduction system 10 for implementing the signal noise reduction method described above. Figure 6 As shown, it includes: Signal acquisition module 1 is used to synchronously acquire a first signal and a second signal that interferes with the first signal from the same source; The mode decomposition module 2 is connected to the signal acquisition module 1 and is used to perform ensemble empirical mode decomposition on the first signal and the second signal respectively to obtain multiple first intrinsic mode function components and a first residual function corresponding to the first signal and multiple second intrinsic mode function components corresponding to the second signal. The permutation entropy calculation module 3 is connected to the mode decomposition module 2 and is used to calculate the permutation entropy values ​​of multiple first intrinsic mode function components; The similarity measurement module 4 is connected to the permutation entropy calculation module 3 and is used to calculate the similarity measurement value between the first intrinsic mode function component and the second intrinsic mode function component. The filtering module 5 is connected to the similarity measurement module 4 and is used to filter valid first intrinsic mode function components based on the permutation entropy value and the similarity measurement value. The signal reconstruction module 6 is connected to the filtering module 5 and is used to reconstruct the signal based on the filtered valid first intrinsic mode function components and the first residual function to obtain the signal corresponding to the denoised first signal.

[0085] This invention provides a signal transmission system for a glass electric melting furnace, such as... Figure 7 As shown, it includes a transmitter 100 and a receiver 200. The transmitter 100 includes the signal denoising system 10 described above, to perform the signal denoising method described above on the first signal.

[0086] In glass electric melting furnaces, electromagnetic interference and thermal noise are not stable Gaussian white noise, but rather complex noise exhibiting amplitude and frequency variations over time, and a nonlinear coupling relationship with the actual temperature signal. Furthermore, the temperature control of the electric melting furnace depends on subtle trend changes in the signal. However, under strong background electromagnetic noise, these subtle characteristics are almost completely masked, resulting in an extremely low signal-to-noise ratio.

[0087] In existing technologies, hardware filtering (such as RC filter circuits) and conventional digital filtering algorithms (such as Fourier transform filtering and wavelet thresholding) are commonly used to address noise (including thermal noise and electromagnetic interference noise) in thermocouple signals. Traditional filtering methods, based on fixed frequency band assumptions, are prone to damaging or distorting the true temperature signal characteristics while filtering out noise, leading to signal distortion. This invention innovatively applies the Ensemble Empirical Mode Decomposition (EEMD) method to the temperature signal processing of glass electric melting furnaces.

[0088] Facing the specific and extremely complex industrial environment of glass electric melting furnaces, this method overcomes the inherent and insurmountable technical limitations of traditional filtering methods. EEMD, as an adaptive signal processing method, automatically decomposes a signal into a series of fixed mode functions (IMFs) based on its local characteristic time scale. Through a sieving process, all signal components from high to low frequencies can be extracted step by step, selectively discarding high-frequency IMF components representing noise, thereby accurately separating the complex noise mixed in the high-frequency modes while completely preserving the true temperature trend information in the mid- and low-frequency modes, overcoming the difficulty of separating strong non-stationary and nonlinear noise. By reconstructing the remaining IMF components representing the true temperature signal, high-precision temperature measurements can be miraculously recovered from a signal almost destroyed by noise, achieving the extraction of weak feature signals under extreme noise backgrounds.

[0089] In glass electric melting furnaces, the main sources of strong electromagnetic interference (such as electrode heating systems and high-power stirring motors) are known and common. This interference couples simultaneously to both thermocouple temperature signals and motor current signals; that is, the operating current signal of motors such as stirrers is also subject to interference from complex electromagnetic fields, and its interference source is highly similar to that of the temperature signal interference source. Therefore, this invention does not process the temperature signal in isolation, but rather constructs a multi-dimensional signal collaborative processing system framework, namely, collaborative denoising based on the "temperature signal-motor current signal".

[0090] Specifically, this invention utilizes the characteristic that thermocouple temperature signals and motor current signals are "noise from the same source but signals are heterogeneous" to simultaneously acquire temperature signals and motor current signals, and perform ensemble empirical mode decomposition (EEMD) in parallel to obtain two sets of intrinsic mode functions (IMFs), namely, multiple first intrinsic mode function components corresponding to the temperature signals and multiple second intrinsic mode function components corresponding to the motor current signals.

[0091] Subsequently, noise mode co-correlation analysis and signal denoising reconstruction were performed on the two sets of intrinsic mode functions (IMFs) to reduce signal errors. Specifically, firstly, the two sets of IMFs were compared and analyzed; secondly, the permutation entropy values ​​of all first IMF components and all second IMF components corresponding to the motor current signal were calculated, and noise IMF components with high permutation entropy values ​​(i.e., strong randomness) were screened and discarded; then, the cross-correlation coefficients of the high-frequency components in the first IMF components and the high-frequency components in the second IMF components were calculated to identify and eliminate common-mode electromagnetic interference noise components.

[0092] When the first intrinsic mode function component and the second intrinsic mode function component are highly similar in waveform, such as the absolute value of the Pearson cross-correlation coefficient... ≥0.7), and both have very high permutation entropy values, such as H NPE When the value is ≥0.55, it is determined that the noise contained in the first intrinsic mode function component is mainly common-mode noise from electromagnetic interference.

[0093] For the first intrinsic mode function component, i.e. common-mode noise, which is identified as highly correlated with the noise mode of the current signal, regardless of its own permutation entropy value, it is discarded during signal reconstruction because it is highly likely to be electromagnetic interference rather than real temperature fluctuation.

[0094] For the first intrinsic mode function component that is not related to the current signal noise mode but has a high arrangement entropy value, it may be thermal noise unique to the thermocouple. The threshold can be set according to actual needs to decide whether to retain or discard it.

[0095] The first intrinsic mode function component with low permutation entropy and mainly containing low-frequency information is determined to be a valid component and is retained during signal reconstruction.

[0096] Finally, all valid first intrinsic mode function components are reconstructed to obtain the denoised temperature signal.

[0097] Furthermore, this invention is not limited to obtaining a more accurate temperature reading; it can also utilize the denoised high-precision signal to achieve closed-loop control of the melting furnace temperature, avoiding overshoot or undershoot of the temperature control system due to signal errors, reducing excessive energy supply or supplementary heating, and significantly reducing electricity or gas energy consumption. At the same time, stable temperature is key to ensuring the uniformity of the molten glass and product quality, reducing the generation of defective and scrap products, improving product quality and yield; reducing drastic fluctuations in the control system, protecting heating elements and refractory materials, and extending equipment life.

[0098] The method and system of this invention are applicable to industrial signal transmission, particularly for high-quality transmission of weak electrical signals inside electric melting furnaces in glass factories. The core function of this method and system is to effectively reduce thermal noise interference experienced by weak electrical signals during transmission, thereby significantly improving signal accuracy and stability. By applying this method, the accuracy of the feedback control system is improved, thereby optimizing control efficiency and achieving the goals of energy conservation and emission reduction.

[0099] Of course, in practical applications, for two signals with other sources of interference, the parallel ensemble empirical mode decomposition method can also be used. For one of the signals to be denoised, the effective IMF components of the signal to be denoised are selected based on the IMF component permutation entropy of the signal to be denoised and the similarity measure between the IMF components of the signal to be denoised and the IMF components of the interference signals, and then the signal to be denoised is reconstructed.

[0100] The above are merely preferred embodiments of the present invention and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should recognize that any equivalent substitutions and obvious changes made using the content of this specification and illustrations should be included within the protection scope of the present invention.

Claims

1. A signal noise reduction method, characterized in that, include: Step S1: Synchronously acquire a first signal and a second signal that interferes with the first signal from the same source; the first signal is the temperature signal of the thermocouple, and the second signal is the motor current signal; Step S2: Perform ensemble empirical mode decomposition on the first signal and the second signal respectively to obtain multiple first intrinsic mode function components and a first residual function corresponding to the first signal, and multiple second intrinsic mode function components corresponding to the second signal. Step S3: Calculate the permutation entropy value of multiple first intrinsic mode function components; Step S4: Calculate the similarity metric between the first intrinsic mode function component and the second intrinsic mode function component, wherein the similarity metric is the Pearson cross-correlation coefficient; Step S5: Filter the valid first intrinsic mode function components based on the permutation entropy value and similarity metric value; Step S5 includes: Step S51: For each first intrinsic mode function component, detect whether a preset condition is met; wherein, the preset condition is that the permutation entropy value is less than a first preset threshold and the similarity metric value with the second intrinsic mode function component is less than a second preset threshold. If the preset conditions are met, the first intrinsic mode function component is retained; If the preset conditions are not met, the first intrinsic mode function component is removed; Step S52: Iterate through each first intrinsic mode function component in all first intrinsic mode function components until all first intrinsic mode function components have been traversed, so as to filter out all valid first intrinsic mode function components. Step S6: Reconstruct the signal corresponding to the denoised first signal by using the selected valid first intrinsic mode function components and the first residual function.

2. The signal noise reduction method according to claim 1, characterized in that, In step S2, the first signal and the second signal are used as signals to be decomposed to perform ensemble empirical mode decomposition, which includes: Step S21: Add a zero-mean Gaussian white noise sequence to the signal to be decomposed to obtain a new sequence; Step S22: Perform empirical mode decomposition on the new sequence to obtain multiple intrinsic mode function components and residual functions; Step S23: Repeat steps S21-S22 a preset number of times, with each time a different white noise sequence is added; Step S24: The mean value of each intrinsic mode function component obtained after empirical mode decomposition for a preset number of times is determined as the final intrinsic mode function component and residual function.

3. The signal noise reduction method according to claim 1, characterized in that, Step S3 includes: Step S31: Reconstruct the phase space for each of the first intrinsic mode function components to obtain a set of reconstruction vectors; Step S32: Rearrange the vector elements in each reconstructed vector in ascending order of their values, and record the symbol sequence formed by the index positions; Step S33: Calculate the relative frequency of each symbol sequence relative to the total number of reconstructed vectors; Step S34: Calculate the permutation entropy value corresponding to each first intrinsic mode function component based on the relative frequency; Step S35: Normalize the permutation entropy value to obtain the normalized permutation entropy value.

4. The signal noise reduction method according to claim 3, characterized in that, The method for calculating the permutation entropy value is as follows: ； in, Indicates relative frequency; k Indicates the symbol sequence index; m Indicates the embedding dimension; Represents the permutation entropy value; The normalized permutation entropy value is: ； in, This represents the normalized permutation entropy value.

5. The signal noise reduction method according to claim 1, characterized in that, Step S4 includes: Step S41: Select one first intrinsic mode function component from multiple first intrinsic mode function components, and simultaneously select one second intrinsic mode function component from multiple second intrinsic mode function components; Step S42: Calculate the covariance of the selected first intrinsic mode function component and the second intrinsic mode function component, and calculate the first standard deviation corresponding to the selected first intrinsic mode function component and the second standard deviation corresponding to the selected second intrinsic mode function component respectively. Step S43: Calculate the Pearson cross-correlation coefficients of the selected first intrinsic mode function component and the second intrinsic mode function component based on the covariance, the first standard deviation, and the second standard deviation, and determine the calculated Pearson cross-correlation coefficients as the similarity measure.

6. The signal noise reduction method according to claim 5, characterized in that, The method for calculating the Pearson cross-correlation coefficient is as follows: ； in, Represents the first intrinsic mode function component Second intrinsic mode function component covariance; This represents the first standard deviation corresponding to the selected first intrinsic mode function component; This represents the second standard deviation corresponding to the selected second intrinsic mode function component; This represents the Pearson cross-correlation coefficient.

7. A signal noise reduction system, characterized in that, For implementing the signal noise reduction method as described in any one of claims 1-6, comprising: The signal acquisition module is used to synchronously acquire a first signal and a second signal that interferes with the first signal from the same source; the first signal is the temperature signal of the thermocouple, and the second signal is the motor current signal; The mode decomposition module, connected to the signal acquisition module, is used to perform ensemble empirical mode decomposition on the first signal and the second signal respectively, to obtain multiple first intrinsic mode function components and a first residual function corresponding to the first signal, and multiple second intrinsic mode function components corresponding to the second signal; The permutation entropy calculation module, connected to the mode decomposition module, is used to calculate the permutation entropy values ​​of multiple first intrinsic mode function components; The similarity measurement module, connected to the permutation entropy calculation module, is used to calculate the similarity measurement value between the first intrinsic mode function component and the second intrinsic mode function component, wherein the similarity measurement value is the Pearson cross-correlation coefficient; A filtering module, connected to a similarity measurement module, is used to filter valid first intrinsic mode function (IMF) components based on permutation entropy and similarity measurement values. For each IMF component, the filtering module checks whether it meets preset conditions. These preset conditions are: the permutation entropy is less than a first preset threshold and the similarity measurement value with a second IMF component is less than a second preset threshold. If the preset conditions are met, the first IMF component is retained; if not, it is discarded. The process iterates through each of all first IMF components until all first IMF components have been traversed, thus obtaining all valid first IMF components. The signal reconstruction module, connected to the filtering module, is used to reconstruct the signal based on the filtered valid first intrinsic mode function components and the first residual function, thereby obtaining the signal corresponding to the denoised first signal.

8. A signal transmission system for a glass electric melting furnace, characterized in that, It includes a transmitting end and a receiving end, wherein the transmitting end is used to implement the signal noise reduction method as described in any one of claims 1-6 on the first signal.

Images