Thermocouple dual-channel degradation characterization driven probabilistic life prediction method and system

Abstract

This invention relates to a probabilistic lifetime prediction method and system driven by dual-channel degradation characterization of thermocouples. The method includes: acquiring online observation data of the thermocouple; constructing a dual-channel degradation characterization of insulation and thermal performance degradation through multi-timescale adaptive feature extraction; determining the reliability of the degradation channel based on feature quality and availability; modulating the characterization based on the degree of inter-channel interference; subsequently constructing a composite degradation score; defining the first moment that meets the preset degradation gating criterion as the first degradation moment; and finally, outputting the probabilistic prediction result of the remaining lifetime of the thermocouple based on the parameters corresponding to this moment and the probabilistic lifetime prediction model, which can be continuously updated with newly added online observation data. This invention can achieve stable determination of the thermocouple degradation initiation and probabilistic prediction of remaining lifetime under limited observation conditions, and has the advantages of strong degradation characterization, stable gating start, reliable prediction results, and adaptability to continuous online updates.

Description

Technical Field

[0001] This application relates to the field of temperature sensor performance monitoring and online lifetime management, specifically to a probabilistic lifetime prediction method and system driven by thermocouple dual-channel degradation characterization. Background Technology

[0002] Thermocouples are widely used for temperature monitoring in industrial processes, high-temperature equipment, and complex operating conditions due to their simple structure, low cost, wide applicable temperature range, and strong environmental adaptability. During long-term service, the output characteristics of thermocouples are affected by various degradation factors, including increased signal fluctuations, spectral anomalies, and intensified high-frequency disturbances caused by insulation degradation, and slower dynamic response due to thermal performance degradation, thus affecting the thermocouple's service performance and lifespan. Existing technologies related to thermocouple lifespan management mainly include fault detection and degradation identification methods based on single abnormal features, overall drift, or changes in loop electrical signals; state assessment methods based on dynamic response parameters; and methods that directly apply general remaining service life prediction models to the sensor degradation process.

[0003] While the aforementioned methods can reflect thermocouple anomalies or performance degradation to some extent, they typically treat different degradation effects together and lack separate characterization of insulation-related time-frequency anomalies and thermally related dynamic response degradation, making it difficult to form targeted degradation inputs for lifetime prediction. At the same time, existing technologies mostly rely on fixed thresholds, single anomaly out-of-bounds events, or empirical triggering conditions for lifetime prediction start-up time, lacking comprehensive constraints on reference health baseline distribution, channel reliability, and inter-channel interference relationships. Therefore, it is difficult to stably determine the first degradation time under limited online observation conditions, further affecting the reliability of the remaining lifetime probability prediction results.

[0004] Therefore, there is a need for a probabilistic lifetime prediction method and system driven by thermocouple dual-channel degradation characterization, which can achieve stable determination of the start time of insulation-thermal dual-channel degradation characterization and lifetime prediction under limited observation conditions, as well as the unified introduction of degradation channel credibility and mutual interference modulation mechanism in degradation evaluation and probabilistic lifetime prediction. Summary of the Invention

[0005] The purpose of this invention is to provide a probabilistic lifetime prediction method and system driven by dual-channel thermocouple degradation characterization, in order to solve the problems in the prior art where the effects of insulation performance degradation and thermal performance degradation are mixed in lifetime prediction, and there is a lack of targeted degradation input for lifetime prediction, resulting in insufficient correlation between degradation features and lifetime evolution and reduced accuracy of remaining lifetime prediction; and to solve the problem that the reliability of degradation channels and the mutual interference relationship between channels are not uniformly incorporated into degradation evaluation and lifetime prediction, resulting in low-quality or interfered features affecting the stability of degradation evaluation and the reliability of lifetime prediction. This invention constructs a dual-channel degradation characterization representing insulation and thermal performance degradation through multi-timescale adaptive feature extraction. A composite degradation score is then built based on degradation channel reliability and inter-channel interference modulation mechanism. Furthermore, the lifetime prediction start time is determined through rolling updates and first-reach degradation gating. Finally, a probabilistic lifetime prediction model is used to output the probabilistic prediction result of the remaining lifetime. This achieves the separate expression of degradation effects, the stable determination of the lifetime prediction start time, and the expression of the uncertainty of the remaining lifetime. This helps reduce the impact of abnormal fluctuations and inter-channel interference on lifetime prediction, improving the accuracy, stability, and reliability of thermocouple remaining lifetime prediction.

[0006] The present invention adopts the following technical solution: In a first aspect, embodiments of this application provide a probabilistic lifetime prediction method driven by thermocouple dual-channel degradation characterization, comprising: Acquire observation data during the online operation of thermocouples, perform time synchronization and rolling window segmentation on the observation data, and construct corresponding short time window measurement sequences and long time window measurement sequences; Based on the short-time-window measurement sequence and the long-time-window measurement sequence, multi-time-scale adaptive feature extraction is performed to construct a dual-channel degradation characterization representing insulation performance degradation and thermal performance degradation; The reliability of each degradation channel is determined based on the quality and availability of the corresponding features, and the dual-channel degradation characterization is modulated according to the degree of mutual interference between channels. Based on the degradation channel confidence, the modulated dual-channel degradation characterization, and the reference baseline feature distribution, a composite degradation score is constructed. Based on the composite degradation score, the moment when the preset degradation gating criterion is first met is taken as the first degradation moment; Based on the modulated dual-channel degradation characterization, degradation channel reliability, and probabilistic lifetime prediction model corresponding to the first degradation time, the probabilistic prediction result of the current remaining lifetime of the thermocouple is obtained.

[0007] Secondly, embodiments of this application provide a probabilistic lifetime prediction system driven by thermocouple dual-channel degradation characterization, comprising: The data acquisition module is used to acquire observation data during the online operation of the thermocouple, perform time synchronization and rolling window segmentation on the observation data, and construct corresponding short time window measurement sequences and long time window measurement sequences. A dual-channel degradation characterization construction module is used to perform multi-timescale adaptive feature extraction based on the short-time-window measurement sequence and the long-time-window measurement sequence to construct a dual-channel degradation characterization representing insulation performance degradation and thermal performance degradation. The channel credibility determination and characterization modulation module is used to determine the credibility of the degradation channel based on the quality and availability of the corresponding features of each degradation channel, and to modulate the dual-channel degradation characterization according to the degree of mutual interference between channels; A composite degradation score construction module is used to construct a composite degradation score based on the degradation channel confidence, the modulated dual-channel degradation characterization, and the reference baseline feature distribution. The first degradation time determination module is used to determine the first degradation time based on the composite degradation score, and to determine the first degradation time as the first degradation time that meets the preset degradation gating criterion. The probabilistic lifetime prediction module is used to obtain the probabilistic prediction result of the current remaining lifetime of the thermocouple based on the modulated dual-channel degradation characterization, degradation channel reliability and probabilistic lifetime prediction model corresponding to the first degradation time.

[0008] Thirdly, embodiments of this application provide an electronic device, including: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method as described in the first aspect.

[0009] Fourthly, embodiments of this application provide a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the method as described in the first aspect.

[0010] The technical solutions provided by the embodiments of this application may include the following beneficial effects: 1. Highly targeted degradation input for lifetime prediction. This invention constructs a dual-channel degradation characterization of insulation performance degradation and thermal performance degradation through multi-timescale adaptive feature extraction, which can reduce the mutual mixing of different degradation effects in the input layer and form a more targeted lifetime prediction input.

[0011] 2. Robustness of degradation evaluation. This invention constructs a composite degradation score by introducing degradation channel credibility and inter-channel interference modulation mechanism, ensuring that low-quality, low-availability, or severely interfered channel features do not indiscriminately enter the degradation evaluation process, thereby improving the stability and reliability of the degradation score.

[0012] 3. Reliable determination of lifetime prediction start time. This invention determines the first degradation time by using a composite degradation score based on a reference baseline distribution, rolling updates, and a continuous triggering mechanism, which can reduce the risk of false triggering caused by weak degradation signals, single anomalies, or characteristic fluctuations.

[0013] 4. The prediction results are highly reliable and suitable for online applications. This invention outputs a probabilistic prediction of the remaining service life and updates it continuously after new observations arrive. It can express prediction uncertainty and is more suitable for online thermocouple life management and predictive maintenance.

[0014] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0015] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0016] Figure 1 The flowchart illustrates the probabilistic lifetime prediction method driven by thermocouple dual-channel degradation characterization provided in this embodiment of the invention.

[0017] Figure 2 This is a schematic diagram of online output temperature data samples from thermocouples provided in an embodiment of the present invention.

[0018] Figure 3 This is a schematic diagram illustrating the characterization of thermocouple insulation performance degradation pathways and thermal performance degradation pathways provided in embodiments of the present invention.

[0019] Figure 4 This is a schematic diagram illustrating the probability prediction results of the remaining service life of thermocouples provided in an embodiment of the present invention.

[0020] Figure 5 A block diagram of a probabilistic lifetime prediction system driven by thermocouple dual-channel degradation characterization provided in an embodiment of the present invention. Detailed Implementation

[0021] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0022] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0023] To make the technical solution of this application clearer, the following is combined with Figures 1 to 5 The specific implementation methods of this application will be described in detail.

[0024] Figure 1 A flowchart of a probabilistic lifetime prediction method driven by thermocouple dual-channel degradation characterization is shown, which may include the following steps: S1: Acquire observation data during the online operation of the thermocouple, perform time synchronization and rolling window segmentation on the observation data, and construct corresponding short time window measurement sequences and long time window measurement sequences; S2: Based on the short-time-window measurement sequence and the long-time-window measurement sequence, perform multi-time-scale adaptive feature extraction to construct a dual-channel degradation characterization representing insulation performance degradation and thermal performance degradation; S3: Determine the reliability of the degradation channel based on the quality and availability of the corresponding features of each degradation channel, and modulate the dual-channel degradation characterization according to the degree of mutual interference between channels; S4: Based on the reliability of the degradation channel, the modulated dual-channel degradation characterization, and the feature distribution of the reference baseline, construct a composite degradation score; S5: Based on the composite degradation score, the moment when the first degradation gating criterion is met is taken as the first degradation moment; S6: Based on the modulated dual-channel degradation characterization, degradation channel reliability, and probabilistic lifetime prediction model corresponding to the first degradation time, the probabilistic prediction result of the current remaining lifetime of the thermocouple is obtained.

[0025] As can be seen from the above embodiments, this invention obtains observation data during the online operation of thermocouples and segments the observation data into rolling windows. Then, based on multi-timescale adaptive feature extraction, time-frequency features characterizing insulation performance degradation are extracted within short time windows, and dynamic response features characterizing thermal performance degradation are identified within long time windows, constructing a dual-channel degradation characterization. Next, the reliability of the degradation channel is determined based on the quality, availability, and inter-channel interference relationship of the corresponding features of each degradation channel, and the dual-channel degradation characterization is modulated. Then, a composite degradation score is constructed based on the degradation channel reliability, the modulated dual-channel degradation characterization, and the feature distribution of the reference baseline, and the first degradation time is determined through the continuously updated composite degradation score. Finally, based on the modulated dual-channel degradation characterization corresponding to the first degradation time, the degradation channel reliability, and the probabilistic lifetime prediction model, the probabilistic prediction result of the thermocouple's current remaining lifetime is obtained and continuously updated after acquiring new online observation data.

[0026] In terms of technical advantages, this invention, through multi-timescale adaptive feature extraction, separately characterizes insulation-related time-frequency anomalies and thermally-related dynamic response degradation in thermocouple outputs, enabling the formation of targeted degradation inputs for lifetime prediction under limited online observation conditions. Unlike existing methods that directly evaluate and predict based on mixed features, overall drift, or a single index, this invention introduces degradation channel credibility and inter-channel interference modulation mechanisms. This ensures that low-quality, low-availability, or severely interfered channel features do not indiscriminately enter the degradation evaluation process, thereby improving the robustness of composite degradation scoring and first-reach degradation gating. Simultaneously, this invention, through composite degradation scoring based on reference baseline distribution, a rolling update mechanism, and a continuous triggering mechanism, can more stably determine the lifetime prediction initiation time, reducing the risk of false triggering caused by single anomalies, random fluctuations, and weak degradation signals.

[0027] In the specific implementation of S1: acquire the observation data during the online operation of the thermocouple, perform time synchronization and rolling window segmentation on the observation data, and construct the corresponding short time window measurement sequence and long time window measurement sequence; Specifically, the system collects thermocouple output signals, cold junction temperature, environmental operating conditions, and service history information. The thermocouple output signals can be thermocouple thermoelectric potential signals or temperature output signals converted by the acquisition device. The environmental operating conditions information can include at least one of ambient temperature, process stage identifiers, temperature rise and fall status, load level, and operating mode. The service history information can include at least one of cumulative service time, cumulative number of thermal cycles, cumulative high-temperature residence time, and historical maintenance identifiers.

[0028] Let the output signal of the thermocouple be The short time window measurement sequence corresponding to the current moment and long window measurement sequences They can be represented as: in, For the first Thermocouple output signal at each sampling time For short time window length, This represents the length of the long window at the current moment.

[0029] The short time window length and the length of the long window The value can be determined based on the sampling frequency, the equivalent thermal inertia of the thermocouple, the rate of change of operating conditions, and the sample size required for dynamic identification. The short-time window is used to extract features sensitive to insulation performance degradation, such as local fluctuations, spectral anomalies, and high-frequency disturbances. The long-time window is used to extract features sensitive to thermal performance degradation, such as dynamic hysteresis, equivalent time constant, and dynamic identification residuals. By assigning the short-time window and the long-time window to the feature update scales of different degradation channels, a multi-scale characterization of insulation performance degradation and thermal performance degradation can be achieved based on the same online observation data. This reduces the confounding effects of short-term noise, gradual changes in operating conditions, and dynamic transition processes on the subsequent dual-channel degradation characterization.

[0030] Figure 2 The diagram shows a sample of output temperature data acquired during the online operation of the thermocouple in this embodiment of the invention. This output temperature data can be used as one of the observation inputs for constructing short-time-window and long-time-window measurement sequences. Cold junction temperature, environmental condition information, and service history information can be collected synchronously at the same time reference and are not shown individually in the figure.

[0031] In the specific implementation of S2: based on the short-time-window measurement sequence and the long-time-window measurement sequence, multi-time-scale adaptive feature extraction is performed to construct a dual-channel degradation characterization representing insulation performance degradation and thermal performance degradation; this step may include the following sub-steps: S21: Within a short time window, perform frequency domain analysis or time-frequency analysis on the short time window measurement sequence. The frequency domain analysis or time-frequency analysis includes at least one of Fourier transform, power spectral density estimation, empirical mode decomposition, and wavelet decomposition. Extract time-frequency feature clusters that characterize the output signal volatility, spectral anomaly, or high-frequency disturbance from the corresponding analysis results, and construct an insulation performance degradation channel characterization accordingly.

[0032] Specifically, let the first The short-time window measurement sequence corresponding to each sampling time is: In a preferred embodiment, empirical mode decomposition can be used to analyze the short-time-window measurement sequence: in, For the short time window measurement sequence, the first The output value of each sampling point For the first The empirical mode decomposition at the sampling time obtained the first sampling time. The intrinsic mode component in the th... The values ​​of each sampling point For the residual term, The number of modal components.

[0033] Considering that thermocouple degradation signals typically exhibit non-stationary and intermittent perturbation characteristics, the preferred method is... Several high-frequency modal components serve as supplementary evidence of insulation performance degradation, and a high-frequency energy ratio is constructed. : in, To prevent extremely small positive numbers with a denominator of zero; The larger the value, the higher the proportion of high-frequency disturbance energy in the total energy, the more unstable the output signal, and the greater the possibility of insulation performance degradation.

[0034] Furthermore, construct the spectral barycenter shift. This is used to reflect the degree of deviation of the main energy distribution of the spectrum from the healthy baseline: in, Indicates the first Each time window corresponds to a frequency Spectral energy at the location, Indicates the frequency band to be analyzed. This represents the mean of the centroid of the reference baseline phase spectrum.

[0035] In addition, the rolling variance ratio and the proportion of abnormal spikes can be constructed to characterize the degree of enhancement of signal fluctuation amplitude relative to the baseline stage and the density of local abnormal disturbances, respectively. in, This represents the variance of the output sequence within a short time window. This represents the variance corresponding to the reference baseline phase. This indicates the number of abnormal peak sample points within a short time window.

[0036] This yields a cluster of time-frequency features representing insulation degradation. : To eliminate the influence of different dimensions, the insulation-related time-frequency feature cluster is normalized: in, This is the mean vector of the characteristic features of the insulation channel reference baseline. It is a diagonal matrix composed of the standard deviations of the insulation channel reference baseline.

[0037] Furthermore, construct a characterization path for insulation performance degradation. : in, This is the weight vector for the feature fusion of the insulating channel.

[0038] S22: Based on at least one of the current operating condition excitation sufficiency, dynamic feature identification residual and output sequence stability, the length of the long time window is adaptively adjusted, the dynamic response identification of the long time window measurement sequence is performed, dynamic feature clusters characterizing the dynamic response capability of the thermocouple are extracted, and a thermal performance degradation channel characterization is constructed accordingly.

[0039] Specifically, let the first The long-time window measurement sequence corresponding to each sampling time is: The sufficiency of excitation can be determined based on the operating conditions. Dynamic feature identification residual and output sequence stability Adjust the length of the long window to improve the adaptability of dynamic response identification under different operating conditions.

[0040] Among them, the sufficiency index of working condition incentives It can be represented as: in, This represents the current range of thermocouple output variation within a long time window. To preset the reference temperature variation range, This indicates that the input value is restricted to... A cutoff function within an interval.

[0041] The dynamic response process is described using a first-order inertial model: in, To fit the temperature, The initial temperature, For steady-state temperature, This is the dynamic time constant.

[0042] Dynamic feature identification residual It can be represented as: in, This represents the number of sampling points used for dynamic identification within a long time window. Indicates the first The output temperature corresponding to each sampling point.

[0043] Output sequence stability index Used to characterize the local stability of the current long-time window output sequence: in, It is a smoothing function.

[0044] Therefore, the update amount of the long window length can be expressed as: in, For window adjustment amount, , , The time window length is adjusted by the following parameters: when the operating condition excitation is insufficient, the identification residual is large, or the output sequence stability is poor, the time window length is increased accordingly to improve the robustness of dynamic response identification; conversely, the time window length is appropriately reduced to improve the online update speed.

[0045] In a preferred embodiment, the equivalent dynamic time constant can be identified using a least squares approach. And further construct the time constant fluctuation. and dynamic lag area : in, For the first The time interval corresponding to a long window.

[0046] This leads to the dynamic characteristic cluster representing thermal performance degradation: To eliminate the influence of different dimensions, the thermally correlated dynamic feature cluster is normalized: in, This represents the mean vector of the thermally relevant dynamic feature clusters at the reference baseline stage. This represents a diagonal matrix composed of the standard deviations of each characteristic.

[0047] Furthermore, a thermal performance degradation pathway characterization was constructed. This allows for the independent characterization of thermal degradation effects, such as slower dynamic response of thermocouples, from the output sequence. in, This is the weight vector for thermal channel feature fusion.

[0048] S23: Construct a dual-channel degradation characterization based on the aforementioned insulation performance degradation channel characterization and thermal performance degradation channel characterization.

[0049] Specifically, no. Dual-channel degradation characterization at each sampling time It can be represented as: Figure 3 An embodiment of the present invention based on Figure 2 The insulation performance degradation channel characterization and thermal performance degradation channel characterization obtained from the online output temperature data are shown.

[0050] In the specific implementation of S3: the reliability of the degradation channel is determined based on the quality and availability of the corresponding features of each degradation channel, and the dual-channel degradation characterization is modulated according to the degree of inter-channel interference; this step may include the following sub-steps: S31: Determine the reliability of the insulation performance degradation channel based on the time-frequency characteristic significance index and the abnormal distortion index.

[0051] Specifically, based on the normalized insulation-related time-frequency feature clusters Constructing a significance index for time-frequency features and abnormal distortion index : in, The significance mapping coefficient vector, This is the vector of distortion mapping coefficients.

[0052] When insulation-related time-frequency anomalies are highly significant within the current time window and do not exhibit severe distortion, the reliability of the insulation performance degradation channel is considered high; conversely, when characteristic jumps are strong and distortions are significant, its reliability is considered low. Therefore, the reliability of the insulation performance degradation channel... It can be represented as: in, , The modulation coefficient, For monotone bounded mapping functions, the Logistic mapping function is preferred to ensure... By simultaneously considering both the saliency of features and the degree of anomalous distortion, the impact of transient interference on the reliability of insulation degradation assessment can be reduced.

[0053] S32: Determine the reliability of thermal performance degradation channels based on the operating condition excitation sufficiency index, dynamic feature identification residuals, and frequency domain contamination index constructed from short time window time-frequency features.

[0054] Specifically, frequency domain pollution indicators It can be represented as: in, This is the vector of frequency domain contamination mapping coefficients.

[0055] Reliability of thermal performance degradation pathway It can be represented as: in, , , The modulation coefficient is denoted as . By incorporating dynamic excitation, identification residuals, and frequency domain contamination into the confidence calculation, over-reliance on thermal channel results can be avoided when identification conditions are insufficient.

[0056] S33: Construct interference factors of thermal performance degradation channels on insulation performance degradation channels based on dynamic transition strength or non-stationarity indices over long time windows; construct interference factors of insulation performance degradation channels on thermal performance degradation channels based on high-frequency contamination indices.

[0057] Specifically, thermal nonstationarity index It can be represented as: in, This is the vector of thermal nonstationarity mapping coefficients.

[0058] The interference factor of thermal property degradation channels on insulation property degradation channels. And the interference factor of insulation degradation channels on thermal degradation channels. This reduces the effects of miscoupling between channels in subsequent characterization modulation. in, and This is the proportionality coefficient. This is a saturation mapping function.

[0059] S34: Construct a channel modulation matrix based on the credibility of the insulation performance degradation channel, the credibility of the thermal performance degradation channel, the interference factor of the insulation performance degradation channel, and the interference factor of the thermal performance degradation channel; use the channel modulation matrix to modulate the dual-channel degradation characterization to obtain the dual-channel degradation characterization after credibility and mutual interference modulation.

[0060] Specifically, the channel modulation matrix It can be represented as: in, and This indicates the degree of self-reliability of each degradation channel. and This indicates the inhibitory effect from the other channel. Characterization of modulated dual-channel degradation. It can be represented as: This modulation process allows for the appropriate weakening of channel characterizations with low reliability or those heavily affected by mutual interference in subsequent degradation assessments.

[0061] In the specific implementation of S4: a composite degradation score is constructed based on the degradation channel confidence, the modulated dual-channel degradation characterization, and the reference baseline feature distribution; this step may include the following sub-steps: S41: Perform the same feature extraction and confidence modulation processing on the long and short window observation data of the reference baseline phase as at the current time to obtain a dual-channel reference representation sample set, thereby determining the reference baseline mean vector and reference baseline covariance matrix.

[0062] Specifically, the observation period during the initial installation of the thermocouple or after calibration confirming it to be in a healthy state can be selected as the reference baseline stage, and its sample index set is denoted as [missing information]. The same feature extraction, channel confidence calculation, and characterization modulation processing are performed on the long and short window observation data during the reference baseline phase, ensuring that the current degradation characterization and the healthy baseline are in the same statistical space, thus obtaining a dual-channel reference characterization sample set. : Reference baseline mean vector and sample covariance matrix It can be represented as: S42: Construct a composite degradation score based on the degree of deviation of the modulated dual-channel degradation characterization from the mean vector of the reference baseline, so as to characterize the comprehensive degradation level at the current moment relative to the reference baseline stage.

[0063] Specifically, the composite degradation score at the current moment can be constructed using the Mahalanobis distance form: when The larger the value, the more significantly the current modulated dual-channel degradation characteristics deviate from the healthy baseline distribution, and the higher the overall degree of degradation.

[0064] In the specific implementation of S5: based on the composite degradation score, the moment when the first degradation gating criterion is met is taken as the first degradation moment; this step may include the following sub-steps: S51: The composite degradation score is recursively smoothed to obtain a smoothed score that is updated on a rolling basis.

[0065] Specifically, first-order exponential smoothing can be preferred to obtain a rolling-up smooth composite degradation score. This reduces the impact of a single abnormal fluctuation on the degradation gating decision. in, This is the smoothing coefficient.

[0066] S52: When the smoothing score is greater than the preset degradation gating threshold in multiple consecutive sampling times, the first sampling time in the multiple consecutive sampling times is defined as the first degradation time.

[0067] Specifically, the preset degradation threshold can be determined based on the statistical distribution of the smoothed composite degradation score at the reference baseline stage. In a preferred embodiment, the preset degradation gating threshold can be expressed as: in, and represents the mean and standard deviation of the smoothed composite degradation score at the reference baseline stage, respectively. The threshold coefficient can be determined based on the statistical distribution of the reference baseline.

[0068] The moment of first degradation can be represented as: in, To continuously trigger the steps, the first degradation time is confirmed only after the gating condition is met at multiple consecutive sampling times, which can reduce false triggering caused by short-term noise or local disturbances.

[0069] In a preferred embodiment, to reduce the risk of false triggering caused by low-quality observations, a minimum confidence condition can be further introduced. When the smaller value of the confidence level in the two channels is not lower than a preset confidence lower limit, the current continuous triggering interval is confirmed to be valid. When the current time reaches the confirmation time of the continuous triggering condition, a probabilistic lifetime prediction initiation condition is formed, and the first time of degradation is recorded as the first sampling time in the continuous triggering interval.

[0070] In the specific implementation of S6: based on the modulated dual-channel degradation characterization, degradation channel reliability, and probabilistic lifetime prediction model corresponding to the first degradation time, the probabilistic prediction result of the current remaining lifetime of the thermocouple is obtained; this step may include the following sub-steps: S61: Based on the modulated dual-channel degradation characterization and degradation channel reliability corresponding to the first degradation time, construct the initial distribution for lifetime prediction.

[0071] Specifically, let the time of first degradation be... The corresponding modulated dual-channel degradation characterization for: Meanwhile, the online observation sequence up to the first degradation moment is recorded as follows: And record the confidence levels of the insulation performance degradation channel and the thermal performance degradation channel at that moment as follows: and Using the dual-channel degradation characterization corresponding to the first degradation moment as the initial state for lifetime prediction allows lifetime extrapolation to begin from the moment when degradation stabilizes.

[0072] In a preferred embodiment, the initial distribution of the lifetime prediction can be taken as a Gaussian distribution: in, The initial covariance matrix for lifetime prediction can be constructed based on the confidence level of the degradation channel as follows: in, and These represent the fundamental variance parameters corresponding to the insulation performance degradation channel and the thermal performance degradation channel at the reference baseline stage, respectively. The lower the confidence level of the degradation channel, the larger the corresponding initial covariance, thus reflecting the uncertainty of the current degradation characterization in lifetime prediction.

[0073] Sampled from the initial distribution of lifetime prediction Initial sample for group lifetime prediction: when At that time, the current lifetime prediction distribution can be constructed based on the updated modulation dual-channel degradation characterization and degradation channel confidence at the current moment, so as to achieve rolling correction of the initial lifetime prediction distribution.

[0074] S62: Based on the probabilistic lifetime prediction model, forward propagation is performed on the samples in the initial lifetime prediction distribution to obtain the future degradation trajectory, and the first termination time corresponding to each sample is determined according to the preset lifetime termination criterion, thereby obtaining the remaining lifetime corresponding to each sample.

[0075] Specifically, based on the current lifespan prediction time The start time; the first trigger time. During subsequent rolling updates Let the th element in the current lifetime prediction distribution be... Group sample : in, Indicates the first The insulation performance degradation state corresponding to the group of samples Indicates the first The thermal performance degradation state corresponding to the group of samples; then the future degradation trajectory can propagate forward in the following manner: in, To propagate forward steps, This is a probabilistic lifetime prediction model composed of state update equations for dual-channel degradation characterization quantities. It refers to one of the following: the average, the final value, or the task-given value, based on the environmental conditions input within the most recent time window before the initial degradation time. For model parameters, This represents the process noise term. By performing forward propagation on multiple sets of samples, a set of future degradation trajectories can be obtained, expressing the uncertainty in lifetime prediction.

[0076] In a preferred embodiment, the preset lifespan termination criterion can be expressed as: in, This indicates the end-of-life threshold for the insulation degradation pathway. This indicates the lifetime termination threshold of the thermal performance degradation path. This represents the life-end threshold of the composite degradation score. Indicates by the first A composite degradation score is constructed from the statistics of the future dual-channel degradation samples relative to the reference baseline.

[0077] Will be satisfied for the first time The future moment as the first The first arrival time of each sample group is determined, and the remaining lifetime of that sample is obtained. in, This represents the sampling time interval.

[0078] S63: Construct a probability prediction result of the current remaining lifespan of the thermocouple based on the remaining lifespan of each sample.

[0079] Specifically, the probabilistic prediction of the current remaining service life of the thermocouple can be expressed as: in, This represents the Dirac function.

[0080] when When the probability prediction result is the first remaining useful life prediction result after the initial degradation time is triggered; when In this case, the probability prediction result is the current remaining useful life prediction result updated based on the newly added online observation data.

[0081] Figure 4 This invention illustrates the prediction results and probability prediction range of the remaining thermocouple lifetime obtained by updating online observation data after triggering lifetime prediction at the first degradation moment according to an embodiment of the present invention.

[0082] In one specific embodiment, this method can be applied to the online life management of thermocouples under continuous thermal service conditions. Figure 2 The thermocouple output temperature time series shown is used as a sample of online observation data, and the thermocouple output signal, cold junction temperature, environmental condition information, and service history information are synchronized according to a unified sampling time; among them, Figure 2 This example only shows thermocouple output temperature as one type of observation data. Cold junction temperature, environmental operating condition information, and service history information can be collected synchronously at the same time reference. The initial installation phase or a stable operating phase confirmed to be in a healthy state after verification is selected as the reference baseline. Short-time window measurement sequences and long-time window measurement sequences are constructed based on a rolling window. Within the short-time window, time-frequency features such as high-frequency energy ratio, spectral centroid shift, rolling variance ratio, and abnormal peak proportion are extracted to construct a characterization of insulation performance degradation channels. Within the long-time window, dynamic features such as equivalent dynamic time constant, time constant fluctuation, dynamic hysteresis area, and dynamic identification residual are extracted to construct a characterization of thermal performance degradation channels. This yields... Figure 3The two degradation channel feature time series are shown. Further, the dual-channel degradation characterization is modulated based on degradation channel confidence and inter-channel interference factor, and a composite degradation score is constructed according to the reference baseline feature distribution. When the smoothed composite degradation score is greater than a preset degradation gating threshold for multiple consecutive sampling times, the first degradation time is determined. Based on the current dual-channel degradation characterization and degradation channel confidence after this time, probabilistic lifetime prediction is performed, and the probability prediction interval is determined according to the quantiles of the remaining lifetime sample set, resulting in the following... Figure 4 The remaining service life prediction mean and prediction range are shown. Through the above implementation method, insulation-related time-frequency anomalies and thermally-related dynamic response degradation can be incorporated into different degradation channels, and service life prediction can be initiated after degradation stabilizes, thereby improving the stability and reliability of online thermocouple service life prediction.

[0083] like Figure 5 As shown, the present invention also provides a probabilistic lifetime prediction system driven by thermocouple dual-channel degradation characterization, the system comprising: Data acquisition module 1 is used to acquire observation data during the online operation of thermocouples, perform time synchronization and rolling window segmentation on the observation data, and construct corresponding short time window measurement sequences and long time window measurement sequences. The dual-channel degradation characterization construction module 2 is used to perform multi-timescale adaptive feature extraction based on the short-time-window measurement sequence and the long-time-window measurement sequence to construct a dual-channel degradation characterization representing insulation performance degradation and thermal performance degradation. The channel credibility determination and characterization modulation module 3 is used to determine the credibility of the degradation channel based on the quality and availability of the corresponding features of each degradation channel, and to modulate the dual-channel degradation characterization based on the degree of mutual interference between channels; The composite degradation score construction module 4 is used to construct a composite degradation score based on the degradation channel confidence, the modulated dual-channel degradation characterization, and the reference baseline feature distribution. The first degradation time determination module 5 is used to determine the first degradation time based on the composite degradation score, and to determine the first time that meets the preset degradation gating criterion. The probabilistic lifetime prediction module 6 is used to obtain the probabilistic prediction result of the current remaining lifetime of the thermocouple based on the modulated dual-channel degradation characterization, degradation channel reliability and probabilistic lifetime prediction model corresponding to the first degradation time.

[0084] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0085] For the system embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0086] Accordingly, this application also provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the above-described thermocouple dual-channel degradation characterization driven probabilistic lifetime prediction method.

[0087] Accordingly, this application also provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the above-described probabilistic lifetime prediction method driven by thermocouple dual-channel degradation characterization.

[0088] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this application are indicated by the claims.

[0089] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A probabilistic lifetime prediction method driven by thermocouple dual-channel degradation characterization, characterized in that, include: Acquire observation data during the online operation of thermocouples, perform time synchronization and rolling window segmentation on the observation data, and construct corresponding short time window measurement sequences and long time window measurement sequences; Based on the short-time-window measurement sequence and the long-time-window measurement sequence, multi-time-scale adaptive feature extraction is performed to construct a dual-channel degradation characterization representing insulation performance degradation and thermal performance degradation; The reliability of each degradation channel is determined based on the quality and availability of the corresponding features, and the dual-channel degradation characterization is modulated according to the degree of mutual interference between channels. Based on the degradation channel confidence, the modulated dual-channel degradation characterization, and the reference baseline feature distribution, a composite degradation score is constructed. Based on the composite degradation score, the moment when the preset degradation gating criterion is first met is taken as the first degradation moment; Based on the modulated dual-channel degradation characterization, degradation channel reliability, and probabilistic lifetime prediction model corresponding to the first degradation time, the probabilistic prediction result of the current remaining lifetime of the thermocouple is obtained.

2. The method according to claim 1, characterized in that, The observation data includes at least thermocouple output signals, cold junction temperature, environmental operating conditions, and service history information.

3. The method according to claim 1, characterized in that, Based on the short-time-window measurement sequence and the long-time-window measurement sequence, multi-time-scale adaptive feature extraction is performed to construct a dual-channel degradation characterization representing insulation performance degradation and thermal performance degradation, including: Within a short time window, frequency domain analysis or time-frequency analysis is performed on the short time window measurement sequence. The frequency domain analysis or time-frequency analysis includes at least one of Fourier transform, power spectral density estimation, empirical mode decomposition, and wavelet decomposition. Time-frequency feature clusters characterizing the output signal volatility, spectral anomalies, or high-frequency disturbances are extracted from the corresponding analysis results, and insulation performance degradation channel characterization is constructed accordingly. Based on at least one of the current operating condition excitation sufficiency, dynamic feature identification residual and output sequence stability, the length of the long time window is adaptively adjusted, the dynamic response identification of the long time window measurement sequence is performed, dynamic feature clusters characterizing the dynamic response capability of the thermocouple are extracted, and a thermal performance degradation channel characterization is constructed accordingly. Based on the aforementioned characterization of insulation performance degradation channels and thermal performance degradation channels, a dual-channel degradation characterization is constructed.

4. The method according to claim 1, characterized in that, The reliability of each degradation channel is determined based on the quality and availability of its corresponding features, and the dual-channel degradation characterization is modulated according to the degree of inter-channel interference, including: The reliability of insulation performance degradation channels is determined based on time-frequency characteristic significance index and abnormal distortion index; The reliability of thermal performance degradation channels is determined based on the operating condition excitation sufficiency index, dynamic feature identification residuals, and frequency domain contamination index constructed from short time window time-frequency features. The interference factor of thermal performance degradation channel on insulation performance degradation channel is constructed based on the dynamic transition strength or non-stationarity index of long time window, and the interference factor of insulation performance degradation channel on thermal performance degradation channel is constructed based on the high frequency pollution index. A channel modulation matrix is ​​constructed based on the reliability of the insulation performance degradation channel, the reliability of the thermal performance degradation channel, the interference factor of the insulation performance degradation channel, and the interference factor of the thermal performance degradation channel; the dual-channel degradation characterization is modulated using the channel modulation matrix to obtain the dual-channel degradation characterization after reliability and mutual interference modulation.

5. The method according to claim 1, characterized in that, Based on the degradation channel confidence level, the modulated dual-channel degradation characterization, and the reference baseline feature distribution, a composite degradation score is constructed, including: The same feature extraction and confidence modulation processing as at the current time is performed on the long and short window observation data of the reference baseline phase to obtain a dual-channel reference representation sample set, thereby determining the reference baseline mean vector and the reference baseline covariance matrix; A composite degradation score is constructed based on the degree of deviation of the modulated dual-channel degradation characterization from the mean vector of the reference baseline, so as to characterize the overall degradation level at the current moment relative to the reference baseline stage.

6. The method according to claim 1, characterized in that, Based on the composite degradation score, the moment when the first degradation threshold criterion is met is taken as the first degradation moment, including: The composite degradation score is recursively smoothed to obtain a smoothed score that is updated on a rolling basis; When the smoothing score is greater than the preset degradation gating threshold in multiple consecutive sampling times, the first sampling time in the multiple consecutive sampling times is defined as the first degradation time.

7. The method according to claim 1, characterized in that, Based on the modulated dual-channel degradation characterization, degradation channel reliability, and probabilistic lifetime prediction model corresponding to the first degradation time, the probabilistic prediction result of the current remaining lifetime of the thermocouple is obtained, including: Based on the modulated dual-channel degradation characterization and degradation channel reliability corresponding to the first degradation time, an initial distribution for lifetime prediction is constructed; Based on the probabilistic lifetime prediction model, the samples in the initial lifetime prediction distribution are forward propagated to obtain the future degradation trajectory, and the first termination time corresponding to each sample is determined according to the preset lifetime termination criterion, thereby obtaining the remaining lifetime corresponding to each sample. Based on the remaining service life of each sample, a probability prediction result of the current remaining service life of the thermocouple is constructed.

8. A probabilistic lifetime prediction system driven by thermocouple dual-channel degradation characterization, characterized in that, include: The data acquisition module is used to acquire observation data during the online operation of the thermocouple, perform time synchronization and rolling window segmentation on the observation data, and construct corresponding short time window measurement sequences and long time window measurement sequences. A dual-channel degradation characterization construction module is used to perform multi-timescale adaptive feature extraction based on the short-time-window measurement sequence and the long-time-window measurement sequence to construct a dual-channel degradation characterization representing insulation performance degradation and thermal performance degradation. The channel credibility determination and characterization modulation module is used to determine the credibility of the degradation channel based on the quality and availability of the corresponding features of each degradation channel, and to modulate the dual-channel degradation characterization according to the degree of mutual interference between channels; A composite degradation score construction module is used to construct a composite degradation score based on the degradation channel confidence, the modulated dual-channel degradation characterization, and the reference baseline feature distribution. The first degradation time determination module is used to determine the first degradation time based on the composite degradation score, and to determine the first degradation time as the first degradation time that meets the preset degradation gating criterion. The probabilistic lifetime prediction module is used to obtain the probabilistic prediction result of the current remaining lifetime of the thermocouple based on the modulated dual-channel degradation characterization, degradation channel reliability and probabilistic lifetime prediction model corresponding to the first degradation time.

9. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-7.

10. A computer-readable storage medium storing computer instructions thereon, characterized in that, When executed by the processor, this instruction implements the steps of the method as described in any one of claims 1-7.

Images