Solid-state relay multi-source degradation state decoupling identification and life prediction method and system

Abstract

This invention relates to the field of solid-state relay monitoring technology, specifically a method and system for decoupling identification and lifetime prediction of multi-source degradation states in solid-state relays. It integrates a solid-state relay degradation analysis framework that incorporates dynamic modeling, digital twins, and fault prediction and health management (PHM) concepts. By establishing the correlation between the degradation state space, degradation contribution space, and lifetime evolution space of internal functional units, it achieves integrated processing of solid-state relay health status perception, fault mechanism identification, and remaining lifetime prediction, outputting degradation state identification results, dominant degradation path identification results, and remaining lifetime prediction results. This solution can achieve decoupling identification of multi-source degradation states, quantify the impact contribution of each internal functional unit on external performance changes, and improve the accuracy of dominant degradation path identification and remaining lifetime prediction.

Description

Technical Field

[0001] This invention relates to the field of solid-state relay monitoring, specifically a method and system for decoupling identification and life prediction of multi-source degradation states of solid-state relays. Background Technology

[0002] A solid-state relay (SSR) is an electronic switching device that uses semiconductor devices to connect and disconnect circuits. It is widely used in industrial automation, power electronic equipment, new energy equipment, and intelligent manufacturing. Internally, it typically includes multiple functional parts such as an input optocoupler unit, a drive circuit unit, a power output unit, and a packaged interconnect structure. These parts work together to perform functions such as signal isolation, drive control, and power output. During long-term operation, the operating state of a SSR is affected by factors such as temperature, current, and voltage. Therefore, monitoring and evaluating its operating status is crucial for ensuring reliable equipment operation.

[0003] Currently, technical solutions for monitoring the operating status and lifespan assessment of solid-state relays typically analyze the health status of the device by collecting external electrical parameters such as on-state voltage drop, on-state resistance, switching response time, and housing temperature. Some solutions obtain performance change data at different aging stages through high-temperature aging tests and establish lifespan prediction models based on the changing patterns of conduction characteristics; some solutions construct degradation trend curves using historical monitoring data and infer the remaining lifespan of the device based on parameter changes; and some solutions employ data-driven modeling methods, inputting the collected electrical characteristics into the prediction model to estimate the future performance trends of the solid-state relay.

[0004] The inventors of this application have discovered that the aforementioned technology still suffers from the following technical problems under long-term high-temperature operation scenarios for solid-state relays: the input optocoupler unit, drive circuit unit, power output unit, and package interconnection structure may all degrade. The degradation of different internal functional units collectively affects external observation characteristics such as on-state voltage drop data, load current data, casing temperature data, and response time, causing multiple degradation mechanisms to exhibit high similarity at the observation level. Simply put, while existing technologies can detect performance degradation, they struggle to determine which internal functional unit is degrading, let alone differentiate the specific contribution of each degradation source to performance changes. This leads to mechanistic confusion and contribution coupling in external observation signals, easily causing misjudgments of degradation states. Furthermore, existing technologies lack a quantitative analysis mechanism for the degradation contribution of different internal functional units, making it difficult to determine the dominant degradation path at the current stage and to establish dynamic evolution relationships between different degradation states, thus hindering accurate prediction of future performance degradation trends and remaining lifespan. Therefore, how to achieve decoupled identification of degradation states, determination of dominant degradation paths, and prediction of remaining lifetime under the condition of simultaneous multi-source degradation has become a key technical problem that urgently needs to be solved in the existing solid-state relay degradation assessment technology. Summary of the Invention

[0005] This application provides a method and system for decoupling identification and lifetime prediction of multi-source degradation states of solid-state relays, which solves the problems of confusion and contribution coupling of external observation signals under the condition of synchronous degradation of multiple internal functional units of solid-state relays in the prior art.

[0006] To achieve the above objectives, the embodiments of this application disclose the following technical solutions:

[0007] On the one hand, this solution discloses a method for decoupling identification and lifetime prediction of multi-source degradation states of solid-state relays, including the following steps:

[0008] Acquiring full-cycle transient response data of solid-state relays during operation serves as the foundational observation data in the solid-state relay fault prediction and health management (PHM) system. This data characterizes the dynamic changes in the equipment's health status during operation. Full-cycle transient response data includes turn-on delay, turn-off delay, conduction ramp rate, conduction voltage drop, load current, housing temperature, and thermal response time constant. This data covers the complete dynamic process of the solid-state relay from input triggering, internal drive, power conduction, to the establishment of the thermal response. Compared to using only a single electrical or temperature parameter for degradation assessment, this method simultaneously preserves the propagation characteristics of degradation behavior across different internal functional units over time, providing a unified observation basis for subsequent degradation state decoupling.

[0009] Time synchronization and feature reconstruction are performed on the full-cycle transient response data. A time causal chain model is established according to the operational sequence of the input optocoupler unit, drive circuit unit, power output unit, and package interconnection structure inside the solid-state relay, and a corresponding composite transient feature set is generated. The time causal chain model is used to describe the propagation process of degradation effects between different internal functional units, thereby forming a dynamic evolution expression of the degradation behavior inside the solid-state relay. The time causal chain model is not a simple data sorting process, but rather utilizes the inherent sequential relationship between the energy transfer path and the control signal propagation path inside the solid-state relay to remap the originally coupled observation features to the corresponding operational stages of the internal functional units, thereby establishing a degradation propagation structure with physical meaning.

[0010] A composite transient feature set is input into a pre-constructed inverse state mapping model. Based on a time-causal chain model, the composite transient feature set is subjected to state inversion processing to obtain degradation state vectors corresponding to the input optocoupler unit, drive circuit unit, power output unit, and package interconnection structure, respectively. These degradation state vectors collectively constitute the health state space of the solid-state relay, used to realize health state estimation and state perception in the PHM system. The inverse state mapping model utilizes the correlation between externally observed features and internal degradation states to achieve a reverse solution from observable phenomena to internal states that cannot be directly measured, thus overcoming the limitation of traditional lifetime assessment relying solely on external performance indicators.

[0011] Based on the degradation state vector, the degradation contribution of each internal functional unit within a solid-state relay to external performance changes is calculated, generating a multi-source degradation contribution set. This set is used to identify the degree of influence of different internal functional units on performance degradation, enabling fault mechanism localization and degradation source tracing. Since the degree of influence of different internal functional units on overall performance changes dynamically varies at different stages of their lifespan, the degradation contribution calculation enables quantitative separation of degradation sources, providing a basis for subsequent identification of dominant degradation paths.

[0012] Based on the rate of change and cumulative contribution value of the degradation contribution of each internal functional unit within the continuous monitoring period, the dominant degradation path of the current stage is determined; degradation evaluation indicators are calculated based on the degradation state vector and the set of multi-source degradation contribution; the degradation state vectors obtained from the continuous monitoring period are used to construct a historical degradation state sequence, and a degradation state evolution model is established by combining the historical degradation state sequence; the degradation state evolution model is used to describe the continuous change law of equipment health status within its life cycle and to characterize the coupling evolution relationship between different degradation mechanisms; among them, the degradation evaluation indicators are used to characterize the current health status, while the historical degradation state sequence is used to describe the degradation development trend, and the combination of the two can simultaneously reflect the current state and the future evolution direction.

[0013] Based on the degradation state evolution model, this application outputs degradation state identification results, dominant degradation path identification results, and remaining lifetime prediction results for solid-state relays, thus forming a closed loop of fault prediction and health management covering health status assessment, fault evolution analysis, and remaining lifetime prediction. This application does not directly utilize historical performance data for lifetime fitting; instead, it achieves consistency between the lifetime prediction process and the actual degradation mechanism through joint modeling of internal degradation states, degradation contribution, and dominant degradation paths, thereby improving prediction reliability and interpretability.

[0014] Furthermore, by establishing a temporal causal chain node sequence through unified time reference correction, the purpose is not merely to achieve multi-source data synchronization, but to construct causal constraints by utilizing the temporal sequence of generation and propagation of various observed features within the solid-state relay. Compared to traditional correlation analysis methods, this application uses a temporal correlation matrix to explicitly express the order of interaction between internal functional units and observed features, subjecting the subsequent inverse state mapping process to physical causal constraints, thereby reducing state misjudgments caused by feature coupling.

[0015] Furthermore, a mapping relationship between the external observation feature space and the internal degradation state space is established using historical monitoring samples corresponding to different aging stages. The purpose is to solve the problem that the internal degradation state cannot be directly measured. Since the degradation processes of multiple functional units inside a solid-state relay are superimposed, it is difficult to accurately locate the source of degradation by relying solely on external performance changes. Therefore, by learning the mapping law between the two spaces through historical monitoring samples, the internal degradation state can be inverted.

[0016] Furthermore, a cross-domain attention mechanism is introduced and a feature weight matrix is ​​constructed to address the issue of inconsistent contributions from different observation features to different degradation state parameters. By dynamically adjusting the weights of each observation feature, observation features with stronger correlation to the target degradation state receive higher contributions, thereby improving the decoupling accuracy of the degradation state.

[0017] Furthermore, a low-dimensional state manifold is constructed through manifold dimensionality reduction, and independent degradation trajectories are extracted. This design is not simply data compression, but rather utilizes neighborhood topological relationships in the sample space to preserve the degradation evolution pattern, enabling the degradation processes corresponding to different internal functional units to form independent evolution paths in the low-dimensional space. Further analysis combining change rate and cumulative contribution value can identify the key degradation paths that truly dominate the overall performance degradation at the current stage.

[0018] Furthermore, a digital twin model of a solid-state relay is introduced to construct a degradation state evolution model, the purpose of which is to establish a virtual model that evolves synchronously with the actual device. Compared with traditional empirical lifetime models that only use historical statistical patterns for prediction, this application uses input optocoupler transmission characteristic parameters, drive circuit parameters, power output unit conduction parameters, and package thermal conduction parameters to jointly describe the internal degradation behavior, so that lifetime prediction is based on the actual degradation mechanism.

[0019] The digital twin model constructs an internal mechanism model based on the input optocoupler transmission characteristic parameters, drive circuit parameters, power output unit conduction parameters, and packaged heat conduction parameters. It achieves virtual-real synchronous evolution through real-time observation data and provides support for degradation state estimation, health assessment, and remaining lifetime prediction.

[0020] Furthermore, an unscented Kalman filter algorithm is employed for state estimation to address the problem of prediction errors accumulating over time in digital twin models. By fusing the composite transient feature set obtained from real-time observations with the predicted degradation state vector, the digital twin model can continuously correct its internal state estimation results, thereby improving long-term prediction stability and accuracy. The unscented Kalman filter algorithm enables online closed-loop correction between the digital twin model and actual equipment observation data, ensuring the digital twin model maintains consistency with the actual degradation process.

[0021] Furthermore, thermal coupling compensation processing is performed on the housing temperature data using a thermal conduction hysteresis model to eliminate the influence of thermal crosstalk between adjacent channels on the target channel temperature measurement results. Due to the significant thermal coupling effect of multi-channel solid-state relays, the directly acquired housing temperature data often includes both its own heat generation and the heat transfer contribution from adjacent channels. This application estimates the thermal conduction influence using a thermal conduction hysteresis model, thereby obtaining compensated temperature characteristic data that accurately reflects the degradation state of the target solid-state relay.

[0022] Furthermore, by constructing a load transient characteristic spectrum, inductive, capacitive, and resistive load characteristics are identified, and the target sampling window is determined accordingly. The core of this design lies in the fact that different load types lead to different transient characteristics in the on-state voltage drop waveform. Using a fixed sampling window can easily introduce measurement deviations caused by differences in load type. Dynamically determining the target sampling window through load fingerprint recognition improves the accuracy of extracting true on-state voltage drop characteristic data.

[0023] On the other hand, this solution discloses a system for decoupling identification and lifetime prediction of multi-source degradation states in solid-state relays, including a contact voltage detection unit, a load current detection unit, a temperature detection unit, a data acquisition module, a feature processing module, a time causal chain construction module, a degradation state inversion module, a contribution analysis module, and a lifetime prediction module. These modules do not operate independently but form a complete data flow loop around the full-cycle transient response data. The contact voltage detection unit, load current detection unit, and temperature detection unit are responsible for acquiring degradation observation information; the data acquisition module performs unified aggregation; the feature processing module performs time-domain feature reconstruction; the time causal chain construction module establishes the internal operational logic; the degradation state inversion module solves the internal state problem; the contribution analysis module identifies the sources of degradation; and the lifetime prediction module predicts future evolution. Through the collaborative work of these modules, integrated processing of solid-state relay internal degradation mechanism identification and lifetime prediction is achieved.

[0024] The system further constitutes a fault prediction and health management (PHM) platform for the full life cycle operation management of solid-state relays, which realizes continuous perception and predictive maintenance of equipment health status through condition monitoring, degradation decoupling, dynamic modeling, digital twin simulation and life prediction.

[0025] Compared with existing technologies, this invention acquires full-cycle transient response data during the operation of a solid-state relay and, by combining a time-causal chain model, an inverse state mapping model, and a degradation state evolution model, decouples and identifies the degradation states corresponding to the input optocoupler unit, drive circuit unit, power output unit, and package interconnection structure. This solves the problems of confusion in external monitoring signal mechanisms, contribution coupling, and difficulty in distinguishing degradation sources when multiple internal functional units degrade simultaneously, as seen in existing technologies. By calculating the degradation contribution of each internal functional unit and determining the dominant degradation path, the invention transforms the previously observable result of "device performance degradation" into an analytical process capable of identifying "which internal functional unit caused the performance degradation, how the degradation developed, and how it will evolve in the future," thereby significantly improving the accuracy and interpretability of degradation diagnosis. Simultaneously, by constructing a digital twin model that includes input optocoupler transmission characteristic parameters, drive circuit parameters, power output unit conduction parameters, and packaged thermal conduction parameters, and combining this with historical degradation state sequences for dynamic modeling, continuous tracking and iterative deduction of the degradation state evolution process are achieved. Real-time observation data is then used to correct the model online, enabling the prediction results to be dynamically updated as the actual state of the device changes. Furthermore, by outputting degradation state identification results, dominant degradation path identification results, and remaining lifetime prediction results, a reliable data foundation and decision-making basis are provided for solid-state relay fault prediction and health management, improving lifetime prediction accuracy and enhancing equipment operational reliability and preventative maintenance capabilities. Attached Figure Description

[0026] Figure 1 This is an overall flowchart of an embodiment of the present invention;

[0027] Figure 2 This is a structural diagram of the time causal chain model and the inverse state mapping model according to an embodiment of the present invention;

[0028] Figure 3 This is a flowchart illustrating the dominant degradation path identification process according to an embodiment of the present invention.

[0029] Figure 4 This is a flowchart illustrating the degradation state evolution and lifetime prediction of an embodiment of the present invention.

[0030] Figure 5 This is a system structure block diagram according to an embodiment of the present invention;

[0031] Figure 6 This is a flowchart of the load fingerprint recognition processing and thermal coupling compensation processing according to an embodiment of the present invention. Detailed Implementation

[0032] 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 a part of the embodiments of the present invention, and not all of them. 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. In the following description, numerous specific details are set forth to provide a comprehensive understanding of the present invention. The present invention may be practiced without some or all of these specific details. In other instances, well-known processes have not been described in detail to avoid unnecessarily obscuring the present invention.

[0033] When used in conjunction with the terms "comprising," "method comprising," or similar language in this specification and appended claims, the singular forms "a," "some," and "the" include plural references unless the context clearly indicates otherwise. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0034] Terminology definition:

[0035] Multi-source degradation refers to the set of degradation characteristics generated by multiple functional units within a solid-state relay during operation, including the degradation state of the input optocoupler unit, the drive circuit unit, the power output unit, and the package interconnect structure. Its purpose is to reflect the impact of different degradation sources on device performance changes.

[0036] The time-causal chain model is used to describe the sequential relationship between the functions of internal units in a solid-state relay and their temporal correlation with externally observed characteristics. Its purpose is to establish the causal basis for the relationship between internal degradation processes and changes in external response.

[0037] The inverse state mapping model is a state inversion model used to establish the correspondence between the external observed feature space and the internal degenerate state space. Its function is to infer the degenerate state information of internal functional units based on the external measurable response.

[0038] The degradation state vector is a set of parameters used to characterize the degree of degradation of the input optocoupler unit, drive circuit unit, power output unit, and package interconnect structure. Its purpose is to achieve a unified quantitative description of the internal degradation state.

[0039] The multi-source degradation contribution set is a set of contributions used to describe the degree of influence of each internal functional unit on external performance changes. Its purpose is to quantify the weight of the impact of different degradation sources on the overall performance degradation.

[0040] The dominant degradation path refers to the degradation trajectory that contributes the most to the performance changes of a solid-state relay at the current degradation stage. Its purpose is to identify the key degradation sources that determine the trend of device lifetime evolution.

[0041] Degradation state evolution models are predictive models used to describe the evolution of degradation states over time. Their role is to characterize the development trend of the degradation process and support lifetime prediction analysis.

[0042] A digital twin model of a solid-state relay is a digital state model corresponding to the actual operating state of the solid-state relay. It is used to synchronously characterize the operating state of the input optocoupler unit, drive circuit unit, power output unit, and package interconnection structure. Its function is to achieve dynamic mapping and evolution prediction of degradation states.

[0043] The load transient characteristic spectrum is a set of features used to characterize the dynamic changes in load current and on-state voltage waveforms. Its purpose is to identify the interference effects of different load types on the on-state voltage drop measurement results.

[0044] Example 1

[0045] This embodiment discloses a method for decoupling and identifying the multi-source degradation state of solid-state relays and predicting their remaining lifetime. Addressing the problem of difficulty in accurate identification due to the overlapping degradation mechanisms of the input optocoupler unit, drive circuit unit, power output unit, and package interconnection structure, the method constructs a time-causal chain model, an inverse state mapping model, and a digital twin evolution model to achieve decoupling of multi-source degradation contributions, identification of dominant degradation paths, and prediction of remaining lifetime. The method includes the following steps:

[0046] Step 1: Acquire full-cycle transient response data and perform thermal coupling compensation and load fingerprint recognition to extract true on-state voltage drop characteristic data;

[0047] In this embodiment, the full-cycle transient response data of the solid-state relay during operation is obtained. The full-cycle transient response data includes the turn-on delay, turn-off delay, conduction ramp, conduction voltage drop, housing temperature data, and thermal response time constant.

[0048] After acquiring the shell temperature data, thermal coupling compensation processing is performed on the shell temperature data. The thermal response time constant is calculated based on the shell temperature change process. Since the heat generated during the operation of adjacent channels will affect the temperature of the target channel, a heat conduction hysteresis relationship model between the target channel and adjacent channels is established. This model is used to characterize the time hysteresis characteristics and spatial attenuation characteristics during the heat transfer process, and calculates the heat conduction influence of adjacent channels in conjunction with the temperature rise change process of adjacent channels.

[0049] Preferably, a thermal coupling influence matrix is ​​established based on the structural distance between the target channel and adjacent channels, the relationship between their installation positions, and the historical temperature rise response characteristics. The thermal contribution value of each adjacent channel to the target channel is then calculated based on the thermal coupling influence matrix.

[0050] Since the heat generated in adjacent channels is affected by both spatial distance attenuation and thermal diffusion time hysteresis when it propagates to the target channel, the temperature rise process in adjacent channels can be regarded as thermal excitation input, and the thermal interference experienced by the target channel can be regarded as thermal response output. Furthermore, based on the characteristic that the thermal influence gradually decays with propagation time during the thermal diffusion process, an exponential hysteresis kernel function is introduced to characterize the dynamic process of thermal diffusion, and a thermal conduction hysteresis relationship model is established in conjunction with the thermal coupling influence matrix.

[0051] First, thermal coupling weighting coefficients are constructed based on the structural distances between the target channel and each adjacent channel:

[0052] ;

[0053] The thermal coupling weights of each adjacent channel satisfy the following:

[0054] ;

[0055] Furthermore, based on the mapping relationship between the temperature rise changes of adjacent channels and the temperature response of the target channel, a heat conduction hysteresis equation is constructed:

[0056] ;

[0057] This yields the compensated temperature characteristic data:

[0058] ;

[0059] Through the above calculations, the temperature influence caused by thermal crosstalk between adjacent channels can be obtained, and external thermal coupling interference can be eliminated accordingly, thereby obtaining compensated temperature characteristic data that can truly reflect the heating state of the target solid-state relay itself.

[0060] in, It represents the total heat conduction influence on the target channel at time t; it is calculated by substituting the temperature rise data of each adjacent channel into the heat conduction hysteresis equation; it is used to characterize the additional temperature rise caused by thermal crosstalk.

[0061] This represents the compensated temperature characteristic data; it is obtained by subtracting the thermal conduction effect from the original shell temperature data; and it is used for subsequent degradation state analysis.

[0062] This represents the original housing temperature data of the target channel at time t; it is acquired in real time by a temperature sampling circuit installed on the surface of the solid-state relay housing.

[0063] N represents the number of adjacent channels involved in the thermal coupling calculation; it is obtained by counting the number of solid-state relay channels actually installed around the target channel.

[0064] This represents the thermal coupling weight coefficient corresponding to the j-th adjacent channel; it is calculated from the structural distance between the target channel and the j-th adjacent channel; it is used to reflect the relative contribution of different adjacent channels to the thermal influence of the target channel.

[0065] This represents the structural distance between the target channel and the j-th adjacent channel; it is obtained based on the equipment installation structure diagram, installation coordinate data, or actual measurement results; it can be calculated using the Euclidean distance between the center positions of the channels.

[0066] τ represents the sampling time variable corresponding to the historical temperature rise data, with a value range from 0 to the current monitoring time t. It is used to accumulate the contribution of the historical temperature rise of adjacent channels to the thermal impact of the current target channel.

[0067] L represents the thermal diffusion characteristic length; it is used to characterize the heat transfer capability under the current installation structure; it is obtained through temperature rise response tests; specifically, constant power heating is applied to multiple channels of known distance, and the thermal diffusion characteristic length is obtained by fitting the temperature response curve.

[0068] Indicates the time of the j-th adjacent channel. The change in temperature is calculated using the following formula:

[0069] ;

[0070] in, For the j-th adjacent channel at time... The real-time temperature is obtained from the corresponding temperature sensor.

[0071] The initial temperature is the temperature at which the j-th adjacent channel begins monitoring; it is obtained from the start time of the monitoring cycle.

[0072] This represents the thermal diffusion hysteresis coefficient corresponding to the j-th adjacent channel; it is used to characterize the response speed during the process of heat transfer from adjacent channels to the target channel; it is obtained through thermal response experiments. The unit is the reciprocal of time. The larger the value, the faster the heat transfer speed and the more rapidly the historical temperature rise effect decays; the smaller the value, the more obvious the thermal inertia effect and the longer the duration of the historical temperature rise effect.

[0073] Specifically, a step power excitation is applied to the j-th adjacent channel, the temperature response curve of the target channel is recorded, and the response time constant corresponding to the final stable temperature rise of the target channel is measured. Then calculate:

[0074] ;

[0075] This represents the thermal response time constant when the thermal influence of the j-th adjacent channel is transferred to the target channel. It is obtained through a step power test, i.e., applying a stable thermal excitation to the j-th adjacent channel, recording the temperature rise curve of the target channel, and determining the time corresponding to when the temperature rise reaches 63.2% of the final stable temperature rise. .

[0076] t represents the current monitoring time.

[0077] The collected shell temperature data is corrected based on the influence of heat conduction to obtain compensated temperature characteristic data.

[0078] After acquiring the on-state voltage signal and load current signal, load fingerprinting processing is performed on the on-state voltage signal. A load transient characteristic spectrum is constructed based on the load current waveform and the on-state voltage waveform, and inductive load characteristics, capacitive load characteristics, and resistive load characteristics are identified based on the load transient characteristic spectrum.

[0079] Specifically, phase relationship features, transient oscillation features, and harmonic distribution features can be extracted from the load current waveform and the conduction voltage waveform to construct a load transient feature spectrum for characterizing the dynamic behavior of the load.

[0080] Since inductive, capacitive, and resistive loads differ significantly in voltage-current phase relationships, transient oscillation energy, and harmonic composition, a load transient characteristic spectrum integrating multidimensional dynamic features is constructed.

[0081] This leads to the load feature vector:

[0082] ;

[0083] in,

[0084] ;

[0085] ;

[0086] Further construct the classification decision function:

[0087] ;

[0088] ;

[0089] in, It is a load characteristic vector; it is formed by combining voltage and current phase difference characteristics, transient oscillation energy characteristics, and harmonic distribution characteristics; it is used to characterize the dynamic behavior of the load.

[0090] u(t) is the conduction voltage waveform; it is obtained by the voltage sampling circuit.

[0091] This is the load current waveform; obtained by the current sampling circuit.

[0092] The voltage-current phase difference characteristic is calculated from the normalized inner product of the conduction voltage waveform and the load current waveform within the target sampling window; it is used to characterize the degree of phase deviation between voltage and current; among which, inductive load corresponds to a larger lag phase difference, capacitive load corresponds to a leading phase difference, and resistive load corresponds to a phase difference close to zero.

[0093] The target sampling window length is determined based on the waveform of the conduction phase before load type identification. The starting point of the sampling window is the conduction start time, and the ending point is the time when the steady-state conduction interval is entered.

[0094] The transient oscillation energy characteristic is obtained by integrating the square of the load current change rate within the target sampling window; it is used to characterize the transient oscillation intensity during the conduction process; the more intense the oscillation, the larger the corresponding Eo value.

[0095] The harmonic distribution characteristics are calculated from the amplitudes of each harmonic obtained by FFT spectrum analysis; it is used to characterize the degree of harmonic distortion of the load current waveform relative to the fundamental wave; the larger the Rh value, the more obvious the nonlinear characteristics of the load.

[0096] k is the highest harmonic order participating in the harmonic analysis; it is determined based on the FFT spectrum analysis results; when the proportion of harmonic energy above the Kth order to the total harmonic energy is lower than a preset threshold, this order is determined as the highest analysis order; preferably, the order is between 10 and 20.

[0097] The amplitude of the k-th harmonic is obtained by performing a fast Fourier transform on the load current waveform within the target sampling window to obtain the amplitude of the corresponding frequency component; it is used to characterize the intensity of the k-th harmonic component.

[0098] The feature center vector corresponding to the m-th type of load is obtained by averaging multiple load feature vectors belonging to the m-th type of load in the historical sample database.

[0099] The number of training samples corresponding to the m-th type of load is obtained from statistics of the historical load sample database.

[0100] This is the load feature vector corresponding to the nth historical sample of the m-th load type; it is calculated from the on-state voltage waveform and load current waveform of the corresponding sample.

[0101] C represents the load type determination result, which is obtained by calculating the cosine similarity between the current load feature vector and the center vectors of various load features; the category with the highest similarity corresponds to the final load type.

[0102] 'm' represents the load category number, used to distinguish different load types; load types include at least inductive loads, capacitive loads, and resistive loads.

[0103] The target sampling window is determined based on the identification results. Among them, inductive loads correspond to the stable stage after the reverse induced potential decays, capacitive loads correspond to the stable stage after the charge-discharge impact ends, and resistive loads correspond to the conduction stability range.

[0104] The actual on-state voltage drop characteristic data is extracted within the target sampling window to characterize the changes in the conduction performance of the power output unit.

[0105] In addition, the thermal response time constant is obtained based on the change process of the housing temperature data to characterize the thermal inertia of the device.

[0106] Ultimately, the turn-on delay, turn-off delay, conduction ramp rate, actual conduction voltage drop characteristic data, compensated temperature characteristic data, and thermal response time constant are combined to form full-cycle transient response data. This provides fundamental data support for subsequent time-causal chain model construction, degradation state inversion processing, and degradation state evolution analysis. The conduction voltage drop and shell temperature data are the raw response data acquired, which are processed by load fingerprinting and thermal coupling compensation to generate actual conduction voltage drop characteristic data and compensated temperature characteristic data, respectively.

[0107] Step 2: Perform time synchronization and feature reconstruction processing and establish a time causal chain model.

[0108] In this embodiment, the full-cycle transient response data obtained in step 1 is subjected to time synchronization processing and feature reconstruction processing. A time causal chain model is established according to the working order of the input optocoupler unit, drive circuit unit, power output unit and package interconnection structure inside the solid-state relay, and the corresponding composite transient feature set is generated.

[0109] The time synchronization process includes uniform time reference correction for turn-on delay, turn-off delay, conduction ramp, actual conduction voltage drop characteristic data, compensated temperature characteristic data, and thermal response time constant, so that each observation characteristic corresponds to the same monitoring cycle.

[0110] After completing the time synchronization process, a time causal chain node sequence is constructed based on the response time corresponding to each feature, and a time causal chain model is established according to the order of function of the input optocoupler unit, the drive circuit unit, the power output unit, and the package interconnection structure.

[0111] Specifically, the time causal chain model is used to describe the causal transmission relationship between the degradation process of internal functional units and the change process of external observed characteristics. The input optocoupler unit corresponds to the turn-on action delay and the turn-off action delay, the drive circuit unit corresponds to the conduction ramp rate, the power output unit corresponds to the conduction voltage drop, and the package interconnect structure corresponds to the shell temperature data and the thermal response time constant.

[0112] Based on the sequential interaction relationships between nodes in the time causal chain, a temporal correlation matrix between internal functional units and external observed features is generated, and the temporal correlation matrix is ​​used as the input constraint for the subsequent inverse state mapping model.

[0113] Furthermore, feature reconstruction processing is performed on each observed feature after time synchronization. The turn-on delay, turn-off delay, conduction ramp slope, true conduction voltage drop feature data, compensated temperature feature data, and thermal response time constant are combined in a unified time order, and the trend feature, rate of change feature, and cross-feature coupling feature are extracted to form a composite transient feature set.

[0114] Finally, the composite transient feature set and the temporal correlation matrix are output together to the inverse state mapping model, providing input data and physical constraints for subsequent degenerate state inversion processing.

[0115] Step 3: Input the composite transient feature set into the inverse state mapping model to perform state inversion and independent degradation trajectory extraction.

[0116] In this embodiment, the composite transient feature set generated in step 2 is input into the pre-constructed inverse state mapping model. Based on the temporal correlation matrix constraints provided by the time causal chain model, the composite transient feature set is subjected to state inversion processing to obtain degradation state vectors corresponding to the input optocoupler unit, the drive circuit unit, the power output unit, and the package interconnection structure, respectively. The degradation state parameter corresponding to the input optocoupler unit is denoted as the input optocoupler degradation state parameter, which is used to characterize the performance degradation degree of the input optocoupler unit during the aging process.

[0117] Preferably, the inverse state mapping model establishes a mapping relationship between the external observation feature space and the internal degradation state space using historical monitoring samples corresponding to different aging stages. The external observation feature space consists of observation features corresponding to turn-on delay, turn-off delay, conduction ramp rate, conduction voltage drop data, load current data, casing temperature data, and thermal response time constant.

[0118] Among them, the historical monitoring samples are derived from the high-temperature aging test process of solid relays, and the full-cycle transient response data collected at different aging time points are combined with the degradation state parameters of the internal functional units at the corresponding stages.

[0119] In this embodiment, the inverse state mapping model is constructed based on the correspondence between the composite transient feature set obtained during historical aging tests and the corresponding degradation state parameters.

[0120] Specifically, full-cycle transient response data of solid-state relays are collected at different aging stages, and after time synchronization processing and feature reconstruction processing, composite transient feature sets corresponding to different monitoring cycles are obtained. At the same time, based on the device disassembly and testing results, performance degradation test results, and changes in internal functional unit parameters, the degradation state parameters of the input optocoupler unit, drive circuit unit, power output unit, and package interconnection structure under the corresponding monitoring cycle are determined, thereby forming a sample correspondence between the composite transient feature set and the degradation state vector.

[0121] Based on the sample correspondence, a nonlinear mapping relationship is established from the composite transient feature set to the degenerate state vector:

[0122] ;

[0123] The mapping function is determined by the correlation between the time delay feature, conduction transition feature, conduction voltage drop feature, and temperature response feature in the composite transient feature set and the degradation state parameters of each internal functional unit.

[0124] Furthermore, the timing correlation matrix generated in step 2 is introduced as a constraint condition to modify the mapping result so that the output degradation state vector satisfies the degradation propagation relationship between the input optocoupler unit, the drive circuit unit, the power output unit, and the package interconnection structure.

[0125] Specifically, the inverse state mapping model uses basis function expansion to establish a nonlinear mapping relationship between the composite transient feature set and the initial degenerate state.

[0126] Since the influence of the degradation state of different internal functional units on the external transient response has nonlinear characteristics, the composite transient feature set is mapped to multiple nonlinear basis function spaces, and the initial degradation state estimation result is obtained by weighted combination of the output results of each basis function.

[0127] The nonlinear mapping function is expressed as:

[0128] ;

[0129] Composite transient feature set Input each nonlinear basis function The corresponding characteristic response values ​​are obtained, and weighting coefficients are assigned according to the degree of influence of different basis function response results on the degradation state estimation results. .

[0130] This represents the output of the k-th nonlinear basis function after performing feature transformation on the composite transient feature set. It is used to extract the nonlinear change law related to the degradation of internal functional units in the composite transient feature set. The nonlinear basis function is determined through the basis function training process based on the mapping samples between the composite transient feature set and the corresponding degradation state parameters collected during the historical aging test. Its form can be selected from polynomial basis functions, radial basis functions, or other function forms that can characterize nonlinear mapping relationships according to the feature data distribution.

[0131] The mapping weight coefficient corresponding to the k-th nonlinear basis function is used to characterize the influence of the corresponding feature change law on the degradation state inversion result. It is obtained through the fitting optimization process between the composite transient feature set in the historical samples and the degradation state parameters of the input optocoupler, the degradation state parameters of the drive circuit, the degradation state parameters of the power output unit, and the degradation state parameters of the package interconnect structure.

[0132] K represents the number of basis functions involved in the nonlinear mapping calculation, which is determined based on the feature dimension and the convergence of the mapping error in historical samples.

[0133] During the process of establishing the mapping relationship, a dimensionality reduction and reconstruction process is performed on the composite transient feature set to eliminate feature redundancy and noise interference, thereby improving the computational efficiency and recognition accuracy in the state inversion process.

[0134] Since multiple degradation sources act simultaneously on external observation features, an inverse state mapping model under the constraint of the temporal correlation matrix is ​​established.

[0135] Based on the mapping relationship between the composite transient feature set and the degenerate state space, the following is constructed:

[0136] ;

[0137] When matrix When the non-singular condition is met, the degenerate state vector is obtained by matrix inversion:

[0138] ;

[0139] Furthermore, a nonlinear mapping function is constructed using basis function expansion:

[0140] After substituting, we get:

[0141] ;

[0142] To ensure that the inverse state mapping equation has a unique solution, the constraint time-series constraint weight coefficient β and the time-series correlation matrix are used. satisfy: ,in, Represents the spectral radius of the matrix.

[0143] The above calculations enable the reverse solution from externally observed features to internally degraded states, thereby obtaining the degraded state vectors corresponding to the input optocoupler unit, drive circuit unit, power output unit, and package interconnection structure.

[0144] This is a degradation state vector, including degradation state parameters for the input optocoupler, the drive circuit, the power output unit, and the package interconnect structure. The input optocoupler degradation state parameters characterize the degree of degradation of the input optocoupler unit, the drive circuit degradation state parameters characterize the degree of degradation of the drive circuit unit, the power output unit degradation state parameters characterize the degree of degradation of the power output unit's conduction performance, and the package interconnect structure degradation state parameters characterize the degree of degradation of the package interconnect structure's thermal conductivity and connectivity performance.

[0145] Input optocoupler degradation state parameters include: photoelectric conversion efficiency and CTR retention rate;

[0146] Degradation parameters of the drive circuit include: drive output current capability and switching response delay;

[0147] The degradation parameters of the power output unit include: the rate of increase of the on-state voltage drop and the rate of change of the on-state resistance;

[0148] Degradation parameters of packaged interconnect structures include: thermal resistance growth rate and contact resistance change rate;

[0149] Fc is a composite transient feature set, including turn-on delay, turn-off delay, conduction ramp, actual conduction voltage drop feature data, compensated temperature feature data, and thermal response time constant after time synchronization. It further includes trend features, rate of change features, and cross-feature coupling features extracted from the above features.

[0150] The feature mapping result represents the state representation obtained by the inverse state mapping model after performing a nonlinear feature transformation on the composite transient feature set. The inverse state mapping model is trained using observed feature data and degraded state data from samples at different aging stages, and is used to establish the correspondence between the external observed feature space and the internal degraded state space.

[0151] This is a temporal correlation matrix, generated by a temporal causal chain model. It describes the temporal dependencies and causal propagation relationships between the input optocoupler unit, drive circuit unit, power output unit, and package interconnect structure and each observed feature. The matrix elements are calculated based on the order of action and correlation strength between the nodes in each temporal causal chain.

[0152] Furthermore, the time-series correlation matrix At is generated by a time-causal chain model. Specifically, based on the actual operational sequence of the input optocoupler unit, drive circuit unit, power output unit, and packaged interconnection structure within the solid-state relay, a corresponding time-causal chain node sequence is established, and the initial connection relationship is determined according to the sequential operational relationship between each node. Subsequently, combining the correlation degree between each observed feature in historical monitoring samples and the degradation state parameters of each internal functional unit, the correlation strength between each node is calculated, and the correlation strength is assigned as a matrix element to the time-series correlation matrix At.

[0153] The correlation strength characterizes the impact of a change in the degradation state of an internal functional unit on subsequent changes in internal functional units or observed characteristics. It can be calculated based on statistical correlation, state propagation frequency, response delay relationship, or causal contribution in historical samples. After normalization, a time-series correlation matrix At is formed, so that the matrix elements simultaneously reflect the internal degradation propagation path and propagation strength.

[0154] By introducing the temporal correlation matrix At, the sequential action relationship in the internal degradation propagation process of solid-state relays can be explicitly introduced into the state inversion process. This ensures that the degradation state vector not only satisfies the observed feature mapping law but also the actual degradation transmission law between internal functional units, thereby improving the physical consistency and interpretability of the degradation state identification results.

[0155] β is the temporal constraint weight coefficient, used to adjust the constraint strength of the temporal correlation matrix on the state inversion result. Its value is determined through the training process of samples at different aging stages, and the optimization objective is to minimize the degradation state inversion error. Furthermore, the temporal constraint weight coefficient β is used to adjust the influence of the temporal correlation matrix At on the degradation state inversion result. Its value is not fixed but obtained through training with historical aging test samples. Specifically, the error between the actual degradation state parameters corresponding to historical samples and the degradation state vector output by the inverse state mapping model is used as the optimization objective. The value of β is continuously adjusted through an iterative optimization process. When the degradation state inversion error reaches its minimum or meets the preset convergence condition, the corresponding β is determined as the final temporal constraint weight coefficient.

[0156] During training, when the value of β is small, the inverse state mapping model mainly relies on the composite transient feature set for state inversion; when the value of β increases, the constraint effect of the temporal correlation matrix At on the state inversion results is enhanced. The β obtained through optimization can establish a balance between observed feature information and temporal causal constraints, thereby avoiding misjudgment of degraded states caused by feature coupling and improving the accuracy of multi-source decoupled identification.

[0157] Preferably, the value of β is determined by minimizing the mean square error between the degraded state vector and the historically calibrated degraded state parameters.

[0158] After state inversion processing, the inverse state mapping model outputs a degradation state vector consisting of the degradation state parameters of the input optocoupler, the degradation state parameters of the drive circuit, the degradation state parameters of the power output unit, and the degradation state parameters of the package interconnect structure.

[0159] It is an identity matrix with the same dimensions as the degenerate state vector, used to ensure that the inverse state mapping equation is solvable.

[0160] The samples from different aging stages include a set of composite transient characteristics and corresponding degradation state data collected from solid-state relays in the initial state, mild degradation state, moderate degradation state, and severe degradation state. The degradation state data is obtained through aging tests, failure analysis results, or device parameter detection results.

[0161] To further enhance the separability between different degradation sources, the inverse state mapping model employs a cross-domain attention mechanism. This mechanism analyzes the correlation between different observed features and different degradation state parameters, thereby achieving adaptive matching between observed features and degradation state parameters. The different observed features refer to the individual observed features in the composite transient feature set.

[0162] Specifically, a feature weight matrix is ​​generated based on the degree of correlation between different observation features and different degradation state parameters. The feature weight matrix is ​​used to represent the importance of each observation feature to each degradation state parameter, with different observation features corresponding to different weight values.

[0163] Since different observation features correspond to different internal functional units with significantly different sensitivities, for example, the turn-on delay mainly reflects the degradation of the input optocoupler, while the conduction voltage drop mainly reflects the degradation of the power output unit, it is necessary to establish a cross-domain attention weight model that simultaneously considers feature correlation and temporal causality constraints.

[0164] The observation features in the external observation feature space are the observation features extracted from the full-cycle transient response data.

[0165] The composite transient feature set is weighted according to the feature weight matrix to obtain a decoupled feature set. This process enhances the feature information related to the target degradation state while suppressing interference from irrelevant features, gradually separating the features corresponding to different degradation sources.

[0166] Based on the low-dimensional state manifold, independent degradation trajectories corresponding to the input optocoupler unit, drive circuit unit, power output unit, and package interconnect structure are extracted.

[0167] Furthermore, corresponding degradation contribution evolution sequences are generated based on independent degradation trajectories, serving as the basis for identifying dominant degradation paths. Specifically, the corresponding contribution evolution sequences can be obtained by analyzing the development trends of each independent degradation trajectory within a continuous monitoring period.

[0168] Subsequently, the rate of change and cumulative contribution value of each contribution evolution sequence are calculated. The rate of change reflects the development speed of the corresponding degradation source, and the cumulative contribution value reflects the long-term impact of the corresponding degradation source on the overall performance degradation.

[0169] The independent degradation trajectory with the largest contribution is determined based on the rate of change and the cumulative contribution value, and this independent degradation trajectory is identified as the dominant degradation path in the current stage.

[0170] Finally, the degraded state vector, independent degraded trajectories, and the identification results of the dominant degraded path at the current stage are output, providing an input basis for subsequent calculation of degraded contribution and establishment of a degraded state evolution model.

[0171] Step 4: Calculate the degradation contribution of each internal functional unit and generate a multi-source degradation contribution set.

[0172] In this embodiment, the degradation contribution of each internal functional unit to the external performance change is calculated based on the degradation state vector obtained in step 3, and a multi-source degradation contribution set is generated.

[0173] Specifically, the degradation state parameters corresponding to the input optocoupler unit, drive circuit unit, power output unit and package interconnection structure are used as inputs. The turn-on delay, turn-off delay, conduction ramp, conduction voltage drop, shell temperature data and thermal response time constant obtained in step 1 are used as performance characterization parameters to establish the correlation between degradation state and performance change, and to calculate the degradation contribution of each internal functional unit.

[0174] Since different internal functional units have different degrees of impact on overall performance degradation, a coupling method of performance sensitivity and degradation degree is used to calculate the degradation contribution.

[0175] Therefore, we get:

[0176] ;

[0177] in, The degradation contribution of the i-th internal functional unit is used to characterize the relative contribution of this internal functional unit to the overall performance degradation.

[0178] The degradation state parameter corresponding to the i-th internal functional unit is obtained by inversion of the inverse state mapping model based on the composite transient feature set, and is used to characterize the current degradation degree of the corresponding internal functional unit.

[0179] P represents the comprehensive performance index, which is formed by integrating the turn-on delay, turn-off delay, conduction ramp, conduction voltage drop, temperature, and thermal response time constant. The comprehensive performance index is obtained by weighted summation after normalizing the turn-on delay, turn-off delay, conduction ramp, actual conduction voltage drop characteristic data, compensated temperature characteristic data, and thermal response time constant. The weights are statistically obtained based on the contribution of each feature to the failure result in historical degradation samples.

[0180] This represents the degradation state parameter corresponding to the k-th internal functional unit, used to characterize the current degradation level of the k-th internal functional unit; its acquisition method is the same as... The parameters are identical and are obtained by inversion from the composite transient feature set using the inverse state mapping model. Different values ​​of k correspond to the degradation state parameters of the input optocoupler unit, drive circuit unit, power output unit, and package interconnect structure, respectively.

[0181] The performance sensitivity corresponding to the i-th degradation state parameter is used to characterize the degree of response of the comprehensive performance index to changes in this degradation state parameter. It is calculated by the relationship between the changes in the comprehensive performance index and the changes in the corresponding degradation state parameter in historical aging samples. Keeping other degradation state parameters constant in the historical aging samples, the change in the comprehensive performance index caused by the change in the i-th degradation state parameter is statistically analyzed, and the performance sensitivity is calculated using a difference method. Preferably, the degradation state parameter Si is normalized, and its value ranges from 0 to 1; where 0 represents no degradation state, and 1 represents the state corresponding to reaching a preset failure threshold.

[0182] This represents the performance sensitivity of the overall performance index P relative to the degradation state parameter Sk of the k-th internal functional unit, used to characterize the degree of influence of changes in the degradation state of the k-th internal functional unit on changes in the overall performance index. Its calculation method is similar to... Similarly, it is obtained by examining the relationship between the changes in comprehensive performance indicators and the changes in corresponding degradation state parameters in historical aging samples.

[0183] To determine the number of internal functional units involved in the contribution calculation, this embodiment includes an input optocoupler unit, a drive circuit unit, a power output unit, and a packaged interconnect structure.

[0184] This represents the total degradation effect of all internal functional units on the overall performance index. It is used to normalize the degradation effect of each internal functional unit so that the sum of each degradation contribution Gi is 1, thereby reflecting the relative contribution ratio of different internal functional units to the overall performance degradation. When the denominator is zero, the degradation contribution of each internal functional unit is initialized to an equal value.

[0185] Subsequently, the degradation contribution of each internal functional unit is normalized, and the normalized degradation contribution is combined to form a multi-source degradation contribution set corresponding to the current monitoring period.

[0186] Preferably, a smoothing process is performed on the degradation contribution obtained from continuous monitoring periods to reduce the impact of short-term fluctuations on the contribution calculation results.

[0187] The final output is a set of multi-source degradation contribution values, which will be used as input data for subsequent degradation state evolution model establishment and lifetime prediction analysis.

[0188] Step 5: Establish a degradation state evolution model and perform dynamic state deduction using a solid-state relay digital twin model.

[0189] In this embodiment, the dominant degradation path at the current stage is determined based on the rate of change and cumulative contribution value of the degradation contribution of each internal functional unit within the continuous monitoring period; degradation evaluation index is calculated based on the degradation state vector and the set of multi-source degradation contribution; the degradation state vector obtained from the continuous monitoring period is used to construct a historical degradation state sequence, and a degradation state evolution model is established by combining the historical degradation state sequence.

[0190] Among them, the degradation state evolution model is used to describe the dynamic change process of the internal degradation state of solid relay over time. It not only reflects the current degradation state, but also predicts the degradation evolution trend in multiple future monitoring cycles.

[0191] First, a time-series analysis is performed on the multi-source degradation contribution set generated in step 4. Since the dominant position of different internal functional units may change in different aging stages, it is necessary to determine the dominant degradation path in the current stage based on the rate of change and cumulative contribution value of the corresponding degradation contribution of each internal functional unit within the continuous monitoring period.

[0192] Specifically, by analyzing the evolution of the degradation contribution of the input optocoupler unit, drive circuit unit, power output unit, and package interconnect structure during a continuous monitoring period, the evolution sequence of each contribution is obtained. Then, the rate of change and cumulative contribution value of each contribution evolution sequence are calculated, and the dominant degradation path of the current stage is determined based on the combined result of the rate of change and the cumulative contribution value.

[0193] Preferably, when the contribution of a certain internal functional unit continues to increase and the cumulative contribution value reaches the maximum of the current stage, the corresponding degradation path is determined as the dominant degradation path of the current stage.

[0194] After determining the dominant degradation path, degradation evaluation indices are calculated based on the degradation state vector and the set of multi-source degradation contributions. These indices are used to comprehensively characterize the current health status of the solid-state relay.

[0195] Specifically, the degradation status parameters of the input optocoupler, the degradation status parameters of the drive circuit, the degradation status parameters of the power output unit, and the degradation status parameters of the package interconnection structure are fused with the corresponding degradation contribution to obtain a degradation evaluation index that can reflect the overall degradation level.

[0196] Subsequently, the degradation state vectors obtained from continuous monitoring cycles are arranged in chronological order to form a historical degradation state sequence. The historical degradation state sequence records the degradation evolution process of the solid-state relay from its initial state to its current state, and can reflect the development law of degradation state of different internal functional units.

[0197] A degradation state evolution model is established based on historical degradation state sequences. The degradation state evolution model includes a digital twin model of solid-state relays.

[0198] Specifically, the digital twin model of the solid-state relay includes input optocoupler transmission characteristic parameters, drive circuit parameters, power output unit conduction parameters, and package thermal conduction parameters.

[0199] During operation, the solid-state relay digital twin model is corrected online using real-time acquired full-cycle transient response data, generating an updated degraded state vector. To improve the accuracy of state estimation, the solid-state relay digital twin model employs an unscented Kalman filter algorithm to fuse real-time observation data with predicted state data, dynamically updating each state parameter.

[0200] After obtaining the updated degradation state vector, the degradation state evolution model iteratively extrapolates the degradation state corresponding to multiple future prediction periods based on the updated degradation state vector.

[0201] Finally, the system outputs the degradation state sequence, degradation evaluation index change trend, and dominant degradation path evolution results for multiple future prediction periods, and uses these as the basic data for remaining lifetime prediction.

[0202] Step 6: Based on the degradation state evolution model, output the degradation state identification results, the dominant degradation path identification results, and the remaining lifetime prediction results.

[0203] In this embodiment, the future health status of the solid-state relay is assessed based on the degradation state sequence, degradation evaluation index change trend, and dominant degradation path evolution results corresponding to the future prediction period obtained in step 5.

[0204] Specifically, the predicted degradation state vectors for each future prediction period are read, and combined with degradation evaluation indicators and dominant degradation path information, the degradation development trend of each internal functional unit is obtained.

[0205] Furthermore, a correspondence is established between the degradation state and the failure boundary. When the predicted degradation state vector approaches the preset failure boundary, the distance index between the current state and the failure boundary is calculated.

[0206] Since the digital twin model of a solid-state relay can output the trajectory of its future degradation state, the distance decay rate between the current predicted state and the failure boundary is used as the basis for lifetime assessment.

[0207] Then, the failure distance function is constructed:

[0208] ;

[0209] The lifespan consumption rate is:

[0210] ;

[0211] Further, the remaining lifespan is obtained:

[0212] ;

[0213] This represents the predicted degradation state parameter corresponding to the i-th type of internal functional unit during the k-th prediction cycle; it is obtained by iterative deduction from the solid-state relay digital twin model combined with real-time observation data; after normalization, its value ranges from 0 to 1, where 0 represents no degradation state and 1 represents reaching the corresponding failure boundary state.

[0214] This represents the failure boundary state parameters corresponding to the i-th type of internal functional unit. These parameters are obtained through solid-state relay life tests. When a sample reaches the failure criteria, the corresponding degradation state parameters are recorded and statistically analyzed. The failure criteria include at least one of the following: on-state voltage drop exceeding a preset threshold, turn-on delay exceeding a preset threshold, turn-off delay exceeding a preset threshold, thermal response time constant exceeding a preset threshold, or the device failing to conduct normally. Statistical analysis is performed on the degradation state vectors corresponding to multiple life test samples reaching the failure criteria, and the failure boundary state vectors are determined using the mean, weighted mean, or probability quantile method.

[0215] M represents the total number of degradation state parameters involved in the lifetime assessment; corresponding to the number of degradation state parameters of the input optocoupler unit, drive circuit unit, power output unit, and package interconnect structure.

[0216] This represents the failure distance index corresponding to the k-th prediction period. It is calculated by the Euclidean distance between the predicted degradation state vector and the failure boundary state vector, and is used to characterize the remaining margin between the current state and the failure state. Since the degradation state vector contains degradation state parameters corresponding to multiple internal functional units, the Euclidean distance is used to comprehensively characterize the overall deviation between the current degradation state and the failure boundary state, thereby avoiding the one-sidedness of evaluation caused by relying solely on a single degradation state parameter for lifetime assessment.

[0217] n represents the prediction step size; the number of consecutive prediction cycles involved in the lifetime consumption rate calculation; determined by the prediction sequence output by the solid-state relay digital twin model.

[0218] The lifetime consumption rate is calculated based on the change in failure distance between the current prediction period and the next n prediction periods, and is used to characterize the speed at which the degradation state approaches the failure boundary.

[0219] Represents the stabilization coefficient; used to avoid calculation anomalies caused by excessively low lifetime attrition rates; determined based on historical lifetime attrition rate statistics; takes a preset minimum lifetime attrition rate threshold; when Less than At that time, with replace Participate in calculation

[0220] The remaining lifetime prediction result can be expressed as remaining time or remaining number of switching cycles.

[0221] Based on the future degradation trajectory and the evolution results of the dominant degradation path, the main sources of degradation in the future lifespan are identified; when the dominant degradation path shifts, the corresponding degradation mechanism determination results are updated simultaneously.

[0222] Final output:

[0223] (1) Degradation state identification results are used to characterize the degradation state of each internal functional unit in the current and future prediction cycles;

[0224] (2) The results of the identification of the dominant degradation path are used to characterize the main sources of degradation in the current and future stages;

[0225] (3) Remaining lifetime prediction results are used to characterize the remaining lifetime time or remaining number of switching cycles before reaching the preset failure threshold.

[0226] The above results can generate a life assessment report and output it to the monitoring platform or maintenance terminal to provide a basis for equipment maintenance decisions.

[0227] Example 2

[0228] This embodiment discloses a system for decoupling identification and lifetime prediction of multi-source degradation states of solid-state relays, including a contact voltage detection unit, a load current detection unit, a temperature detection unit, a data acquisition module, a feature processing module, a time causal chain construction module, a degradation state inversion module, a contribution analysis module, and a lifetime prediction module. The modules are connected via a data acquisition interface and an internal data processing link. The contact voltage detection unit, load current detection unit, and temperature detection unit respectively acquire electrical response data and thermal response data during the operation of the solid-state relay and transmit the acquired data to the data acquisition module. The data acquisition module synchronously acquires the multi-source detection signals, forming a basic data stream containing dynamic electrical characteristics and temperature characteristics. This data stream is then processed sequentially by the feature processing module, the time causal chain construction module, the degradation state inversion module, and the contribution analysis module. Finally, the lifetime prediction module outputs the degradation state identification result, the dominant degradation path, and the remaining lifetime prediction result.

[0229] In this embodiment, the contact voltage detection unit collects on-state voltage drop data, the load current detection unit collects load current data, and the temperature detection unit collects housing temperature data, and inputs the above data into the data acquisition module. In this embodiment, the preset temperature aging environment is not a fixed temperature value, but a temperature loading condition determined based on the material system, packaging structure, and target degradation stage of the solid-state relay under test.

[0230] Specifically, based on the rated operating temperature range, allowable junction temperature range, and high-temperature aging test requirements of the solid-state relay, the aging test temperature range is determined, and the target aging temperature is selected within the temperature range.

[0231] Preferably, the preset temperature is used to accelerate the degradation process of the input optocoupler unit, drive circuit unit, power output unit and package interconnection structure, so that the full-cycle transient response data corresponding to different degradation stages can be collected.

[0232] The data acquisition module receives on-state voltage drop data, load current data, and casing temperature data. Simultaneously, it acquires turn-on delay, turn-off delay, and conduction ramp data, and transmits this multi-source data to the feature processing module according to a unified time reference. The feature processing module calculates the thermal response time constant based on the casing temperature data changes, performs time synchronization processing on data from different sampling sources, and executes feature reconstruction processing to generate full-cycle transient response data. This processing transforms the multi-source data, which originally contained sampling time deviations and thermal coupling interference, into degradation characterization data with a unified temporal relationship, providing stable input for subsequent degradation state analysis.

[0233] The time-causal chain construction module receives the full-cycle transient response data output by the feature processing module and establishes a time-causal chain model based on the operational sequence of the input optocoupler unit, drive circuit unit, power output unit, and packaged interconnect structure. This module correlates and maps the external observation features corresponding to different internal functional units to generate a composite transient feature set, and transmits the temporal constraint information containing the operational relationships of the functional units to the degradation state inversion module.

[0234] The degradation state inversion module performs state inversion processing on the composite transient feature set based on the constraints provided by the time-causal chain model. It converts external observation features into a degradation state vector composed of input optocoupler degradation state parameters, drive circuit degradation state parameters, power output unit degradation state parameters, and package interconnect structure degradation state parameters. By utilizing time-causal constraints, it avoids state aliasing caused by multiple degradation sources simultaneously affecting the observed signal, thus improving the accuracy of internal degradation state identification.

[0235] The contribution analysis module receives the degradation state vector output by the degradation state inversion module, calculates the degradation contribution based on the degree of influence of the degradation state parameters of different internal functional units on the overall performance change, and determines the dominant degradation path in the current stage. This module converts the degradation contribution relationship of each internal functional unit into a multi-source degradation contribution set, and forms a historical degradation state sequence from the contribution change process, providing a basis for degradation evolution for the lifetime prediction module.

[0236] The lifetime prediction module receives the dominant degradation path and historical degradation state sequence output by the contribution analysis module, combines them with the degradation state evolution law to predict future degradation trends, and calculates the remaining lifetime prediction result of the solid-state relay before reaching the failure boundary. By associating the degradation state change process with the dominant degradation path, the lifetime prediction not only reflects the overall degradation degree, but also can locate the main sources of lifetime consumption.

[0237] In an optional configuration, the feature processing module not only fuses the electrical response data but also combines the housing temperature data collected by the temperature detection unit to eliminate the thermal coupling effects of adjacent channels, enabling subsequent degradation analysis to be based on the target solid-state relay's own thermal response characteristics. The time causal chain construction module further utilizes the operational sequence between the input optocoupler unit, drive circuit unit, power output unit, and package interconnection structure to transform the traditional detection method that relies solely on external performance changes into an internal degradation source decoupling analysis method.

[0238] Furthermore, the degradation state inversion module and the contribution analysis module form a continuous data processing link. By calculating the degradation contribution of different internal functional units through degradation state vectors, the system can distinguish the dominant degradation sources under different aging stages. The lifetime prediction module performs dynamic deduction based on this dominant degradation path, realizing a complete closed loop from "performance change detection" to "degradation mechanism identification" and then to "remaining lifetime prediction," improving the accuracy and interpretability of the solid-state relay reliability assessment process.

[0239] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them; although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications can still be made to the specific implementation methods of the present invention or equivalent substitutions can be made to some technical features without departing from the spirit of the technical solutions of the present invention, and all such modifications should be covered within the scope of the technical solutions claimed in the present invention.

Claims

1. A method for decoupling identification and lifetime prediction of multi-source degradation states of solid-state relays, characterized in that, Includes the following steps: Acquire the full-cycle transient response data of the solid-state relay during operation. The full-cycle transient response data includes turn-on delay, turn-off delay, conduction ramp, conduction voltage drop data, load current data, housing temperature data, and thermal response time constant. Time synchronization and feature reconstruction are performed on the full-cycle transient response data. A time causal chain model is established according to the working order of the input optocoupler unit, drive circuit unit, power output unit and package interconnection structure inside the solid-state relay, and the corresponding composite transient feature set is generated. The composite transient feature set is input into a pre-constructed inverse state mapping model, and the composite transient feature set is subjected to state inversion processing based on the time causal chain model to obtain the degraded state vectors corresponding to the input optocoupler unit, the drive circuit unit, the power output unit, and the package interconnection structure, respectively. The degradation contribution of each internal functional unit of the solid-state relay to external performance changes is calculated based on the degradation state vector, and a multi-source degradation contribution set is generated. Based on the rate of change and cumulative contribution of the degradation contribution of each internal functional unit within the continuous monitoring period, the dominant degradation path at the current stage is determined; degradation evaluation indicators are calculated based on the degradation state vector and the set of multi-source degradation contribution; the degradation state vectors obtained from the continuous monitoring period are used to construct a historical degradation state sequence, and a degradation state evolution model is established by combining the historical degradation state sequence. Based on the degradation state evolution model, the output solid-state relay degradation state identification results, dominant degradation path identification results, and remaining lifetime prediction results are presented.

2. The method for decoupling identification and lifetime prediction of multi-source degradation states of solid-state relays according to claim 1, characterized in that: Time synchronization processing includes uniform time reference correction for turn-on delay, turn-off delay, conduction ramp, conduction voltage drop data, shell temperature data, and thermal response time constant, and constructs a time causal chain node sequence based on the response time corresponding to each feature. The time-causal chain model generates a time-series correlation matrix between the internal functional units of a solid-state relay and the observed features corresponding to the full-cycle transient response data by recording the sequential action relationships between each node in the time-causal chain. The temporal correlation matrix serves as the input constraint for the inverse state mapping model.

3. The method for decoupling identification and lifetime prediction of multi-source degradation states of solid-state relays according to claim 1, characterized in that: The inverse state mapping model establishes a mapping relationship between the external observation feature space and the internal degradation state space by using historical monitoring samples corresponding to multiple different aging stages, and performs dimensionality reduction and reconstruction on the composite transient feature set; The inverse state mapping model consists of the degradation state parameters of the input-output optocoupler, the degradation state parameters of the drive circuit, the degradation state parameters of the power output unit, and the degradation state parameters of the package interconnect structure.

4. The method for decoupling identification and lifetime prediction of multi-source degradation states of solid-state relays according to claim 3, characterized in that: The inverse state mapping model employs a cross-domain attention mechanism; A feature weight matrix is ​​generated based on the degree of correlation between different observation features and different degradation state parameters; The composite transient feature set is weighted according to the feature weight matrix to obtain the decoupled feature set.

5. The method for decoupling identification and lifetime prediction of multi-source degradation states of solid-state relays according to claim 4, characterized in that: Manifold dimensionality reduction is performed on the decoupled feature set to construct a low-dimensional state manifold based on the neighborhood topological relationships in the sample space. Based on the low-dimensional state manifold, independent degradation trajectories corresponding to the input optocoupler unit, drive circuit unit, power output unit, and package interconnect structure are extracted; The degradation contribution of each internal functional unit is calculated based on the independent degradation trajectory, and the corresponding degradation contribution evolution sequence is formed according to the time sequence of the monitoring cycle, which serves as the basis for identifying the dominant degradation path. Calculate the rate of change and cumulative contribution value of each degradation contribution evolution sequence; The independent degradation trajectory with the largest contribution is determined based on the rate of change and the cumulative contribution value, and this independent degradation trajectory is identified as the dominant degradation path in the current stage.

6. The method for decoupling identification and lifetime prediction of multi-source degradation states of solid-state relays according to claim 1, characterized in that: Degradation state evolution models include digital twin models of solid-state relays; The digital twin model of a solid-state relay includes input optocoupler transmission characteristic parameters, drive circuit parameters, power output unit conduction parameters, and package thermal conduction parameters; The digital twin model of the solid-state relay is corrected online by collecting real-time full-cycle transient response data, and an updated degradation state vector is generated. The degradation state evolution model iteratively extrapolates the degradation state corresponding to multiple future prediction periods based on the updated degradation state vector.

7. The method for decoupling identification and lifetime prediction of multi-source degradation states of solid-state relays according to claim 6, characterized in that: The digital twin model of the solid-state relay uses the unscented Kalman filter algorithm for state estimation; By using the composite transient feature set collected and processed in the current monitoring cycle as the observation input, and fusing it with the predicted degradation state vector derived from the current degradation state vector by the solid-state relay digital twin model, the state parameters corresponding to each internal functional unit are updated. Output the predicted degenerate state vector for the next time step.

8. The method for decoupling identification and lifetime prediction of multi-source degradation states of solid-state relays according to claim 1, characterized in that: After acquiring the shell temperature data, thermal coupling compensation processing is performed on the shell temperature data; By establishing a model of the heat conduction hysteresis relationship between the target channel and adjacent channels, the heat conduction influence of adjacent channels is calculated. The collected shell temperature data is corrected based on the influence of heat conduction to obtain compensated temperature characteristic data.

9. The method for decoupling identification and lifetime prediction of multi-source degradation states of solid-state relays according to claim 1, characterized in that: Acquire the on-state voltage signal and load current signal, and perform load fingerprint recognition processing on the on-state voltage signal; The system collects load current timing data and conduction voltage timing data and constructs a load transient characteristic spectrum. Based on the load transient characteristic spectrum, it identifies inductive load characteristics, capacitive load characteristics, and resistive load characteristics, and determines the target sampling window according to the transient response process corresponding to different load types. Extract the true conduction voltage drop characteristic data within the target sampling window.

10. A system for executing the method for decoupling identification and lifetime prediction of multi-source degradation states of solid-state relays according to any one of claims 1 to 9, characterized in that, include: The contact voltage detection unit is used to collect the on-state voltage drop data of the solid-state relay under test; The load current detection unit is used to collect the load current data of the solid-state relay under test. Temperature detection unit, used to collect the housing temperature data of the solid-state relay under test; The data acquisition module is used to receive on-state voltage drop data, load current data, and housing temperature data, and to acquire the turn-on delay, turn-off delay, and conduction ramp data of the solid-state relay. The feature processing module is used to calculate the thermal response time constant based on the shell temperature data change process and generate full-cycle transient response data. It is also used to perform time synchronization processing and feature reconstruction processing on the full-cycle transient response data. The time causal chain construction module is used to construct a time causal chain model based on the operational sequence of the input optocoupler unit, drive circuit unit, power output unit, and package interconnection structure, and generate a composite transient feature set. The degradation state inversion module is used to perform state inversion processing on the composite transient feature set based on the time causal chain model, and generate degradation state vectors; The contribution analysis module is used to calculate the degradation contribution of each internal functional unit based on the degradation state vector and determine the dominant degradation path. The lifetime prediction module is used to generate remaining lifetime prediction results based on the dominant degradation path and historical degradation state sequence.

Images