A digital twin-based LED driver power supply fault prediction and management system
The LED driver power supply fault prediction and management system, which combines digital twin technology and narrowband frequency-locked filters, solves the problems of high cost, susceptibility to interference, and poor accuracy in existing EMI performance monitoring technologies. It achieves online monitoring and early warning throughout the entire life cycle, ensuring the reliability of EMC safety regulations.
Patent Information
- Application Number
- CN202611079076.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-20
- Publication Date
- 2026-11-10
AI Technical Summary
Existing EMI performance monitoring of LED driver power supplies relies on laboratory type testing and periodic offline spot checks, which cannot achieve online status monitoring and degradation early warning throughout the entire life cycle. Furthermore, existing online monitoring solutions are costly, susceptible to interference, have poor measurement accuracy, and cannot adapt to different power grid conditions and the characteristics of LED driver switching noise.
A fault prediction and management system for LED driver power supplies based on digital twins is adopted. Through a front-end narrowband frequency locking and noise floor acquisition module, a pure mechanism digital twin modeling module, a deterministic attenuation comparison and residual generation module, combined with static threshold early warning of component tolerance boundaries, the system can achieve accurate monitoring and early warning of EMI filters.
It enables low-cost, miniaturized online deployment, ensures the consistency and repeatability of measurement results, provides reliable EMC safety management, and can provide early warnings before filter degradation, avoiding economic losses caused by excessive conducted emissions.
Smart Images

Figure CN123066431A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of LED driver power supply and EMC safety monitoring technology, and in particular to an LED driver power supply fault prediction and management system based on digital twin. Background Technology
[0002] With the widespread adoption of LED lighting technology in municipal lighting, industrial lighting, and commercial complexes, existing EMI performance control methods rely entirely on laboratory type testing and periodic offline spot checks. This is a typical post-event control model, which cannot achieve online status monitoring and early warning of degradation throughout the product's entire lifecycle. Furthermore, the full-band EMI receivers used in traditional testing are bulky and expensive, making it impossible to integrate them into the board-level circuitry of LED driver power supplies, thus hindering large-scale online deployment of mass-produced products.
[0003] Currently, the few publicly available EMI online monitoring solutions generally adopt the technical route of broadband high-speed sampling and FFT spectrum analysis. In order to cover the full frequency band specified by the conducted emission standard, it is necessary to configure a high sampling rate AD chip of at least 60MS / s and a high computing power DSP processor. This not only significantly increases the hardware cost and operating power consumption, but also does not meet the low cost and miniaturization design requirements of LED driver power supplies. Furthermore, it is extremely susceptible to high-frequency switching interference from the main power circuit of the driver power supply, making it difficult to guarantee measurement stability and anti-interference capability.
[0004] Meanwhile, existing solutions mostly adopt the method of direct parallel sampling at the AC input terminal, which cannot shield the impact of large-scale fluctuations in the external power grid impedance on the measurement results. The measurement results under different application scenarios and different power grid conditions are extremely inconsistent. Furthermore, they have not been specifically optimized for the harmonic distribution characteristics of LED driver switch noise, and cannot accurately isolate the interference of normal operating conditions such as load fluctuations and input voltage changes, as well as the impact of component aging on the filter attenuation characteristics. The measurement accuracy cannot meet the requirements for long-term online monitoring.
[0005] Therefore, a fault prediction and management system for LED driver power supplies based on digital twins is proposed to address the aforementioned problems. Summary of the Invention
[0006] The purpose of this invention is to provide a digital twin-based LED driver power supply fault prediction and management system to solve the above-mentioned problems.
[0007] To achieve the above objectives, the present invention adopts the following technical solution: A digital twin-based LED driver power supply fault prediction and management system includes: The front-end narrowband frequency locking and noise floor acquisition module is configured to extract the common-mode and differential-mode noise from the AC input terminal of the LED driver power supply, convert it into the DC level of the noise energy at the corresponding frequency point, and obtain the real-time noise level at the target frequency point through sampling. The pure mechanism digital twin modeling module is configured to calculate the ideal attenuation of the target frequency under nominal conditions and the tolerance limit attenuation under the factory tolerance limit of the components respectively. The deterministic attenuation comparison and residual generation module is configured to input real-time noise level and ideal attenuation. After processing, it is combined with the initial reference noise level of the new state that is fixed when the equipment is first powered on to deduce the current actual attenuation. The attenuation residual after removing fluctuations under normal operating conditions is obtained by calculation. The static threshold warning module based on component tolerance boundaries is configured to calculate a fixed physical warning threshold based on the input tolerance limit attenuation and the ideal attenuation. The EMI filter safety failure warning signal is triggered to be output only when the attenuation residual of a consecutive preset number of samples is greater than the warning threshold.
[0008] Preferably, the miniaturized impedance stabilization network of the front-end narrowband frequency locking and noise floor acquisition module is built into the AC input terminal of the LED driver power supply. It adopts a simplified high-pass impedance matching topology and processes the common-mode noise current and differential-mode noise current of the AC input terminal in sequence through the DC blocking subunit, impedance matching subunit, and high-speed differential operation subunit. This process shields the influence of external power grid impedance fluctuations on noise measurement and ensures the repeatability of noise measurement results.
[0009] Preferably, the front-end narrowband frequency locking and noise floor acquisition module further includes: The hardware-level comb bandpass filter array contains several analog bandpass filters. The center frequency of each analog bandpass filter is strictly locked at a higher harmonic of a preset integer multiple of the switching frequency of the main switch of the LED driver power supply. No broadband sampling is performed; only narrowband filtering is performed on the noise at the corresponding frequency point. The high-frequency AC signal output after filtering is converted into a DC level representing the noise floor energy in the corresponding narrowband by the effective value envelope detector unit. Sampling is completed by the low sampling rate analog-to-digital converter unit, realizing the dimensionality reduction of the sampling rate requirement.
[0010] Preferably, in the pure mechanism digital twin modeling module, independent two-port models are established for the common-mode channel and the differential-mode channel respectively: Differential mode channel model: Component cascade sequence is: parallel Series common-mode inductor leakage inductance Ldm, parallel ; The ABCD matrix of each component is as follows: The ABCD matrix of the parallel capacitor is as follows: The ABCD matrix of the series inductor is By multiplying the ABCD matrices of the three components in concatenated order, the total ABCD matrix of the entire differential channel can be obtained; where, To open the path for transfer admittance; This is the short-circuit transfer impedance; Common-mode channel model: Component cascade order is: parallel Series common mode inductors, common mode inductance Lcm, parallel common mode inductors By using the same matrix concatenation method as the differential mode channel, the total ABCD matrix of the common mode channel can be obtained.
[0011] Preferably, the process of obtaining the ideal attenuation includes: Insertion loss can be calculated using the voltage ratio: ,in This is the noise voltage at the filter input. This represents the noise voltage at the filter output. A positive attenuation value indicates that the signal is attenuated. When the output of a two-port network is connected to a matching load At that time, the output voltage With input voltage The relationship is ,in This refers to the output port current phasor of a two-port network. For the A parameter of the ABCD transfer matrix; For the B parameter of the ABCD transfer matrix; Therefore, the transfer function , where A and B are elements of the two-port total ABCD matrix; angular frequency of the target frequency point Substitute into the transfer function, Calculate the magnitude of the transfer function for the target monitoring frequency. Thus, the ideal insertion loss under nominal conditions can be obtained: The negative sign is used to convert the attenuation amount into a positive value.
[0012] Preferably, the real-time attenuation calculation process is as follows: When the device is powered on for the first time, the initial noise reference value is read. : Continuous data collection for a preset duration Take the average value as the initial baseline value. ; For real-time data collection Perform condition compensation, including: Load compensation: The compensation coefficient is obtained through calibration before leaving the factory. ,in Rated output current, The output current is sampled in real time, and the compensated noise level is: ; Input voltage compensation: The compensation coefficient is obtained through calibration before leaving the factory. ,in The rated input voltage, The effective value of the AC input voltage is sampled in real time, and the compensated noise level is: ; Initially, the filter is in a completely new state, and the attenuation is equal to the ideal attenuation. The output noise at this time is The current output noise is the compensated noise. Therefore, the actual attenuation is: .
[0013] Preferably, the process of calculating the residual amount of attenuation after stripping normal operating condition fluctuations includes: The formula for calculating the residual is: ; The use of absolute values is to cover abnormal situations, including abnormally increased attenuation caused by component soldering errors or parameter offsets. In actual operation, filter aging will only lead to a decrease in attenuation. The residual is the decrease in the amount of decay.
[0014] Preferably, the tolerance extreme value calculation logic is as follows: Define the tolerance range of the components, select the limit value that results in the minimum attenuation among all component tolerances, combine them into the parameter set for the worst operating condition, and substitute it into the digital twin model of the pure mechanism digital twin modeling module. The calculated attenuation is the tolerance limit attenuation. .
[0015] Preferably, the step of triggering the EMI filter safety failure warning signal output when the attenuation residual of a consecutive preset number of samples is greater than the warning threshold specifically includes: Calculate the warning threshold: ; If and only if samples are taken consecutively for a preset number of times At that time, a precise safety regulation failure warning is triggered; After the warning is triggered, the MCU performs the following actions: First, the status indicator light for controlling the drive power supply flashes red to alert on-site maintenance personnel. Secondly, the warning information, residual values, and current operating parameters are uploaded to the background management system through the communication interface, and the system notifies the replacement of the filter or the entire drive power supply.
[0016] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. This invention reduces the sampling computing power requirement and hardware cost by combining hardware-level narrowband frequency-locked filtering with effective value envelope detection. It can be directly integrated into the control board of LED driver power supply to achieve online deployment throughout the product lifecycle. At the same time, through miniaturized impedance stabilization network design, it shields the interference of external power grid impedance fluctuations on the measurement results, ensuring the consistency and repeatability of measurement results under different application scenarios and different power grid conditions.
[0017] 2. This invention sets fixed warning thresholds based on the physical boundaries of component factory tolerances. Each warning judgment has a clear physical meaning and traceability, eliminating the risk of misjudgment and omission caused by manual experience thresholds. The warning results meet the mandatory requirements of EMC safety management for determinism and interpretability. It can provide early warnings before filter degradation leads to compliance risks, avoid economic and brand losses caused by excessive conducted emissions of products, and provide reliable technical support for the safety management of LED driver power supplies throughout their entire life cycle. Attached Figure Description
[0018] Further details, features, and advantages of this application are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which: Figure 1 This is a system structure diagram of the present invention. Detailed Implementation
[0019] Several embodiments of this application will now be described in more detail with reference to the accompanying drawings to enable those skilled in the art to implement this application. This application may be embodied in many different forms and for various purposes and should not be limited to the embodiments set forth herein. These embodiments are provided to make this application thorough and complete, and to fully convey the scope of this application to those skilled in the art. The embodiments described do not limit this application.
[0020] Unless otherwise defined, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It will be further understood that terms such as those defined in commonly used dictionaries shall be interpreted as having a meaning consistent with their meaning in the relevant field and / or the context of this specification, and shall not be interpreted in an idealized or overly formal sense unless expressly defined herein.
[0021] Example 1 Its specific implementation method is combined with the appendix Figure 1 Please provide a detailed explanation.
[0022] Appendix Figure 1This invention provides a block diagram of a fault prediction and management system for LED driver power supplies based on digital twins. The diagram shows the connection between the front-end narrowband frequency locking and noise floor acquisition module and the static threshold early warning module based on component tolerance boundaries, and marks the main functional interaction flow of each module.
[0023] In this embodiment, it includes: Module 1: Front-end narrowband frequency locking and noise floor acquisition module, configured with a miniaturized impedance stabilization network, a hardware-level comb bandpass filter array, an RMS envelope detector unit, and a low-speed AD sampling unit connected in sequence. The miniaturized impedance stabilization network extracts the common-mode and differential-mode noise from the AC input of the LED driver power supply. After narrowband filtering by the hardware-level comb bandpass filter array with the center frequency locked at the target high harmonic of the LED driver power supply main switch frequency, the noise energy is converted into DC level of the corresponding frequency point by the RMS envelope detector unit. The real-time noise level of the target frequency point is obtained by low-speed AD sampling and output to the deterministic attenuation comparison and residual generation module. This module is the physical measurement foundation of the entire system. Its core task is to accurately extract the output noise energy of a specific frequency point of the EMI filter without using an expensive full-band EMI receiver or consuming high computing resources. All circuits are integrated on the control board of the LED driver power supply and are completely electrically isolated from the main power circuit.
[0024] The output of this module is the DC level of noise energy corresponding to a specific harmonic frequency. It is directly input into module three for calculation. The coupling accuracy with the other three modules directly determines the measurement error of the entire system.
[0025] Details of the Mini-LISN (Mini-Impedance Stabilization Network): The core function of this section is to shield the measurement results from the impact of external power grid impedance fluctuations, ensuring repeatability of measurement results under different application scenarios and power grid conditions. This is a crucial prerequisite for achieving measurement consistency and is completely different from existing technologies that directly sample in parallel at the AC input. For the commonly used 220V / 50Hz AC input scenarios in China and the commonly used 110V / 60Hz scenarios overseas, the Mini-LISN adopts a simplified dual-branch high-pass impedance matching topology, completely abandoning the bulky power frequency inductors and load switching circuits of traditional benchtop LISNs, retaining only the impedance stabilization function at the target monitoring frequency. First, connect a 630VAC rated X7R material and 100nF safety DC blocking capacitor to the live and neutral wire input terminals of the AC input. This capacitor has an impedance of over 30MΩ to 50Hz / 60Hz power frequency signals, which can completely isolate the power frequency high voltage from entering the measurement circuit. At the same time, it has an impedance of less than 10Ω to high frequency noise signals above 300kHz, ensuring that the noise signal at the target frequency point enters the measurement branch without loss.
[0026] Each DC blocking capacitor is connected in series with a 50Ω high-frequency thin-film resistor with 1% accuracy, 25ppm / ℃ temperature coefficient, and 0402 package. This resistor is an impedance matching element for the target frequency point, ensuring that the input impedance of the entire measurement branch is stable at 50Ω±1% in the range of 300kHz~500kHz, unaffected by the external power grid impedance (which is usually 0.1Ω~100Ω fluctuation).
[0027] Common-mode and differential-mode noise separation is achieved using a high-speed differential operational amplifier with a bandwidth of 1GHz and a slew rate of 1000V / μs. The two 50Ω matching resistors are connected to the non-inverting and inverting inputs of the differential operational amplifier, respectively. The common-mode output of the operational amplifier is the sum of the two signals, corresponding to the common-mode noise voltage; the differential-mode output is the difference between the two signals, corresponding to the differential-mode noise voltage. The operational amplifier is powered by a 3.3V high-precision LDO. A 100nF NPO MLCC and a 1μF FX7R MLCC decoupling capacitor are connected in parallel next to the power supply pins. The wiring length of the decoupling capacitor from the pin does not exceed 1mm to prevent power supply noise from interfering with the measurement link.
[0028] Factory calibration procedure: After the entire unit is assembled, a 325kHz, 1Vpp common-mode standard signal is injected into the AC input using a standard signal generator. The 0.5% accuracy adjustable resistor (resistance range 1kΩ~10kΩ) of the operational amplifier feedback loop is adjusted so that the voltage at the common-mode output of the operational amplifier is 100mVpp. After calibration, the adjustable resistor is fixed with conformal coating to ensure that the calibration parameters will not deviate due to vibration or temperature changes. The differential-mode channel is calibrated using the same method by injecting a 325kHz, 1Vpp differential-mode standard signal. After calibration, the measurement linearity of the entire Mini-LISN is better than ±0.1dB in the 300kHz~500kHz range.
[0029] Wiring requirements: All signal lines of the Mini-LISN shall be routed using microstrip lines with a characteristic impedance of 50Ω. The line width shall be set to 0.2mm according to the PCB board thickness (usually 1.6mm). The routing path shall be free of vias and right-angle bends. The creepage distance from the main power circuit shall be no less than 8mm and the clearance shall be no less than 6mm to avoid high-frequency interference from the main power circuit from entering the measurement link.
[0030] Details of a hardware-level comb bandpass filter (Comb-BPF) array: This section differs from existing broadband sampling combined with FFT analysis schemes by using pure hardware filtering to achieve accurate selection of the target frequency, thereby reducing the computational power requirements for subsequent AD sampling from the root.
[0031] First, clarify the selection logic of the target monitoring frequency: The switching frequency of the main switching transistor (MOSFET) of the LED driver power supply is usually set to 60kHz~100kHz. More than 90% of the energy of EMI noise is concentrated at integer multiples of the switching frequency harmonics. At the same time, the attenuation characteristics of the low-frequency band (2nd to 4th harmonics, usually 120kHz~400kHz) of the EMI filter are mainly determined by the inductance of the common mode inductor. The aging rate of the common mode inductor is extremely low. After 10 years of operation, the inductance change is less than 5%, and the impact on the attenuation is less than 0.3dB. The attenuation characteristics of the 5th and higher harmonics are mainly determined by the high-frequency characteristics of the X and Y safety capacitors. The aging of safety capacitors (dielectric drying, capacitance reduction, and ESR increase) has a 3 to 5 times greater impact on the attenuation of the high-frequency band than the low-frequency band. Therefore, selecting the 5th and 7th harmonics as the target monitoring frequencies can maximize the sensitivity of aging detection.
[0032] In this embodiment, the switching frequency is used. For example, corresponding to the 5th harmonic 7th harmonic If the switching frequency is adjusted, only the center frequency of the filter needs to be adjusted accordingly, and the entire topology can be completely reused.
[0033] The comb-shaped bandpass filter array consists of two independent fourth-order Chebyshev bandpass filters, corresponding to the two target frequencies of 325kHz and 455kHz, respectively. The specific implementation details are as follows: Topology selection: A fourth-order Chebyshev bandpass topology is adopted, with a passband ripple set to 0.1dB and a -3dB bandwidth set to ±5kHz. This can cover the temperature drift range of ±1% of the switching frequency (the temperature drift of industrial-grade MOSFET switching frequency is usually no more than ±500Hz), and can suppress interference signals outside the passband by more than 40dB, ensuring that only noise energy at the target frequency can pass through the filter.
[0034] Component selection: All capacitors are NPO MLCCs with 1% accuracy and a temperature coefficient of ±30ppm / ℃, ensuring that the temperature drift of the center frequency does not exceed ±100Hz / ℃; all inductors are high-frequency wire-wound power inductors. Value ≥100@300kHz, precision 5%, avoid Too low a value leads to excessive passband insertion loss and decreased selectivity. Taking the 325kHz channel as an example, the first-stage resonant circuit consists of a 1nF NPO capacitor in series with a 220nH inductor and a 270pF NPO capacitor in parallel. The second-stage resonant circuit uses the exact same parameters as the first stage. The two stages are cascaded to achieve fourth-order filtering characteristics. For the 455kHz channel, the parameters of the resonant components can be adjusted accordingly, and the rest of the design is completely the same.
[0035] Layout and calibration: The inductors of each filter are placed vertically, with a minimum distance of 5mm between any two inductors. The PCB space beneath the inductors is completely cleared to prevent eddy current losses from affecting the inductors. The value decreases and the filtering characteristics shift. The input and output terminals of the filter are connected with 50Ω microstrip lines, which are completely impedance matched with the preceding Mini-LISN and subsequent RMS detection circuits to avoid measurement errors caused by signal reflection.
[0036] At the factory, a signal generator is used to calibrate the center frequency of the filter: a sweep signal corresponding to the input point is injected, and the 0.5pF~10pF ceramic adjustable capacitor connected in parallel in the first stage resonant circuit is adjusted so that the deviation between the lowest insertion loss point of the filter and the target frequency point does not exceed ±1kHz. After calibration, the adjustable capacitor is fixed with conformal coating.
[0037] The value is a dimensionless core parameter that describes the ratio of energy stored to energy consumed by resonant components such as inductors and capacitors. Essentially, it characterizes the energy loss characteristics of the component. The higher the value, the smaller the active power loss and the better the frequency selectivity of the component.
[0038] RMS envelope detection and AD sampling: The core function of this section is to convert the narrowband high-frequency AC signal output by the filter into a DC level, thereby reducing the dimensionality required for high sampling rates and completely eliminating the need for high-bandwidth, high-sampling-rate AD chips.
[0039] First, let's clarify the dimensionality reduction logic: Existing technologies require a sampling rate of at least 60MS / s to achieve sampling analysis of conducted emissions across the entire 150kHz~30MHz frequency band. If the radiated emission frequency band is covered, a sampling rate of several hundredMHz is required. This not only results in high cost and power consumption but also necessitates additional DSP resources for FFT calculations. In contrast, this solution uses a comb-shaped BPF to filter out all signals outside the target frequency, retaining only the AC signal at a single frequency. Therefore, it only needs to be converted to the corresponding DC level through RMS detection. Subsequent sampling only needs to be performed on the DC signal, reducing the sampling rate to 100Hz, a dimensionality reduction of more than six orders of magnitude.
[0040] The specific implementation details are as follows: The RMS detector circuit uses a dedicated high-frequency RMS detector chip with a bandwidth covering 100kHz / 1GHz, linearity of ±0.5dB, and an input range of 1mVrms / 1Vrms. The output-to-input RMS ratio is 1V / Vrms, meaning that a 100mVrms AC signal input will output a 100mV DC level. A 100nF NPO DC blocking capacitor is connected in series at the input port of the detector chip to prevent DC bias from the preceding stage from entering the detector chip and causing measurement errors. A 100nF MLCC and a 1μF tantalum capacitor are connected in parallel at the power supply port for decoupling. A 1μF tantalum capacitor and a 100nF MLCC are connected in parallel at the output port to form a low-pass filter network with a time constant of 10ms. This effectively filters out high-frequency ripple and keeps pace with changes in noise energy.
[0041] The AD sampling uses a 12-bit precision SAR ADC with a built-in 2.048V high-precision voltage reference, a temperature drift of 2ppm / ℃, and a sampling rate of 100Hz, meaning the DC level of the detector output is sampled every 10ms. A first-order RC low-pass filter with a cutoff frequency of 10Hz is set at the AD input front end to further filter out high-frequency interference on the line. The sampled data is preprocessed: after 10 consecutive samples, the maximum and minimum values are removed, and the average of the remaining 8 samples is taken as the noise level at the current frequency. This avoids measurement errors caused by transient pulse interference. Preprocessing It is stored directly in the MCU's RAM, waiting for module three to call and calculate.
[0042] Module 2: Pure Mechanism Digital Twin Modeling Module, configured as a two-port network analytical model that incorporates the nominal parameters and high-frequency parasitic parameters of the passive components of the EMI filter. It calculates the ideal attenuation at the target frequency under the nominal state and the tolerance limit attenuation under the factory tolerance limit of the components. The ideal attenuation is output to the deterministic attenuation comparison and residual generation module, and the two types of attenuation are output to the static threshold early warning module based on the component tolerance boundary. This module is the virtual mapping core of the entire system. It is built entirely on classical electromagnetics and circuit theory, without any black-box components such as data-driven, machine learning, or neural networks. All parameters come from the nominal values of the components. It can be used directly without any training or calibration. It runs on the 32-bit Cortex-M0 core MCU built into the LED driver and only needs to occupy less than 1KB of Flash space. The computing power consumption is negligible.
[0043] The output of this module is the ideal attenuation and tolerance limit attenuation at the target frequency. These are directly input into modules three and four for calculation, forming a virtual-real comparison coupling relationship with the physical measurement module.
[0044] Transmission line and two-port network modeling (ABCD matrix): The core of this section is to establish an analytical two-port model of the EMI filter and to realize the rapid calculation of the transfer function based on the cascade characteristics of the ABCD matrix, which is completely different from the equivalent circuit simulation and fitting modeling schemes of existing technologies.
[0045] First, let's clarify the standard topology of EMI filters: Mainstream LED driver front-end EMI filters all use a combination topology of two-stage X capacitors, common-mode inductors, and Y capacitors. The component order from the AC input terminal to the main power rectifier terminal is as follows: First-stage differential-mode X capacitor (Connected in parallel between the live and neutral wires), common-mode inductor Lcm (two windings connected in series with the live and neutral wires respectively), common-mode Y capacitor (Fire wire to ground) (Neutral wire to ground), second-stage differential mode X capacitor (Connected in parallel between the live wire and the neutral wire).
[0046] Since the transmission paths of common-mode noise and differential-mode noise are completely different, independent two-port models are established for the common-mode channel and the differential-mode channel respectively: Differential-mode channel model: Differential-mode noise is noise with out-of-phase connection between the live and neutral wires. The X capacitor attenuates differential-mode noise, the leakage inductance of the common-mode inductor attenuates differential-mode noise, and the Y capacitor has no effect on differential-mode noise. Therefore, the component cascading order for the differential-mode channel is: parallel connection. Series common-mode inductor leakage inductance Ldm, parallel .
[0047] The ABCD matrix of each component is as follows: The ABCD matrix of the parallel capacitor is as follows: The ABCD matrix of the series inductor is The total ABCD matrix of the entire differential channel can be obtained by multiplying the ABCD matrices of the three components in cascaded order (when the output of the previous stage is connected to the input of the next stage, the total matrix is the product of the previous stage matrix and the next stage matrix, and the order cannot be reversed).
[0048] For open-circuit transfer admittance, the complex frequency domain admittance of the capacitor characterizes the current transfer characteristics of the parallel capacitors and is a complex quantity; The short-circuit transfer impedance is the complex frequency domain impedance of the inductor, and the voltage transfer characteristic of the series inductor is a complex quantity.
[0049] Common-mode channel model: Common-mode noise is noise where the live and neutral wires are in phase with ground. Y capacitors attenuate common-mode noise, the common-mode inductance of the common-mode inductor attenuates common-mode noise, and X capacitors have no effect on common-mode noise. Therefore, the component cascading order for a common-mode channel is: parallel connection. Series common mode inductors, common mode inductance Lcm, parallel common mode inductors By using the same matrix concatenation method as the differential mode channel, the total ABCD matrix of the common mode channel can be obtained.
[0050] The nominal parameters of the EMI filter used in this embodiment for the 100W LED driver are: =220nFX2 safety capacitor, =100nFX2 safety capacitor, Lcm common mode inductance 10mH, leakage inductance Ldm=100uH = =2.2nFY1 safety capacitor, all parameters are from the nominal values in the component manufacturer's datasheet, can be directly entered into the model, no additional measurement is required.
[0051] Introduction of high-frequency parasitic parameters: First, we need to clarify the influence mechanism of parasitic parameters: the high-frequency characteristics of ideal capacitors and inductors differ significantly from those of actual components. Taking capacitor X as an example, its equivalent circuit is an ideal capacitor in series with equivalent series resistance (ESR) and equivalent series inductance (ESL). When the operating frequency is close to the self-resonant frequency (SRF) of the capacitor, the inductive reactance of ESL will cancel the capacitive reactance of the capacitor, resulting in an increase in the total impedance of the capacitor and a decrease in attenuation capability. If parasitic parameters are not considered, the model calculation error at the 325kHz frequency point will exceed 2dB, which is completely unable to meet the warning requirements.
[0052] This module writes the high-frequency parasitic parameters of all passive components into the model as fixed variables. All parameters are derived from the nominal values of the component manufacturers, eliminating the need for on-site measurement. Parasitic parameters of safety capacitors: The typical ESL of X2 safety capacitor is 5nH and the typical ESR is 10mΩ@100kHz. The typical ESL of Y1 safety capacitor is 2nH and the typical ESR is 5mΩ@1MHz. The temperature coefficients of capacitance, ESR and ESL are all less than 50ppm / ℃. The variation in the operating range of -40℃ to 125℃ does not exceed 5%, and the impact on the model is less than 0.2dB, which can be ignored.
[0053] Parasitic parameters of common mode inductors: The typical value of the parasitic capacitance of the winding of a common mode inductor is 10pF. This capacitance will cause the inductance of the common mode inductor to decrease in the high frequency range and reduce its attenuation capability. It is also written into the model as a fixed parameter.
[0054] The component model is updated after introducing parasitic parameters as follows: The equivalent circuit of the actual capacitor is a series ESR, a series ESL, and a parallel ideal capacitor. The ABCD matrices of the three components are cascaded to obtain the two-port matrix of the actual capacitor. The equivalent circuit of the actual common-mode inductor is a series ideal inductor and a parallel parasitic capacitor. The ABCD matrices of the two components are cascaded to obtain the two-port matrix of the actual inductor. Substituting the updated component matrices into the previous channel matrix yields the accurate two-port model considering parasitic parameters.
[0055] Detailed analysis of ideal attenuation calculation: The core of this section is to calculate the ideal insertion loss under nominal conditions using the transfer function of a two-port model.
[0056] First, let's define insertion loss: Insertion loss is the logarithm of the ratio of the power received at the load end after the filter is connected to the circuit to the power received at the load end when the filter is not connected, expressed in dB. Since the load impedance of the measurement circuit is fixed at 50Ω, the insertion loss can be calculated using the voltage ratio. ,in This is the noise voltage at the filter input. This represents the noise voltage at the filter output. A positive attenuation value indicates that the signal has been attenuated.
[0057] Derivation of the transfer function based on the two-port ABCD matrix: When the output of the two-port network is connected to a matched load When the resistance is 50Ω, the output voltage is... With input voltage The relationship is ,in This refers to the output port current phasor of a two-port network. For the A parameter of the ABCD transfer matrix; For the B parameter of the ABCD transfer matrix; Therefore, the transfer function , where A and B are elements of the two-port total ABCD matrix.
[0058] Calculation of ideal attenuation: The angular frequency of the target frequency point... Substitute the values into the transfer function and calculate the magnitude of the transfer function. Thus, the ideal insertion loss under nominal conditions can be obtained: ; The negative sign is used to convert the attenuation to a positive value. For example, when the transfer function magnitude is 0.1, the ideal attenuation is 20dB. 20 is the voltage-to-decibels conversion fixed coefficient (dimensionless, which is the standard coefficient for converting voltage ratios to decibels in circuit engineering). The frequency value of the target monitoring point; Two calculation methods are available: One method is to include all target frequencies before shipment. The calculations are pre-calculated and stored in the MCU's Flash memory, eliminating the need for real-time calculations and further reducing computing power requirements. Another method is to perform real-time calculations each time the computer is powered on. Since the calculations only involve complex number multiplication and modulo operations, a 32-bit Cortex-M0 core MCU can complete the calculations in just over 10 clock cycles. The calculation accuracy of the two methods is exactly the same.
[0059] In this embodiment, the ideal attenuation at the 325kHz frequency point is 20dB, and the ideal attenuation at the 455kHz frequency point is 25dB.
[0060] Module 3: Deterministic Attenuation Comparison and Residual Generation Module. It is configured to input real-time noise level and ideal attenuation. After processing to eliminate load and input voltage fluctuation interference through operating condition compensation, it derives the current actual attenuation by combining the initial reference noise level of the new state solidified when the equipment is first powered on. The attenuation residual after removing normal operating condition fluctuations is obtained by direct algebraic subtraction and output to the static threshold early warning module based on component tolerance boundary.
[0061] This module is the core logic layer connecting physical measurements and the virtual model. It is responsible for performing algebraic operations on the actual measurement results from Module 1 and the model output from Module 2, removing the influence of operating condition fluctuations, and obtaining the deterministic deviation purely caused by filter aging. The output of this module is the attenuation residual. The information is directly input into Module 4 to participate in the early warning determination.
[0062] Details of real-time attenuation projection: The core of this section is to derive the actual attenuation of the current filter by measuring the noise level in real time. The key premise is that the noise energy at the target frequency generated by the main power circuit remains stable throughout its entire lifespan. The switching noise of the main switch transistor mainly comes from the switching process. (The amplitude of current change per unit time) The voltage change per unit time is determined by parameters such as input bus voltage, main inductor inductance, MOSFET gate drive resistance, and output capacitor. The aging rate of these parameters over 10 years does not exceed 10%, and the corresponding noise energy change is less than 0.5dB. Compared with the attenuation change of more than 2dB caused by capacitor aging, it can be completely ignored. Therefore, there is no need to measure the noise at the input of the filter. The change in attenuation can be derived from the change in the noise at the output.
[0063] First, initial reference acquisition is performed: When the device is powered on for the first time (either the first power-on after factory installation or the first power-on after user installation), the MCU first reads the initial noise reference value stored in the EEPROM. If the value is empty (factory default), it is considered as the first power-on, and the reference acquisition process will start automatically. Reference acquisition conditions: The MCU controls the LED driver to work at 100% rated load. The load is judged by sampling the output current to determine whether it reaches more than 90% of the rated value. If it does not reach the rated value, the acquisition is delayed and the user is notified by the status indicator light. This ensures that the noise energy of the main power is at a stable maximum value during the acquisition. At the same time, the AC input voltage is sampled to ensure that the input voltage is within ±5% of the rated value to avoid the reference value deviation caused by the input voltage fluctuation.
[0064] Baseline acquisition process: Continuous acquisition for 1 minute Take the average value as the initial baseline value. It is stored in EEPROM and only needs to be sampled once during its entire life cycle, and it is not lost when power is off.
[0065] Then, real-time attenuation calculation is performed: During normal operation, in order to eliminate the influence of load fluctuations and input voltage fluctuations on the measurement results, the real-time acquired data is first... Perform working condition compensation: Load compensation: The noise energy of the main power is linearly related to the output current, and the compensation coefficient is obtained through calibration before leaving the factory. ,in Rated output current, The output current is sampled in real time, and the compensated noise level is: ; Input voltage compensation: The noise energy of the main power is linearly related to the DC bus voltage, and the DC bus voltage is directly proportional to the effective value of the AC input voltage. The compensation coefficient is obtained through calibration before leaving the factory. ,in The rated input voltage, The effective value of the AC input voltage is sampled in real time, and the compensated noise level is: ; After double compensation The operating fluctuation is less than 0.3dB, eliminating the influence of load and input voltage changes. The derivation logic for the current actual attenuation is as follows: In the initial state, the filter is in a completely new state, and the attenuation is equal to the ideal attenuation. The output noise at this time is The current output noise is the compensated noise. Therefore, the actual attenuation is: ; If the current noise is twice as high as the initial reference ( If the attenuation decreases by 6dB, it is consistent with the actual physical process.
[0066] Details of algebraic residual calculation: The core of this section is to obtain the aging deviation after removing all normal fluctuations, using a direct algebraic subtraction method.
[0067] The formula for calculating the residual is: ; The use of absolute values is to cover all abnormal situations, including abnormally increased attenuation due to component soldering errors or parameter offsets. In actual operation, filter aging will only lead to a decrease in attenuation. The residual is the decrease in the amount of decay.
[0068] The portion of the residual within 0.3dB is a normal deviation caused by operating condition fluctuations and measurement errors. The portion exceeding 0.3dB is entirely due to the aging and parameter degradation of the safety capacitor, without any other interfering factors. For example, when the capacitance of capacitor X decreases by 20%, the attenuation at the 325kHz frequency point will decrease by about 2dB, corresponding to a residual of 2dB, which is much larger than the normal fluctuation range and can accurately identify the aging state.
[0069] Module 4: Static threshold warning module based on component tolerance boundary, configured to calculate a fixed physical warning threshold based on the input tolerance limit attenuation and the ideal attenuation, and trigger the output of EMI filter safety failure warning signal if and only if the attenuation residual of a consecutive preset number of samples is greater than the warning threshold. This module is the final output of the entire system. Its core feature is that the warning threshold is set entirely based on the physical tolerance boundary of the components, without any human setting, statistical fitting, or probability judgment. The certainty of the warning result is extremely high, and the false judgment rate is less than one in a million.
[0070] Tolerance extreme value calculation: The core of this section is to determine the physical baseline of the filter attenuation, that is, the attenuation of all components under the worst possible limits of the factory tolerance, which is the safety margin baseline of the factory design.
[0071] First, clarify the tolerance range of the components: all electronic components have a clearly defined tolerance range when they leave the factory. This is the range of parameter fluctuations promised by the manufacturer. Components that exceed this range are unqualified products.
[0072] The component tolerances involved in this solution are as follows: X capacitor capacitance tolerance ±20%, ESR tolerance +50%, ESL tolerance +20%; Y capacitor capacitance tolerance ±10%, ESR tolerance +50%, ESL tolerance +20%; common mode inductor inductance tolerance ±5%, parasitic capacitance tolerance +30%.
[0073] The calculation logic for the tolerance extreme value is as follows: Select the limit value that results in the minimum attenuation among all component tolerances, combine them into the parameter set for the worst operating condition, substitute it into the digital twin model of Module 2, and the calculated attenuation is the tolerance limit attenuation. .
[0074] For example, the X capacitor capacitance is set to a lower limit of -20%, the ESR to an upper limit of +50%, and the ESL to an upper limit of +20%. The Y capacitor capacitance is set to a lower limit of -10%, and the common mode inductance is set to a lower limit of -5%. The filter attenuation under this parameter combination is the minimum value of all qualified new products. If the actual attenuation is lower than this value, it means that the filter parameters have exceeded the factory tolerance range and have deteriorated.
[0075] In this embodiment, after substituting the above limiting parameters into the model, the 325kHz frequency point is calculated. 455kHz frequency This value has been verified in practice: for example, 100 brand-new prototypes were selected, and after adjusting the component parameters to the tolerance limit, conducted emission tests were performed. The conducted emission values of all prototypes just met the standard limits of EN55015 and GB / T17743, which is the physical bottom line for safety compliance.
[0076] Once the tolerance extreme value is calculated, it is stored in the MCU's Flash memory and remains unchanged throughout its entire lifespan, requiring no subsequent adjustments.
[0077] Triggering judgment logic: The core of this section is to achieve deterministic early warning by comparing the residual with the threshold. There are no extra judgment conditions, and the logic is simple and reliable.
[0078] First, calculate the warning threshold: ; In this embodiment, Threshold=2dB for the 325kHz frequency point and Threshold=2dB for the 455kHz frequency point. Independent thresholds are used for the two frequency points. As long as the residual of either frequency point exceeds the threshold, an early warning will be triggered, further improving the detection coverage.
[0079] The warning trigger condition is: if and only if there are 10 consecutive samples (i.e., 10 consecutive seconds, with module 3 calculating the residual once every second). When the time is right, a precise safety regulation failure warning is triggered. The addition of continuous sampling is to avoid false triggering caused by abnormal conditions such as transient pulse interference and power grid flicker. Verification shows that the false judgment rate of this logic is less than one in a million.
[0080] After the warning is triggered, the MCU performs the following actions: First, it controls the status indicator light of the drive power supply to flash red to alert on-site maintenance personnel; second, it uploads the warning information, residual values, and current operating parameters to the background management system through built-in communication interfaces such as RS485, Bluetooth, and Zigbee, clearly informing that the EMI filter has aged and deteriorated, and the attenuation has exceeded the safety margin limit. At this time, the conducted emissions of the equipment are very likely to exceed the safety standard limits, and the filter or the entire drive power supply needs to be replaced in time to avoid compliance risks.
[0081] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.
[0082] It should be noted that, in this document, the use of relational terms such as "first" and "second" is merely for distinguishing one entity or operation from another, and does not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.
[0083] It should be understood that in the various embodiments of this application, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.
[0084] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0085] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0086] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0087] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.
[0088] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
[0089] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.
Claims
1. A fault prediction and management system for LED driver power supplies based on digital twins, characterized in that, include: The front-end narrowband frequency locking and noise floor acquisition module is configured to extract the common-mode and differential-mode noise from the AC input terminal of the LED driver power supply, convert it into the DC level of the noise energy at the corresponding frequency point, and obtain the real-time noise level at the target frequency point through sampling. The pure mechanism digital twin modeling module is configured to calculate the ideal attenuation of the target frequency under nominal conditions and the tolerance limit attenuation under the factory tolerance limit of the components respectively. The deterministic attenuation comparison and residual generation module is configured to input real-time noise level and ideal attenuation. After processing, it is combined with the initial reference noise level of the new state that is fixed when the equipment is first powered on to deduce the current actual attenuation. The attenuation residual after removing fluctuations under normal operating conditions is obtained by calculation. The static threshold warning module based on component tolerance boundaries is configured to calculate a fixed physical warning threshold based on the input tolerance limit attenuation and the ideal attenuation. The EMI filter safety failure warning signal is triggered to be output only when the attenuation residual of a consecutive preset number of samples is greater than the warning threshold.
2. The LED driver power supply fault prediction and management system based on digital twin according to claim 1, characterized in that, The miniaturized impedance stabilization network of the front-end narrowband frequency locking and noise floor acquisition module is built into the AC input terminal of the LED driver power supply. It adopts a simplified high-pass impedance matching topology and processes the common-mode noise current and differential-mode noise current of the AC input terminal in sequence through the DC blocking sub-unit, impedance matching sub-unit, and high-speed differential operation sub-unit. It shields the influence of external power grid impedance fluctuations on noise measurement and ensures the repeatability of noise measurement results.
3. The LED driver power supply fault prediction and management system based on digital twin according to claim 2, characterized in that, The front-end narrowband frequency locking and noise floor acquisition module also includes: The hardware-level comb bandpass filter array contains several analog bandpass filters. The center frequency of each analog bandpass filter is strictly locked at a higher harmonic of a preset integer multiple of the switching frequency of the main switch of the LED driver power supply. No broadband sampling is performed; only narrowband filtering is performed on the noise at the corresponding frequency point. The high-frequency AC signal output after filtering is converted into a DC level representing the noise floor energy in the corresponding narrowband by the effective value envelope detector unit. Sampling is completed by the low sampling rate analog-to-digital converter unit, realizing the dimensionality reduction of the sampling rate requirement.
4. The LED driver power supply fault prediction and management system based on digital twin according to claim 1, characterized in that, In the pure mechanism digital twin modeling module, independent two-port models are established for the common-mode channel and the differential-mode channel respectively: Differential mode channel model: Component cascade sequence is: parallel Series common-mode inductor leakage inductance Ldm, parallel ; The ABCD matrix of each component is as follows: The ABCD matrix of the parallel capacitor is as follows: The ABCD matrix of the series inductor is By multiplying the ABCD matrices of the three components in concatenated order, the total ABCD matrix of the entire differential channel can be obtained; where, To open the path for transfer admittance; This is the short-circuit transfer impedance; Common-mode channel model: Component cascade order is: parallel Series common mode inductors, common mode inductance Lcm, parallel common mode inductors By using the same matrix concatenation method as the differential mode channel, the total ABCD matrix of the common mode channel can be obtained.
5. The LED driver power supply fault prediction and management system based on digital twin according to claim 4, characterized in that, The process of obtaining the ideal attenuation includes: Insertion loss can be calculated using the voltage ratio: ,in This is the noise voltage at the filter input. This represents the noise voltage at the filter output. A positive attenuation value indicates that the signal is attenuated. When the output of a two-port network is connected to a matching load At that time, the output voltage With input voltage The relationship is ,in This refers to the output port current phasor of a two-port network. For the A parameter of the ABCD transfer matrix; For the B parameter of the ABCD transfer matrix; Therefore, the transfer function , where A and B are elements of the two-port total ABCD matrix; angular frequency of the target frequency point Substitute into the transfer function, Calculate the magnitude of the transfer function for the target monitoring frequency. Thus, the ideal insertion loss under nominal conditions can be obtained: The negative sign is used to convert the attenuation amount into a positive value.
6. The LED driver power supply fault prediction and management system based on digital twin according to claim 1, characterized in that, The real-time attenuation calculation process is as follows: When the device is powered on for the first time, the initial noise reference value is read. : Continuous data collection for a preset duration Take the average value as the initial baseline value. ; For real-time data collection Perform condition compensation, including: Load compensation: The compensation coefficient is obtained through calibration before leaving the factory. ,in Rated output current, The output current is sampled in real time, and the compensated noise level is: ; Input voltage compensation: The compensation coefficient is obtained through calibration before leaving the factory. ,in The rated input voltage, The effective value of the AC input voltage is sampled in real time, and the compensated noise level is: ; Initially, the filter is in a completely new state, and the attenuation is equal to the ideal attenuation. The output noise at this time is The current output noise is the compensated noise. Therefore, the actual attenuation is: .
7. The LED driver power supply fault prediction and management system based on digital twin according to claim 6, characterized in that, The process of calculating the residual amount of attenuation after stripping normal operating condition fluctuations includes: The formula for calculating the residual is: ; The use of absolute values is to cover abnormal situations, including abnormally increased attenuation caused by component soldering errors or parameter offsets. In actual operation, filter aging will only lead to a decrease in attenuation. The residual is the decrease in the amount of decay.
8. The LED driver power supply fault prediction and management system based on digital twin according to claim 1, characterized in that, The logic for calculating tolerance extrema is as follows: Define the tolerance range of the components, select the limit value that results in the minimum attenuation among all component tolerances, combine them into the parameter set for the worst operating condition, and substitute it into the digital twin model of the pure mechanism digital twin modeling module. The calculated attenuation is the tolerance limit attenuation. .
9. A fault prediction and management system for LED driver power supplies based on digital twins according to claim 8, characterized in that, An EMI filter safety failure warning signal is triggered only when the attenuation residuals from a consecutive preset number of samples are all greater than the warning threshold. Specifically, this includes: Calculate the warning threshold: ; If and only if samples are taken consecutively for a preset number of times At that time, a precise safety regulation failure warning is triggered; After the warning is triggered, the MCU performs the following actions: First, the status indicator light for controlling the drive power supply flashes red to alert on-site maintenance personnel. Secondly, the warning information, residual values, and current operating parameters are uploaded to the background management system through the communication interface, and the system notifies the replacement of the filter or the entire drive power supply.
