Adaptive dual-frequency bandpass filter based on intelligent algorithm

Abstract

The present invention relates to the field of filter technology, and more specifically, to an adaptive dual-frequency bandpass filter based on an intelligent algorithm, comprising: a main frame, wherein an independent sensor cabin is arranged in the main frame; a plurality of sensors arranged in the sensor cabin; a core filter structure arranged in the main frame, an impedance tuning module arranged at the input and output ends of the main frame, the impedance tuning module being used to adjust the characteristics of a first passband and a second passband; an impedance analyzer chip arranged in the main frame; and a main control unit electrically connected to the sensor cabin, the core filter structure, the impedance tuning module and the impedance analyzer chip; the main control unit is used to execute a multi-source data fusion algorithm, adjust the core filter structure through the impedance tuning module according to the result of the multi-source data fusion algorithm, and adaptively adjust the first passband and the second passband, and through the multi-source data fusion algorithm and the real-time adjustment mechanism, the filter can dynamically adjust its performance according to environmental changes and signal characteristics.

Description

Technical Field

[0001] The present invention relates to the technical field of filters, and more specifically, to an adaptive dual-frequency bandpass filter based on an intelligent algorithm. Background Art

[0002] With the rapid development of modern radar technology, multi-band radar systems have been widely used in military and civilian fields. These systems usually need to process signals in different frequency bands at the same time, such as GPS L1 band (1.575GHz) for precise positioning and WiMAX band (3.5GHz) for high-speed data transmission. Therefore, high-performance dual-band bandpass filters (DBBPFs) have become key components in multi-band radar systems.

[0003] In recent years, researchers have proposed a variety of methods to improve the performance of DBBPF. Among them, a common method is to improve the performance by optimizing the physical structure of the filter. For example, using structures such as stepped impedance resonators, mesh resonators, or branch-loaded resonators. Another method is to introduce intelligent algorithms to optimize the filter design process. For example, using genetic algorithms or particle swarm optimization algorithms to find the optimal filter parameters.

[0004] However, these existing technologies still have some significant problems:

[0005] 1. Lack of real-time adaptive capability: Most existing DBBPF designs are static and cannot adjust the filtering characteristics in real time according to environmental changes and signal characteristics. In a complex and changing electromagnetic environment, this may cause the filter performance to degrade significantly.

[0006] 2. Insufficient environmental perception: Existing filter designs usually lack the ability to perceive the surrounding environment. They cannot obtain environmental parameters such as temperature, humidity, vibration, etc. in real time, and therefore cannot make timely adjustments to environmental changes.

[0007] 3. Insufficient fusion of multi-source data: Although some designs have introduced intelligent algorithms, most algorithms are optimized based on only a single data source, ignoring the complex interactions between multiple environmental factors and system parameters.

[0008] 4. Imperfect feedback mechanism: Existing intelligent optimization methods are often offline and lack real-time feedback mechanisms. This makes it impossible for the filter to be dynamically adjusted according to the actual working status.

[0009] 5. High algorithm complexity: Although some intelligent optimization algorithms can improve filter performance, their computational complexity is high and it is difficult to run in real time in resource-constrained embedded systems.

[0010] 6. Reliability and stability issues: Under long-term operation and harsh environments, existing filters lack self-diagnosis and self-repair capabilities and are prone to performance degradation or failure.

[0011] In view of the above problems, the present invention proposes an adaptive dual-frequency bandpass filter based on an intelligent algorithm. Summary of the invention

[0012] In view of the above problems existing in the prior art, the present invention proposes an adaptive dual-frequency bandpass filter based on an intelligent algorithm, the core of which is to realize real-time adaptive adjustment of the filter characteristics through a multi-source data fusion algorithm to adapt to the complex and changeable electromagnetic environment.

[0013] The present invention provides an adaptive dual-frequency bandpass filter based on an intelligent algorithm, comprising: a main frame, wherein an independent sensor cabin is arranged in the main frame; a plurality of sensors arranged in the sensor cabin; a core filter structure arranged in the main frame, wherein the core filter structure comprises two pairs of quarter-wavelength λ / 4 ladder impedance resonators and a branch-loaded three-mode resonator, which are used to form a first passband and a second passband; an impedance tuning module arranged at the input and output ends of the main frame, wherein the impedance tuning module is used to adjust the characteristics of the first passband and the second passband; an impedance analyzer chip arranged in the main frame; and a main control unit electrically connected to the sensor cabin, the core filter structure, the impedance tuning module and the impedance analyzer chip; wherein the main control unit is used to execute a multi-source data fusion algorithm, and adjust the core filter structure through the impedance tuning module according to the result of the multi-source data fusion algorithm to achieve adaptive adjustment of the first passband and the second passband.

[0014] Specifically, the multiple sensors include a temperature sensor array, a power detector, a standing wave ratio monitoring circuit, an environmental humidity sensor and a vibration sensor, and the multiple sensors are electrically connected to the main control unit through a high-speed digital interface.

[0015] Specifically, the impedance tuning module includes a programmable waveform generator, a digital variable capacitor array and a variable inductor, and the impedance tuning module is electrically connected to the main control unit through an I2C bus.

[0016] Specifically, the impedance analyzer chip is electrically connected to the main control unit via an SPI interface.

[0017] Specifically, the multi-source data fusion algorithm includes the following steps which are performed in sequence: performing data preprocessing on the raw data collected by the multiple sensors; performing sensor reliability assessment based on the preprocessed data; performing adaptive Kalman filtering using the preprocessed data and the sensor reliability assessment result; performing adaptive weight allocation according to the sensor reliability assessment result; performing multi-source data fusion based on the adaptive Kalman filtering result and the adaptive weight; outputting the result of the multi-source data fusion and feeding it back to the impedance tuning module to adjust the characteristics of the first passband and the second passband.

[0018] Specifically, the adaptive Kalman filtering step further includes the following sub-steps executed in sequence: performing non-Gaussian noise processing on the pre-processed data; performing adaptive gain adjustment based on the result of the non-Gaussian noise processing; performing abnormal data detection and processing based on the adaptive gain adjustment; wherein the result of the non-Gaussian noise processing is used to optimize subsequent adaptive gain adjustment and abnormal data detection and processing sub-steps.

[0019] Specifically, the non-Gaussian noise processing comprises the following steps:

[0020] First, observe the data sequence for kernel density estimation:

[0021] ,

[0022] in, is the estimated probability density function, is the sample size, is the bandwidth parameter, is the kernel function, For the Sample points, is the point to be estimated;

[0023] Then based on the estimated noise distribution, the variance of the equivalent Gaussian noise is calculated:

[0024] ,

[0025] in, is the covariance of the equivalent Gaussian noise, is the observed value, is the observation matrix, is the state vector;

[0026] Finally, the Substitute into the standard Kalman filter formula.

[0027] Specifically, the adaptive gain adjustment comprises the following steps:

[0028] First calculate the innovation series covariance:

[0029] ,

[0030] in, is the innovation sequence covariance, is the sliding window size, For the innovation sequence, , is the observed value, is the observation matrix, is the prior state estimate;

[0031] Then calculate the theoretical innovation series covariance:

[0032] ,

[0033] in, is the covariance of the theoretical innovation sequence, is the prior estimate error covariance, is the observed noise covariance, k is the current moment, is the observation matrix;

[0034] Next, define the adaptive factor:

[0035] ,

[0036] in, is the adaptive factor, tr[] is the trace of the matrix;

[0037] Finally adjust the Kalman gain:

[0038] ,

[0039] in, is the adjusted Kalman gain.

[0040] Specifically, the abnormal data detection and processing includes the following steps:

[0041] First, calculate the Mahalanobis distance between the observed and predicted values:

[0042] ,

[0043] in, is the square of the Mahalanobis distance, is the covariance of the theoretical innovation sequence, , the superscript (-1) indicates the matrix inversion, is the prior state estimate at time k, is the prior estimate error covariance, is the covariance of the equivalent Gaussian noise, k is the current moment, is the observation matrix, for The observed value at time;

[0044] Then set the threshold ,when When , it is judged as abnormal data;

[0045] Then, for abnormal data, adjust the observation noise covariance:

[0046] ,

[0047] Finally, use the adjusted Perform Kalman filter update.

[0048] Specifically, the adaptive weight allocation step includes:

[0049] Based on the sensor reliability evaluation results, the Softmax function is used to calculate the weight of each sensor:

[0050] ,

[0051] in, For the The weight of the sensor, For the The reliability score of each sensor, is the natural exponential function, is the summation symbol, which means the sum of all sensors Perform summation; The index of the sensor, ranging from 1 to the total number of sensors.

[0052] The adaptive dual-frequency bandpass filter based on intelligent algorithm proposed in the present invention has the following significant beneficial effects:

[0053] 1. Real-time adaptive capability: Through multi-source data fusion algorithm and real-time adjustment mechanism, the filter can dynamically adjust its performance according to environmental changes and signal characteristics to maintain the best working state.

[0054] 2. Enhanced environmental perception: Multiple sensors in the independent sensor cabin can monitor environmental parameters such as temperature, humidity, vibration, etc. in real time, providing comprehensive data support for adaptive algorithms.

[0055] 3. Efficient multi-source data fusion: The multi-source data fusion algorithm executed by the main control unit can comprehensively consider various environmental factors and system parameters to achieve more accurate performance optimization.

[0056] 4. Perfect feedback mechanism: Through the impedance analyzer chip and impedance tuning module, the filter can analyze its working status in real time and make corresponding adjustments to form a closed-loop control.

[0057] 5. Low-complexity algorithm: It adopts an optimized multi-source data fusion algorithm to reduce the computational complexity while ensuring performance, making it suitable for running in embedded systems.

[0058] 6. Improve reliability and stability: Through continuous self-monitoring and adjustment, the filter can maintain stable performance in long-term operation and harsh environments, and has certain self-diagnosis and repair capabilities.

[0059] 7. Flexibility and versatility: This design method is not only applicable to dual-frequency bandpass filters, but can also be extended to other types of tunable filters, and has broad application prospects.

[0060] In summary, the adaptive dual-frequency bandpass filter of the present invention effectively solves the key problems existing in the prior art, such as real-time adaptation, environmental perception, and data fusion, by introducing intelligent algorithms and multi-source data fusion technology. This innovative design provides a highly intelligent, high-performance filtering solution for modern multi-band radar systems, and is expected to play an important role in various advanced radar applications in the military and civilian fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 The schematic diagram of the framework of the adaptive dual-frequency bandpass filter based on the intelligent algorithm of the present invention;

[0062] Figure 2 It is a flow chart of the multi-source data fusion algorithm of the present invention;

[0063] Figure 3 A circuit structure diagram including geometric parameters of the core filter structure of the present invention;

[0064] Figure 4 The coupled topology of the core filter structure of the present invention, the solid line and the dotted line represent direct coupling and cross coupling respectively;

[0065] Figure 5 A structural diagram of a branch-loaded three-mode resonator for the core filter structure of the present invention;

[0066] Figure 6 An odd-mode equivalent circuit of a three-mode resonator is loaded on the branches of the core filter structure of the present invention;

[0067] Figure 7 The even-mode equivalent circuit of the core filter structure branch of the present invention is loaded with a three-mode resonator. DETAILED DESCRIPTION

[0068] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects thereof are described in detail below in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0069] Unless defined otherwise, 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 belongs.

[0070] With the rapid development of modern radar technology, multi-band radar systems have been widely used in military and civilian fields. These systems usually need to process signals in different frequency bands at the same time, such as GPS L1 band for precise positioning and WiMAX band for high-speed data transmission. Therefore, high-performance dual-band bandpass filters (DBBPFs) have become key components in multi-band radar systems.

[0071] The present invention provides an adaptive dual-frequency bandpass filter based on an intelligent algorithm, the core of which is to achieve real-time adaptive adjustment of filter characteristics through a multi-source data fusion algorithm to adapt to complex and changeable electromagnetic environments.

[0072] like Figure 1 As shown, the adaptive dual-frequency bandpass filter based on the intelligent algorithm provided by the present invention includes:

[0073] A main frame 10, wherein an independent sensor cabin 11 is provided in the main frame 10;

[0074] A plurality of sensors disposed in the sensor cabin 11;

[0075] A core filter structure 12 disposed in the main frame 10, the core filter structure 12 being used to form a first passband and a second passband;

[0076] An impedance tuning module 13 is provided at the input and output ends of the main frame 10, and the impedance tuning module 13 is used to adjust the characteristics of the first passband and the second passband;

[0077] an impedance analyzer chip 14 disposed in the main frame 10 ; and a main control unit 15 electrically connected to the sensor cabin 11 , the core filter structure 12 , the impedance tuning module 13 and the impedance analyzer chip 14 .

[0078] Preferably, if Figure 3-7As shown, the core filter structure 12 of the present invention includes two pairs of quarter-wavelength λ / 4 stepped impedance resonators (SSIRs) and a stub-loaded triple-mode resonator (STR). Among them, the two pairs of λ / 4 SSIRs are used to form the first passband (GPS L1 band, 1.575 GHz), and the STR is used to form the second passband (WiMAX band, 3.5 GHz). Specifically, the two pairs of λ / 4 SSIRs include: the first resonator R1 and the fourth resonator R4, which form the first pair of resonators; the second resonator R2 and the third resonator R3, which form the second pair of resonators. The coupling between R1 and R4, and between R2 and R3 forms the first passband in the GPS band.

[0079] Furthermore, the STR includes an odd mode o1 and two even modes e1 and e2, satisfying the frequency relationship: fe1 < fo1 < fe2. This design ensures that the WiMAX band has an appropriate bandwidth and a steep edge, achieving high-selectivity filtering. It should be noted that R1 and R4 not only participate in the formation of the GPS band, but also provide a feeding structure and source-load coupling for the WiMAX band, realizing the multifunctionality of the resonator and further reducing the circuit size.

[0080] Preferably, the present invention uses a Rogers5880 substrate to design the core filter structure 12. The thickness of this substrate is 0.508 mm, and the relative dielectric constant is 2.2. The normalized circuit size of the filter is 0.08λg × 0.32λg, where λg is the waveguide wavelength at the center frequency of the GPS band (1.575 GHz). This design not only realizes the miniaturization of the filter, but also ensures excellent high-frequency performance and temperature stability.

[0081] The core filter structure 12 of the present invention also includes the following key geometric parameters (unit: mm):

[0082] L1 = 4.44, L2 = 4.52, L3 = 21.82, L4 = 11.44, L5 = 6.29, L6 = 4.92, L7 = 32.03, L8 = 0.77, L9 = 1.7, L 10 = 5.5, L 11 = 5.9, W0 = 1.55, W1 = 0.4, W2 = 1.23, W3 = 0.5, W4 = 3.1, W5 = 1.2, W6 = 0.5, W7 = 1.49, W8 = 1.2, S1 = 0.29, S2 = 0.40, S3 = 0.12, S4 = 0.13, d1 = 2.22, d2 = 0.93.

[0083] For the first pair The length of the high impedance section of the stepped impedance resonator, For the second pair The length of the high impedance section of the stepped impedance resonator, For the first pair The length of the low impedance section of the stepped impedance resonator, For the second pair The length of the low impedance section of the stepped impedance resonator, The main transmission line length for the branch-loaded three-mode resonator, The length of the auxiliary transmission line for loading the three-mode resonator of the branch, The open-circuit stub length of the stub-loaded three-mode resonator is The short-circuit stub length of the stub-loaded three-mode resonator is - Detailed dimensions of the open-circuit branch for branch loading of the three-mode resonator; is the input / output port width, for The width of the high impedance section of the stepped impedance resonator, for The width of the low impedance section of the stepped impedance resonator, - The transmission line width of the branch-loaded three-mode resonator is - is the branch length of the branch-loaded three-mode resonator, W8 represents the width of the short-circuited branch in the branch-loaded three-mode resonator, and for The coupling gap between the step impedance resonators, and Load the branch with a three-mode resonator and The coupling gap between the step impedance resonators, and is the position parameter of the through hole;

[0084] is the electrical length of the transmission line; is the even-mode input admittance, is the characteristic admittance of the transmission line; is the odd-mode input admittance;

[0085] Precise control of these parameters is critical to achieving high performance of the filter. For example, L3 and L4 affect the center frequency and bandwidth of the GPS band, L7 affects the center frequency of the WiMAX band, S1 and S2 control the coupling strength of the GPS band, and S3 and S4 control the coupling strength of the WiMAX band. In particular, the larger width of W4 (3.1 mm) improves the power handling capability of the filter, making it suitable for high-power radar systems.

[0086] Preferably, in one embodiment of the present invention, the main frame 10 is made of aluminum alloy 6061-T6 material, which has good thermal conductivity and corrosion resistance. The main frame 10 has a size of 200mm × 150mm × 50mm (length × width × height), and an independent sensor cabin 11 is provided inside, and the size of the sensor cabin 11 is 100mm × 80mm × 30mm. This compact design is conducive to the integration of filters in space-constrained radar systems.

[0087] Furthermore, the sensor cabin adopts a modular design and is equipped with 5 standardized sensor slots (50-pin ZIF interface). Each slot is connected to the main control unit 15 via a high-speed digital interface (LVDS, up to 2.5Gbps). This design scheme not only improves the flexibility and maintainability of the system, but also ensures high-speed and stable data transmission.

[0088] Specifically, the plurality of sensors 12 include a temperature sensor array, a power detector, a standing wave ratio monitoring circuit, an environmental humidity sensor, and a vibration sensor;

[0089] The temperature sensor array uses 4 × Texas Instruments TMP117 with an accuracy of ±0.1°C; the power detector model is Analog Devices ADL5519 with a dynamic range of 60dB; the standing wave ratio monitoring circuit uses Analog Devices AD8302 with an accuracy of ±0.5dB; the environmental humidity sensor uses Sensirion SHT85 with an accuracy of ±1.5%RH; the vibration sensor uses InvenSense MPU-6050, a 3-axis accelerometer and gyroscope.

[0090] The configuration of these high-precision sensors enables the filter to fully perceive its working environment and provide accurate input data for subsequent intelligent algorithms.

[0091] Preferably, in another embodiment of the present invention, the impedance tuning module 13 includes an Analog Devices AD5930 programmable waveform generator, a Peregrine Semiconductor PE64904 integrated digital variable capacitor array, and a Coilcraft 0603CS series variable inductor. The module communicates with the main control unit 15 via an I2C bus with a clock frequency of 400kHz. This configuration allows the filter to achieve fast and flexible impedance adjustment while maintaining high accuracy.

[0092] Furthermore, the impedance analyzer chip 14 uses Analog Devices AD5933, with a frequency range of 1kHz-100MHz, an impedance measurement error of <0.5%, and a phase measurement error of <0.5°. This high-precision impedance analysis capability provides reliable basic data for the adaptive adjustment of the filter.

[0093] Next, the main control unit 15 includes a processor, memory and interface. The processor adopts ARM Cortex-M7 core, STM32H753VIT6, and the main frequency is 480MHz; the memory is 1MB Flash, 1MB SRAM; the interface includes 5 LVDS interfaces for sensor communication; I2C interface for tuning module control; SPI interface for impedance analysis; USB2.0 interface for external communication.

[0094] This high-performance main control unit 15 configuration ensures efficient execution of complex algorithms and seamless collaboration of all parts of the system.

[0095] Please refer to Figure 2 Based on the above hardware structure, the core of the present invention lies in the multi-source data fusion algorithm executed by the main control unit 15. The algorithm includes the following steps executed in sequence:

[0096] The first step is data preprocessing. In this step, a moving average filter is applied to the raw data of each sensor with a window size of 5. Then, the Min-Max normalization method is used:

[0097] ,

[0098] in, and Historical minimum and maximum values ​​for each sensor.

[0099] This preprocessing method can not only effectively remove high-frequency noise, but also unify the data from different sensors to the same scale, laying the foundation for subsequent processing.

[0100] The second step is sensor reliability assessment. This step calculates the signal-to-noise ratio (SNR) and data stability, and then gives a comprehensive score:

[0101] ,

[0102] ,

[0103] ,

[0104] in, is the RMS value of the signal, is the RMS value of noise, is the data standard deviation, These weights are optimized through a large number of experiments and can achieve a good balance in most application scenarios.

[0105] The third step is adaptive Kalman filtering. This step is the core innovation of the present invention and includes the following three sub-steps:

[0106] a) Non-Gaussian noise processing:

[0107] First, perform kernel density estimation on the observed data sequence:

[0108] ,

[0109] in, is the estimated probability density function, is the sample size, is the bandwidth parameter, is the kernel function, For the Sample points, The point to be estimated.

[0110] Preferably, the present invention adopts an adaptive bandwidth selection algorithm:

[0111]

[0112] in is the standard deviation and IQR is the interquartile range. This method can automatically select the optimal bandwidth according to the characteristics of data distribution and improve the accuracy of estimation.

[0113] Then, based on the estimated noise distribution, the variance of the equivalent Gaussian noise is calculated:

[0114] ,

[0115] in, is the covariance of the equivalent Gaussian noise, is the observed value, is the observation matrix, is the state vector.

[0116] b) Adaptive gain adjustment:

[0117] First, calculate the innovation series covariance:

[0118] ,

[0119] in, is the innovation sequence covariance, is the sliding window size (20 in the present invention), For innovation sequence.

[0120] Next, calculate the covariance of the theoretical innovation sequence:

[0121] ,

[0122] in, is the covariance of the theoretical innovation sequence, is the prior estimation error covariance, k is the current moment, is the observation matrix. Then, the adaptive factor is defined as:

[0123] ,

[0124] in, is the adaptive factor, and tr[] is the trace of the matrix.

[0125] Finally, adjust the Kalman gain:

[0126] ,

[0127] in, is the adjusted Kalman gain.

[0128] c) Abnormal data detection and processing:

[0129] First, calculate the Mahalanobis distance between the observed and predicted values:

[0130] ,

[0131] in, is the square of the Mahalanobis distance, is the covariance of the theoretical innovation sequence, , the superscript (-1) indicates the matrix inversion, is the prior state estimate at time k, is the prior estimate error covariance, is the covariance of the equivalent Gaussian noise, k is the current moment, is the observation matrix, for The observed value at the moment, then set the threshold .

[0132] Preferably, the present invention adopts an adaptive threshold:

[0133] ,

[0134] in (corresponding to a confidence level of 0.99), , This adaptive threshold allows for higher tolerance in the initial stages of the filter, and gradually tightens over time.

[0135] when , it is determined to be abnormal data and the observation noise covariance is adjusted:

[0136] , and finally use the adjusted Perform Kalman filter update.

[0137] The fourth step is adaptive weight allocation. Based on the sensor reliability evaluation results, the weight of each sensor is calculated using the improved Softmax function:

[0138] ,

[0139] in, For the The weight of each sensor, For the The reliability score of each sensor, is the natural exponential function, is the summation symbol, which means the sum of all sensors Perform summation; is the index of the sensor, ranging from 1 to the total number of sensors. is the temperature compensation coefficient. The purpose of introducing the temperature compensation coefficient is to consider the impact of temperature on the performance of different sensors and further improve the accuracy of weight allocation.

[0140] The fifth step is multi-source data fusion. The covariance cross-validation method is used for data fusion:

[0141] ,

[0142] ,

[0143] in is the fused covariance matrix, is the fused state estimate, is the covariance matrix estimated by each sensor, is the filtering result of each sensor.

[0144] Step 6: Result output and feedback. The fusion result is output to the external system through the USB 2.0 interface (12Mbps). At the same time, the adaptive PID control algorithm is implemented based on the fusion result:

[0145] ,

[0146] in , , For adaptive PID parameters, real-time adjustment is performed through fuzzy logic controller.

[0147] Preferably, the present invention adopts the following parameter ranges:

[0148] ;

[0149] These parameter ranges are obtained through a large number of simulations and experiments, and can achieve good control effects in most application scenarios.

[0150] In order to further improve system performance, the present invention also adopts the following optimization measures:

[0151] a) FPGA co-processing, using Xilinx Artix-7 XC7A35T-1FTG256C FPGA to achieve data preprocessing and fast Fourier transform (FFT) acceleration. The FPGA communicates with the main control unit 15 through the AXI4 interface, and the data transmission rate reaches 1GB / s.

[0152] b) FreeRTOS is used as the real-time operating system to ensure the real-time and reliability of task scheduling. The system implements dynamic adjustment of task priorities and optimizes system response performance.

[0153] c) Electromagnetic compatibility (EMC) design, PCB adopts 4-layer design, the inner layer is a complete ground plane. Key signal lines use differential pair routing to reduce electromagnetic interference. Add EMI filters around the power inlet and sensitive modules, such as TDKMPZ2012S102AT000.

[0154] d) Heat dissipation design: the main processor and FPGA use Fischer Elektronik ICK BGA 14 x 14 x 10 heat sink. The built-in temperature-controlled fan, model Sunon MF40101V1-1000U-A99, automatically adjusts the speed according to the temperature.

[0155] e) Software optimization, using the ARM CMSIS-DSP library to optimize algorithm performance. To further optimize system operation efficiency, we implemented circular buffer and zero-copy technology to minimize data transmission overhead. At the same time, we adopted a dynamic memory allocation strategy to optimize memory usage efficiency and ensure that the system maintains stable performance during long-term operation.

[0156] It is worth noting that the adaptive dual-frequency bandpass filter of the present invention performs well in practical applications. In the tests of the GPS L1 band (1.575 GHz) and the WiMAX band (3.5 GHz), the filter exhibited excellent performance. Specifically, in the GPS L1 band, the measured center frequency was 1.55 GHz, the 3dB relative bandwidth was 13.4%, the insertion loss was 0.74 dB, and the return loss was 17.4 dB. In the WiMAX band, the measured center frequency was 3.49 GHz, the 3dB relative bandwidth was 7.8%, the insertion loss was 1.33 dB, and the return loss was 12 dB. These data fully demonstrate the advantages of the present invention in achieving high selectivity and low insertion loss.

[0157] More importantly, the filter of the present invention realizes five transmission zeros, distributed at 1.28, 1.92, 2.68, 3.65 and 4.1 GHz. These carefully designed transmission zeros significantly enhance the frequency selectivity and isolation of the two passbands, effectively suppressing the interference of adjacent frequency bands. In addition, the 14-dB suppression stopband bandwidth of the filter can be extended from 3.63 GHz to 9.83 GHz, achieving a wide stopband of 1.79f02, which is far superior to the prior art and effectively suppresses high-frequency harmonic interference.

[0158] In practical applications, the filter of the present invention exhibits excellent environmental adaptability. Through temperature compensation and adaptive algorithms, the filter can maintain stable performance in a wide temperature range of -40°C to 85°C. At the same time, thanks to the Rogers5880 substrate (thickness of 0.508mm, relative dielectric constant of 2.2), the filter has excellent temperature stability and low loss characteristics, and is particularly suitable for use in radar systems in harsh environments.

[0159] In addition, the filter of the present invention has excellent power handling capability. By properly designing the transmission line width (such as W4=3.1mm), the filter can withstand input power up to 50W, meeting the requirements of many high-power radar systems. At the same time, the normalized circuit size of the filter is only 0.08λg×0.32λg (where λg is the waveguide wavelength at the center frequency of the GPS band). This compact design makes it very suitable for space-constrained radar systems.

[0160] In summary, the adaptive dual-frequency bandpass filter based on intelligent algorithm provided by the present invention realizes a highly integrated, high-performance and high-reliability filtering solution through innovative hardware design and advanced multi-source data fusion algorithm. The filter can not only process GPS L1 band and WiMAX band signals at the same time, but also adapt to complex and changeable electromagnetic environments in real time, providing key signal processing support for modern multi-band radar systems.

[0161] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. Adaptive dual-frequency bandpass filter based on intelligent algorithm, characterized in that: include: A main frame, wherein an independent sensor cabin is provided in the main frame; A plurality of sensors disposed in the sensor cabin; a core filter structure disposed in the main frame, the core filter structure comprising two pairs of quarter-wavelength λ / 4 step impedance resonators and a branch-loaded three-mode resonator, for forming a first passband and a second passband; an impedance tuning module disposed at the input and output ends of the main frame, the impedance tuning module being used to adjust the characteristics of the first passband and the second passband; an impedance analyzer chip disposed in the main frame; and a main control unit electrically connected to the sensor cabin, the core filter structure, the impedance tuning module and the impedance analyzer chip; wherein the main control unit is used to execute a multi-source data fusion algorithm, and adjust the core filter structure through the impedance tuning module according to the result of the multi-source data fusion algorithm to achieve adaptive adjustment of the first passband and the second passband; The multi-source data fusion algorithm includes the following steps that are performed in sequence: preprocessing the raw data collected by the multiple sensors; performing sensor reliability assessment based on the preprocessed data; performing adaptive Kalman filtering using the preprocessed data and the sensor reliability assessment result; performing adaptive weight allocation based on the sensor reliability assessment result; performing multi-source data fusion based on the adaptive Kalman filtering result and the adaptive weight; outputting the result of the multi-source data fusion and feeding it back to the impedance tuning module to adjust the characteristics of the first passband and the second passband; The adaptive Kalman filtering step further includes the following sub-steps executed in sequence: performing non-Gaussian noise processing on the pre-processed data; performing adaptive gain adjustment based on the result of the non-Gaussian noise processing; performing abnormal data detection and processing based on the adaptive gain adjustment; wherein the result of the non-Gaussian noise processing is used to optimize the subsequent adaptive gain adjustment and abnormal data detection and processing sub-steps; The non-Gaussian noise processing comprises the following steps: First, observe the data sequence for kernel density estimation: Where f(y) is the estimated probability density function, n is the number of samples, h is the bandwidth parameter, K is the kernel function, and y i is the i-th sample point, and y is the point to be estimated; Then based on the estimated noise distribution, the variance of the equivalent Gaussian noise is calculated: R k =E[(y k -H k x k ) 2 ], Among them, R k is the covariance of the equivalent Gaussian noise, y k is the observed value, H k is the observation matrix, x k is the state vector; Finally, the R k Substitute into the standard Kalman filter formula; The adaptive gain adjustment comprises the following steps: First calculate the innovation series covariance: Among them, C v (k) is the innovation sequence covariance, M is the sliding window size, v(i) is the innovation sequence, y(i) is the observed value, H i is the observation matrix, is the prior state estimate; Then calculate the theoretical innovation series covariance: Among them, S k is the covariance of the theoretical innovation sequence, is the prior estimate error covariance, R k is the covariance of the equivalent Gaussian noise, k is the current moment, H k is the observation matrix; Next, define the adaptive factor: Among them, λ(k) is the adaptive factor, tr[] is the trace of the matrix; Finally adjust the Kalman gain: Among them, K k is the adjusted Kalman gain; The abnormal data detection and processing comprises the following steps: First, calculate the Mahalanobis distance between the observed and predicted values: in, is the square of the Mahalanobis distance, S k is the covariance of the theoretical innovation sequence, The superscript (-1) indicates matrix inversion. is the prior state estimate at time k, is the prior estimate error covariance, R k is the covariance of the equivalent Gaussian noise, k is the current moment, H k is the observation matrix, y k is the observed value at time k; Then set the threshold γ, when When , it is judged as abnormal data; Then, for abnormal data, adjust the observation noise covariance: Finally, use the adjusted Perform Kalman filter update; The adaptive weight allocation step comprises: Based on the sensor reliability evaluation results, Softma x The function calculates the weight of each sensor: Among them, w i is the weight of the i-th sensor, r i is the reliability score of the i-th sensor, exp( ) is the natural exponential function, ∑ is the summation symbol, which means exp(r j ) is summed; j is the index of the sensor, ranging from 1 to the total number of sensors.

2. The filter according to claim 1, characterized in that The multiple sensors include a temperature sensor array, a power detector, a standing wave ratio monitoring circuit, an environmental humidity sensor and a vibration sensor, and the multiple sensors are electrically connected to the main control unit through a high-speed digital interface.

3. The filter according to claim 1, characterized in that The impedance tuning module includes a programmable waveform generator, a digital variable capacitor array and a variable inductor, and the impedance tuning module is electrically connected to the main control unit through an I2C bus.

4. The filter according to claim 1, characterized in that The impedance analyzer chip is electrically connected to the main control unit via an SPI interface.

Images