MEMS sensor signal conversion method and system based on cycle-by-cycle noise reduction

Abstract

The application discloses a MEMS sensor signal conversion method and system based on period-by-period noise reduction, and relates to the technical field of signal noise reduction.The method comprises the following steps: in a first sampling period, a first signal sequence of a target device collected by a MEMS sensor is acquired, noise filtering is performed according to preset noise reduction parameters to obtain a first filtered signal sequence and a first noise signal sequence; a first running parameter sequence of the target device is collected, a second predicted running parameter sequence and a second predicted noise signal sequence in a second sampling period are predicted according to the first noise signal sequence; noise reduction parameter optimization in the second sampling period is performed according to the second predicted noise signal sequence to obtain second optimized noise reduction parameters; noise reduction control is performed in the second sampling period according to the second optimized noise reduction parameters, and period-by-period noise reduction signal conversion is continuously performed.The application solves the technical problem of poor signal noise reduction effect in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of signal denoising technology, and in particular to a MEMS sensor signal conversion method and system based on cycle-by-cycle denoising. BACKGROUND

[0002] In the industrial field, equipment denoising is a common technical problem faced by equipment. Traditional denoising techniques mostly use fixed filtering or static parameter configuration, which is difficult to adapt to real-time changes in working conditions. Fixed signal conversion schemes are prone to insufficient denoising or excessive suppression when working conditions change suddenly, and are difficult to meet the real-time needs of high dynamic scenarios. In addition, existing methods usually rely on historical data statistics, and have limited tracking ability for instantaneous sudden noise and rapid working condition migration, resulting in poor transient denoising effect. More critically, there is a strong coupling relationship between noise characteristics and equipment operating state, and a dynamic balance needs to be achieved between noise suppression and equipment operation. SUMMARY

[0003] The present application provides a MEMS sensor signal conversion method and system based on cycle-by-cycle denoising, which is used to solve the technical problem of poor signal denoising effect in the prior art.

[0004] In view of the above problems, the present application provides a MEMS sensor signal conversion method and system based on cycle-by-cycle denoising.

[0005] In a first aspect, the present application provides a MEMS sensor signal conversion method based on cycle-by-cycle denoising, comprising:

[0006] In a first sampling period, a first signal sequence of a target device collected by a MEMS sensor is obtained, and a first denoising signal sequence and a first noise signal sequence are obtained by denoising according to a preset denoising parameter;

[0007] A first operating parameter sequence of the target device is collected, and a second predicted operating parameter sequence and a second predicted noise signal sequence in a second sampling period are predicted according to the first noise signal sequence;

[0008] According to the second predicted noise signal sequence, a denoising parameter optimization in the second sampling period is performed to obtain a second optimized denoising parameter, wherein the optimization parameter is configured according to the deviation between the second predicted operating parameter sequence and the second predicted noise signal sequence;

[0009] According to the second optimized denoising parameter, denoising control is performed in the second sampling period, and cycle-by-cycle denoising signal conversion is continued.

[0010] In a second aspect, the present application provides a MEMS sensor signal conversion system based on cycle-by-cycle denoising, comprising:

[0011] The first signal acquisition module is configured to acquire a first signal sequence of the target device collected by the MEMS sensor in a first sampling period, and obtain a first noise filtering signal sequence and a first noise signal sequence by noise filtering according to a preset noise reduction parameter.

[0012] The second signal prediction module is configured to acquire a first operating parameter sequence of the target device, and predict a second predicted operating parameter sequence and a second predicted noise signal sequence in a second sampling period according to the first noise signal sequence.

[0013] The optimization parameter acquisition module is configured to optimize the noise reduction parameter in the second sampling period according to the second predicted noise signal sequence, and obtain a second optimized noise reduction parameter, wherein the optimization parameter is configured according to the deviation between the second predicted operating parameter sequence and the second predicted noise signal sequence.

[0014] The noise reduction control module is configured to perform noise reduction control in the second sampling period according to the second optimized noise reduction parameter, and continue to perform the cycle-by-cycle noise signal conversion.

[0015] The one or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0016] The present application provides a MEMS sensor signal conversion method and system based on cycle-by-cycle noise reduction, which significantly improves the noise reduction effect under dynamic conditions through a closed-loop optimization mechanism dynamically coupling the device operating state and noise prediction. Compared with the traditional method, the technical solution provided by the present application significantly overcomes the static limitations and response lag defects of fixed parameter noise reduction, and achieves the technical effect of low delay, low cost and high robustness of noise suppression in a dynamic environment. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0018] Figure 1 The flowchart of the MEMS sensor signal conversion method based on cycle-by-cycle noise reduction provided by the embodiments of the present application is shown.

[0019] Figure 2 The structure diagram of the MEMS sensor signal conversion system based on cycle-by-cycle noise reduction provided by the embodiments of the present application is shown.

[0020] In the drawings, the components represented by the numbers are described as follows:

[0021] The first signal acquisition module 100, the second signal prediction module 200, the optimization parameter acquisition module 300, and the noise reduction control module 400. DETAILED DESCRIPTION

[0022] The application provides a MEMS sensor signal conversion method and system based on periodic noise reduction, which is used to solve the problem of poor signal noise reduction effect in the prior art.

[0023] The technical solutions in the embodiments of the application will be clearly and completely described in combination with the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the application.

[0024] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server comprising a series of steps or units does not have to be limited to the clearly listed steps or units, but can include other steps or modules that are not clearly listed or inherent to the process, method, product or device.

[0025] Embodiment one, as shown in the application provides a MEMS sensor signal conversion method based on periodic noise reduction, wherein the method comprises: Figure 1

[0026] S10: In a first sampling period, a first signal sequence of a target device collected by a MEMS sensor is acquired, and a first noise reduction signal sequence and a first noise signal sequence are obtained by noise reduction according to a preset noise reduction parameter.

[0027] In a dynamic working condition environment, the traditional method directly uses fixed parameters for noise reduction due to the lack of a reliable noise reference, which leads to the fact that the preset parameters cannot adapt to the actual noise characteristics of the current device, and the initial noise reduction may be insufficient or excessive, and the noise signal and the useful signal are not effectively separated, which makes it difficult to build a data basis for subsequent optimization.

[0028] The step S10 in the method provided in the embodiments of the application comprises:

[0029] In a first sampling period, a first signal sequence of a target device collected by a MEMS sensor is acquired;

[0030] The first signal sequence is subjected to noise reduction processing to obtain a first noise reduction signal sequence and a first noise signal sequence.

[0031] ​In the embodiment of the present application, in the first sampling period, a MEMS sensor is used to collect a first signal sequence of a target device. The first sampling period is exemplarily set to 60 minutes, and the first signal sequence of the target device is collected by using the MEMS sensor in 30 minutes. The target device refers to a device that needs to be optimized for noise reduction, which can be a generator that generates noise during operation. The MEMS sensor is a functional device for realizing micro-electro-mechanical system sensing and signal processing, and has various structures and functions. Optionally, a MEMS sensor capable of collecting vibration signals is used to collect vibration signals of the target device, and the vibration signals are integrated in time sequence to obtain the first signal sequence.

[0032] The first signal sequence is filtered to obtain a first filtered noise signal sequence and a first noise signal sequence. Specifically, the first signal sequence is filtered by using a filter, for example, the filter noise parameter of the filter is set to 110% of the basic frequency of the normal operation of the device, so as to filter out the high-frequency noise generated during the operation of the device. The basic frequency of the normal operation of the device can be obtained from the specific design parameters of the device. For example, when the basic frequency of the normal operation of the device is 1000 Hz, the filter noise parameter is set to 1100 Hz, and the filtered first filtered noise signal sequence and the filtered first noise signal sequence are output.

[0033] The initial filtering is completed by using the preset noise reduction parameter, and the first noise signal sequence is separated synchronously, which provides accurate input for subsequent dynamic optimization. The initial filtering can ensure signal availability, and the separated noise sequence accurately represents the current working condition interference characteristics, thereby establishing a reliable benchmark for noise evolution prediction.

[0034] S20: collecting a first operation parameter sequence of the target device, and predicting a second predicted operation parameter sequence and a second predicted noise signal sequence in a second sampling period according to the first noise signal sequence.

[0035] The prior art ignores the strong coupling relationship between the device operation parameter and the noise, and only relies on the static prediction of the historical noise data, so that the noise prediction associated with the working condition change may lag behind the actual interference evolution, and the noise characteristics caused by the sudden working condition change cannot be predicted, resulting in inaccurate configuration of the noise reduction parameter in the next period.

[0036] The step S20 in the method provided in the embodiment of the present application comprises:

[0037] collecting a first operation parameter sequence of the target device in the first sampling period;

[0038] inputting the first operation parameter sequence and the first noise signal sequence into a noise signal predictor to obtain a second predicted operation parameter sequence and a second predicted noise signal sequence in a second sampling period by prediction output;

[0039] The training step of the noise signal predictor comprises:

[0040] According to the operation monitoring data of the same device, a sample first operation parameter sequence set and a sample first noise signal sequence set are collected, and a sample second operation parameter sequence set and a sample second noise signal sequence set in the next period are collected.

[0041] A noise signal predictor is constructed based on machine learning.

[0042] The sample first operation parameter sequence set, the sample first noise signal sequence set, the sample second operation parameter sequence set, and the sample second noise signal sequence set are used to supervise the training of adjusting the parameters of the noise signal predictor until the test converges to complete the training.

[0043] In the embodiments of the present application, a first operation parameter sequence of the target device in a first sampling period is collected, which may include operation power (obtained by extracting the operation log of the target device), operation temperature (obtained by a temperature sensor), and the like.

[0044] According to the operation monitoring data of the same device, a sample first operation parameter sequence set and a sample first noise signal sequence set are collected, and a sample second operation parameter sequence set and a sample second noise signal sequence set in the next period are collected. The period is exemplarily set to 120 minutes. The operation parameter sequence may be selected from operation power and operation temperature, and the noise signal sequence is obtained by collecting the signal sequence and using a filter to filter noise. The operation parameter sequence is integrated into an operation parameter sequence set, and the noise signal parameter is integrated into a noise signal sequence set, to obtain the sample first operation parameter sequence set, the sample first noise signal sequence set, the sample second operation parameter sequence set, and the sample second noise signal sequence set.

[0045] Based on machine learning, a noise signal predictor is constructed, which is preferably constructed in a 3-layer structure, wherein the input layer is used to receive the first operation parameter sequence and the first noise signal sequence, the hidden layer contains 64 nodes and uses a ReLU function for activation, and the output layer outputs the predicted second operation parameter sequence and the second noise signal sequence. The sample first operation parameter sequence set, the sample first noise signal sequence set, the sample second operation parameter sequence set, and the sample second noise signal sequence set are used to supervise the training of adjusting the parameters of the noise signal predictor until the test converges, for example, the second predicted operation parameter sequence and the second predicted noise signal error output by inputting the operation parameter sequence are both within ±2%, that is, the training of the noise signal predictor is completed.

[0046] The first operation parameter sequence and the first noise signal sequence are input into the noise signal predictor to predict and output a second predicted operation parameter sequence and a second predicted noise signal sequence in a second sampling period.

[0047] A dynamic correlation model of working conditions and noise is constructed by jointly predicting the running parameters and the noise sequence of the equipment, wherein the running parameter sequence reflects the trend of the state transition of the equipment in real time, so that the noise prediction is in line with the actual working conditions; the coupling prediction based on the current noise characteristics and the working conditions can significantly improve the predictability of sudden interference, and provides a core basis for the proactive optimization of the noise reduction parameters.

[0048] S30: According to the second predicted noise signal sequence, the noise reduction parameter optimization in the second sampling period is performed to obtain a second optimized noise reduction parameter, wherein the optimization parameter is configured according to the deviation between the second predicted running parameter sequence and the second predicted noise signal sequence.

[0049] The traditional noise reduction optimization method has limitations because it does not associate the collaborative deviation of the predicted working conditions and noise: single-dimensional noise suppression can easily cause equipment operation problems, and the static optimization target cannot adapt to the balance requirement of noise suppression and equipment operation in a dynamic scene.

[0050] The step S30 in the method provided by the embodiment of the present application comprises:

[0051] Randomly configuring a noise reduction parameter;

[0052] Calculating a running parameter deviation degree of the second predicted running parameter sequence and the first running parameter sequence;

[0053] Calculating a noise signal deviation degree of the second predicted noise signal sequence and the first noise signal sequence;

[0054] Calculating a control deviation degree of the noise reduction parameter and a preset noise reduction parameter as a transient distortion degree;

[0055] According to the running parameter deviation degree, the noise signal deviation degree and the transient distortion degree, a cost fitness degree is calculated and obtained;

[0056] According to the running parameter deviation degree, the noise signal deviation degree and the transient distortion degree, a cost fitness degree is calculated and obtained, comprising:

[0057] Calculating the sum of the running parameter deviation degree and the noise signal deviation degree;

[0058] Calculating the similarity between the sum of the running parameter deviation degree and the noise signal deviation degree and the transient distortion degree to obtain the cost fitness degree;

[0059] According to the second predicted noise signal sequence and the noise reduction parameter, a noise reduction fitness degree is analyzed and obtained;

[0060] According to the second predicted noise signal sequence and the noise reduction parameter, a noise reduction fitness degree is analyzed and obtained, comprising:

[0061] obtaining a noise reduction analyzer, wherein the noise reduction analyzer is trained by using a set of sample noise signal sequences, a set of sample noise reduction parameters and a set of labeled sample noise reduction fitnesses, each of the sample noise reduction fitnesses including a ratio of a signal-to-noise ratio after noise reduction to a preset signal-to-noise ratio;

[0062] inputting the second predicted noise signal sequence and the noise reduction parameter into the noise reduction analyzer, and outputting an obtained noise reduction fitness;

[0063] configuring a noise reduction weight and a cost weight according to the operating parameter deviation degree and the noise signal deviation degree, and performing weighted calculation on the noise reduction fitness and the cost fitness to obtain a noise reduction fitness;

[0064] The method further includes: configuring a noise reduction weight and a cost weight according to the operating parameter deviation degree and the noise signal deviation degree, and performing weighted calculation on the noise reduction fitness and the cost fitness to obtain a noise reduction fitness.

[0065] The method further includes: calculating a cost weight according to the operating parameter deviation degree and the noise signal deviation degree, and calculating a noise reduction weight.

[0066] The method further includes: calculating a cost weight according to the operating parameter deviation degree and the noise signal deviation degree, and calculating a noise reduction weight.

[0067] The method further includes: calculating a cost weight according to the operating parameter deviation degree and the noise signal deviation degree, and calculating a noise reduction weight.

[0068] The method further includes: calculating a cost weight according to the operating parameter deviation degree and the noise signal deviation degree, and calculating a noise reduction weight.

[0069] The method further includes: configuring a noise reduction weight and a cost weight according to the operating parameter deviation degree and the noise signal deviation degree, and performing weighted calculation on the noise reduction fitness and the cost fitness to obtain a noise reduction fitness.

[0070] iteratively optimizing the noise reduction parameter to obtain a second optimized noise reduction parameter, wherein the second optimized noise reduction parameter has a maximum noise reduction fitness in the optimization.

[0071] In the embodiments, a random number generator is used to randomly generate and configure the noise reduction parameter, for example, the noise reduction parameter is frequency noise reduction, and the parameter is 108% of the basic frequency of the device in normal operation, for example, the basic frequency of the device in normal operation is 1000 Hz, and the noise reduction parameter is 1080 Hz.

[0072] An arithmetic mean of the operating parameters of the second predicted operating parameter sequence and the first operating parameter sequence is calculated, and a deviation degree is calculated according to the arithmetic mean of the operating parameters of the second predicted operating parameter sequence and the first operating parameter sequence. Specifically, the operating parameter deviation degree = (arithmetic mean of the second predicted operating parameter sequence - arithmetic mean of the first operating parameter sequence) ÷ arithmetic mean of the first operating parameter sequence, with a unit of %. For example, the arithmetic mean of the operating power in the second predicted operating parameter sequence is 120 W, and the arithmetic mean of the operating power in the first operating parameter sequence is 150 W, then the operating parameter deviation degree of the operating power = (120 - 150) ÷ 120 = 25%. The operating parameter deviation degree is equal to the arithmetic mean of the deviation degrees of a plurality of operating parameters in the operating parameter, for example, the operating parameter deviation degree = (deviation degree of the operating power in the operating parameter + deviation degree of the operating temperature in the operating parameter) ÷ 2. The operating parameter deviation degree is a value representing the accuracy of the predicted operating parameter, and the smaller the operating parameter deviation degree, the higher the accuracy of the second predicted operating parameter sequence, and the more reliable.

[0073] An arithmetic mean of the noise signals of the second predicted noise signal sequence and the first noise signal sequence is calculated, and a deviation degree is calculated according to the arithmetic mean of the noise signals of the second predicted noise signal sequence and the first noise signal sequence. Specifically, the noise signal deviation degree = (arithmetic mean of the second predicted noise signal sequence - arithmetic mean of the first noise signal sequence) ÷ arithmetic mean of the first noise signal sequence, with a unit of %. For example, the arithmetic mean of the second predicted noise signal sequence is 1100 Hz, and the arithmetic mean of the first noise signal sequence is 1200 Hz, then the noise signal deviation degree = (1100 - 1200) ÷ 1200 = 8.3%. Similarly, the smaller the noise deviation degree, the higher the accuracy of the predicted noise signal sequence.

[0074] A control deviation degree of the noise reduction parameter and the preset noise reduction parameter is calculated as a transient distortion degree. The transient distortion degree = (noise reduction parameter - preset noise reduction parameter) ÷ preset noise reduction parameter, with a unit of %. For example, the noise reduction parameter is 1080 Hz, and the preset noise reduction parameter is 1100 Hz, then the transient distortion degree = (1080 - 1100) ÷ 1100 = 1.8%. The transient distortion degree is a value reflecting the deviation of the noise reduction parameter and the preset noise reduction parameter, and the larger the transient distortion degree, the greater the possibility that the noise reduction parameter deviates and does not conform to the current working condition.

[0075] According to the operating parameter deviation degree, the noise signal deviation degree, and the transient distortion degree, a cost adaptation degree is calculated and obtained.

[0076] The sum of the operating parameter deviation degree and the noise signal deviation degree is calculated.

[0077] The similarity between the sum of the operating parameter deviation degree and the noise signal deviation degree and the transient distortion degree is calculated to obtain a cost fitness. Optionally, the cost fitness = 1 ÷ |the sum of the operating parameter deviation degree and the noise signal deviation degree - the transient distortion degree|. When the sum of the operating parameter deviation degree and the noise signal deviation degree - the transient distortion degree = 0, the cost fitness is 1. The greater the cost fitness, the smaller the cost of the transient distortion adjustment noise reduction parameter, and the better the adaptation to the target device and the current working condition of the target device under the condition of smaller operating parameter change cost.

[0078] A noise reduction analyzer is obtained. For example, the noise reduction analyzer is constructed using machine learning, for example, using a 3-layer structure, the input layer receives the noise signal sequence and the noise reduction parameter, the hidden layer uses 16 nodes and uses the ReLU function for activation, and the output layer outputs the analyzed noise reduction fitness.

[0079] The noise reduction analyzer is trained using a set of sample noise signal sequences, a set of sample noise reduction parameters, and a set of labeled sample noise reduction fitnesses. Each sample noise reduction fitness includes the ratio of the signal-to-noise ratio after noise reduction to the preset signal-to-noise ratio. For example, the preset signal-to-noise ratio can be set to 20 dB, and the noise reduction fitness = the signal-to-noise ratio after noise reduction ÷ the preset signal-to-noise ratio. The greater the noise reduction fitness, the higher the signal-to-noise ratio after noise reduction, the stronger the signal relative to the noise, and the better the noise reduction effect.

[0080] Training is performed until the noise reduction analyzer converges, for example, the input noise signal sequence, the noise reduction parameter, and the output noise reduction fitness accuracy rate are above 90%, i.e., the noise reduction analyzer training is complete.

[0081] The second predicted noise signal sequence and the noise reduction parameter are input into the noise reduction analyzer, and the noise reduction fitness is obtained.

[0082] The ratio of the operating parameter deviation degree, the noise signal deviation degree, the maximum historical operating parameter deviation, and the maximum noise signal deviation is calculated, and the average is calculated to obtain a deviation coefficient as a cost weight. The cost weight = (the operating parameter deviation degree ÷ the maximum historical operating parameter deviation + the noise deviation degree ÷ the maximum historical noise deviation) ÷ 2. The cost weight is the weight of the cost fitness. The greater the cost weight, the higher the requirement for operating parameter stability, the more sensitive to the behavior of noise reduction optimization with operating parameter as the cost, and the greater weight given to the cost fitness.

[0083] According to the cost weight, a noise reduction weight is calculated. The noise reduction weight = 1 - the cost weight.

[0084] The noise reduction fitness and the cost fitness are weighted and calculated using the noise reduction weight and the cost weight to obtain the noise reduction fitness. The noise reduction fitness = the noise reduction fitness × the noise reduction weight + the cost fitness × the cost weight.

[0085] The random configuration of the noise reduction parameter is repeated, the noise reduction parameter is iteratively optimized, and a second optimized noise reduction parameter is obtained after convergence, wherein the second optimized noise reduction parameter has the largest noise reduction fitness in the optimization. The convergence condition is exemplarily set as that no better noise reduction fitness appears for 20 iterations, and the noise reduction parameter corresponding to the current best noise reduction fitness is output as the second optimized noise reduction parameter.

[0086] The deviation of the predicted operating parameter from the noise sequence is introduced as a basis for configuring the optimization parameter, a multi-objective dynamic balance is achieved, and through fusion of multi-dimensional parameters such as noise deviation, operating condition deviation, and transient distortion, the noise reduction parameter is real-time self-adaptive tuned between interference intensity suppression and equipment operating state, to ensure optimal coordination of noise reduction effect and equipment operation under complex operating conditions.

[0087] S40: performing noise reduction control in the second sampling period according to the second optimized noise reduction parameter, and continuing to perform the periodic noise reduction signal conversion.

[0088] In the embodiment of the application, the noise reduction signal conversion is performed in the second sampling period according to the second optimized noise reduction parameter, and the periodic noise reduction signal conversion is continued, for example, the noise reduction signal conversion is performed in the third sampling period according to the same method.

[0089] By executing the prediction-optimization-control process periodically, a continuously adaptive dynamic noise reduction system is constructed, the noise reduction parameter is updated based on the latest operating condition and noise prediction in each period, so that the noise reduction process always fits the real-time state of the equipment, and the closed-loop iterative mechanism can accumulate errors.

[0090] Embodiment two, as Figure 2 shown, based on the same inventive concept as the MEMS sensor signal conversion method based on periodic noise reduction provided in embodiment one, the embodiment of the application further provides a MEMS sensor signal conversion system based on periodic noise reduction, comprising:

[0091] The first signal acquisition module 100 is configured to acquire a first signal sequence of a target device collected by a MEMS sensor in a first sampling period, and obtain a first filtered noise signal sequence and a first noise signal sequence by filtering according to a preset noise reduction parameter.

[0092] The second signal prediction module 200 is configured to acquire a first operating parameter sequence of the target device, and predict a second predicted operating parameter sequence and a second predicted noise signal sequence in a second sampling period according to the first noise signal sequence.

[0093] The optimization parameter acquisition module 300 is configured to optimize the noise reduction parameter in the second sampling period according to the second predicted noise signal sequence, and obtain a second optimized noise reduction parameter, wherein the optimization parameter is configured according to the deviation of the second predicted operating parameter sequence and the second predicted noise signal sequence.

[0094] The noise reduction control module 400 is used to perform noise reduction control in the second sampling period according to the second optimized noise reduction parameters, and to continue to perform cycle-by-cycle noise reduction signal conversion.

[0095] In one embodiment, the first signal acquisition module 100 is further configured to:

[0096] During the first sampling period, the first signal sequence of the target device acquired by the MEMS sensor is obtained;

[0097] The first signal sequence is subjected to noise filtering to obtain a first noise-filtered signal sequence and a first noise signal sequence.

[0098] In one embodiment, the second signal prediction module 200 is further configured to:

[0099] Collect the first operating parameter sequence of the target device within the first sampling period;

[0100] The first operating parameter sequence and the first noise signal sequence are input into the noise signal predictor, and the prediction output is used to obtain the second predicted operating parameter sequence and the second predicted noise signal sequence within the second sampling period.

[0101] The training steps of the noise signal predictor include:

[0102] Based on the operational monitoring data of the same equipment, a first set of sample operating parameters and a first set of sample noise signals are collected, and a second set of sample operating parameters and a second set of sample noise signals are collected in the next cycle.

[0103] Building a noise signal predictor based on machine learning;

[0104] Using the first set of sample operating parameters, the first set of sample noise signals, the second set of sample operating parameters, and the second set of sample noise signals, supervised training adjusts the parameters of the noise signal predictor until the test converges and training is completed.

[0105] In one embodiment, the optimization parameter acquisition module 300 is further configured to:

[0106] Randomly configure noise reduction parameters;

[0107] Calculate the deviation of the operating parameters between the second predicted operating parameter sequence and the first operating parameter sequence;

[0108] Calculate the noise signal deviation between the second predicted noise signal sequence and the first noise signal sequence;

[0109] The control deviation between the noise reduction parameters and the preset noise reduction parameters is calculated and used as the transient distortion.

[0110] The cost fitness is calculated based on the deviation of the operating parameters, the deviation of the noise signal, and the transient distortion.

[0111] The cost fitness is calculated based on the operating parameter deviation, noise signal deviation, and transient distortion, including:

[0112] Calculate the sum of the deviation of the operating parameters and the deviation of the noise signal;

[0113] The similarity between the sum of the operating parameter deviation and the noise signal deviation and the transient distortion is calculated to obtain the cost fitness.

[0114] Based on the second predicted noise signal sequence and noise reduction parameters, the noise reduction fitness is obtained through analysis;

[0115] The noise reduction fitness is obtained by analyzing the second predicted noise signal sequence and noise reduction parameters, including:

[0116] A noise reduction analyzer is obtained, wherein the noise reduction analyzer is trained using a set of sample noise signal sequences, a set of sample noise reduction parameters, and a set of labeled sample noise reduction fitness. Each sample noise reduction fitness includes the ratio of the noise reduction signal-to-noise ratio to the preset signal-to-noise ratio.

[0117] The second predicted noise signal sequence and noise reduction parameters are input into the noise reduction analyzer, and the noise reduction fitness is output.

[0118] Based on the deviation of the operating parameters and the deviation of the noise signal, the noise reduction weight and the cost weight are configured, and the noise reduction fitness and the cost fitness are weighted and calculated to obtain the noise reduction fitness.

[0119] Specifically, the noise reduction weight and cost weight are configured based on the operating parameter deviation and noise signal deviation, and the noise reduction fitness and cost fitness are weighted and calculated to obtain the noise reduction fitness, including:

[0120] Based on the deviation of the operating parameters and the deviation of the noise signal, the cost weight is calculated, and the noise reduction weight is obtained.

[0121] Specifically, based on the deviation of the operating parameters and the deviation of the noise signal, the cost weight is calculated, and the noise reduction weight is obtained, including:

[0122] Calculate the ratio of the deviation of the operating parameters and the deviation of the noise signal to the maximum deviation of the historical operating parameters and the maximum deviation of the noise signal, and calculate the mean to obtain the deviation coefficient, which is used as the cost weight;

[0123] The noise reduction weights are calculated based on the cost weights.

[0124] The noise reduction fitness is obtained by weighting the noise reduction fitness and the cost fitness using the aforementioned noise reduction weight and cost weight.

[0125] The noise reduction parameters are iteratively optimized to obtain the converged second optimized noise reduction parameters, wherein the second optimized noise reduction parameters have the largest noise reduction fitness during the optimization.

[0126] In summary, the embodiments of this application have at least the following technical effects:

[0127] This application proposes a MEMS sensor signal conversion method and system based on cycle-by-cycle noise reduction. Through a closed-loop optimization mechanism that dynamically couples device operating status with noise prediction, it significantly improves noise reduction performance under dynamic operating conditions. Specifically, by collecting device operating parameters cycle-by-cycle and jointly predicting noise signals, a dynamic correlation between operating conditions and noise is constructed. This ensures that noise reduction parameter optimization is always synchronized with the actual evolution of the device's state, avoiding insufficient noise reduction or excessive signal suppression due to sudden changes in operating conditions. Secondly, an online optimization mechanism for noise reduction parameters based on predicted noise sequences is introduced to balance noise suppression intensity and device operating status in real time, maintaining a stable signal-to-noise ratio even in complex interference environments. Finally, a noise reduction analyzer intelligently evaluates the noise reduction parameters, combining the dynamic ratio of cost weights and noise reduction weights to ensure that the optimization direction always aligns with the core requirements of the current operating condition. Compared to traditional methods, the technical solution provided in this application significantly overcomes the static limitations and response lag defects of fixed noise reduction parameters, achieving low-latency, low-cost, and highly robust noise suppression in dynamic environments.

[0128] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0129] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0130] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for dynamic gain control of a MEMS sensor based on cycle-by-cycle noise reduction, characterized in that, The method comprises: In the first sampling period, the MEMS sensor collects a first signal sequence of the target device, and a first noise signal sequence and a first filtered signal sequence are obtained by filtering according to a preset gain noise reduction parameter; A first running parameter sequence of the target device is collected, and a second predicted running parameter sequence and a second predicted noise signal sequence in a second sampling period are predicted according to the first noise signal sequence and the first running parameter sequence; The gain noise reduction parameter in the second sampling period is optimized according to the second predicted noise signal sequence, and a second optimized gain noise reduction parameter is obtained, wherein the optimization parameter is configured according to the deviation between the second predicted running parameter sequence and the second predicted noise signal sequence, and the optimization parameter comprises: Randomly configuring the gain noise reduction parameter; Calculating the running parameter deviation degree of the second predicted running parameter sequence and the first running parameter sequence; Calculating the noise signal deviation degree of the second predicted noise signal sequence and the first noise signal sequence; Calculating the control deviation degree of the gain noise reduction parameter and the preset gain noise reduction parameter as the transient distortion degree; According to the running parameter deviation degree, the noise signal deviation degree and the transient distortion degree, the cost fitness is calculated and obtained; According to the second predicted noise signal sequence and the gain noise reduction parameter, the noise reduction fitness is analyzed and obtained; According to the running parameter deviation degree, the noise signal deviation degree, the noise reduction weight and the cost weight, the noise reduction fitness and the cost fitness are weighted and calculated to obtain the gain fitness; The gain noise reduction parameter is iteratively optimized to obtain the second optimized gain noise reduction parameter after convergence, wherein the second optimized gain noise reduction parameter has the maximum gain fitness in the optimization; According to the second optimized gain noise reduction parameter, the noise reduction gain control is performed in the second sampling period, and the periodic noise reduction gain control is continuously performed.

2. The method of claim 1, wherein, In the first sampling period, the MEMS sensor collects a first signal sequence of the target device, and a first noise signal sequence and a first filtered signal sequence are obtained by filtering according to a preset gain noise reduction parameter; In the first sampling period, the MEMS sensor collects a first signal sequence of the target device; The first signal sequence is filtered to obtain the first filtered signal sequence and the first noise signal sequence.

3. The cycle-by-cycle noise reduction based MEMS sensor dynamic gain control method of claim 1, wherein, The first running parameter sequence of the target device is collected, and a second predicted running parameter sequence and a second predicted noise signal sequence in a second sampling period are predicted according to the first noise signal sequence, comprising: The first running parameter sequence of the target device in the first sampling period is collected; The first running parameter sequence and the first noise signal sequence are input into a noise signal predictor to predict and output a second predicted running parameter sequence and a second predicted noise signal sequence in a second sampling period.

4. The method of claim 3, wherein, The training step of the noise signal predictor comprises: According to the running monitoring data of the same device, a sample first running parameter sequence set and a sample first noise signal sequence set are collected, and a sample second running parameter sequence set and a sample second noise signal sequence set in the next period are collected; A noise signal predictor is constructed based on machine learning; The parameters of the noise signal predictor are supervised and trained by using the sample first running parameter sequence set, the sample first noise signal sequence set, the sample second running parameter sequence set and the sample second noise signal sequence set until the test converges to complete the training.

5. The cycle-by-cycle noise reduction based MEMS sensor dynamic gain control method of claim 1, wherein, According to the running parameter deviation degree, the noise signal deviation degree and the transient distortion degree, a cost fitness is calculated and obtained, including: The sum of the running parameter deviation degree and the noise signal deviation degree is calculated; The sum of the running parameter deviation degree and the noise signal deviation degree and the similarity of the transient distortion degree are calculated to obtain the cost fitness.

6. The cycle-by-cycle noise reduction based MEMS sensor dynamic gain control method according to claim 1, wherein, According to the second predicted noise signal sequence and the gain noise reduction parameter, a noise reduction fitness is analyzed and obtained, including: A noise reduction analyzer is obtained, wherein the noise reduction analyzer is trained by using a sample noise signal sequence set, a sample gain noise reduction parameter set and a labeled sample noise reduction fitness set, and each sample noise reduction fitness includes the ratio of the signal-to-noise ratio after noise reduction to the preset signal-to-noise ratio; The second predicted noise signal sequence and the gain noise reduction parameter are input into the noise reduction analyzer to output and obtain the noise reduction fitness.

7. The cycle-by-cycle noise reduction based MEMS sensor dynamic gain control method according to claim 1, wherein, According to the running parameter deviation degree, the noise signal deviation degree, the noise reduction weight and the cost weight are configured, and the noise reduction fitness and the cost fitness are weighted and calculated to obtain a gain fitness, including: According to the running parameter deviation degree and the noise signal deviation degree, the cost weight is calculated, and the noise reduction weight is calculated and obtained; The noise reduction fitness and the cost fitness are weighted and calculated by using the noise reduction weight and the cost weight to obtain the gain fitness.

8. The cycle-by-cycle noise reduction based MEMS sensor dynamic gain control method according to claim 7, characterized in that, According to the running parameter deviation degree and the noise signal deviation degree, the cost weight is calculated, and the noise reduction weight is calculated and obtained, including: The ratio of the running parameter deviation degree, the noise signal deviation degree, the maximum value of the historical running parameter deviation and the maximum value of the noise signal deviation is calculated, and the mean value is calculated to obtain a deviation coefficient as the cost weight; According to the cost weight, the noise reduction weight is calculated and obtained.

9. A MEMS sensor dynamic gain control system based on cycle-by-cycle noise reduction, characterized in that, The system is used to implement the MEMS sensor dynamic gain control method based on periodic noise reduction according to any one of claims 1-8, and the system comprises: A first signal acquisition module is configured to acquire a first signal sequence of a target device collected by a MEMS sensor in a first sampling period, and to filter noise to obtain a first filtered signal sequence and a first noise signal sequence according to a preset gain noise reduction parameter; A second signal prediction module is configured to acquire a first running parameter sequence of the target device, and to predict a second predicted running parameter sequence and a second predicted noise signal sequence in a second sampling period according to the first noise signal sequence and the first running parameter sequence; An optimized parameter acquisition module is configured to optimize a gain noise reduction parameter in the second sampling period according to the second predicted noise signal sequence to obtain a second optimized gain noise reduction parameter, wherein the optimized parameter is configured according to the deviation of the second predicted running parameter sequence and the second predicted noise signal sequence, including: The gain noise reduction parameter is randomly configured; A running parameter deviation degree of the second predicted running parameter sequence and the first running parameter sequence is calculated; A noise signal deviation degree of the second predicted noise signal sequence and the first noise signal sequence is calculated; a control bias degree of the gain noise reduction parameter and a preset gain noise reduction parameter is calculated as a transient distortion degree; a cost fitness degree is calculated according to the operation parameter bias degree, the noise signal bias degree and the transient distortion degree; a noise reduction fitness degree is analyzed according to the second predicted noise signal sequence and the gain noise reduction parameter; a gain fitness degree is obtained by weighting calculation of the noise reduction fitness degree and the cost fitness degree according to the operation parameter bias degree, the noise signal bias degree, a noise reduction weight and a cost weight; the gain noise reduction parameter is iteratively optimized to obtain a second optimized gain noise reduction parameter which has the largest gain fitness degree in the optimization; a noise reduction gain control module is configured to perform noise reduction gain control in the second sampling period according to the second optimized gain noise reduction parameter, and continue to perform the periodic noise reduction gain control.

Images