Multi-parameter cooperative detection high-precision instrument data processing method and system

Abstract

The embodiment of the application discloses a kind of high-precision instrument and meter data processing method and system of multi-parameter coordination detection, it is related to instrument and meter and data processing technical field.The method comprises: the measured parameter of target equipment and environmental interference parameter are synchronously collected by multi-source sensor array, and time series dataset is formed;Dynamic coupling analysis model is constructed, and time series dataset is converted into spatial correlation matrix using gram angle field;Variable weight collaborative decision model is established based on long short-term memory network, and the contribution weight of the data collected in final detection result is dynamically adjusted according to real-time working condition;Hybrid calibration algorithm is used to online calibrate sensor output;The compensation amount of environmental interference parameter is calculated by Bayesian network;High-precision detection data is output.The application can map time-domain parameters into two-dimensional space features in real time, so that the pressure-temperature coupling coefficient recognition accuracy is improved, and the multi-parameter collaborative decision error is reduced.

Description

Technical Field

[0001] This invention relates to the field of instrumentation and data processing technology, and in particular to a high-precision instrumentation data processing method and system for multi-parameter collaborative detection. Background Technology

[0002] In the power sector, accurate instrument data processing is crucial for ensuring equipment reliability and measurement accuracy. Currently, high-precision power instruments such as single-phase and three-phase multi-functional standard energy meters and single-phase and three-phase field calibrators are widely used measuring instruments in power systems. Due to the high measurement accuracy and multiple measurement ranges of these instruments, the data processing process is complex.

[0003] Common data processing methods for electronic instruments include data acquisition, data transmission, data storage, and data analysis. Through data processing, we can better understand the data generated during experiments, draw useful conclusions, and make reasonable decisions. Therefore, mastering commonly used data processing methods for electronic instruments is essential.

[0004] Existing instrument and meter data processing methods have the following drawbacks:

[0005] 1. Independent processing of single parameters: Traditional instruments process multiple parameters separately, ignoring the physical coupling relationship between parameters;

[0006] 2. Static compensation model: Existing compensation algorithms are mostly based on fixed correction coefficients, which cannot adapt to dynamic changes in working conditions;

[0007] 3. Error accumulation: The lack of an error propagation suppression mechanism during multi-sensor data fusion leads to a decrease in system-level accuracy. Summary of the Invention

[0008] In view of this, embodiments of the present invention provide a high-precision instrument data processing method and system for multi-parameter collaborative detection, in order to solve the problems of existing instrument data processing methods that ignore the physical coupling relationship between parameters, cannot adapt to dynamic changes in working conditions, and cause a decrease in system-level accuracy due to error accumulation.

[0009] On one hand, the present invention provides a high-precision instrument data processing method for multi-parameter collaborative detection, comprising:

[0010] Step S1: Simultaneously collect N measured parameters and M environmental interference parameters of the target device through a multi-source sensor array to form a time series dataset;

[0011] Step S2: Construct a dynamic coupling analysis model, use Gram angle field to convert the time series dataset into a spatial correlation matrix, and quantify the nonlinear coupling strength between parameters;

[0012] Step S3: Establish a variable-weight collaborative decision-making model based on a long short-term memory network, and dynamically adjust the contribution weight of the data collected in step S1 in the final detection result according to the real-time operating conditions;

[0013] Step S4: The measured parameters of the final sensor output after processing are calibrated online using a hybrid calibration algorithm, which combines state estimation by Kalman filtering with parameter optimization by particle swarm optimization.

[0014] Step S5: Calculate the compensation amount of environmental interference parameters through a Bayesian network to correct the output value of the corresponding sensor, and feed the compensation result back to the variable weight collaborative decision-making model;

[0015] Step S6: Output the high-precision detection data obtained after processing in step S5 above.

[0016] Furthermore, step S2 includes:

[0017] Step S2.1: Normalize the time series of each parameter to eliminate dimensional differences;

[0018] Step S2.2: Map the one-dimensional time series to a two-dimensional matrix using Gram angle field, preserving temporal correlation;

[0019] Step S2.3: Use a convolutional neural network to extract cross-parameter coupling features from the matrix and generate an N×N correlation matrix.

[0020] Furthermore, in step S2.3, the update frequency of the correlation matrix is ​​positively correlated with the rate of change of the parameters, specifically satisfying the following condition:

[0021] ;

[0022] Where α is the preset sensitivity coefficient, x_i is the instantaneous value of the i-th parameter, and the vertical line represents the condition separator.

[0023] Furthermore, step S3 includes:

[0024] Step S3.1: Establish a bidirectional LSTM network that includes an attention mechanism;

[0025] Step S3.2: Set the forgetting factor γ = 0.9-0.99 to dynamically attenuate the weighting effect of historical data;

[0026] Step S3.3: Output the real-time weight distribution of each parameter using the Softmax function.

[0027] Furthermore, step S4 includes:

[0028] Step S4.1: Eliminate the first-order error of sensor noise using Kalman filtering;

[0029] Step S4.2: Use an improved particle swarm optimization algorithm to perform second-order nonlinear error compensation.

[0030] Furthermore, in step S4.2, the inertia weight ω is calculated using the formula ω=ω_max-(ω_max-ω_min)·(k / K). 2 Adaptive adjustment, where k is the current iteration number, K is the total iteration number, ω_max is the maximum value of the inertia weight, and ω_min is the minimum value of the inertia weight.

[0031] Furthermore, step S5 includes:

[0032] Step S5.1: Construct a directed acyclic graph of environmental interference parameters and measurement errors, with the conditional probabilities of nodes obtained through training on historical data;

[0033] Step S5.2: Use the Markov chain Monte Carlo method to perform posterior probability inference and output the optimal compensation amount.

[0034] Furthermore, step S1 includes:

[0035] A dynamic time warping algorithm is applied to the sensor data to solve the timing misalignment problem caused by asynchronous sampling, and the alignment accuracy reaches 1 / 100 of the sampling period.

[0036] On the other hand, the present invention provides a high-precision instrument data processing system for multi-parameter collaborative detection, comprising:

[0037] The multi-parameter synchronous acquisition module is used to synchronously acquire N measured parameters and M environmental interference parameters of the target device through a multi-source sensor array to form a time series dataset.

[0038] The coupling analysis module is used to construct a dynamic coupling analysis model, using Gram angle field to convert the time series dataset into a spatial correlation matrix and quantify the nonlinear coupling strength between parameters.

[0039] The dynamic decision-making module is used to establish a variable weight collaborative decision-making model based on a long short-term memory network, and dynamically adjust the contribution weight of the data collected in the multi-parameter synchronous acquisition module in the final detection result according to the real-time working conditions.

[0040] The online calibration module is used to calibrate the measured parameters of the final sensor output after processing using a hybrid calibration algorithm, which combines state estimation by Kalman filtering with parameter optimization by particle swarm optimization.

[0041] The environmental compensation module is used to calculate the compensation amount of environmental interference parameters through a Bayesian network to correct the output value of the corresponding sensor, and to feed the compensation result back to the variable weight collaborative decision-making model.

[0042] The data output module is used to output the high-precision detection data obtained after processing by the environmental compensation module.

[0043] Furthermore, the dynamic decision-making module adopts a pipelined architecture, including:

[0044] First-stage pipeline: Parameter weight calculation, cycle 5ns;

[0045] Second-stage pipeline: Environmental compensation is superimposed, with a cycle of 3ns;

[0046] Third-level pipeline: Data validity verification, cycle 2ns.

[0047] The present invention has the following beneficial effects:

[0048] The high-precision instrument data processing method and system for multi-parameter collaborative detection of this invention, by setting a correlation matrix combining Gram angle field (GAF) and convolutional neural network (CNN), maps time-domain parameters into two-dimensional spatial features in real time, thereby improving the recognition accuracy of pressure-temperature coupling coefficient and reducing multi-parameter collaborative decision-making errors. A Long Short-Term Memory (LSTM) network with a forgetting factor (γ=0.95) is used to calculate parameter weights in real time, reducing the delay in weight adjustment during sudden changes in operating conditions. By deploying a Bayesian network and a Markov chain Monte Carlo (MCMC) inference engine, environmental compensation is calculated in real time, reducing pressure measurement deviations caused by steam density fluctuations and improving the signal-to-noise ratio (SNR) under electromagnetic interference (EMI). By employing a Dynamic Time Warping (DTW) hardware accelerator to align asynchronous data in real time, the phase difference between flow and pressure signals is reduced, improving the timeliness of data fusion. A three-stage pipeline architecture is implemented through a Field Programmable Gate Array (FPGA) to execute LSTM inference in real time. This solves the problems of existing instrument data processing methods, such as instruments processing multiple parameters separately, ignoring the physical coupling between parameters, compensation algorithms based on fixed correction coefficients that cannot adapt to dynamic changes in operating conditions, and the lack of an error propagation suppression mechanism when fusing multi-sensor data, which leads to a decrease in system-level accuracy. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a flowchart illustrating the high-precision instrument data processing method for multi-parameter collaborative detection according to the present invention.

[0051] Figure 2 This is a schematic diagram of the structure of the high-precision instrument data processing system for multi-parameter collaborative detection according to the present invention. Detailed Implementation

[0052] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0053] It should be understood that the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0054] On the one hand, embodiments of the present invention provide a high-precision instrument data processing method for multi-parameter collaborative detection, such as... Figure 1 As shown, it includes the following steps:

[0055] S1. Simultaneously collect N measured parameters and M environmental interference parameters of the target device through a multi-source sensor array to form a time series dataset;

[0056] In this step, the target equipment refers to the industrial equipment, instruments, or systems that need to be monitored or controlled, such as: industrial scenarios: pressure pipelines, reaction vessels, power transformers, etc.; other fields: environmental monitoring stations, smart meters, etc. The measured parameters (N) refer to the key physical or chemical quantities that need to be directly measured during the operation of the target equipment, usually related to the core functions or state of the equipment, such as: industrial pressure pipelines: pressure (P), temperature (T), flow rate (F), etc.; electrical equipment: voltage, current, power factor, etc. The environmental interference parameters (M) refer to external environmental factors that may affect the measurement accuracy of the measured parameters, requiring simultaneous monitoring and compensation for their interference, such as: industrial scenarios: ambient temperature, humidity, electromagnetic interference (EMI), steam density, vibration, etc.; general interference: sensor noise, power fluctuations, sampling timing deviations, etc. Where N and M are both integers greater than 1.

[0057] As an optional embodiment, step S1 includes:

[0058] The Dynamic Time Warping (DTW) algorithm is applied to multi-source sensor data to solve the timing misalignment problem caused by asynchronous sampling, and the alignment accuracy reaches 1 / 100 of the sampling period.

[0059] S2. Construct a dynamic coupling analysis model and use the Gramian Angular Field (GAF) to convert the time series dataset into a spatial correlation matrix to quantify the nonlinear coupling strength between parameters.

[0060] As an optional embodiment, this step may include:

[0061] S2.1 Normalize the time series (data) of each parameter to eliminate dimensional differences;

[0062] S2.2 Maps a one-dimensional time series (i.e., the time series of each parameter) to a two-dimensional matrix using Gram angle field, preserving time-domain correlation;

[0063] S2.3 Convolutional Neural Networks (CNN) are used to extract cross-parameter coupling features from the matrix and generate an N×N correlation matrix.

[0064] Preferably, the update frequency of the spatial correlation matrix is ​​positively correlated with the rate of change of the parameters, specifically satisfying the following conditions:

[0065]

[0066] Where α is the preset sensitivity coefficient, x_i is the instantaneous value of the i-th parameter, and the vertical line represents the condition separator.

[0067] S3. Establish a variable-weight collaborative decision-making model based on Long Short-Term Memory (LSTM) network, and dynamically adjust the contribution weight of the data collected in step S1 in the final detection result according to the real-time working conditions.

[0068] As an optional embodiment, this step may include:

[0069] S3.1 Establish a bidirectional LSTM network that includes an attention mechanism;

[0070] Specifically, the real-time weight distribution output by the bidirectional LSTM network, normalized by Softmax, is directly used for subsequent multi-parameter data fusion (step S5), determining the contribution ratio of each parameter in the final detection result. The environmental interference compensation calculated by the Bayesian network in step S5 is fed back to the Bi-LSTM model (step S3), dynamically adjusting the weights to counteract the impact of environmental interference. The sensor data optimized by the hybrid calibration algorithm in step S4 is used as input to the LSTM, improving the accuracy of its weight calculation.

[0071] S3.2 Set the forgetting factor γ = 0.9-0.99 to dynamically attenuate the weighting effect of historical data;

[0072] S3.3 The real-time weight distribution of each parameter is output through the Softmax function.

[0073] In practical implementation, parameter importance can be used as training labels. Supervised learning enables the LSTM to learn to output raw scores related to parameter importance. Softmax property: Ensures output values ​​are in the (0,1) interval and sum to 1, meeting the basic requirements of weight allocation. Dynamic adaptability: The LSTM can dynamically adjust the output scores based on changes in input data, thus reflecting the changes in the relative importance of each parameter under different operating conditions.

[0074] In step S3.3, the sum of the real-time weights of each parameter is 1, that is, Σw_i=1, where w_i is the real-time weight of the i-th parameter.

[0075] Specifically, the working principle of LSTM is as follows:

[0076] (1) Network structure

[0077] Bidirectional design: It includes forward and backward LSTM layers, which scan time series data from the past to the future and from the future to the past respectively, to fully capture context dependencies.

[0078] Forward LSTM: Processing historical data;

[0079] Backward LSTM: Processes reverse data.

[0080] Attention mechanism: Assign importance scores to different time steps to highlight key periods (such as moments of sudden temperature changes).

[0081] (2) Dynamic weight calculation

[0082] Forgetting factor (γ=0.9-0.99): controls the rate of decay of historical memory. For example, γ=0.95 means that 95% of historical information is retained, which can adapt to gradual and sudden changes.

[0083] Softmax output: Maps the LSTM hidden states to a weight vector. ),satisfy This ensures that the weights are normalized.

[0084] (3) Synergy with physical coupling

[0085] Input features: The spatial correlation matrix (quantizing the nonlinear coupling strength between parameters) from GAF-CNN is received as an auxiliary basis for weight decision.

[0086] S4. A hybrid calibration algorithm is used to calibrate the measured parameters of the final sensor output after processing. The hybrid calibration algorithm combines state estimation by Kalman filtering with parameter optimization by Particle Swarm Optimization (PSO).

[0087] As an optional embodiment, this step may include:

[0088] S4.1 First-order error elimination of sensor noise is achieved using Kalman filtering;

[0089] S4.2 An improved particle swarm optimization algorithm is used for second-order nonlinear error compensation.

[0090] In step S4.2, preferably, the inertia weight ω is calculated according to the formula ω=ω_max-(ω_max-ω_min)·(k / K). 2 Adaptive adjustment, where k is the current iteration number, K is the total iteration number, ω_max is the maximum value of the inertia weight, and ω_min is the minimum value of the inertia weight.

[0091] S5. Calculate the compensation amount of environmental interference parameters through Bayesian network to correct the output value of the corresponding sensor, and feed the compensation result back to the variable weight collaborative decision-making model.

[0092] In this step, the following processes are performed simultaneously:

[0093] 1) Directly correct the output value of the corresponding sensor;

[0094] 2) As a feature input to the variable weight collaborative decision-making model, the weight distribution coefficients of each parameter are dynamically adjusted.

[0095] As an optional embodiment, this step may include:

[0096] S5.1 Construct a directed acyclic graph (DAG) of environmental disturbance parameters and measurement errors, with the conditional probabilities of nodes obtained through training on historical data;

[0097] S5.2 uses the Markov Chain Monte Carlo (MCMC) method for posterior probabilistic inference to output the optimal compensation amount.

[0098] In step S5.2, the preferred optimal compensation amount is ΔY = argmax P(Y|X_E), where X_E is the environmental parameter vector, argmax is a function that calculates the parameters (set) of a function, P represents the conditional probability, and ΔY is the compensation amount for the measured parameter Y.

[0099] In the specific implementation of steps S2-S5 above, the nonlinear coupling strength is as follows: In dynamic coupling modeling, time-series data (such as pressure and temperature) are converted into a two-dimensional matrix using GAF, preserving temporal correlation. Cross-parameter features are extracted using CNN to generate a coupling coefficient matrix. An LSTM network analyzes the temporal relationship of parameters and outputs weights based on the coupling matrix. Kalman filtering eliminates first-order noise from the sensor, the improved PSO algorithm compensates for second-order nonlinear errors, and a Bayesian network compensates for the input environmental parameters X_E (such as temperature and steam density), outputting the optimal compensation amount ΔY=argmax P(Y|X_E).

[0100] The collaborative workflow of these three elements, taking industrial pressure pipeline monitoring as an example:

[0101] Data acquisition: Simultaneous measurement of pressure (P), temperature (T), flow rate (F), ambient temperature, and steam density.

[0102] Coupling analysis: GAF+CNN revealed that the coupling coefficient between P and T is 0.72.

[0103] Weighting: When there is a sudden temperature change, LSTM increases the weight of T from 0.3 to 0.4 and decreases the weight of P to 0.5.

[0104] Compensation calculation: The Bayesian network outputs a pressure compensation of ΔP = 0.23 MPa based on the change in steam density.

[0105] Final output: Based on the combined weights and compensation, output the corrected high-precision pressure value.

[0106] S6. Output the high-precision detection data obtained after processing in step S5.

[0107] Compared with existing technologies, this invention provides a high-precision instrument data processing method for multi-parameter collaborative detection. It offers advantages such as improved data processing accuracy, reduced response speed, and enhanced anti-interference capabilities. It addresses the problems of existing instrument data processing methods, which treat multiple parameters discretely, neglecting the physical coupling between parameters; compensation algorithms often rely on fixed correction coefficients, failing to adapt to dynamic changes in operating conditions; and the lack of error propagation suppression mechanisms during multi-sensor data fusion, leading to a decrease in system-level accuracy. The technical solution of this invention has the following beneficial effects:

[0108] 1. This high-precision instrument data processing method for multi-parameter collaborative detection maps time-domain parameters into two-dimensional spatial features in real time by setting a correlation matrix combining Gram angle field (GAF) and CNN, thereby improving the recognition accuracy of pressure-temperature coupling coefficient and reducing the error of multi-parameter collaborative decision-making.

[0109] 2. This high-precision instrument data processing method for multi-parameter collaborative detection uses an LSTM network with a forgetting factor (γ=0.95) to calculate parameter weights in real time, reducing the weight adjustment delay when operating conditions change abruptly. By deploying a Bayesian network and an MCMC inference engine, it calculates environmental compensation in real time, reducing pressure measurement deviation caused by steam density fluctuations and improving the signal-to-noise ratio (SNR) under electromagnetic interference (EMI).

[0110] 3. This high-precision instrument data processing method for multi-parameter collaborative detection adopts a dynamic time warping (DTW) hardware accelerator to align asynchronous data in real time, reducing the phase difference between flow and pressure signals and improving the timeliness of data fusion. It implements a three-stage pipeline architecture through FPGA (Field Programmable Gate Array) and executes LSTM inference in real time.

[0111] Implementation Case: Multi-parameter Monitoring System for Industrial Pressure Pipelines 1. Technical Background

[0113] A waste-to-energy plant in Zhejiang Province needs to monitor the pressure, temperature, and flow parameters of high-pressure pipelines in real time. Traditional instruments have an error of ±2.5% under steam disturbance conditions, which cannot meet the requirements for safe production.

[0114] 2. Implementation Steps

[0115] 2.1 Data Acquisition

[0116] Sensor configuration: piezoelectric pressure sensor (range 0-10MPa, accuracy 0.1%), PT100 temperature sensor (-50~300℃, ±0.5℃), vortex flow meter (accuracy class 0.5).

[0117] Synchronous sampling: The three channels are synchronized using the IEEE 1588 protocol (deviation < 1 μs).

[0118] 2.2 Dynamic Coupling Modeling

[0119] Convert the time-series data (sampling rate 1kHz) of pressure (P), temperature (T), and flow rate (F) into a Gram matrix:

[0120] # Example code snippet

[0121] from pyts.image import GramianAngularField

[0122] gaf = GramianAngularField(image_size=100)

[0123] X_gaf = gaf.fit_transform(XT) # X is the standardized data of P / T / F

[0124] Features are extracted using a dynamic coupling analysis model to obtain the correlation matrix:

[0125]

[0126] 2.3 Adaptive Compensation

[0127] Dynamic adjustment of LSTM weights (128 hidden nodes):

[0128] The steady-state weights are: P(0.6) / T(0.3) / F(0.1).

[0129] When there is a sudden temperature change (ΔT>5℃ / s): it is automatically adjusted to P(0.5) / T(0.4) / F(0.1).

[0130] Environmental compensation: The pressure measurement deviation caused by the change in steam density is calculated by Bayesian network, and the compensation amount ΔP = 0.23 MPa (when the steam density is from 2.1 to 3.5 kg / m³).

[0131] 3. Performance Verification

[0132] Performance verification is shown in Table 1:

[0133] Table 1

[0134]

[0135] On the other hand, embodiments of the present invention provide a high-precision instrument data processing system for multi-parameter collaborative detection, such as... Figure 2 As shown, it includes:

[0136] The multi-parameter synchronous acquisition module 10 is used to synchronously acquire N measured parameters and M environmental interference parameters of the target device through a multi-source sensor array to form a time series dataset; in specific implementation, this module may include a high-precision ADC circuit and an anti-aliasing filter.

[0137] The coupling analysis module 20 is used to construct a dynamic coupling analysis model, which uses Gram angle field to convert the time series dataset into a spatial correlation matrix and quantifies the nonlinear coupling strength between parameters. In specific implementation, this module can be configured to perform GAF transformation and CNN feature extraction.

[0138] The dynamic decision module 30 is used to establish a variable weight collaborative decision model based on a long short-term memory network, and dynamically adjust the contribution weight of the data collected in the multi-parameter synchronous acquisition module 10 in the final detection result according to the real-time operating conditions; in specific implementation, this module can be an LSTM accelerator implemented with an integrated FPGA.

[0139] The online calibration module 40 is used to perform online calibration of the measured parameters of the final detection output of the sensor after processing using a hybrid calibration algorithm. The hybrid calibration algorithm combines state estimation by Kalman filtering with parameter optimization by particle swarm optimization.

[0140] The environmental compensation module 50 is used to calculate the compensation amount of environmental interference parameters through a Bayesian network to correct the output value of the corresponding sensor, and to feed the compensation result back to the variable weight collaborative decision-making model; in specific implementation, this module may include a multi-dimensional sensor array of temperature / humidity / electromagnetic field.

[0141] The data output module 60 is used to output the high-precision detection data obtained after processing by the environmental compensation module 50. In specific implementations, this module can support Modbus / TCP and IEEE 1588 protocols.

[0142] The apparatus of this embodiment can be used to perform Figure 1 The technical solutions of the method embodiments shown are similar in principle and in effect, and will not be described again here.

[0143] Preferably, the dynamic decision-making module 60 adopts a pipelined architecture, including:

[0144] First-stage pipeline: Parameter weight calculation, cycle 5ns;

[0145] Second-stage pipeline: Environmental compensation is superimposed, with a cycle of 3ns;

[0146] Third-level pipeline: Data validity verification, cycle 2ns.

[0147] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps described in any of the above method embodiments.

[0148] Embodiments of the present invention also provide an application program that is executed to implement the method provided in any embodiment of the present invention.

[0149] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0150] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on its differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments. For ease of description, the above devices are described by dividing them into various functional units / modules. Of course, in implementing this invention, the functions of each unit / module can be implemented in one or more software and / or hardware.

[0151] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0152] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A high-precision instrument data processing method for multi-parameter collaborative detection, characterized in that, include: Step S1: Simultaneously collect N measured parameters and M environmental interference parameters of the target device through a multi-source sensor array to form a time series dataset; Step S2: Construct a dynamic coupling analysis model, use Gram angle field to convert the time series dataset into a spatial correlation matrix, and quantify the nonlinear coupling strength between parameters; Step S3: Establish a variable-weight collaborative decision-making model based on a long short-term memory network, and dynamically adjust the contribution weight of the data collected in step S1 in the final detection result according to the real-time operating conditions; Step S4: The measured parameters of the final sensor output after processing are calibrated online using a hybrid calibration algorithm, which combines state estimation by Kalman filtering with parameter optimization by particle swarm optimization. Step S5: Calculate the compensation amount of environmental interference parameters through a Bayesian network to correct the output value of the corresponding sensor, and feed the compensation result back to the variable weight collaborative decision-making model; Step S6: Output the high-precision detection data obtained after processing in step S5; Specifically, the construction of the dynamic coupling analysis model in step S2 includes: Step S2.1: Normalize the time series of each parameter to eliminate dimensional differences; Step S2.2: Map the one-dimensional time series to a two-dimensional matrix using Gram angle field, preserving temporal correlation; Step S2.3: Use a convolutional neural network to extract cross-parameter coupling features from the matrix and generate an N×N correlation matrix; In step S2.3, the update frequency of the correlation matrix is ​​positively correlated with the rate of change of the parameters, specifically satisfying the following conditions: ； Where α is the preset sensitivity coefficient, x_i is the instantaneous value of the i-th parameter, and the vertical line represents the condition separator.

2. The high-precision instrument data processing method for multi-parameter collaborative detection according to claim 1, characterized in that, Step S3 includes: Step S3.1: Establish a bidirectional LSTM network that includes an attention mechanism; Step S3.2: Set the forgetting factor γ = 0.9-0.99 to dynamically attenuate the weighting effect of historical data; Step S3.3: Output the real-time weight distribution of each parameter using the Softmax function.

3. The high-precision instrument data processing method for multi-parameter collaborative detection according to claim 1, characterized in that, Step S4 includes: Step S4.1: Eliminate the first-order error of sensor noise using Kalman filtering; Step S4.2: Use an improved particle swarm optimization algorithm to perform second-order nonlinear error compensation.

4. The high-precision instrument data processing method for multi-parameter collaborative detection according to claim 3, characterized in that, In step S4.2, the inertial weight ω is calculated using the formula ω=ω_max-(ω_max-ω_min)·(k / K). 2 Adaptive adjustment, where k is the current iteration number, K is the total iteration number, ω_max is the maximum value of the inertia weight, and ω_min is the minimum value of the inertia weight.

5. The high-precision instrument data processing method for multi-parameter collaborative detection according to claim 1, characterized in that, Step S5 includes: Step S5.1: Construct a directed acyclic graph of environmental interference parameters and measurement errors, with the conditional probabilities of nodes obtained through training on historical data; Step S5.2: Use the Markov chain Monte Carlo method to perform posterior probability inference and output the optimal compensation amount.

6. The high-precision instrument data processing method for multi-parameter collaborative detection according to any one of claims 1-5, characterized in that, Step S1 includes: A dynamic time warping algorithm is applied to multi-source sensor data to solve the timing misalignment problem caused by asynchronous sampling, achieving an alignment accuracy of 1 / 100 of the sampling period.

7. A high-precision instrument data processing system for multi-parameter collaborative detection, characterized in that, include: The multi-parameter synchronous acquisition module is used to synchronously acquire N measured parameters and M environmental interference parameters of the target device through a multi-source sensor array to form a time series dataset. The coupling analysis module is used to construct a dynamic coupling analysis model, using Gram angle field to convert the time series dataset into a spatial correlation matrix and quantify the nonlinear coupling strength between parameters. The dynamic decision-making module is used to establish a variable weight collaborative decision-making model based on a long short-term memory network, and dynamically adjust the contribution weight of the data collected in the multi-parameter synchronous acquisition module in the final detection result according to the real-time working conditions. The online calibration module is used to calibrate the measured parameters of the final sensor output after processing using a hybrid calibration algorithm, which combines state estimation by Kalman filtering with parameter optimization by particle swarm optimization. The environmental compensation module is used to calculate the compensation amount of environmental interference parameters through a Bayesian network to correct the output value of the corresponding sensor, and to feed the compensation result back to the variable weight collaborative decision-making model. The data output module is used to output the high-precision detection data obtained by the environmental compensation module. Specifically, the construction of the dynamic coupling analysis model in the coupling analysis module includes: Step S2.1: Normalize the time series of each parameter to eliminate dimensional differences; Step S2.2: Map the one-dimensional time series to a two-dimensional matrix using Gram angle field, preserving temporal correlation; Step S2.3: Use a convolutional neural network to extract cross-parameter coupling features from the matrix and generate an N×N correlation matrix; In step S2.3, the update frequency of the correlation matrix is ​​positively correlated with the rate of change of the parameters, specifically satisfying the following conditions: ； Where α is the preset sensitivity coefficient, x_i is the instantaneous value of the i-th parameter, and the vertical line represents the condition separator.

8. The high-precision instrument data processing system for multi-parameter collaborative detection according to claim 7, characterized in that, The dynamic decision-making module adopts a pipelined architecture, including: First-stage pipeline: Parameter weight calculation, cycle 5ns; Second-stage pipeline: Environmental compensation is superimposed, with a cycle of 3ns; Third-level pipeline: Data validity verification, cycle 2ns.

Images