CN-121995829-A - Transmitter system based on multi-sensor fusion and edge AI and self-diagnosis method

CN121995829ACN 121995829 ACN121995829 ACN 121995829ACN-121995829-A

Abstract

The invention discloses a transmitter system based on multi-sensor fusion and edge AI and a self-diagnosis method, belonging to the technical field of industrial process control instruments, wherein the industrial intelligent transmitter system comprises a multi-sensor unit, a signal acquisition unit, a microcontroller, an AI coprocessor and a dual-mode communication unit; the system comprises a multi-sensing unit, a signal acquisition unit, a microcontroller, an AI coprocessor and a dual-mode communication unit, wherein the multi-sensing unit is used for synchronously acquiring original data, the signal acquisition unit is used for converting the original data into digital signals, the microcontroller is used for processing the digital signals and performing environment self-adaptive compensation to obtain compensated sensor data, the AI coprocessor is used for performing feature extraction and fault probability reasoning on the compensated sensor data to realize real-time self-diagnosis, and the dual-mode communication unit is used for transmitting compensated process control signals and diagnosis information in a multichannel mode. The system has the functions of environment self-adaptive compensation, real-time self-diagnosis and the like, and the problems of parameter drift and fault response delay in an industrial environment are remarkably improved.

Inventors

FANG XUDONG
TIAN BIAN
LIU HUAMING
ZHANG BEI
ZHANG ZHENGYI
WU CHEN
LI KE
SUN HAO
FANG ZIYAN
ZHANG ZHONGKAI

Assignees

西安交通大学

Dates

Publication Date: 20260508
Application Date: 20260130

Claims (10)

1. A multi-sensor fusion and edge AI-based transmitter system, comprising: The multi-sensor unit comprises a main sensor and an auxiliary sensor, and is used for synchronously acquiring original data, wherein the original data comprises main process parameters of an industrial process acquired by the main sensor and environment parameters acquired by the auxiliary sensor; The signal acquisition unit is connected with the multi-sensing unit and used for filtering and amplifying the acquired original data and converting the acquired original data into digital signals; the microcontroller is connected with the signal acquisition unit and is used for processing the digital signals and running a multi-sensor fusion algorithm to perform environment self-adaptive compensation so as to obtain compensated sensor data; The AI coprocessor is connected with the microcontroller and is used for running a preset neural network model, and performing feature extraction and fault probability reasoning on the compensated sensor data to realize real-time self-diagnosis; the dual-mode communication unit is connected with the microcontroller and the AI coprocessor and is used for transmitting the diagnosis result and the compensated sensor data in a multichannel mode.
2. The multi-sensor fusion and edge AI-based transmitter system of claim 1, wherein the multi-sensor fusion algorithm is: in the formula, In order to compensate for the post-primary sensor data, Is primary sensor original data, alpha is a temperature compensation coefficient, beta is a vibration influence factor, T is an integral upper limit, tau is an integral variable, T is temperature, alpha (tau) is a vibration acceleration time domain signal function, Is the reference temperature.
3. The transmitter system based on multi-sensor fusion and edge AI of claim 2, wherein the temperature compensation coefficient is obtained by establishing a mapping relationship between temperature and temperature compensation coefficient by a cubic polynomial, the cubic polynomial being: Wherein c 0 、c 1 、c 2 、c 3 is a constant coefficient.
4. The transmitter system based on multi-sensor fusion and edge AI of claim 1, wherein the primary sensor is a pressure sensor or a flow sensor, and the secondary sensor comprises at least two of a temperature sensor, a vibration sensor and a humidity sensor.
5. The transmitter system based on multi-sensor fusion and edge AI according to claim 1, wherein the signal acquisition unit comprises a signal conditioning circuit and a 24-bit sigma-delta analog-to-digital converter, the signal conditioning circuit comprises a filtering and amplifying circuit for performing preliminary filtering, noise reduction and amplification on the original multi-source data signal, and the 24-bit sigma-delta analog-to-digital converter is used for converting the analog signal preprocessed by the signal conditioning circuit into a digital signal and transmitting the digital signal to the microcontroller.
6. The transmitter system based on multi-sensor fusion and edge AI of claim 1, wherein the neural network model is a ResNet-8 model of INT8 quantization, the ResNet-8 model comprising an input layer, a convolution layer, a max-pooling layer, four residual blocks, a global average pooling layer, a full connection layer, and an output layer connected in sequence; The input layer receives the feature vector of 128 dimension multiplied by 3 channels, 3 channels respectively correspond to pressure, temperature and vibration data, the number of channels of the four residual blocks is 16, 32 and 32, and the output layer outputs probability distribution of three working conditions of normal working conditions, sensor drift, circuits or mechanical faults.
7. The transmitter system based on the multi-sensor fusion and the edge AI, which is disclosed in claim 1, is characterized in that the dual-mode communication unit comprises two modes of wired communication and wireless communication, wherein the wired communication realizes 4-20 mA standard analog signal output and is compatible with HART protocol output, and the wireless communication adopts LoRa WAN output.
8. The self-diagnosis method of the industrial intelligent transmitter is characterized by comprising the following steps of: S1, synchronously acquiring original data of a plurality of paths of sensors, and carrying out signal conditioning and analog-to-digital conversion to obtain digital signals; s2, denoising and filtering the digital signal, and running a multi-sensor fusion algorithm to perform environment self-adaptive compensation; S3, extracting time domain features and frequency domain features of the compensated data to generate a multi-channel feature vector; S4, inputting the multichannel feature vector into a trained ResNet-8 model, and outputting fault probability; and S5, performing local alarm or data uploading according to the fault probability value.
9. The industrial intelligent transmitter self-diagnostic method of claim 8, wherein the ResNet-8 model training process comprises the steps of: SA1, collecting synchronous pressure, temperature and vibration three-channel time sequence data in the operation of a plurality of industrial transmitters, and respectively constructing a training set and a verification set which comprise known fault modes; SA2, carrying out data enhancement and channel normalization processing on time sequence data in the training set and the verification set, extracting time domain features and frequency domain features, and generating corresponding multi-channel feature vector samples; SA3, initializing ResNet-8 model parameters, initializing weights by adopting He normal distribution on a convolution layer, initializing a batch normalization layer scaling factor gamma to be 1, and initializing an offset factor beta to be 0; And SA4, training ResNet-8 models by adopting a cross entropy loss function and an Adam optimization algorithm, setting a gradient clipping threshold, optimizing by adopting a self-adaptive moment estimation algorithm in the training process, and implementing early shutdown based on verification set loss so as to save the optimal models.
10. The industrial intelligent transmitter self-diagnosis method according to claim 8, wherein the trained ResNet-8 model is subjected to INT8 quantization, layer fusion and channel pruning three-level optimization before deployment.

Description

Transmitter system based on multi-sensor fusion and edge AI and self-diagnosis method Technical Field The invention belongs to the technical field of industrial process control instruments, and particularly relates to a transmitter system based on multi-sensor fusion and edge AI and a self-diagnosis method. Background In the field of industrial process monitoring, such as petrochemical industry, electric power, etc., transmitters are the core for acquiring key process parameters (e.g., pressure, temperature, flow, etc.). The main functions of the traditional transmitter are signal conversion and transmission of single parameters, and the following significant problems exist: (1) Environmental sensitivity, namely, complex and changeable industrial field environments (temperature fluctuation, vibration, electromagnetic interference and the like), which easily causes drift (parameter drift) of sensor measurement values and influences measurement accuracy and reliability. (2) The fault diagnosis is lagged, the fault diagnosis of the traditional transmitter often depends on a background system or periodic maintenance, the response delay is large, sudden faults or performance degradation are difficult to find and locate in time, and potential safety hazards exist. (3) The information island is characterized in that single sensor information is limited, the equipment or process state is difficult to comprehensively reflect, and process control signals and equipment state diagnosis information are usually transmitted in a mixed mode, so that the efficiency is low, and mutual interference is possible. (4) The intelligent degree is low, the localized data processing and decision making capability is lacking, the complex analysis and compensation are carried out by depending on an upper system, the real-time performance is poor, and the network load is increased. In the prior art, although intelligent elements are introduced, the method still has the defects in the aspects of effectively solving the influence of environment dynamic coupling, realizing high-real-time localization self-diagnosis, optimizing an information transmission framework and the like. Therefore, there is a need for an intelligent transmitter system that incorporates multiple sensors, achieves adaptive compensation of the environment, and performs real-time fault diagnosis. Disclosure of Invention The invention provides a transmitter system based on multi-sensor fusion and edge AI and a self-diagnosis method thereof, which have the functions of environment self-adaptive compensation, real-time self-diagnosis and the like and remarkably improve the problems of parameter drift and fault response delay in an industrial environment. In order to achieve the above purpose, the present invention adopts the following technical scheme: in a first aspect, the present invention provides a multi-sensor fusion and edge AI based transmitter system comprising: The multi-sensor unit comprises a main sensor and an auxiliary sensor, and is used for synchronously acquiring original data, wherein the original data comprises main process parameters of an industrial process acquired by the main sensor and environment parameters acquired by the auxiliary sensor; The signal acquisition unit is connected with the multi-sensing unit and used for filtering and amplifying the acquired original data and converting the acquired original data into digital signals; the microcontroller is connected with the signal acquisition unit and is used for processing the digital signals and running a multi-sensor fusion algorithm to perform environment self-adaptive compensation so as to obtain compensated sensor data; The AI coprocessor is connected with the microcontroller and is used for running a preset neural network model, and performing feature extraction and fault probability reasoning on the compensated sensor data to realize real-time self-diagnosis; the dual-mode communication unit is connected with the microcontroller and the AI coprocessor and is used for transmitting the diagnosis result and the compensated sensor data in a multichannel mode. Further, the multi-sensor fusion algorithm is: in the formula, In order to compensate for the post-primary sensor data,Is primary sensor original data, alpha is a temperature compensation coefficient, beta is a vibration influence factor, T is an integral upper limit, tau is an integral variable, T is temperature, alpha (tau) is a vibration acceleration time domain signal function,Is the reference temperature. Further, the temperature compensation coefficient is obtained by establishing a mapping relation between the temperature and the temperature compensation coefficient through a cubic polynomial, wherein the cubic polynomial is as follows: Wherein c 0、c1、c2、c3 is a constant coefficient. Further, the main sensor is a pressure sensor or a flow sensor, and the auxiliary sensor comprises at least two of a temperature sensor, a vibration