CN-121982853-A - Radio frequency coaxial electric wire outer conductor trimming alarm system and method

CN121982853ACN 121982853 ACN121982853 ACN 121982853ACN-121982853-A

Abstract

The application relates to the technical field of communication, discloses a trimming alarm system and method for an outer conductor of a radio frequency coaxial cable, and aims to solve the problems of lag monitoring in the trimming process, weak abnormality identification capability, lack of an active alarm mechanism and low defect detection rate caused by unfused multisource signals in the prior art. The method comprises the steps of collecting physical deformation and electromagnetic scattering parameter signals in real time through a multi-mode sensor, extracting characteristics such as edge flatness deviation, incision width fluctuation, stress concentration index, impedance continuity mutation and the like, generating abnormal characteristic vectors through abnormal characteristic fusion neural networks, calculating risk scores based on a Gaussian process risk assessment model, triggering grading alarm according to the scores and pushing the grading alarm to a man-machine terminal, and recording a data closed-loop optimization model. By the technical scheme, the application realizes high-sensitivity defect detection, dynamic adaptation process drift, low false alarm rate and graded linkage alarm, and remarkably improves the quality and production reliability of products.

Inventors

TAN FANGWEI
CHEN LI
XU LONGHAI
Qin Zhikun

Assignees

成都大唐线缆有限公司

Dates

Publication Date: 20260505
Application Date: 20260122

Claims (10)

1. A radio frequency coaxial cable outer conductor trimming alarm system, comprising: the sensing module comprises a laser displacement sensor, a strain gauge array and a vector network analyzer, wherein the laser displacement sensor is arranged along the movement track of a trimming cutter, the strain gauge array is stuck to the surface of the outer conductor near the trimming area, and the vector network analyzer is connected to the two ends of a coaxial cable through a coupling probe; The feature extraction module comprises an abnormal feature fusion neural network with a double-branch structure, wherein the first branch is a one-dimensional convolution neural network and is used for processing normalized physical deformation feature sequences, the second branch is a full-connection network and is used for processing normalized electromagnetic feature vectors, and the two branch outputs are spliced in feature dimensions and then are subjected to nonlinear fusion through two layers of full-connection layers to output 128-dimensional abnormal feature vectors; The trimming state risk assessment module is used for calculating a risk score of the current trimming state based on the trimming abnormal feature vector and in combination with a preset process tolerance threshold model, wherein the process tolerance threshold model is constructed by adopting dynamic Gaussian process regression; The alarm decision and execution module is used for triggering a grading alarm signal and outputting the grading alarm signal to the man-machine interaction terminal according to a comparison result of the risk score and a preset alarm threshold; and the data storage and feedback optimization module is used for recording historical trimming state data, alarm events and operator response behaviors and carrying out online iterative optimization on the process tolerance threshold model based on the recorded data.
2. The trimming alarm system for the outer conductor of the radio frequency coaxial line according to claim 1, wherein the sampling frequency of the laser displacement sensor is not lower than 10kHz, the measuring precision is +/-2 mu m, the strain gauge arrays are symmetrically arranged in a bridge connection mode by 4 high-sensitivity strain gauges, the working frequency range of the vector network analyzer covers 0.5GHz to 6GHz, and the S parameter measuring dynamic range is not lower than 90dB.
3. The trimming alarm system for the outer conductor of the radio frequency coaxial line according to claim 1, wherein the synchronous data acquisition of the multi-mode sensor array is realized through a hardware trigger signal, the synchronous precision is better than 10 microseconds, and the physical deformation and the electromagnetic characteristic data are ensured to be aligned strictly on a time axis.
4. The radio frequency coaxial line outer conductor trimming alarm system according to claim 1, wherein the mean function of the process tolerance threshold model is determined by the mean value of abnormal feature vectors of historical qualified trimming samples, the covariance function is a radial basis function, and the kernel parameters are updated on line by maximizing an edge likelihood function.
5. The trimming alarm system for the outer conductor of the radio frequency coaxial line according to claim 1, wherein the alarm decision and execution module sets two levels of alarm thresholds, wherein the primary alarm threshold is set to be the average value of risk scores of historical qualified samples plus 2 times of standard deviation, corresponding to yellow early warning, the advanced alarm threshold is set to be 1.5 times of the primary alarm threshold, corresponding to red stop signals, and the power source of the trimming equipment is automatically cut off and a station is locked.
6. The trimming alarm system for the outer conductor of the radio frequency coaxial line according to claim 1, wherein the data storage and feedback optimization module adopts an incremental learning mechanism, a model is triggered to be finely tuned after 50 groups of tagged trimming event data are accumulated, a loss function of the abnormal feature fusion neural network is a weighted cross entropy, weights are dynamically adjusted according to the inverse of the number of samples in each category, and a covariance matrix of a Gaussian process risk assessment model is updated through a sliding window mechanism, wherein the window length is 200 qualified samples.
7. An alarm method applied to the radio frequency coaxial line outer conductor trimming alarm system as claimed in any one of claims 1 to 6, characterized by comprising the steps of: Step S110, acquiring physical deformation signals and electromagnetic scattering parameter signals of the radio frequency coaxial cable outer conductor in the trimming process in real time through a multi-mode sensor array deployed at a trimming station; Step S120, performing time domain-frequency domain joint analysis on the physical deformation signal, extracting edge flatness deviation, incision width fluctuation and local stress concentration index, and performing S-parameter analysis on the electromagnetic scattering parameter signal, and extracting the external conductor impedance continuity mutation characteristic and return loss abnormal peak value; step S130, after normalizing the edge flatness deviation, the incision width fluctuation, the local stress concentration index, the impedance continuity abrupt change characteristic and the return loss abnormal peak value, inputting a pre-trained abnormal characteristic fusion neural network, and outputting a 128-dimensional trimming abnormal characteristic vector; Step S140, inputting the trimming abnormal feature vector into a risk assessment model based on a Gaussian process, and calculating a risk score of the current trimming state by combining a dynamically updated process tolerance threshold; Step S150, judging whether the risk score exceeds a preset primary alarm threshold or a high-level alarm threshold, triggering a yellow early warning signal if the risk score exceeds the primary alarm threshold, triggering a red shutdown signal if the risk score exceeds the high-level alarm threshold, and pushing the alarm signal to a man-machine interaction terminal; Step S160, recording original sensing data, abnormal feature vectors, risk scores, alarm levels and subsequent manual rechecking results of the trimming event, and constructing a closed-loop feedback data set for carrying out online fine adjustment on parameters of the abnormal feature fusion neural network and the Gaussian process risk assessment model.
8. The method for trimming an rf coaxial cable outer conductor according to claim 7, wherein in step S120, the notch profile data obtained by the laser displacement sensor is fourier transformed to extract high frequency component energy quantification, the notch width fluctuation is calculated by sliding window standard deviation of difference values measured by the upstream and downstream laser displacement sensors, and the local stress concentration index is defined as the ratio of the maximum principal strain to the average principal strain in the strain gauge array.
9. The method according to claim 7, wherein in step S140, the covariance matrix of the gaussian process risk assessment model is dynamically updated by a sliding window mechanism, and the window contains up to 200 qualified samples.
10. The method for trimming and alarming the outer conductor of the radio frequency coaxial cable according to claim 7, wherein in the step S160, a weighted cross entropy loss function is adopted for fine tuning of the model, the abnormal sample weight is the inverse number of the abnormal sample weight, and the mean function and the covariance matrix of the Gaussian process model are updated on line based on a closed loop feedback data set with a manual review tag, so that the adaptability of the system to process drift and material batch difference is ensured.

Description

Radio frequency coaxial electric wire outer conductor trimming alarm system and method Technical Field The invention belongs to the technical field of communication, and particularly relates to a trimming alarm system and method for an outer conductor of a radio frequency coaxial cable. Background The radio frequency coaxial cable is used as a key carrier for high-frequency signal transmission, is widely applied to the fields of communication, radar, broadcast television, test measurement and the like, and the structure of the radio frequency coaxial cable is generally composed of an inner conductor, a dielectric layer, an outer conductor and an outer sheath, wherein the outer conductor not only plays a role in shielding electromagnetic interference, but also plays a decisive role in signal integrity and transmission stability. In the production and processing process of the radio frequency coaxial cable, the molding quality of the outer conductor directly influences the product performance, especially in the trimming process, if problems such as offset cutting, continuous cutting, burrs or excessive cutting occur, the shielding efficiency is reduced, the standing wave ratio is deteriorated, even the signal is leaked, and equipment malfunction or system failure can be caused when serious. The trimming process of the outer conductor of the radio frequency coaxial line is a key link in the manufacture of the cable, and aims to accurately remove redundant parts of the outer conductor on the premise of not damaging an inner layer medium and the inner conductor so as to realize end face regularity and electrical connection reliability. The process generally depends on the control of a mechanical cutter matched with a vision or position sensor, and high requirements are put on machining precision, real-time feedback and abnormal response capability. The problems that monitoring means are lagged, abnormality identification capability is weak, an active alarm mechanism is lacking and the like in the trimming process of an outer conductor of a radio frequency coaxial line in the prior art generally exist, firstly, most systems only rely on preset travel or time to control cutting actions and cannot sense whether the trimming state meets standards in real time, secondly, the online evaluation capability of trimming quality is lacking, normal trimming and abnormal working conditions such as cutter abrasion, material deformation or positioning offset are difficult to distinguish, and even if trimming failure or outer conductor damage occurs, the systems cannot interrupt the flow in real time and trigger effective alarm to cause defective products to flow into subsequent working procedures, and finally, physical signals (such as vibration, current and image characteristics) in the trimming process are not subjected to multi-dimensional fusion analysis in the prior art, and a high-robustness abnormal discrimination model is difficult to construct. The defects are particularly prominent in a high-speed continuous production line, and mass quality accidents are very easy to cause, so that a system and a method for monitoring the trimming of the radio frequency coaxial electric wire outer conductor, which can sense in real time, intelligently judge and actively alarm, are needed. Disclosure of Invention The invention aims to make up the defects of the prior art, and provides a trimming alarm system and a trimming alarm method for an outer conductor of a radio frequency coaxial cable, which can effectively solve the problems in the background art. In order to achieve the aim, the invention provides the technical scheme that the trimming alarm system for the radio frequency coaxial electric wire outer conductor comprises a trimming state sensing module for sensing the trimming state of the radio frequency coaxial electric wire outer conductor, and is used for collecting the physical deformation and electromagnetic characteristic change data of the radio frequency coaxial electric wire outer conductor in the trimming process in real time; the system comprises a physical deformation and electromagnetic property change data acquisition module, a trimming abnormal feature extraction module, a trimming state risk assessment module, an alarm decision and execution module, a data storage and feedback optimization module, a human-computer interaction terminal, a data storage and feedback optimization module, a processing module and a processing module, wherein the physical deformation and electromagnetic property change data acquisition module is used for carrying out multidimensional feature fusion and abnormal pattern recognition to generate a trimming abnormal feature vector; On the other hand, the trimming alarm method for the radio frequency coaxial electric wire outer conductor comprises the following specific steps: in step S110 of the process, acquiring physical deformation signals and electromagnetic scattering parameter sig