Remote automatic metering system and method for belt scale

Abstract

The present application belongs to the technical field of metrology system, and particularly relates to a remote automatic metering system and method for a belt scale, which comprises a high-frequency state spectrum acquisition module for generating a high-dimensional time-series state spectrum; a phase space reconstruction and manifold analysis module for outputting a stability deviation index; an abnormal pattern decoupling and attribution diagnosis module for identifying a dominant physical quantity and attributing a physical fault root cause; a nonlinear matrix generation module for constructing a dynamic correction matrix to analyze a high-fidelity metering result; and a reference adaptive evolution module for continuously tracking an evolution trajectory of the dynamic correction matrix. The present application adopts a technical scheme of multi-source data fusion and dynamics manifold analysis for state evaluation, deviation decoupling attribution for fault diagnosis, and combination of nonlinear dynamic correction and reference adaptive evolution, and can realize dynamic compensation of metering errors and autonomous learning and adaptation of system states, and improve the metering accuracy and long-term stability of the belt scale under complex working conditions.

Claims

1. Belt scale remote automatic metering system, characterized in that, The method comprises the following steps: a high-frequency state spectrum acquisition module is used to acquire and integrate multi-source heterogeneous data in real time from a belt scale sensor array, including weighing, speed, vibration and belt tension, to generate a high-dimensional time series state spectrum; wherein the weighing data, speed data, vibration data and belt tension data are aligned and synchronized on the timestamp to form a time-aligned multi-source data package; a phase space reconstruction and manifold analysis module is used to map a reference dynamic attractor in an initial calibration stage, and reconstruct the high-dimensional time series state spectrum into a real-time running trajectory through delay embedding, and output a stability deviation index by calculating the geometric manifold distance between the reference dynamic attractor and the real-time running trajectory; an abnormal mode decoupling and attribution diagnosis module is used to calculate a deviation vector from the real-time running trajectory and the reference dynamic attractor and perform multi-dimensional component orthogonal decomposition when the stability deviation index exceeds a preset deviation threshold, to identify a dominant physical quantity and attribute a physical fault root cause; a nonlinear matrix generation module is used to dynamically derive a parameterized correction operator according to the physical fault root cause, and perform real-time parameterization on the parameterized correction operator using the stability deviation index, to construct a dynamic correction matrix, and perform online transformation on the high-dimensional time series state spectrum using the dynamic correction matrix to analyze a high-fidelity measurement result; a reference adaptive evolution module is used to continuously track the evolution trajectory of the dynamic correction matrix, and when it is determined that the dynamic correction matrix converges to a stable non-zero state, a re-calibration process is started to iteratively update the reference dynamic attractor; the phase space reconstruction and manifold analysis module comprises: a reference attractor construction unit is used to perform phase space reconstruction processing on the high-dimensional time series state spectrum in the initial calibration stage to establish a reference dynamic attractor; a real-time trajectory reconstruction unit is used to reconstruct the real-time high-dimensional time series state spectrum through delay embedding to generate a real-time running trajectory; a deviation index calculation unit is used to calculate the geometric manifold distance between the reference dynamic attractor and the real-time running trajectory to generate a stability deviation index; the abnormal mode decoupling and attribution diagnosis module comprises: a deviation vector analysis unit is used to lock an instantaneous state point on the real-time running trajectory and calculate the Euclidean space vector between the instantaneous state point and the neighboring point of the reference dynamic attractor to generate a deviation vector when the stability deviation index exceeds the deviation threshold; a dominant component identification unit is used to perform principal component analysis on the deviation vector, project it onto the orthogonal physical quantity basis defined by the multi-source heterogeneous data, and identify the dominant physical quantity according to the contribution degree projection component; a fault root cause reasoning unit is used to perform time series evolution feature reasoning based on the dominant physical quantity to output a physical fault root cause; the abnormal mode decoupling and attribution diagnosis module is also used to: establish a cross-space mapping relationship between the deviation vector and the time-aligned multi-source data package to generate a deviation correlation graph; locate and extract a related original data segment from the time-aligned multi-source data package according to the deviation correlation graph to generate a physical deviation trace signal; the nonlinear matrix generation module comprises: an operator structure deduction unit is used to analyze the physical fault root cause into a disturbance term, and deduce a parameterized correction operator according to the influence of the disturbance term; A real-time parameter filling unit is configured to calculate undetermined coefficients in a parameterized correction operator structure based on a stability deviation index, and generate an instantiated correction operator; A matrix transformation and analysis unit is configured to express the instantiated correction operator as a dynamic correction matrix, and perform online transformation on the high-dimensional time-series state spectrum using the dynamic correction matrix to analyze a high-fidelity measurement result; The reference adaptive evolution module includes: A convergence state identification unit is configured to continuously track an evolution trajectory of the dynamic correction matrix, and generate a recalibration trigger signal when the dynamic correction matrix converges to a stable non-zero state; A historical deviation induction and weighting strategy generation unit is configured to respond to the recalibration trigger signal, statistically analyze the physical deviation traceability signal, identify main physical quantities contributing to the deviation, and generate a reference update weight vector; A weighted iterative update unit is configured to start a recalibration process, apply the reference update weight vector to perform weighted phase space reconstruction on the high-dimensional time-series state spectrum, and iteratively update a reference dynamic attractor.

2. The remote automatic weighing system of claim 1, wherein, The high-frequency state spectrum acquisition module includes: A data acquisition unit is configured to acquire weighing data through a weighing sensor, speed data through a speed sensor, vibration data through a vibration sensor, and belt tension data through a tension sensor; A time series synchronization unit is configured to align and synchronize the weighing data, speed data, vibration data, and belt tension data in time stamp to form a time-aligned multi-source data packet; A state spectrum generation unit is configured to perform dimension expansion and standardization processing on the time-aligned multi-source data packet to generate a high-dimensional time-series state spectrum.

3. The remote automatic weighing system of claim 1, wherein, The reference dynamic attractor includes a topological center manifold and a probability boundary envelope, wherein: The topological center manifold is used to calibrate the core dynamic behavior path of the belt scale under ideal fault-free working conditions; The probability boundary envelope is used to define the acceptable random range of the real-time running trajectory fluctuating around the topological center manifold.

4. The remote automatic weighing system of claim 1, wherein, Analyzing the high-fidelity measurement result includes: Converting the instantiated correction operator into a dynamic correction matrix, and projecting the high-dimensional time-series state spectrum into a low-dimensional correction subspace using the dynamic correction matrix to generate a correction state vector; Extracting the material net weight scalar component from the correction state vector to output the high-fidelity measurement result.

5. Belt scale remote automatic weighing method, based on the belt scale remote automatic weighing system according to any one of claims 1-4, characterized in that the steps It includes: Real-time acquisition and integration of multi-source heterogeneous data from a belt scale sensor array, including weighing, speed, vibration, and belt tension, to generate a high-dimensional time-series state spectrum; In the initial calibration stage, map the reference dynamic attractor, and reconstruct the high-dimensional time-series state spectrum into a real-time running trajectory through delay embedding, and output the stability deviation index by calculating the geometric manifold distance between the reference dynamic attractor and the real-time running trajectory; When the stability deviation index exceeds the preset deviation threshold, calculate the deviation vector of the real-time running trajectory and the reference dynamic attractor and perform multi-dimensional component orthogonal decomposition to identify the dominant physical quantity and attribute the physical fault root cause; According to the physical fault root cause, dynamically derive the parameterized correction operator, and use the stability deviation index to real-time parameterize the parameterized correction operator to construct a dynamic correction matrix, and use the dynamic correction matrix to perform online transformation on the high-dimensional time-series state spectrum to analyze the high-fidelity measurement result; The evolution track of the dynamic correction matrix is continuously tracked, and when it is determined that the dynamic correction matrix converges to a stable non-zero state, a re-calibration process is started, and the reference dynamic attractor is iteratively updated.

Images