Detection output method, device and equipment of weighing equipment and storage medium

Abstract

The invention relates to the technical field of weighing equipment detection, and discloses a detection output method, device and equipment for weighing equipment and a storage medium, and the method comprises the steps: carrying out the preprocessing and analog-to-digital conversion of original output signals of a plurality of weighing sensors in the weighing equipment, and obtaining synchronous digital signal data; bayesian multi-channel signal fusion and anomaly detection are carried out based on the synchronous digital signal data, and multi-channel fusion signal data and an anomaly detection identifier are obtained; performing inter-channel correlation analysis and adaptive weight distribution according to the abnormal detection identifier and the multi-channel fusion signal data to obtain a multi-channel weighted fusion result and a calibration compensation parameter; and carrying out real-time fault isolation and multi-channel working state identification on the multi-channel weighted fusion result based on the calibration compensation parameter, and generating a weighing detection result, the weight distribution can be automatically optimized according to the real-time working state and the signal quality of each weighing sensor channel, and the accuracy and the stability of the measurement of the weighing equipment are improved.

Description

Technical Field

[0001] The present invention relates to the field of weighing equipment detection technology, and in particular to a detection output method, device, equipment and storage medium for weighing equipment. Background Art

[0002] Traditional weighing equipment typically utilizes single-channel or simple multi-channel signal processing. These processing methods are susceptible to electromagnetic interference, temperature fluctuations, mechanical vibration, and other factors in complex industrial environments, resulting in unstable measurement accuracy. Existing multi-channel weighing systems often rely on simple signal averaging or weighted averaging for data fusion. These methods lack in-depth analysis of inter-channel correlations and are unable to effectively identify and process abnormal signals. Consequently, when a sensor fails, the reliability of the entire system is significantly reduced.

[0003] Current weighing equipment generally suffers from insufficient real-time performance during signal processing. This is particularly true for fault detection and status identification, which often rely on post-analysis or manual judgment, failing to achieve true real-time fault isolation and automatic compensation. Traditional anomaly detection methods, often based on simple threshold determinations, lack comprehensive utilization of signal statistical characteristics and historical data, and are prone to false positives and missed negatives, impacting the overall performance and reliability of the system. Summary of the Invention

[0004] The present invention provides a detection output method, device, equipment and storage medium for weighing equipment. The present invention can automatically optimize weight distribution according to the real-time working status and signal quality of each weighing sensor channel, thereby improving the accuracy and stability of weighing equipment measurement.

[0005] In a first aspect, the present invention provides a detection and output method for a weighing device, the detection and output method for a weighing device comprising:

[0006] Preprocess and convert the original output signals of the multi-channel weighing sensors in the weighing equipment into analog-to-digital signals to obtain synchronous digital signal data;

[0007] performing Bayesian multi-channel signal fusion and anomaly detection based on the synchronized digitized signal data to obtain multi-channel fused signal data and anomaly detection identification;

[0008] Performing inter-channel correlation analysis and adaptive weight allocation based on the anomaly detection identifier and the multi-channel fusion signal data to obtain a multi-channel weighted fusion result and calibration compensation parameters;

[0009] Real-time fault isolation and multi-channel working state identification are performed on the multi-channel weighted fusion result based on the calibration compensation parameters to generate a weighing detection result.

[0010] In a second aspect, the present invention provides a detection and output device for a weighing device, the detection and output device for the weighing device comprising:

[0011] A preprocessing module is used to preprocess and perform analog-to-digital conversion on the original output signals of the multi-channel weighing sensors in the weighing equipment to obtain synchronized digital signal data;

[0012] A multi-channel signal fusion module, configured to perform Bayesian multi-channel signal fusion and anomaly detection based on the synchronized digitized signal data to obtain multi-channel fused signal data and anomaly detection identification;

[0013] A weight allocation module is used to perform inter-channel correlation analysis and adaptive weight allocation based on the anomaly detection identifier and the multi-channel fusion signal data to obtain a multi-channel weighted fusion result and calibration compensation parameters;

[0014] A generation module is used to perform real-time fault isolation and multi-channel working state identification on the multi-channel weighted fusion result based on the calibration compensation parameter to generate a weighing detection result.

[0015] The third aspect of the present invention provides a computer device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the computer device executes the above-mentioned detection output method of the weighing device.

[0016] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned detection and output method of the weighing device.

[0017] The technical solution provided by the present invention uses independent channel isolation amplifiers for parallel isolation and amplification processing, effectively eliminating mutual interference between channels. Combined with low-pass filtering and synchronous sampling and holding technology, it ensures the synchronization and consistency of the signals of each channel. The Bayesian recursive update and prior and posterior probability fusion processing can fully utilize the statistical characteristics of historical data and the likelihood information of the current observation value to achieve more accurate and stable multi-channel signal fusion, and has stronger anti-interference ability than the traditional simple averaging method. Through a multi-channel state threshold judgment, abnormal state identification, asynchronous change point detection and statistical verification multi-layer detection mechanism, it can timely detect and accurately identify various abnormal situations, realize automatic removal of faulty channels and weight redistribution, and ensure that the system can still operate normally when some sensors fail. Based on the cross-channel calibration matrix and signal-to-noise ratio calculation, the adaptive weight distribution coefficient is dynamically adjusted, and the weight distribution can be automatically optimized according to the real-time working status and signal quality of each channel, thereby improving the accuracy and stability of the overall measurement. The integrated zero drift compensation and temperature drift synchronous compensation functions are realized through the sliding average filter and multi-layer judgment structure to achieve comprehensive compensation for multiple influencing factors, effectively suppress the impact of environmental changes on measurement accuracy, and extend the calibration cycle of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0019] Figure 1 Schematic diagram of the steps of the detection and output method of the weighing device in an embodiment of the present invention;

[0020] Figure 2 Schematic diagram of the structure of the detection and output device of the weighing equipment in an embodiment of the present invention;

[0021] Figure 3 It is a schematic block diagram of the structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] Embodiments of the present invention provide a detection output method, apparatus, device and storage medium for a weighing device. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , an embodiment of the detection output method of the weighing device in the embodiment of the present invention includes:

[0024] Step S1, preprocessing and analog-to-digital conversion are performed on the original output signals of the multi-channel weighing sensors in the weighing equipment to obtain synchronous digital signal data;

[0025] It is understandable that the execution subject of the present invention can be a detection output device of a weighing device, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0026] Specifically, the original output signals of multiple weighing sensors are obtained from the weighing equipment. These signals are weak voltage signals with amplitudes ranging from microvolts to millivolts. At the initial stage of signal processing, a high-precision independent channel isolation amplifier is introduced to perform parallel isolation and amplification of each signal. Each channel is independently configured with a set of amplifier modules. The gain of the amplifier is fixedly set to 1000 times. The input impedance must be higher than 10MΩ to reduce the input signal source load effect. At the same time, the common mode rejection ratio is required to be greater than 120dB to minimize the common mode interference and crosstalk effects that occur, ensuring the signal clarity and independence during the multi-channel amplification process, and outputting multi-channel isolated amplified signal data. A low-pass filter is configured in series after each channel of the above-mentioned isolated amplified signal. The filter cutoff frequency is designed to be 100Hz, thereby effectively filtering out noise components with frequencies higher than the working bandwidth, and obtaining multi-channel filtered signal data processed by low-pass filtering. To ensure signal time-domain consistency and synchronization in the subsequent analog-to-digital conversion process, a synchronous sample-and-hold circuit is deployed after the filtering module. This circuit utilizes a high-speed latch structure with a hold time accuracy of 10 nanoseconds, ensuring simultaneous sampling of multiple signals within the same microsecond range. This outputs multi-channel sample-and-hold signal data. Based on this multi-channel sample-and-hold signal data, an automatic range switching and signal conditioning mechanism is introduced to adapt to the output characteristics of load cells with varying ranges. By monitoring the input signal's amplitude in real time, it dynamically switches between different amplification factors and input voltage ranges, covering a range from ±10mV to ±10V, achieving adaptive conditioning across a wide dynamic range. The signal conditioning module also automatically compensates for bias voltage, corrects baseline drift, and mitigates temperature effects, improving signal stability and consistency. The module then outputs multi-channel pre-processed signal data. This multi-channel pre-processed signal data is fed into the synchronous sampling module of the time-division multiplexer. Using a high-speed analog switch chip, switching times are controlled to less than 1 microsecond, minimizing temporal jitter during channel switching. The sampling system uses a high-performance 24-bit integrator-differential analog-to-digital converter for analog-to-digital conversion. The sampling frequency is set to 10kHz and is centrally scheduled via an FPGA control platform to ensure that each analog-to-digital converter shares the same low-jitter clock source. Clock jitter is controlled within 100 picoseconds, achieving strictly synchronized sampling between channels. Oversampling technology is introduced during the analog-to-digital conversion process, with an oversampling rate set to 256 times. A digital filter is used for noise shaping, effectively improving the accuracy and signal fidelity of the analog-to-digital conversion. Data is transmitted via a high-speed data bus with a bus bandwidth of 1Gbps, ensuring real-time and stable transmission of large-scale parallel data. The resulting synchronized digitized signal data exhibits high resolution, high synchronization, and low noise.

[0027] Multi-channel pre-processed signal data is input into a multi-channel time-division multiplexer. The time-division multiplexer utilizes high-speed analog switching technology. Each signal path is independently controlled, switching within a microsecond, effectively preventing signal distortion caused by switching delays. Through multiple high-speed analog switches, the system can time-sequence the analog signals from multiple channels, generating stable time-division multiplexed signal data. This time-division multiplexed signal data is synchronously sampled using a shared clock source. The system utilizes a highly stable, low-jitter unified clock source with jitter levels controlled to less than 100 picoseconds, ensuring that all channels complete sampling at the same time, eliminating data timing errors caused by clock skew. This shared clock source provides unified timing control for all sampling paths, ensuring that each sampling point is highly aligned on the time axis, and outputting synchronized sampled signal data. This synchronized sampled signal data is then fed into a multi-channel analog-to-digital converter for parallel digitization. The analog-to-digital converter (ADC) utilizes a 24-bit integrator-differential architecture, featuring a high sampling frequency and a high effective number of bits. The sampling frequency for each channel is set to 10kHz, and multiple ADC modules operate in parallel under unified scheduling by the FPGA, ensuring spatial multi-channel parallelism and strict temporal synchronization of data acquisition. Furthermore, to improve signal quality, an oversampling mechanism is introduced during the ADC conversion process, with an oversampling ratio set to 256x. This effectively suppresses quantization noise. Through oversampling and noise shaping techniques, the signal resolution and dynamic range are enhanced, resulting in high-precision raw digital signal data. The raw digital signal data is then digitally filtered using an adaptive finite impulse response filter structure. The filter coefficients are dynamically adjusted based on the spectral characteristics of the input signal, effectively suppressing high-frequency noise and interference signals while preserving the true characteristics of the signal waveform, ensuring high fidelity of the weighing signal in both the frequency and time domains. The filtered output signal is the synchronized digitized signal data.

[0028] Step S2: performing Bayesian multi-channel signal fusion and anomaly detection based on the synchronized digitized signal data to obtain multi-channel fused signal data and anomaly detection identification;

[0029] Specifically, the inter-channel correlation matrix is ​​calculated for the synchronized digitized signal data. By statistically analyzing the mean, standard deviation, and covariance information between the signals of each channel, a standardized correlation matrix is ​​established. Each element in the matrix quantifies the degree of linear correlation between any two channels, resulting in the channel correlation matrix data. Based on this, multivariable coupling transfer modeling is performed in conjunction with the synchronized digitized signal data. By defining the load-output response characteristics, an n-dimensional coupling transfer function matrix is ​​constructed. Each element of the matrix characterizes the dynamic response characteristics between the input and output of each channel. Based on the channel correlation matrix data and coupling transfer parameters, a Bayesian inference mechanism is used for recursive updating and prior and posterior probability fusion. Using historical data or the data distribution obtained during the training phase as the prior probability distribution, the likelihood function is calculated based on the current synchronized digitized observation data. The posterior probability distribution is then updated according to the Bayesian formula. Through this recursive update method, the system's estimate of the true signal state is continuously corrected and optimized, and the initial fused signal data is output. Multi-channel state threshold determination and abnormal state identification are performed on the initial fused signal data. Based on five pre-defined key threshold parameters, including a zero-point drift threshold of ±0.1% FS, a linearity threshold of 0.02% FS, a full-scale threshold of 95% FS, a fault determination threshold of ±5% FS, and a saturation detection threshold of 98% FS, the fused signal is subjected to real-time multi-channel state determination to identify zero-point anomalies, linearity deviations, overload faults, or signal saturation, thereby generating an anomaly detection flag. Furthermore, to capture sudden changes in the signal, asynchronous change point detection is performed on the initial fused signal data. Statistical methods such as the chi-squared test are used to monitor changes in the signal mean and variance in real time. A 99% confidence level is set. When a change point is detected, it is considered that the system has experienced asynchronous interference or a localized channel fault. To enhance detection reliability, statistical verification is combined to ensure the significance and robustness of change detection, avoiding false positives and negatives. Through these steps, multi-channel fused signal data is ultimately obtained.

[0030] Prior probability distributions are calculated based on channel correlation matrix data. By statistically analyzing the correlation matrix between channel signals, the inter-channel covariance characteristics are extracted, and a probabilistic model describing the joint distribution relationship of multiple channels is established. This statistical dependence is converted into prior probability parameters for Bayesian analysis. The prior probability of each channel includes the mean and variance information of the individual channel, as well as the relevant coupling characteristics with other channels. This ensures that the prior probability distribution accurately reflects the joint behavior of the multi-channel system in the unobserved state, resulting in the Bayesian prior probability parameters. Observation values ​​and likelihood functions are calculated for the synchronized digitized signal data based on the coupling transfer parameters. Based on the coupling transfer function matrix, the observed signal of each channel is dynamically mapped to the input and output characteristics of other channels, constructing an observation model that reflects the current input-output relationship of the system. By comparing the actual observed data with the model's predicted output, observation residuals are generated. Using these residuals, a corresponding likelihood function is constructed based on the Gaussian distribution assumption. The likelihood function describes the probability of the current observed data occurring under given coupling transfer model parameters, resulting in a specific likelihood function calculation result that directly reflects the consistency between the observed data and the system's prior expectations. The Bayesian prior probability parameters and the calculated likelihood function are input into the Bayesian formula to calculate the posterior probability. According to Bayes' theorem, the prior probability is multiplied by the likelihood function and then normalized to generate an updated posterior probability distribution. This posterior distribution incorporates both historical system state information and current real-time observations, enabling dynamic correction and optimization of the multi-channel weighing signal state. This process continuously updates the credibility assessment of each channel signal in real time while continuously inputting new observation data, effectively tracking changes in the system state. The posterior probability distribution data is then recursively weighted and multi-channel probability fusion is performed. During the recursive weight update process, the fusion weights of each channel signal are adjusted based on the posterior probability distribution. High-confidence channels are assigned higher fusion weights, while low-confidence or abnormal channels have their weights reduced, ultimately reducing their weights to zero if a severe anomaly is detected. Multi-channel probability fusion then performs a weighted superposition of the channel signals based on the updated weights to generate initial fused signal data with optimal overall estimation performance. During the fusion process, the smoothing factor and historical state memory mechanism are introduced to enhance the stability and anti-interference ability of signal fusion, and finally the initial fusion signal data is output.

[0031] Step S3: performing inter-channel correlation analysis and adaptive weight allocation based on the abnormality detection identifier and the multi-channel fusion signal data to obtain a multi-channel weighted fusion result and calibration compensation parameters;

[0032] Specifically, a cross-calibration matrix is ​​constructed for the multi-channel fused signal data based on anomaly detection indicators. The anomaly detection indicators provide a basis for determining the current health status and potential anomaly characteristics of each channel. Based on this, all channel signals calibrated as normal or with high confidence are selected as references. An n×n-dimensional cross-calibration matrix is ​​constructed to quantify the mapping relationship between the channels. Each element in the matrix represents the linear mapping coefficient between the target channel and the reference channel. The calibration matrix is ​​solved using the least squares method by minimizing the sum of squared residuals of the fused signal to determine the optimal calibration coefficients, thereby obtaining the cross-calibration matrix parameters. Based on the cross-calibration matrix parameters, the signal-to-noise ratio (SNR) and signal stability analysis are performed on the multi-channel fused signal data. The SNR calculation measures the power ratio of the effective component to the noise component of each channel signal, reflecting signal clarity and interference immunity. The signal stability analysis evaluates the signal's volatility and consistency in the time domain based on the rate of change of the signal's short-term mean and standard deviation. By combining the SNR and stability indicators, the health of each channel is comprehensively assessed, and adaptive weight allocation coefficients are calculated based on the evaluation results. The weighting coefficients are designed to assign higher weights to channels with higher signal-to-noise ratios and more stable signals, while channels with high signal noise or poor stability have their weights reduced accordingly. This results in an adaptive weight allocation strategy that dynamically responds to changes in system state. Combined with these adaptive weighting coefficients, zero drift and temperature drift compensation are performed to address systematic errors in the fused signal. Zero drift compensation utilizes a sliding average filter to track and correct the zero baseline of each channel signal in real time. The filter window is set to 100 sampling points, effectively eliminating zero drift caused by long-term operation. Temperature drift compensation utilizes an ambient temperature monitoring module. Based on real-time temperature data, each channel signal is temperature compensated according to a set temperature drift coefficient (controlled to ±0.01% / °C). This ensures that the weighing system maintains high accuracy and consistent output despite varying ambient temperatures. The compensated parameters, known as calibration compensation parameters, incorporate comprehensive correction information for zero and temperature drift. Based on the updated adaptive weighting coefficients, the multi-channel fused signal data is dynamically reweighted to generate the final multi-channel weighted fusion result. During the dynamic reallocation process, changes in channel signal quality are monitored in real time, and the weight coefficient of each channel is recalculated and adjusted according to the preset update period (such as 1 millisecond) to ensure that the fusion output can reflect the actual load information to the greatest extent at any time.

[0033] Step S4: performing real-time fault isolation and multi-channel working state identification on the multi-channel weighted fusion result based on the calibration compensation parameters to generate a weighing detection result.

[0034] Specifically, based on calibrated compensation parameters, a multi-channel weighted fusion result is used to identify the state of a single channel using a multi-layer decision structure. Based on the fused data after compensation parameter calibration, the output signal of each channel is analyzed separately. According to a preset operating state classification standard, the specific operating state of each channel is identified, including zero-point region, linear region, saturation region, overload region, and fault region. A comprehensive judgment is made based on indicators such as zero-point drift, output linearity, saturation, and signal amplitude change. The identification process adopts a multi-layer decision structure. The first layer performs preliminary screening based on a single indicator. The second layer further refines the decision by combining multiple indicators. The resulting single-channel operating state data clearly identifies whether each channel is in a normal or abnormal state, and subdivides the type and severity of the abnormality. Once the single-channel operating state data is obtained, it is used for inter-channel cross-validation and global consistency testing. The inter-channel cross-validation compares the consistency of the responses of each channel under the same load based on the correlation characteristics between different channels. A 2σ validation threshold is set: if a channel deviates by more than two standard deviations from the majority of other channels, it is identified as a potential fault channel. At the same time, a global consistency check is performed. Based on the statistical characteristics of all channel signals, the overall coordination and stability of the system are examined. Statistical deviation, homogeneity of variance tests, and consistency coefficient analysis are used to determine whether the system has widespread or hidden anomalies. Combining the cross-validation and consistency check results, faulty channel identification data is output to locate the anomaly channel and clarify the fault attributes. Based on the identified faulty channel data, the multi-channel weighted fusion results are automatically removed. Based on the faulty channel identification results, the fusion weight of the anomaly-detected channel is reduced to zero, and the weights are redistributed based on the remaining functioning channels to ensure the continuity and reliability of the fused output data. At this point, the basic requirement of the redundancy detection mechanism must be met: out of n channels, at least n-2 channels must remain in normal operation to form a valid redundant detection fusion result. The introduction of the redundancy mechanism significantly enhances the system's robustness in the event of channel failures, preventing the risk of overall weighing failure due to a single channel failure. The redundant detection fusion results are then combined to perform operating status classification and identification. By performing a multi-dimensional analysis of the changing trends, signal amplitude ranges, and statistical characteristics of the fused data, a comprehensive assessment is made of the specific operating state of the current weighing system, such as static load, dynamic load, overload risk, or no-load, generating multi-channel operating state identification data. This data includes not only the quantitative results of the current weighing load, but also the health status of each channel, abnormality records, and an overall system health assessment, ensuring the comprehensiveness and accuracy of the weighing test results. Based on this multi-channel operating state identification data, a weighing test result is generated. This test result includes real-time weighing values, working status reports for each channel, a list of faulty channels, statistical characteristics of the fused signal, and system health indicators, reflecting the operating status and performance level of the weighing system under the current operating conditions.

[0035] In this embodiment of the present invention, independent channel isolation amplifiers are used for parallel isolation and amplification processing, effectively eliminating mutual interference between channels. Low-pass filtering and synchronous sampling and holding techniques are combined to ensure the synchronization and consistency of signals in each channel. Bayesian recursive updating and prior and posterior probability fusion processing are used to fully utilize the statistical characteristics of historical data and the likelihood information of current observations, achieving more accurate and stable multi-channel signal fusion. Compared with traditional simple averaging methods, this method has stronger anti-interference capabilities. A multi-layer detection mechanism, including multi-channel state threshold determination, abnormal state identification, asynchronous change point detection, and statistical verification, can promptly detect and accurately identify various abnormal conditions, automatically remove faulty channels, and redistribute weights, ensuring that the system can still operate normally even when some sensors fail. Based on the inter-channel cross-calibration matrix and signal-to-noise ratio calculation, dynamic adjustment of adaptive weight distribution coefficients is achieved, automatically optimizing weight distribution based on the real-time operating status and signal quality of each channel, improving the accuracy and stability of the overall measurement. Integrated zero drift compensation and temperature drift synchronous compensation functions are implemented. Through a sliding average filter and multi-layer determination structure, comprehensive compensation for multiple influencing factors is achieved, effectively suppressing the impact of environmental changes on measurement accuracy and extending the calibration cycle of the equipment.

[0036] In a specific embodiment, the process of executing step S1 may specifically include the following steps:

[0037] Obtaining original output signals of multiple weighing sensors in a weighing device, and inputting the original output signals into independent channel isolation amplifiers for parallel isolation amplification to obtain multi-channel isolation amplified signal data;

[0038] Performing low-pass filtering on the multi-channel isolated amplified signal data to obtain multi-channel filtered signal data, and performing synchronous sampling and holding on the multi-channel filtered signal data to obtain multi-channel sampled and held signal data;

[0039] Automatic range switching and signal conditioning are performed based on the multi-channel sample-and-hold signal data to obtain multi-channel pre-processed signal data;

[0040] The multi-channel pre-processed signal data is subjected to time division multiplexer synchronous sampling and analog-to-digital conversion to obtain synchronous digitized signal data.

[0041] Specifically, the raw output signals of multiple load cells are obtained from within the weighing equipment. These signals are extremely low-amplitude analog voltage signals, ranging in magnitude from microvolts to millivolts, and are highly susceptible to external electromagnetic interference, thermal drift, and common-mode interference. Therefore, special attention is paid to signal integrity and interference resistance during the initial processing stage of the signal chain. To this end, the system design adopts an independent channel isolation amplifier architecture, inputting each load cell signal into a separate isolation amplification channel. Each isolation amplifier channel is equipped with a dedicated low-noise preamplifier module and features high input impedance (greater than 10MΩ) and an extremely high common-mode rejection ratio (greater than 120dB) to ensure that cross-channel signal interference and common-mode noise are effectively suppressed even in complex electromagnetic environments. The amplifier's gain is set at 1000 times and is designed with a wide dynamic range, ensuring linear amplification and high-fidelity transmission of the signal, regardless of the amplitude of the load cell output signal. The isolation amplifier utilizes a differential input structure and uses magnetic or optical isolation to achieve electrical isolation between signal channels, eliminating noise interference introduced by ground loops, thereby generating multi-channel isolated amplified signal data. The multi-channel isolated amplified signal data undergoes low-pass filtering, with a dedicated low-pass filter configured in series after each channel. The low-pass filter utilizes either a high-order passive filter circuit or an active filter topology, with a precisely set cutoff frequency of 100Hz. This effectively suppresses noise components above the effective signal frequency band while maintaining the integrity and accuracy of the signal waveform within the operating frequency band. The filtered signal reduces amplitude jitter and signal distortion introduced by high-frequency interference sources or power supply noise in the system, and outputs the multi-channel filtered signal data. To ensure high-precision synchronous sampling of each channel's signal during the subsequent analog-to-digital conversion stage, the multi-channel filtered signal data undergoes synchronous sample-and-hold processing. The sample-and-hold circuit design utilizes a high-speed latch mechanism, with each signal channel equipped with an independent sample-and-hold module. This latch and hold process can complete signal latching and holding within 10 nanoseconds, effectively eliminating errors caused by signal fluctuations or sampling delays. The synchronous sample-and-hold circuit utilizes a unified trigger clock control, ensuring that all channels complete sampling operations at the same time. This ensures highly synchronized data on the timeline, eliminates microsecond-level deviations between channels, and generates multi-channel sample-and-hold signal data. Based on the synchronously latched multi-channel sample-and-hold signal data, an automatic range switching and signal conditioning module is introduced to accommodate output amplitude fluctuations caused by load changes in the weighing sensor. The automatic range switching mechanism monitors the signal amplitude in real time and dynamically adjusts the front-end amplifier gain or input attenuation ratio based on the set dynamic range threshold. Covering a range from ±10mV to ±10V, this ensures that the signal remains in the optimal operating range even under varying load conditions, avoiding measurement errors caused by signal overflow or reduced quantization accuracy.At the same time, the signal conditioning module automatically eliminates DC offsets, corrects baseline drift, and performs front-end temperature compensation to ensure output signal stability and consistency under various operating conditions. This series of processes produces stable, reliable, and moderately amplitude multi-channel pre-processed signal data. This multi-channel pre-processed signal data is input into a time-division multiplexer (TDM) for synchronous sampling and analog-to-digital conversion. To maintain high speed and synchronization in the signal path, the system incorporates a multi-channel high-speed TDM architecture. The TDM is based on a high-speed analog switch array, with channel switching times controlled to less than 1 microsecond. This reduces delays in switching between channels and avoids synchronization errors caused by sampling sequence misalignment. The multi-channel signals are sequentially arranged through the TDM, forming a stable, continuous, multi-channel TDM signal stream. To achieve high-precision digitization, the TDM signal data is input into a multi-channel A / D converter module, utilizing a 24-bit integral-differential high-resolution ADC with a sampling frequency of 10 kHz per channel. To eliminate quantization noise and improve signal accuracy, an oversampling mechanism is introduced during the analog-to-digital conversion process. The oversampling rate is set to 256 times, and combined with digital noise shaping filtering, it effectively suppresses quantization error and increases the effective number of bits of the signal. All ADC channels share a unified, highly stable clock source with clock jitter less than 100 picoseconds. This ensures consistent sampling times across channels at the hardware level and avoids sampling asynchrony caused by time base drift. After analog-to-digital conversion, the raw digital signal data is digitally filtered using an adaptive finite impulse response filter. The filter coefficients are dynamically adjusted based on the actual spectral characteristics of the input signal to remove any high-frequency interference and background noise, while maximally preserving the signal's time-domain waveform and frequency-domain characteristics. The final output is synchronized digitized signal data.

[0042] In a specific embodiment, the step of performing time division multiplexing synchronous sampling and analog-to-digital conversion on the multi-channel pre-processed signal data to obtain synchronized digitized signal data may specifically include the following steps:

[0043] Inputting the multi-channel pre-processed signal data into a multi-channel time division multiplexer for high-speed analog switch switching to obtain time division multiplexed signal data;

[0044] Performing synchronous sampling of the time-division multiplexed signal data using a shared clock source to obtain synchronously sampled signal data;

[0045] The synchronous sampling signal data is input into a multi-channel analog-to-digital converter for parallel digital conversion and oversampling to obtain original digital signal data, and digital filtering is performed based on the original digital signal data to obtain synchronous digitized signal data.

[0046] Specifically, the multi-channel pre-processed signal data is input into a high-speed, multi-channel time-division multiplexer (TDM) module. The TDM utilizes high-speed analog switch array technology, integrating low-on-resistance, high-switching-speed MOSFET switching elements. Switching times are controlled to less than 1 microsecond, minimizing inter-channel delay and noise coupling during multi-channel signal switching. By precisely controlling the switching timing, each pre-processed signal is sequentially arranged in time, forming a continuous and stable TDM signal stream. This signal stream follows a fixed time sequence, ensuring that multi-channel signals are free of channel misalignment, overlap, or loss during sampling, thereby generating TDM signal data. To ensure high-precision synchronous sampling after multiplexing, the TDM signal data is synchronously sampled using a shared clock source. The system uses a high-stability, low-jitter crystal oscillator module as a unified clock source, resulting in a stable output clock frequency and jitter below 100 picoseconds, providing a precise and consistent clock signal for subsequent analog-to-digital conversion. All channel sampling units share a unified clock source, ensuring simultaneous sampling triggering. This eliminates time skew caused by asynchronous sampling clocks across channels and prevents temporal asynchrony across multiple channels. Synchronous sampling applies a unified timestamp to the time-division multiplexed signal stream and latches data at a controlled clock cycle. This allows accurate capture of even microsecond-level signal changes, resulting in synchronized sampled signal data with high temporal precision and high synchronization. The synchronized sampled signal data is input into a multi-channel analog-to-digital converter module for parallel digitization and oversampling. The analog-to-digital converter utilizes a 24-bit, high-resolution, integrator-differential architecture, with each channel equipped with an independent analog-to-digital conversion unit. The sampling frequency is set to 10kHz, meeting the requirements for high-precision weighing signal capture. Each analog-to-digital converter operates in parallel under a unified clock, ensuring time alignment across channels and avoiding delays and skew caused by sampling order. In addition to regular sampling, the analog-to-digital converter utilizes oversampling technology, with a 256x oversampling rate, to improve the effective resolution and dynamic range of the signal, resulting in the raw digital signal data. This raw digital signal data is then digitally filtered. The digital filtering module utilizes an adaptive finite impulse response (FIR) filter design. The filter coefficients are dynamically optimized and adjusted based on the input signal's spectral and noise characteristics. This filter eliminates high-frequency interference while maintaining the signal's waveform authenticity. The FIR filter's linear phase characteristics prevent signal distortion and phase shift, ensuring the time consistency and frequency accuracy of the filtered signal. The digital filter utilizes multi-decimation and multi-interpolation techniques, combined with an oversampling rate parameter, to dynamically adjust the filter order and bandwidth based on signal changes. This allows for flexible response to variations in frequency components, effectively removing factors such as ambient noise, power supply ripple, and random jitter during the sampling process. This results in synchronized digitized signal data.

[0047] In a specific embodiment, the process of executing step S2 may specifically include the following steps:

[0048] Performing inter-channel correlation matrix calculation on the synchronous digitized signal data to obtain channel correlation matrix data, and performing multivariable coupling transfer on the synchronous digitized signal data to establish coupling transfer parameters;

[0049] Performing Bayesian recursive updating and prior and posterior probability fusion based on the channel correlation matrix data and the coupling transfer parameters to obtain initial fused signal data;

[0050] Multi-channel state threshold determination and abnormal state identification are performed on the initial fusion signal data to obtain an abnormality detection identifier, and asynchronous change point detection and statistical verification are performed on the initial fusion signal data to obtain multi-channel fusion signal data.

[0051] Specifically, the inter-channel correlation matrix is ​​calculated for the synchronized digitized signal data. By analyzing the joint variation characteristics of each channel signal in the time domain, statistical methods are used to calculate the correlation coefficient between each pair of channel signals. These correlation coefficients form a symmetric correlation matrix. Each element in the matrix represents the degree of linear correlation between the corresponding two channel signals, with values ​​ranging from -1 to 1. A correlation coefficient close to 1 indicates positive correlation, close to -1 indicates negative correlation, and close to 0 indicates no significant linear relationship. This correlation matrix quantifies the dependency structure between multi-channel signals and reveals the coordinated variation patterns of signals during the weighing process caused by factors such as changes in external load distribution, mechanical structure vibration, or electrical coupling. This generates channel correlation matrix data. Based on the channel correlation matrix data and the synchronized digitized signal data, multivariable coupling transfer modeling is performed. An n-dimensional coupling transfer function matrix is ​​established, in which each element represents the influence of one channel's signal input on another channel's signal output in the frequency or time domain. Using experimental calibration or system identification techniques, based on historical sampling data, the mapping characteristics between the load input and signal response between each channel are identified. A multivariable dynamic model of the system is constructed, and transfer parameters reflecting the dynamic coupling characteristics between the channels are extracted. The coupling transfer parameters describe the behavior of different channels under complex factors such as force response, temperature drift, and resonant excitation, and can also reflect potential signal crosstalk, mechanical linkage, or electrical interference. Based on the channel correlation matrix data and the coupling transfer parameters, a Bayesian recursive update method is used for signal fusion. Based on historical statistical data and prior knowledge, a preliminary prior probability distribution for each channel signal is established. This prior distribution reflects the predicted signal state in the absence of new observations. Combined with the current synchronized digitized observation signal, the coupling transfer model calculates the expected output of the observation. Based on the deviation between the actual observation and the expected output, a likelihood function is constructed to quantify the probability of the current observation under the given model. Using the Bayesian formula, the prior probability and likelihood function are combined and normalized to obtain a posterior probability distribution. The posterior probability more accurately reflects the uncertainty and confidence of the current signal state. Through this recursive update mechanism, with each new sample arrival, the system uses the previous posterior distribution as the new prior distribution and updates the signal estimate in real time, thereby obtaining dynamically evolving initial fused signal data. The initial fusion signal is subjected to multi-channel state threshold determination and abnormal state identification. A series of key performance thresholds are preset, including zero drift threshold ±0.1% FS, linearity threshold 0.02% FS, full-scale threshold 95% FS, fault determination threshold ±5% FS, and saturation detection threshold 98% FS. The fusion output of each channel is tested in real time.By comparing the signal with a threshold, the system identifies abnormal signal channels that exceed the specified limits. Based on the type and severity of the anomaly, an anomaly detection indicator is generated. This indicator clearly indicates the abnormal channel number, anomaly type (such as overload, drift, saturation, or open circuit), the severity of the anomaly, and recommended remedial measures. Furthermore, to promptly detect potential sudden changes or hidden faults in the signal, asynchronous change point detection is performed on the initial fused signal data. This asynchronous change point detection relies on statistical methods, such as using the moving mean and standard deviation rate of change to monitor signal trends, or employing methods such as the Chi-square test and Cumulative Sum Chart (CUSUM) to determine the significance of signal changes. A statistical discrimination threshold is set at 99% confidence. When a signal change outside the preset control range is detected, the system determines the presence of an asynchronous change point, indicating possible load mutation, sensor anomaly, or environmental interference. To avoid false positives and false negatives, asynchronous change point detection is validated using multi-metric analysis and a multi-scale sliding window technique to ensure the accuracy and robustness of change point identification. Based on asynchronous change point detection, combined with statistical verification methods, the statistical characteristics of the change points are analyzed to ensure that the detected changes have sufficient statistical significance and practical engineering significance, avoiding misjudgments caused by random noise fluctuations or short-term signal jitter. Multi-channel fused signal data is output.

[0052] Among them, the asynchronous change point detection and statistical verification processing are performed on the initial fusion signal data according to the anomaly detection identification data to obtain multi-channel fusion signal data, including: establishing a sliding time window data caching mechanism based on the initial fusion signal data, and performing time series feature extraction and trend analysis processing on historical signal data to obtain multi-dimensional time series feature parameter data; constructing an autoregressive moving average model and parameter estimation processing according to the multi-dimensional time series feature parameter data, establishing a dynamic prediction model for each channel signal, and obtaining signal prediction model parameter data; inputting the signal prediction model parameter data into a Kalman filter for state prediction and residual calculation processing to obtain signal prediction value data and prediction residual data; performing cumulative sum control chart analysis and dynamic threshold adaptive adjustment processing based on the prediction residual data to establish a predictive fault detection mechanism to obtain predictive anomaly detection identification data; performing multi-layer anomaly fusion verification and statistical significance test processing according to the anomaly detection identification data and the predictive anomaly detection identification data to obtain multi-channel fusion signal data.

[0053] In a specific embodiment, the execution step performs Bayesian recursive updating and a priori-posteriori probability fusion processing based on the channel correlation matrix data and the coupling transfer parameter to obtain the initial fused signal data may specifically include the following steps:

[0054] Performing a priori probability distribution calculation based on the channel correlation matrix data to obtain Bayesian prior probability parameters;

[0055] performing observation value calculation and likelihood function calculation on the synchronous digitized signal data according to the coupling transfer parameter to obtain a likelihood function calculation result;

[0056] Inputting the Bayesian prior probability parameter and the likelihood function calculation result into the Bayesian formula to perform posterior probability calculation to obtain posterior probability distribution data;

[0057] Recursive weight updating and multi-channel probability fusion are performed on the posterior probability distribution data to obtain initial fused signal data.

[0058] Specifically, the prior probability distribution is calculated based on the channel correlation matrix data. The channel correlation matrix data reflects the statistical correlation and mutual influence between multi-channel signals. By performing eigendecomposition and covariance analysis on the correlation matrix, the statistical characteristics of each channel and its combination are established. First, by jointly modeling the mean vector and covariance matrix of each channel signal, a Gaussian model representing the multi-channel joint distribution is obtained. This Gaussian distribution model can describe the amplitude distribution characteristics of a single channel and reflect the changing trend of the joint probability density across multiple channels. Based on this, the mean vector and covariance matrix parameters are extracted using maximum likelihood estimation or empirical matrix inference methods, and then an uncertainty model representing the initial state of the entire multi-channel system is constructed. This model serves as the prior probability distribution in Bayesian inference. After normalization, the Bayesian prior probability parameters are obtained, including the mean estimate, covariance estimate, and initial confidence intervals. Observation value calculation and likelihood function calculation are performed on the synchronized digitized signal data based on the coupling transfer parameter. The coupling transfer parameter is derived from the input-output relationship modeling of the multi-channel system and specifically describes the dynamic response relationship and mutual coupling characteristics between the different channel signals. Based on these parameters, the synchronized digitized signal data at the current moment is mapped through a transfer function to obtain the expected observation value for each channel under the ideal model. The actual observed signal data is compared with the model expected value, and the observation residual is calculated. Assuming that the observation noise follows a Gaussian distribution, a likelihood function is constructed for each channel under this assumption. The likelihood function is defined as the joint probability density function of the actual observed signal data under given transfer parameters and an assumed noise model, reflecting the probability of the current observation data occurring under the assumed model. By combining the likelihood functions of each channel, a comprehensive likelihood function for the entire multi-channel system is formed, thereby obtaining a complete likelihood function calculation result. The prior probability parameters and the likelihood function calculation result are input into the standard Bayesian formula to calculate the posterior probability distribution. According to Bayes' theorem, the posterior probability distribution is proportional to the product of the prior probability and the likelihood function. A normalization factor is used to ensure that the result remains a valid probability distribution. During the calculation process, the prior probability density function and the likelihood function are point-by-point multiplied across each parameter space to produce an unnormalized posterior distribution. The normalization factor is then calculated by integrating the entire parameter space, completing the standardization of the posterior probability distribution. The resulting posterior probability distribution data integrates historical information (prior knowledge) with current observations (real-time signals). It is a statistically optimal estimate of the current system state and possesses the capabilities of self-adaptation, dynamic adjustment, and recursive updating. Recursive weight updates and multi-channel probability fusion are performed on the posterior probability distribution data to generate the initial fused signal data.The recursive weight update mechanism dynamically adjusts the credibility of each channel based on the posterior distribution. By calculating the marginal variance or entropy of the posterior distribution, it quantifies the signal stability and information content of each channel at the current moment. Channels with greater information content and lower variance are assigned higher fusion weights, while those with lower variance are assigned lower weights. Anomalous channels are given very low weights or even zeroed. A sliding window mechanism and exponentially weighted moving average (EWMA) are introduced during the weight update process to ensure smooth and timely weight changes, preventing drastic weight fluctuations under sudden load changes or external interference, thereby improving system stability and robustness. After the weight update is completed, the updated weight coefficients of each channel are combined to perform a weighted fusion of the channel signals to generate a multi-channel probabilistic fusion result. The fusion process uses a Bayesian weighting rule, taking into account inter-channel correlation, observation error covariance, and historical evolution trends, to achieve adaptive optimal fusion. The fusion result preserves the effective information of each channel while suppressing the adverse effects of single-channel anomalies, noise, and drift, significantly improving the signal-to-noise ratio, accuracy, and stability of the fused signal.

[0059] In a specific embodiment, the process of executing step S3 may specifically include the following steps:

[0060] Based on the anomaly detection identifier, a cross-calibration matrix is ​​constructed between channels and a least squares method is used to solve the multi-channel fusion signal data to obtain cross-calibration matrix parameters;

[0061] Performing signal-to-noise ratio calculation and signal stability analysis on the multi-channel fusion signal data according to the cross-calibration matrix parameters to obtain an adaptive weight allocation coefficient;

[0062] Performing zero drift compensation and temperature drift synchronous compensation in combination with the adaptive weight distribution coefficient to obtain calibration compensation parameters;

[0063] Dynamically redistribute the weights of the multi-channel fusion signal data based on the adaptive weight distribution coefficients to obtain a multi-channel weighted fusion result.

[0064] Specifically, based on anomaly detection flag information, a cross-channel cross-calibration matrix is ​​constructed and solved using the least squares method for multi-channel fusion signal data. The anomaly detection flag provides information about each channel's current operating status, degree of anomaly, and reliability. Based on these flags, channel signals are classified as healthy channels and suspected abnormal channels, and a cross-calibration system is constructed using the healthy channels as a benchmark. The calibration matrix is ​​designed to eliminate systematic deviations and drift between channel signals through linear transformations, improving overall signal consistency. Specifically, it takes the form of an n×n-dimensional calibration matrix, where the diagonal elements represent the calibration factors for each channel, and the off-diagonal elements represent the cross-influence coefficients between channels. By minimizing the sum of squared errors between each channel's fusion signal and the healthy reference signal, the objective function is established using the least squares method, and the optimal cross-calibration matrix parameters are obtained through matrix operations. The cross-calibration matrix parameters are used to conduct a deeper analysis of the multi-channel fusion signal data, including signal-to-noise ratio calculation and signal stability analysis. The signal-to-noise ratio (SNR) is calculated based on the ratio of the effective signal power to the noise power after calibration for each channel. Signal power is calculated using the root mean square (RMS) value, while noise power is calculated by extracting high-frequency noise components through residual analysis and spectral analysis, and calculating their energy. A high SNR indicates greater channel signal reliability and lower noise interference, making it a key indicator of channel health. Signal stability analysis, combined with sliding window technology, measures the fluctuation range and dynamic characteristics of each channel's signal based on the short-term mean shift and standard deviation change of the time series. Highly stable channels exhibit lower mean shift and smaller variance fluctuations over time. By normalizing and integrating the SNR and stability indicators, adaptive weight allocation coefficients are derived for each channel. These coefficients dynamically reflect the health and signal reliability of each channel under different operating conditions. Zero drift and temperature drift compensation are simultaneously performed using the adaptive weight allocation coefficients to generate more accurate calibration compensation parameters. Zero-drift compensation addresses signal baseline drift caused by long-term use or environmental changes. A sliding average filter technique is used to dynamically estimate the zero-point offset based on sampled data from each channel over a period of time. This is then weighted and corrected using adaptive weights, thereby tracking and correcting zero-point variations across each channel in real time. To mitigate the effects of short-term fluctuations on the compensation results, the sliding window length is set to 100 to 200 sampling points, and a smoothing factor is introduced to update the filter. Synchronous temperature drift compensation dynamically corrects temperature-induced signal offsets by monitoring ambient temperature changes in real time and combining the temperature drift coefficients of each channel (less than ±0.01% / °C). Temperature drift compensation is performed not only within a single channel but also across multiple channels, ensuring high consistency and high-precision output even in the face of drastic changes in ambient temperature. The combined zero-drift and temperature drift compensation yields highly adaptive and environmentally adaptable calibration parameters, effectively improving the long-term stability and environmental robustness of the weighing system.Based on the updated adaptive weight distribution coefficients, the multi-channel fusion signal data is dynamically redistributed. During the dynamic weight redistribution process, the system dynamically adjusts the weight coefficient of each channel according to the real-time signal quality changes. The weight update period is set to 1 millisecond to ensure that the system can quickly respond to dynamic changes such as load mutations, signal anomalies or external interference. Channels with high health, stable signals and high signal-to-noise ratios are given greater weights, while abnormal channels that are detected to have drift, saturation or increased noise have their weights reduced. If necessary, their weights are reduced to zero to completely eliminate the impact on the final fusion result. In order to prevent system oscillations due to too frequent weight updates, weighted averaging and exponential smoothing mechanisms are introduced in the dynamic weight update process to smoothly control weight changes and ensure the continuity and stability of the fusion output signal. Through the above steps, the multi-channel weighted fusion result is finally generated.

[0065] Among them, the inter-channel cross-calibration matrix is ​​constructed and the least squares method is used to solve the multi-channel fusion signal data based on the abnormal detection identifier to obtain cross-calibration matrix parameters, including: establishing a multi-sensor collaborative calibration benchmark library based on the abnormal detection identifier, and performing historical calibration data backtracking and trend fitting analysis on the multi-channel fusion signal data to obtain calibration benchmark reference data; cross-comparison and consistency deviation calculation processing are performed on each weighing sensor channel according to the calibration benchmark reference data, and an inter-channel mutual verification matrix is ​​established to obtain inter-channel verification matrix parameters; multivariate linear regression and residual analysis processing are performed based on the inter-channel verification matrix parameters to identify the systematic deviation and random deviation of each channel to obtain deviation characteristic parameter data; the deviation characteristic parameter data is input into the iterative least squares optimization algorithm to perform multi-objective function solution and constraint optimization processing to obtain optimized cross-calibration coefficients; matrix reconstruction and numerical stability verification processing are performed on the optimized cross-calibration coefficients to generate cross-calibration matrix parameters.

[0066] In a specific embodiment, the process of executing step S4 may specifically include the following steps:

[0067] Performing single-channel state identification of a multi-layer determination structure on the multi-channel weighted fusion result based on the calibration compensation parameter to obtain single-channel working state data;

[0068] Performing inter-channel cross-validation and global consistency check based on the single-channel working status data to obtain fault channel identification data;

[0069] Automatically removing the faulty channel from the multi-channel weighted fusion result based on the faulty channel identification data to obtain a redundant detection fusion result;

[0070] The redundant detection fusion result is combined to perform working state classification and identification to obtain multi-channel working state identification data, and a weighing detection result is generated according to the multi-channel working state identification data.

[0071] Specifically, based on the latest updated calibration compensation parameters, a multi-channel weighted fusion result is used to carry out single-channel state identification with a multi-layer judgment structure. The calibration compensation parameters cover information such as zero drift correction, temperature drift compensation, and dynamic weight adjustment, which can effectively improve the stability and accuracy of each channel signal. Based on these compensated fusion data, the system adopts a hierarchical judgment mechanism to identify the working state of a single channel. The first layer of judgment distinguishes the basic working range and compares the output amplitude of each channel with the preset range. The signal is divided into five basic states: zero point area, linear working area, saturation area, overload area, and fault area. A preliminary judgment is made based on whether the zero drift exceeds ±0.1% FS, whether the linearity error is less than 0.02% FS, whether the full-scale response reaches 95% FS, and whether signal saturation (exceeding 98% FS) or drift is out of control (exceeding ±5% FS). Building on the first layer, the second layer incorporates dynamic characteristic analysis. By analyzing the short-term mean, standard deviation, and spectral characteristics of channel signals, it identifies dynamic anomalies such as vibration interference, sudden impact loads, or electronic noise. Change point detection is performed using statistical methods such as the Chi-square test and CUSUM control charts. If the amplitude and rate of signal change exceed a set threshold, a dynamic anomaly is identified. The third layer focuses on long-term stability, analyzing trends in the signal mean and variance over a period of time to detect potential chronic failure risks caused by aging, temperature effects, or structural looseness. These three layers of assessment generate single-channel operating status data for each channel. This data includes information about the current operating range and identifies the anomaly type, severity, and trend. Cross-channel cross-validation and global consistency testing are performed based on this single-channel operating status data. During the cross-validation phase, the response consistency of each channel under the same load conditions is compared based on the inter-channel correlation matrix and historical coupling model. Covariance analysis and cross-correlation functions are used to determine the output similarity between any two channels. Specifically, 2σ is set as the judgment threshold. When the output difference between a channel and its adjacent channels exceeds two standard deviations and this difference persists within a certain time window, the channel is considered abnormal. To enhance the robustness of verification, a multi-channel redundant verification mechanism is adopted. Rather than relying on the judgment results of a single channel, the output of most healthy channels is combined to construct a reference output, which is then compared with the channel under test. This avoids misjudgments caused by local anomalies or short-term interference. Global consistency testing further performs a system-level health assessment based on the statistical characteristics of all channels. By analyzing the mean, variance, and spectral energy distribution of all channel signals, the consistency and coordination of the overall output are tested, with particular attention paid to whether there are large-scale abnormal drift, systematic offset, or spectral anomaly clustering. Statistical methods such as variance homogeneity tests, normality tests (such as the KS test), and overall consistency coefficients (such as Cronbach's Alpha coefficient) are used to ensure that the overall system output meets stable and reliable measurement characteristic standards.Finally, combining the results of cross-validation and global consistency checks, the system outputs faulty channel identification data, clearly indicating the faulty channel number, fault type (such as drift, noise, open circuit, or saturation), and recommended handling strategies (such as removal or weight reduction). After faulty channel identification is complete, the multi-channel weighted fusion results are automatically removed based on the identified data. This removal strategy employs a rapid response mechanism. When a channel is detected as faulty, its fusion weight is immediately adjusted to zero, ensuring that its signal no longer participates in subsequent data fusion. The remaining functioning channels are proportionally redistributed to maintain the stability and accuracy of the overall fused signal. The system requires that at least n-2 channels out of n channels remain functional to meet the basic reliability requirements of the redundant detection fusion mechanism. After faulty channel removal and weight redistribution, the redundant detection fusion results are used to classify the operating state. By analyzing the amplitude, change trend, and dynamic characteristics of the fused signal, combined with historical load models and environmental conditions, the current operating state is subdivided into static load, dynamic load, overload risk, no-load, and abnormal risk. The classification and recognition process utilizes machine learning-assisted pattern recognition methods, such as support vector machines, random forests, or neural network-based multi-classifier models, combining current signal characteristics with historical label data for discrimination, significantly improving the accuracy and robustness of state recognition. Each state provides a quantitative estimate of the current load level, along with a confidence score and anomaly risk assessment. Based on multi-channel operating state recognition data, weighing test results are generated. These results include real-time weighing values, current load status, channel health assessment, a list of faulty channels, an overall system health score, and long-term trend forecasts.

[0072] In this embodiment, the single-channel state identification of the multi-channel weighted fusion result based on the calibration compensation parameters is performed on the multi-channel weighted fusion result with a multi-layer judgment structure to obtain single-channel working state data, including: establishing a multi-channel parallel data processing architecture based on the calibration compensation parameters, and performing key threshold parameter initialization setting processing on the multi-channel weighted fusion result to obtain zero drift threshold parameters, linearity threshold parameters, full-scale threshold parameters and fault judgment threshold parameters; performing real-time signal amplitude detection and linear characteristic analysis processing on each weighing sensor channel in the multi-channel weighted fusion result according to the zero drift threshold parameters and the linearity threshold parameters to obtain channel signal characteristic analysis data ; Based on the channel signal characteristic analysis data and the full-scale threshold parameter, multi-level interval judgment and working state classification and identification processing are performed, and the working state of each weighing sensor channel is divided into a zero-point working area, a linear working area, a saturation working area, an overload working area and a fault working area to obtain channel working state classification data; according to the fault judgment threshold parameter, the channel working state classification data is triggered by a real-time fault isolation mechanism and automatically identified as an abnormal channel to obtain fault isolation identification data; the channel working state classification data and the fault isolation identification data are input into a multi-channel state fusion algorithm for parallel verification and consistency verification processing to obtain single-channel working state data.

[0073] The above describes the detection output method of the weighing device in the embodiment of the present invention. The following describes the detection output device of the weighing device in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, a detection output device of a weighing device includes:

[0074] A preprocessing module is used to preprocess and perform analog-to-digital conversion on the original output signals of the multi-channel weighing sensors in the weighing equipment to obtain synchronized digital signal data;

[0075] A multi-channel signal fusion module, configured to perform Bayesian multi-channel signal fusion and anomaly detection based on the synchronized digitized signal data to obtain multi-channel fused signal data and anomaly detection identification;

[0076] A weight allocation module is used to perform inter-channel correlation analysis and adaptive weight allocation based on the anomaly detection identifier and the multi-channel fusion signal data to obtain a multi-channel weighted fusion result and calibration compensation parameters;

[0077] A generation module is used to perform real-time fault isolation and multi-channel working state identification on the multi-channel weighted fusion result based on the calibration compensation parameter to generate a weighing detection result.

[0078] Through the collaborative efforts of these components, independent channel isolation amplifiers provide parallel isolation and amplification processing, effectively eliminating inter-channel interference. Low-pass filtering and synchronous sample-and-hold techniques ensure synchronization and consistency across all channel signals. Bayesian recursive updating and prior-posterior probability fusion leverage the statistical characteristics of historical data and the likelihood of current observations, achieving more accurate and stable multi-channel signal fusion. This approach offers enhanced interference immunity compared to traditional simple averaging methods. A multi-layered detection mechanism, encompassing multi-channel state threshold determination, abnormal state identification, asynchronous change point detection, and statistical validation, enables timely and accurate identification of various anomalies, automatically removing faulty channels and redistributing weights, ensuring system operation even when some sensors fail. Based on an inter-channel cross-calibration matrix and signal-to-noise ratio calculation, adaptive weight allocation coefficients are dynamically adjusted, automatically optimizing weight allocation based on each channel's real-time operating status and signal quality, improving overall measurement accuracy and stability. Integrated zero drift compensation and temperature drift compensation, combined with a sliding average filter and multi-layered decision structure, comprehensively compensate for multiple influencing factors, effectively mitigating the impact of environmental changes on measurement accuracy and extending the device's calibration cycle.

[0079] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3 As shown. The computer device includes a processor, memory, display screen, input device, network interface and database connected via a system bus. The processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the above method is implemented.

[0080] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0081] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which implements the above-described method when executed by a processor. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0082] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Among them, any reference to memory, storage, database, or other media provided by the present invention and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAM bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM.

[0083] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0084] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc., various media that can store program code.

[0085] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A detection output method for a weighing device, characterized in that: include: Pre-process and convert the original output signals of the multi-channel weighing sensors in the weighing equipment into analog-to-digital signals to obtain synchronous digital signal data; performing Bayesian multi-channel signal fusion and anomaly detection based on the synchronized digitized signal data to obtain multi-channel fused signal data and anomaly detection identification; Performing inter-channel correlation analysis and adaptive weight allocation based on the anomaly detection identifier and the multi-channel fusion signal data to obtain a multi-channel weighted fusion result and calibration compensation parameters; Real-time fault isolation and multi-channel working state identification are performed on the multi-channel weighted fusion result based on the calibration compensation parameters to generate a weighing detection result.

2. The detection output method of a weighing device according to claim 1, characterized in that: The method of preprocessing and analog-to-digital converting the original output signals of the multi-channel weighing sensors to obtain synchronous digital signal data includes: Obtaining original output signals of multiple weighing sensors in a weighing device, and inputting the original output signals into independent channel isolation amplifiers for parallel isolation amplification to obtain multi-channel isolation amplified signal data; Performing low-pass filtering on the multi-channel isolated amplified signal data to obtain multi-channel filtered signal data, and performing synchronous sampling and holding on the multi-channel filtered signal data to obtain multi-channel sampled and held signal data; Automatic range switching and signal conditioning are performed based on the multi-channel sample-and-hold signal data to obtain multi-channel pre-processed signal data; The multi-channel pre-processed signal data is subjected to time division multiplexer synchronous sampling and analog-to-digital conversion to obtain synchronous digitized signal data.

3. The detection output method of the weighing equipment according to claim 2, characterized in that: The step of performing time division multiplexer synchronous sampling and analog-to-digital conversion on the multi-channel pre-processed signal data to obtain synchronous digitized signal data includes: Inputting the multi-channel pre-processed signal data into a multi-channel time division multiplexer for high-speed analog switch switching to obtain time division multiplexed signal data; Performing synchronous sampling of the time-division multiplexed signal data using a shared clock source to obtain synchronously sampled signal data; The synchronous sampling signal data is input into a multi-channel analog-to-digital converter for parallel digital conversion and oversampling to obtain original digital signal data, and digital filtering is performed based on the original digital signal data to obtain synchronous digitized signal data.

4. The detection output method of a weighing device according to claim 1, characterized in that: The performing of Bayesian multi-channel signal fusion and anomaly detection based on the synchronized digitized signal data to obtain multi-channel fused signal data and anomaly detection identification includes: Performing inter-channel correlation matrix calculation on the synchronous digitized signal data to obtain channel correlation matrix data, and performing multivariable coupling transfer on the synchronous digitized signal data to establish coupling transfer parameters; Performing Bayesian recursive updating and prior and posterior probability fusion based on the channel correlation matrix data and the coupling transfer parameters to obtain initial fused signal data; Multi-channel state threshold determination and abnormal state identification are performed on the initial fusion signal data to obtain an abnormality detection identifier, and asynchronous change point detection and statistical verification are performed on the initial fusion signal data to obtain multi-channel fusion signal data.

5. The detection output method of a weighing device according to claim 4, characterized in that: The Bayesian recursive update and prior and posterior probability fusion processing based on the channel correlation matrix data and the coupling transfer parameter to obtain initial fused signal data includes: Calculate the prior probability distribution based on the channel correlation matrix data to obtain Bayesian prior probability parameters; performing observation value calculation and likelihood function calculation on the synchronous digitized signal data according to the coupling transfer parameter to obtain a likelihood function calculation result; Inputting the Bayesian prior probability parameter and the likelihood function calculation result into the Bayesian formula to perform posterior probability calculation to obtain posterior probability distribution data; Recursive weight updating and multi-channel probability fusion are performed on the posterior probability distribution data to obtain initial fused signal data.

6. The detection output method of a weighing device according to claim 1, characterized in that: The performing of inter-channel correlation analysis and adaptive weight allocation based on the anomaly detection identifier and the multi-channel fusion signal data to obtain a multi-channel weighted fusion result and calibration compensation parameters includes: Based on the anomaly detection identifier, a cross-calibration matrix is ​​constructed between channels and a least squares method is used to solve the multi-channel fusion signal data to obtain cross-calibration matrix parameters; Performing signal-to-noise ratio calculation and signal stability analysis on the multi-channel fusion signal data according to the cross-calibration matrix parameters to obtain an adaptive weight allocation coefficient; Performing zero drift compensation and temperature drift synchronous compensation in combination with the adaptive weight distribution coefficient to obtain calibration compensation parameters; Dynamically redistribute the weights of the multi-channel fusion signal data based on the adaptive weight distribution coefficients to obtain a multi-channel weighted fusion result.

7. The detection output method of a weighing device according to claim 1, characterized in that: The performing real-time fault isolation and multi-channel working state identification on the multi-channel weighted fusion result based on the calibration compensation parameter to generate a weighing detection result includes: Performing single-channel state identification of a multi-layer determination structure on the multi-channel weighted fusion result based on the calibration compensation parameter to obtain single-channel working state data; Performing inter-channel cross-validation and global consistency check based on the single-channel working status data to obtain fault channel identification data; Automatically removing the faulty channel from the multi-channel weighted fusion result based on the faulty channel identification data to obtain a redundant detection fusion result; The redundant detection fusion result is combined to perform working state classification and identification to obtain multi-channel working state identification data, and a weighing detection result is generated according to the multi-channel working state identification data.

8. A detection output device for a weighing device, characterized in that: Used to execute the detection and output method of a weighing device according to any one of claims 1 to 7, the detection and output device of the weighing device comprises: A preprocessing module is used to preprocess and perform analog-to-digital conversion on the original output signals of the multi-channel weighing sensors in the weighing equipment to obtain synchronized digital signal data; A multi-channel signal fusion module, configured to perform Bayesian multi-channel signal fusion and anomaly detection based on the synchronized digitized signal data to obtain multi-channel fused signal data and anomaly detection identification; A weight allocation module is used to perform inter-channel correlation analysis and adaptive weight allocation based on the anomaly detection identifier and the multi-channel fusion signal data to obtain a multi-channel weighted fusion result and calibration compensation parameters; A generation module is used to perform real-time fault isolation and multi-channel working state identification on the multi-channel weighted fusion result based on the calibration compensation parameter to generate a weighing detection result.

9. A computer device, characterized in that: The device comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the detection output method of the weighing device according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by a processor, the processor is caused to execute the detection and output method of the weighing device according to any one of claims 1 to 7.

Images