A method and system for real-time monitoring of the state of the contents of a plastic packaging drum integrated with multiple sensors

Abstract

The application provides a kind of integrated multi-sensor plastic packaging barrel content state real-time monitoring method and system. Among them, the application collects the weight of barrel and content in real time by integrating weighing sensor in load-bearing support, combines cantilever structure sensor to collect barrel wall vibration sound signal, and generates frequency domain feature map by frequency offset processing through noise analysis and selection of resonance frequency band;Optimized data is generated by using weight data to calibrate reference value and fusing energy gradient compensation of frequency domain map, and state analysis data is extracted by time alignment;Based on the spatial distribution of acoustic sensor array, extract liquid level characteristics, and finally output real-time state markers of content overflow, leakage or liquid level anomaly through time sequence learning mechanism.The application realizes accurate real-time monitoring and early warning of plastic packaging barrel content state (including overflow, leakage or liquid level anomaly) through multi-sensor data fusion and intelligent analysis.

Description

Technical Field

[0001] This application relates to the field of industrial Internet of Things (IoT) sensing and monitoring technology, and in particular to a method and system for real-time monitoring of the contents of plastic packaging barrels that integrate multiple sensors. Background Technology

[0002] In warehousing scenarios in industries such as chemicals and food, plastic packaging drums often store liquid or semi-solid contents (such as corrosive liquids and high-viscosity materials). These scenarios urgently require real-time monitoring of the contents' status (overflow, leakage, or abnormal liquid level) to prevent safety accidents, material losses, and environmental pollution. Due to the non-rigid nature of the plastic drum walls, environmental noise interference (such as mechanical vibration and manual operation noise), and the dynamic changes in the physical properties of the contents, traditional monitoring methods struggle to achieve high-precision, interference-resistant real-time status identification.

[0003] The current mainstream solution uses ultrasonic level sensors for non-contact monitoring: an ultrasonic probe is deployed on the outside of the tank, emitting sound waves into the tank and receiving the reflected signals. The liquid level is calculated by determining the round-trip time difference of the sound waves. This solution filters out some environmental noise through algorithms and uses a multi-probe array to improve spatial coverage, partially solving the problem of single sensors being susceptible to interference.

[0004] While existing ultrasonic liquid level monitoring technology achieves non-contact measurement through sound wave reflection, it suffers from three core problems: the multiple reflections and attenuation effects of sound waves from the plastic barrel wall cause severe signal distortion, especially when the viscosity of the contents changes dynamically, significantly reducing the accuracy of liquid level feature extraction; high-frequency mechanical vibrations and sudden noise interference in the storage environment are difficult to filter out effectively, and existing algorithms cannot distinguish between real liquid level reflection signals and noise, resulting in a high false alarm rate; and single-dimensional acoustic data lacks correlation with weight changes, making it impossible for the system to identify slow weight loss caused by leakage and false liquid level fluctuations caused by barrel vibration, thus failing to meet the monitoring stability and reliability requirements for industrial-grade real-time early warning. Summary of the Invention

[0005] This application provides a method and system for real-time monitoring of the contents of plastic packaging barrels that integrate multiple sensors, in order to solve the problems of data distortion and lack of dynamic calibration capability of weight sensing caused by signal attenuation and noise interference from the barrel wall in the existing technology.

[0006] In a first aspect, this application provides a method for real-time monitoring of the state of the contents of a plastic packaging drum integrating multiple sensors, including:

[0007] An embedded weighing sensor is integrated on the load-bearing bracket at the storage location of the plastic packaging drum. Based on the embedded weighing sensor, real-time weight data of the drum's own weight and the contents are continuously collected.

[0008] Based on the cantilever structure sensor on the outer wall of the plastic packaging barrel, the real-time sound signal of the barrel wall vibration is collected. According to the noise distribution characteristics of the storage environment, the inherent resonance frequency band of the plastic packaging barrel is selected. Frequency shift analysis is performed on the real-time sound signal to generate a frequency domain feature map containing the characteristics of liquid level reflected sound waves.

[0009] Based on the real-time weight data, the suspension reference value is adjusted and calibrated. The resonant cavity is compensated by combining the energy gradient characteristics of the frequency domain feature map to generate optimized frequency domain data. The real-time weight data and the optimized frequency domain data are fused by time alignment to generate state analysis data describing the relationship between the liquid level fluctuation of the contents and the weight of the barrel.

[0010] Based on the spatial distribution of the piezo-acoustic resonance sensor array in the plastic packaging barrel, neighborhood stacking detection is performed on the state analysis data to extract liquid level features characterizing the liquid level height.

[0011] The liquid level features are input into a time-series state-based learning mechanism, which outputs a status marker reflecting whether the contents of the plastic packaging barrel are overflowing, leaking, or have an abnormal liquid level, thus forming a real-time monitoring method for the contents of plastic packaging barrels that integrates multiple sensors.

[0012] Optionally, based on the real-time weight data, the calibration suspension reference value is adjusted, and resonant cavity compensation is performed by combining the energy gradient characteristics of the frequency domain feature map to generate optimized frequency domain data, including:

[0013] Extract the total weight value from the real-time weight data, calculate the adjustment factor based on the difference between the total weight value and the suspension reference value, and update the suspension reference value based on the calculated adjustment factor.

[0014] Based on the energy value data in the frequency domain feature map, the rate of energy change between adjacent frequency points is calculated to generate energy gradient features;

[0015] Based on the updated suspension reference value and the energy gradient characteristics, the energy anomaly region caused by barrel resonance is identified, and a compensation offset is generated.

[0016] The compensation offset is applied to the corresponding frequency point of the frequency domain feature map, and the energy value of the frequency domain feature map is adjusted at the same time. The adjusted frequency domain feature map is then output as optimized frequency domain data.

[0017] Optionally, by time-aligned fusion of the real-time weight data and the optimized frequency domain data, state analysis data describing the correlation between the liquid level fluctuation of the contents and the weight of the container is generated, including:

[0018] Assign timestamps based on the absolute time point of data acquisition to real-time weight data and optimized frequency domain data, and generate time-corresponding data point pairs through timestamp matching and interpolation;

[0019] The weight value and the energy value of the key frequency point are extracted from the data point pair and combined into a multi-dimensional data vector.

[0020] Based on the multidimensional data vector, the covariance mode of weight value change and energy value change is analyzed, and the correlation matrix between the liquid level fluctuation of the contents and the weight of the container is generated.

[0021] The rows of the correlation matrix are output as time series, and the columns of the correlation matrix represent combined parameters of weight and frequency domain features as state analysis data.

[0022] Optionally, based on the noise distribution characteristics of the storage environment, the inherent resonant frequency band of the plastic packaging drum is selected, and frequency shift analysis is performed on the real-time sound signal to generate a frequency domain feature map containing the characteristics of liquid level reflected sound waves, including:

[0023] Collect background noise signals from the storage environment and analyze the energy distribution of the background noise signals;

[0024] Based on the physical properties of plastic packaging barrels, the frequency range in which the barrel produces the maximum vibration when it is empty is determined through preliminary experiments to be the inherent resonant frequency band. Combining the inherent resonant frequency band and the energy distribution, the low-noise resonant frequency band is selected as the target frequency band.

[0025] The real-time audio signal is divided into time windows, and frequency transformation is applied to each window to convert the time-domain signal into a frequency-domain representation.

[0026] Within the target frequency band, the difference between the peak frequency position and the preset reference frequency in the frequency components is detected and calculated, and the difference is used as the offset relative to the reference frequency.

[0027] Based on the frequency domain representation and the offset mapping, a two-dimensional coordinate system diagram is generated, and the offset abrupt change region is marked to output a frequency domain feature map.

[0028] Optionally, based on the spatial distribution of the piezo-acoustic resonant sensor array of the plastic packaging drum, neighborhood stacking detection is performed on the state analysis data to extract liquid level features characterizing the liquid level height, including:

[0029] Based on the physical installation location of the piezo-acoustic resonant sensor array, the sensors are divided into multiple neighborhood groups containing spatially adjacent sensors according to the sensor spacing and barrel geometry.

[0030] Extract the data elements corresponding to each sensor from the state analysis data, and stack the data elements of all sensors in each neighborhood group to form a multidimensional stacked matrix.

[0031] Calculate the distribution pattern of energy values ​​with height in the stacking matrix and identify the location of energy peaks;

[0032] Based on the mapping between the energy peak position and the sensor height, the liquid level height value is derived as a liquid level feature.

[0033] Optionally, the liquid level characteristics are input into a time-series state-based learning mechanism, which outputs a status marker reflecting whether the contents of the plastic packaging drum are overflowing, leaking, or have an abnormal liquid level, thus constituting a real-time monitoring method for the state of the contents of a plastic packaging drum integrating multiple sensors, including:

[0034] Collect historical liquid level characteristic data and status labels including category labels such as overflow, leakage or abnormal liquid level, and construct a training dataset;

[0035] The training dataset is used to train a time-series state-based learning mechanism, which is input as a liquid level feature time series and internal parameters are optimized to learn pattern changes.

[0036] The real-time liquid level feature time series is input into the trained learning mechanism, and the state probability distribution at each time point is calculated through internal parameters.

[0037] The highest probability category is selected as the output state label based on the state probability distribution.

[0038] Optionally, based on the energy value data in the frequency domain feature map, the rate of energy change between adjacent frequency points is calculated to generate energy gradient features, including:

[0039] Extract the energy value sequence of the frequency domain feature map in ascending order of frequency;

[0040] Based on the energy value data of the energy value sequence, the absolute difference between the current point and the previous point is calculated for each frequency point, and a unit frequency change sequence is generated by combining the frequency span.

[0041] A judgment threshold is set for each unit frequency change. If the value exceeds the threshold, a flag bit is set at the corresponding frequency position to form a corresponding flag sequence.

[0042] The energy gradient feature is formed by combining the unit frequency change sequence and the label sequence.

[0043] Secondly, this application provides a real-time monitoring system for the state of contents of a plastic packaging drum integrating multiple sensors, including:

[0044] The data acquisition module is used to integrate an embedded weighing sensor on the load-bearing bracket at the storage location of the plastic packaging drum, and to continuously acquire real-time weight data of the drum's own weight and the contents based on the embedded weighing sensor.

[0045] The generation module is used to collect real-time sound signals of barrel wall vibration based on the cantilever structure sensor on the outer wall of the plastic packaging barrel, select the inherent resonant frequency band of the plastic packaging barrel according to the noise distribution characteristics of the storage environment, perform frequency shift analysis on the real-time sound signal, and generate a frequency domain feature map containing the characteristics of liquid level reflected sound waves.

[0046] The generation module is also used to adjust the calibration suspension reference value based on the weight data, perform resonant cavity compensation by combining the energy gradient characteristics of the frequency domain feature map, generate optimized frequency domain data, and fuse the real-time weight data and the optimized frequency domain data by time alignment to generate state analysis data describing the relationship between the liquid level fluctuation of the contents and the weight of the barrel.

[0047] The detection module is used to perform neighborhood stacking detection on the state analysis data based on the spatial distribution of the piezo-acoustic resonance sensor array of the plastic packaging barrel, and extract liquid level features characterizing the liquid level height.

[0048] The output module is used to input the liquid level features into a time-series state-based learning mechanism and output a status marker reflecting whether the contents of the plastic packaging barrel are overflowing, leaking, or have an abnormal liquid level, thus forming a real-time monitoring method for the contents of plastic packaging barrels that integrates multiple sensors.

[0049] Thirdly, this application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to realize a real-time monitoring method for the state of contents of a plastic packaging barrel integrating multiple sensors as described in the first aspect above.

[0050] Fourthly, this application provides a computer storage medium storing a computer program, which, when executed by a computer, implements a method for real-time monitoring of the state of contents in a plastic packaging barrel with integrated multi-sensor technology as described in the first aspect.

[0051] This application utilizes an embedded weighing sensor to continuously collect real-time weight data of the tank and its contents, enabling quantitative monitoring of slow leakage or weight loss. Combined with a cantilever structure sensor, frequency offset analysis is performed based on the noise distribution to select the inherent resonant frequency band, suppressing environmental interference and accurately extracting the acoustic characteristics of liquid level reflection. Furthermore, the weight data is used to dynamically calibrate the suspension reference value, and the energy gradient of the coupled frequency domain feature map is used for resonant cavity compensation, eliminating interference from the tank wall material and generating optimized frequency domain data with higher physical consistency. Then, weight and acoustic data are fused through time alignment to accurately correlate liquid level fluctuations with weight changes. Neighborhood stacking detection is performed based on the spatial distribution of the piezoelectric sensor array, enhancing the spatial robustness of liquid level height characteristics. Finally, the time-series liquid level characteristics are input into a learning mechanism to dynamically identify multi-state evolution modes of overflow, leakage, and abnormal liquid levels, achieving high-precision real-time early warning.

[0052] Furthermore, a dynamic adjustment factor is generated by extracting the total weight value from real-time weight data and calculating its difference from the suspension reference value, thereby updating the suspension reference value. Simultaneously, energy gradient features are generated by calculating the energy change rate of adjacent frequency points based on the frequency domain feature map. Combining the updated reference value and energy gradient features, an energy anomaly region caused by barrel resonance is identified, generating a frequency domain compensation offset. Finally, the offset is applied to the corresponding frequency points and energy values ​​of the frequency domain feature map to adjust the output optimized frequency domain data. Through the coupled compensation mechanism of dynamic weight calibration and acoustic gradient features, the system accurately distinguishes between barrel resonance noise and the actual liquid level reflection signal, effectively solving the signal distortion problem caused by sound wave attenuation from the plastic barrel wall, significantly improving the signal-to-noise ratio and physical consistency of the frequency domain data, providing high-confidence input for multi-source data fusion, and overcoming the limitations of existing single-dimensional acoustic monitoring that suffers from misjudgment due to barrel interference.

[0053] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 A flowchart of a method for real-time monitoring of the contents of a plastic packaging drum integrating multiple sensors, provided in this application, is shown.

[0056] Figure 2 A schematic diagram of a scenario is shown, illustrating a method for real-time monitoring of the contents of a plastic packaging drum integrating multiple sensors, as provided in this application.

[0057] Figure 3 A schematic diagram of the structure of a real-time monitoring system for the contents of a plastic packaging barrel integrating multiple sensors, provided in this application, is shown.

[0058] Figure 4 A schematic diagram of the structure of a computing device provided in this application is shown. Detailed Implementation

[0059] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0060] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a chronological order, nor do they limit "first" and "second" to different types.

[0061] In the field of real-time monitoring of the contents of plastic packaging drums, existing ultrasonic liquid level detection technology faces a fundamental contradiction: the multiple reflections and attenuation effects of the plastic drum wall on sound waves cause severe distortion of the liquid surface echo signal, especially with high-viscosity contents; traditional filtering algorithms fail when high-frequency mechanical vibration noise from the storage environment intrudes into the monitoring frequency band, leading to a sharp deterioration in the signal-to-noise ratio of the liquid level characteristics; more critically, the single acoustic data dimension is completely disconnected from the dynamic changes in the weight of the drum, such as leakage and weight loss, making it impossible for the system to distinguish between actual liquid level anomalies and drum vibration noise interference. The root of this contradiction lies in the lack of a collaborative processing mechanism for signal physical distortion, environmental noise coupling, and multi-physical quantity correlation mechanisms in existing technologies, resulting in a high false alarm rate in industrial scenarios and a serious lag in the identification of slow leaks.

[0062] To address the signal distortion, noise misjudgment, and state fragmentation defects in ultrasonic liquid level monitoring, this application proposes a multi-sensor spatiotemporal fusion monitoring method: First, an embedded weighing sensor captures the real-time weight changes of the tank and its contents. Simultaneously, a cantilever structure sensor, combined with noise distribution, selects the inherent resonant frequency band to perform frequency shift analysis, generating an anti-noise frequency domain feature map. Then, based on the weight data, the suspended reference value is dynamically calibrated, and the energy gradient characteristics of the coupled frequency domain feature map are used for resonant cavity compensation to eliminate tank wall interference and output optimized frequency domain data. Next, the weight changes and acoustic fluctuations are fused through time alignment to construct physical correlation state analysis data. Finally, based on the spatial distribution of the piezo-acoustic resonant sensor array, robust liquid level features are extracted through neighborhood stacking detection, and a time-series learning mechanism is input to dynamically identify overflow, leakage, and liquid level anomalies, achieving collaborative decision-making among multiple physical quantities.

[0063] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0064] Figure 1 This application provides a flowchart of a method for real-time monitoring of the contents of a plastic packaging drum integrating multiple sensors, as shown in the embodiments of this application. Figure 1 As shown, the method includes:

[0065] 101. An embedded weighing sensor is integrated on the load-bearing bracket at the storage location of the plastic packaging drum, and real-time weight data of the drum's own weight and the contents are continuously collected based on the embedded weighing sensor.

[0066] In the above solution, the embedded weighing sensor refers to a miniature measuring device installed inside the load-bearing point of the storage rack. Real-time weight data refers to the total weight of the barrel itself plus its contents, continuously collected by the sensor, with a collection frequency down to the second level. The load-bearing rack refers to a metal support structure specifically designed to support plastic packaging barrels in the warehouse, which must meet the technical requirements of load-bearing strength and sensor integration.

[0067] In this embodiment, the pressure of the barrel is first sensed by a strain gauge sensor embedded inside the support, which converts physical deformation into an electrical signal. For example, after placing a standard barrel, the slight deformation of the support causes the sensor to generate a 10 millivolt voltage signal. This electrical signal is then amplified by a signal amplifier, increasing the weak voltage to the standard measurement range. For example, the 10 millivolt signal is amplified to 2.5 volts, and then converted into a digital signal by an analog-to-digital converter at a rate of 10 times per second. Next, a moving average filtering technique is used to eliminate instantaneous interference. After continuously collecting 20 sets of data, outliers with deviations exceeding 3% are automatically removed, and the effective data are averaged to generate a stable reading. Finally, the data is processed using a preset weight calculation formula, which includes two key parameters: sensor coefficient K and support self-weight compensation value B. For example, when coefficient K is 500 kg per volt and the measured voltage is 2.5 volts, 500 multiplied by 2.5 equals 1250 kg, then the support self-weight of 50 kg is added, resulting in a total output weight of 1300 kg. The system automatically deducts the barrel weight and displays the net weight of the contents.

[0068] In practical applications, in the raw material storage area B of chemical company A, the load-bearing supports were modified to integrate four C-type weighing sensors (each with a capacity of 2 tons). When a new batch of raw material drums was stored in storage location D, the system automatically captured an initial total weight of 1050 kg. During continuous monitoring, on the fifth day, it detected that the weight had continuously decreased to 280 kg, triggering an E-level inventory warning. The administrator viewed the trend curve through the F terminal and promptly arranged for replenishment. In another case, in the flammable goods storage area of ​​warehouse G, the H-type supports, equipped with temperature compensation function, promptly detected an abnormal weight caused by a slow leak in a solvent drum, avoiding potential safety hazards.

[0069] This solution enables real-time automatic monitoring of the weight of stored materials, accurately tracking the real-time consumption of contents; automatically identifying abnormal weight changes to prevent material loss and leakage risks; directly connecting weight data to the inventory management system to improve supply chain response efficiency; eliminating the possibility of contamination caused by manual opening and inspection; and maintaining the original warehouse space layout with its compact embedded design.

[0070] 102. Based on the cantilever structure sensor on the outer wall of the plastic packaging barrel, the real-time sound signal of the barrel wall vibration is collected. According to the noise distribution characteristics of the storage environment, the inherent resonance frequency band of the plastic packaging barrel is selected. Frequency shift analysis is performed on the real-time sound signal to generate a frequency domain feature map containing the characteristics of liquid level reflected sound waves.

[0071] Optionally, step 102 may specifically include the following steps:

[0072] 1021. Collect background noise signals from the storage environment and analyze the energy distribution of the background noise signals;

[0073] 1022. Based on the physical properties of plastic packaging barrels, the frequency range in which the barrel produces the maximum vibration when it is empty is determined through preliminary experiments as the inherent resonance frequency band. Combining the inherent resonance frequency band and the energy distribution, the low-noise resonance frequency band is selected as the target frequency band.

[0074] 1023. Divide the real-time audio signal into time windows, apply frequency transformation to each window, and convert the time-domain signal into a frequency-domain representation;

[0075] 1024. Within the target frequency band, detect and calculate the difference between the peak frequency position of the frequency component and the preset reference frequency, and use the difference as the offset relative to the reference frequency;

[0076] 1025. Based on the frequency domain representation and the offset mapping, a two-dimensional coordinate system diagram is generated, and the offset abrupt change region is marked, and the frequency domain feature map is output.

[0077] In the above scheme, the cantilever structure sensor refers to an L-shaped vibration detection device installed on the outside of the bucket wall, which senses the micro-vibrations of the bucket wall through a cantilever beam. The real-time sound signal refers to the electrical signal sequence converted from the mechanical vibration of the bucket body by the sensor. The inherent resonant frequency band refers to the specific frequency range in which the plastic bucket produces the strongest vibration when struck while empty. The frequency domain characteristic map is a visual spectrum that presents the frequency distribution and offset of the sound signal in a two-dimensional form, with frequency on the horizontal axis and offset on the vertical axis.

[0078] In this embodiment, firstly, in step 1021, a sound level meter is used to collect background noise signals from the warehouse environment, for example, collecting 30 seconds of background noise signals. The noise energy distribution is analyzed using a Fast Fourier Transform, revealing, for example, a high-energy peak of fan interference in the 50-60Hz frequency band. Secondly, in step 1022, in a laboratory setting, empty plastic packaging barrels of the same model are subjected to a tapping test based on their physical properties. A spectrum analyzer is used to determine their inherent resonant frequency band, such as 120-130Hz. Combining this with the noise energy distribution from step 1021, a frequency band with lower noise energy, such as 120-130Hz, is selected as the target frequency band. Next, in step 1023, the real-time collected 2-second sound signal is divided into 10 0.2-second time windows. A Fourier Transform is applied to each window to convert the time-domain signal into a frequency-domain representation, generating a spectrum. For example, one window's spectrum shows a main peak at 125Hz. Next, in step 1024, the main peak frequency value is detected within the target frequency band, and its offset from the preset reference frequency of 125Hz is calculated. The frequency offset is calculated as: Δf = fpeak − fref, where fpeak refers to the measured peak frequency and fref refers to the empty bucket reference frequency. For example, when the peak frequency is detected as 118Hz and the reference frequency is 115Hz, the offset Δf = 118 − 115 = 3Hz. Finally, in step 1025, the frequency domain representation and the offset are mapped into a two-dimensional coordinate system. For example, the spectral data of 10 time windows and the offset are mapped to a two-dimensional coordinate system: the horizontal axis is the frequency range of 120-130Hz, and the vertical axis is the offset value. Regions with sudden increases in offset are marked, such as a sudden change from +1Hz to +5Hz, and a color-coded frequency domain feature map is output.

[0079] In practical application, in Zone B of Company A's chemical warehouse, a Type C cantilever sensor was installed on the outer wall of a 200-liter plastic drum. The system first collected ambient noise data to determine that 80-90Hz was the high-noise zone. Through preliminary experiments, the inherent resonant frequency band of this drum type was measured to be 110-120Hz, so 110-120Hz was selected as the target frequency band. During monitoring, when the liquid level in one drum dropped to 30%, the system detected a shift in the peak frequency from 115Hz to 118Hz. The characteristic graph showed that the shift suddenly increased from +2Hz to +5Hz, triggering a low liquid level warning.

[0080] This solution enables non-contact liquid level monitoring, avoiding the safety risks of traditional open-tank testing; it improves signal recognition reliability through noise adaptive filtering; frequency offset characteristics can sensitively reflect liquid level changes; visual feature maps intuitively display liquid dynamics; and the cantilever design does not damage the tank structure.

[0081] 103. Adjust the calibration suspension reference value based on the real-time weight data, perform resonant cavity compensation by combining the energy gradient characteristics of the frequency domain feature map, generate optimized frequency domain data, and fuse the real-time weight data and the optimized frequency domain data through time alignment to generate state analysis data describing the relationship between the liquid level fluctuation of the contents and the weight of the barrel.

[0082] Optionally, step 103 may specifically include the following steps:

[0083] 1031. Extract the total weight value from the real-time weight data, calculate the adjustment factor based on the difference between the total weight value and the suspension reference value, and update the suspension reference value based on the calculated adjustment factor;

[0084] 1032. Based on the energy value data in the frequency domain feature map, calculate the rate of energy change between adjacent frequency points and generate energy gradient features;

[0085] Specifically, step 1032 may include the following process: extracting the energy value sequence of the frequency domain feature map in ascending order of frequency; based on the energy value data of the energy value sequence, calculating the absolute difference between the energy value of the current point and the previous point for each frequency point, and generating a unit frequency change sequence in combination with the frequency span; setting a judgment threshold for each unit frequency change, and setting a marker bit at the corresponding frequency position if the value is higher than the threshold, thus forming a corresponding marker sequence; and combining the unit frequency change sequence and the marker sequence to form an energy gradient feature.

[0086] 1033. Based on the updated suspension reference value and the energy gradient characteristics, identify the energy anomaly region caused by barrel resonance and generate a compensation offset.

[0087] 1034. Apply the compensation offset to the corresponding frequency point of the frequency domain feature map, adjust the energy value of the frequency domain feature map, and output the adjusted frequency domain feature map as optimized frequency domain data.

[0088] 1035. Assign timestamps based on the absolute time point of data acquisition to real-time weight data and optimized frequency domain data, and generate time-corresponding data point pairs through timestamp matching and interpolation;

[0089] 1036. Extract the weight value and the energy value of the key frequency point from the data point pair and combine them into a multi-dimensional data vector;

[0090] 1037. Based on the multidimensional data vector, analyze the covariance mode of weight value change and energy value change, and generate a correlation matrix between the fluctuation of the liquid level of the contents and the weight of the container.

[0091] 1038. Output the rows of the correlation matrix to represent the time series, and the columns of the correlation matrix to represent the combined parameters of weight and frequency domain features as state analysis data.

[0092] In the above scheme, the suspended reference value refers to the reference frequency value in the empty barrel state, used as a reference point for detecting frequency offset. The adjustment factor is a calibration coefficient calculated based on real-time weight changes, used to correct the reference value. The energy gradient characteristic is a quantified index of the rate of energy change between adjacent frequency points in the frequency domain graph, including unit change and anomaly markers. The compensation offset is a correction value used to offset barrel resonance interference. Optimized frequency domain data refers to the frequency distribution data after resonance interference compensation. The timestamp is the absolute time stamp of the data acquisition moment. The correlation matrix is ​​a combined data table of weight values ​​and frequency domain characteristic values ​​arranged in a time series, used to analyze the correlation between liquid level and weight.

[0093] In this embodiment, the total weight value in the current real-time weight data is first obtained through step 1031. For example, when the system detects that the barrel weight has changed to 800 kg while the initial reference weight is 500 kg, an adjustment factor is calculated based on the difference between the total weight value and the suspension reference value. The adjustment factor is calculated as (total weight value - suspension reference value) / suspension reference value. The calculated adjustment factor is the difference between 800 and 500 divided by 500, which equals 0.6. The suspension reference value is updated based on the calculated adjustment factor. For example, the product of the adjustment factor and the original suspension reference value is added to the original suspension reference value. If the original suspension reference value is 100 Hz, then 100 x 0.6 + 100 = 160, resulting in a new reference frequency of 160 Hz. This process is similar to recalibrating the starting scale of the measuring ruler based on the water volume. Next, step 1032 analyzes the energy value data in the frequency domain feature map to calculate the energy change trend between adjacent frequency points. For example, if the energy value increases from 85 dB to 90 dB in the 124 Hz to 125 Hz interval, the change per Hz is calculated as 90 minus 85 divided by 1, which equals 5 dB per Hz. When the calculated change exceeds the preset threshold of 3 dB per Hz, an anomaly marker is set at the 125 Hz position, similar to placing a warning flag at an abnormal temperature point. Then, step 1033 calculates the compensation offset, where the excess ratio = (measured change rate - threshold) ÷ threshold, and the compensation offset = new reference × excess ratio. Combining the new reference frequency of 160 Hz and the 125 Hz marker, and based on the excess ratio of 5 minus 3 divided by 3 equals 0.67, the compensation offset of 160 multiplied by 0.67 is approximately a negative correction value of 107 Hz. This is equivalent to customizing a noise reduction scheme for areas with excessive vibration. Then, in step 1034, optimized data is generated. The compensation offset is applied to the corresponding frequency point of the frequency domain feature map, and the energy value of the frequency domain feature map is adjusted. The adjusted frequency domain feature map is output as optimized frequency domain data. A downward correction of 107 Hz is applied to the 125 Hz frequency point to 18 Hz, and the energy value of this point is adjusted proportionally from 90 dB to 85 dB to obtain a more realistic acoustic distribution. Subsequently, in step 1035, timestamps based on the absolute time point of data acquisition are assigned to the real-time weight data and optimized frequency domain data. Data point pairs corresponding to the time are generated through timestamp matching and interpolation to achieve time synchronization. The weight data at 14:30:00 is marked with a timestamp. For the missing acoustic data at 14:30:02, the average of 82 dB at 14:30:01 and 84 dB at 14:30:03 is used to fill in the missing data, ensuring complete alignment of the time axis data. Subsequently, in step 1036, the weight value and the energy value of the key frequency point are extracted from the data point pair, and the data units are combined into a multi-dimensional data vector. Taking 14:30:00 as an example, the weight value of 790 kg, the corrected frequency of 18 Hz, and the energy value of 85 dB are integrated into a three-dimensional data packet, forming a weight-frequency-energy combination entry similar to a product label.Further, in step 1037, based on the aforementioned multidimensional data vector, the covariance pattern between weight and energy value changes is analyzed, revealing a linkage pattern. Three consecutive monitoring points show that the energy level corresponds to 83 dB for a weight of 800 kg, 85 dB for 790 kg, and 87 dB for 780 kg. The system automatically records a stable relationship where the energy increases by 2 dB for every 10 kg decrease in weight. Similar to discovering the correlation between water temperature fluctuations and water vapor generation, a correlation matrix is ​​generated between the fluctuation of the liquid level in the contents and the weight of the container. Finally, in step 1038, a time series table of the correlation matrix is ​​output. The horizontal header contains four parameters: time point, weight value, frequency value, and energy value. The vertical column records monitoring data by minute, forming a directly analyzable state change matrix.

[0094] In practical applications, in the organic solvent storage area of ​​warehouse A, a type B sensor monitors type C plastic drums. When the weight of a drum increases from an initial 500kg to 800kg, the system automatically adjusts the reference frequency from 100Hz to 130Hz. An abnormal energy gradient was detected at a frequency of 125Hz (the rate of change between adjacent points reached 5dB / Hz). After applying a -3Hz compensation, the energy value at this point was corrected from 90dB to 87dB. The time-aligned fusion module combined the weight data of 790kg at 14:30 with the corrected frequency domain data and found that the energy value at 124Hz continuously increased as the weight decreased. A correlation matrix was generated showing a pattern of a 4dB increase in energy value for every 100kg decrease in weight.

[0095] This solution achieves dynamic calibration and compensation of weight and acoustic data, eliminating the interference of barrel resonance on the measurement; it integrates multi-source data to improve the accuracy of liquid level monitoring; it reveals the physical relationship between content consumption and vibration characteristics through time correlation analysis; it provides multi-dimensional decision-making basis for inventory early warning; and its adaptive benchmark adjustment mechanism adapts to the detection needs of different filling states.

[0096] 104. Based on the spatial distribution of the piezo-acoustic resonance sensor array of the plastic packaging barrel, perform neighborhood stacking detection on the state analysis data to extract liquid level features characterizing the liquid level height;

[0097] Optionally, step 104 may specifically include the following steps:

[0098] 1041. Based on the physical installation position of the piezo-acoustic resonance sensor array, the sensor is divided into multiple neighborhood groups containing spatially adjacent sensors according to the sensor spacing and barrel geometry.

[0099] 1042. Extract the data elements corresponding to each sensor from the state analysis data, and for each neighborhood group, stack the data elements of all sensors in the neighborhood group to form a multidimensional stacked matrix.

[0100] 1043. Calculate the distribution pattern of energy values ​​with height in the stacking matrix and identify the location of energy peaks;

[0101] 1044. Based on the mapping between the energy peak position and the sensor height, the liquid level height value is derived as a liquid level feature.

[0102] In the above scheme, the piezoelectric acoustic resonance sensor array refers to a group of vibration detection devices installed at different heights on the barrel wall. A neighborhood group refers to a cluster of adjacent detection units divided according to the spatial location of the sensors. The multidimensional stacking matrix is ​​a three-dimensional table that integrates data from multiple sensors according to their spatial location. The energy peak position refers to the spatial height point where the sound wave energy is strongest. The liquid level height value refers to the actual position of the liquid surface calculated from the energy peak.

[0103] In this embodiment, the sensors are first automatically grouped based on their spatial location in step 1041. The system reads the geometric parameters of the barrel and the physical installation coordinates of the piezo-acoustic resonance sensor array, and divides the sensors into multiple neighborhood groups containing spatially adjacent sensors according to a preset neighborhood radius threshold. For example, if 10 sensors are installed on the wall of a 1.2-meter-high cylindrical barrel, the system divides the neighborhoods with a radius of 30 centimeters: the bottom group includes sensors 1 and 2 at heights of 0.2 meters and 0.4 meters, the middle group includes sensors 3 to 5 at heights of 0.6 meters, 0.8 meters, and 1.0 meters, and the top group includes sensors 9 and 10 at heights of 1.1 meters and 1.2 meters. This process uses a nearest neighbor algorithm to ensure that the distance between sensors within a group does not exceed the threshold. Next, a neighborhood data stacking matrix is ​​constructed in step 1042. Time slice data is extracted from the state analysis data, and parameters are integrated according to the sensor number order within each neighborhood group. For example, data from the central group at 14:30: Sensor 3 at 0.6 meters recorded a key frequency of 120 Hz and an energy value of 85 dB; Sensor 4 at 0.8 meters recorded 118 Hz and 90 dB; and Sensor 5 at 1.0 meters recorded 122 Hz and 88 dB. Data elements from all sensors within the stacked neighborhood group are arranged in ascending order of height to generate a corresponding 3×2 multidimensional stacked matrix. Then, step 1043 calculates the energy distribution and identifies peak positions. Numerical analysis and interpolation are performed on the multidimensional stacked matrix. For example, the average energy value of the central group is calculated as follows: 85 + 90 + 88 = 263 dB, divided by 3, yielding approximately 87.7 dB. Cubic spline interpolation generates a continuous curve from 0.6 to 1.0 meters in height: 85 dB at 0.6 meters, a peak of 90 dB at 0.8 meters, and 88 dB at 1.0 meters. Combined with sensor height mapping, 0.8 meters is determined as the characteristic position of the liquid surface. The process establishes a peak validity verification rule: a peak energy value exceeding twice the standard deviation of the mean is considered valid. Finally, the liquid level height value is derived and the error is labeled in step 1044. Based on the mapping between the energy peak position and the sensor height, compensation parameters for the tank structure are applied. For example, with sensor #4 installed at a height of 0.8 meters, a liquid sound velocity compensation value of 0.025 meters is added to obtain a reference liquid level of 0.825 meters. Through error propagation model synthesis: sensor positioning error ±0.005 meters, interpolation calculation error ±0.01 meters, compensation coefficient error ±0.008 meters, the final output liquid level characteristic is 0.825 meters ±0.015 meters.

[0104] In a practical application, six sensors were deployed on a 0.8-meter-high Type B storage tank in the raw material warehouse of chemical company A. The system was divided into four neighborhood groups: bottom, lower middle, middle, and top. Sensor data from the middle group showed an energy value of 90 dB at 0.4 meters from the bottom of the tank, peaking at 95 dB at 0.5 meters, and dropping to 90 dB at 0.6 meters, indicating the liquid level was at a height of 0.5 meters, with an error range of ±0.03 meters.

[0105] This solution enhances the ability to capture regional features through spatial grouping; enables multi-sensor collaborative analysis through data stacking; accurately maps liquid surface position using energy peak location; eliminates structural errors through height compensation mechanism; and improves measurement reliability through full-process error control.

[0106] 105. Input the liquid level characteristics into a time-series state-based learning mechanism, and output a state marker reflecting whether the contents of the plastic packaging barrel are overflowing, leaking, or have an abnormal liquid level, thereby constituting a real-time monitoring method for the state of the contents of a plastic packaging barrel that integrates multiple sensors.

[0107] Optionally, step 105 may specifically include the following steps:

[0108] 1051. Collect historical liquid level characteristic data and status labels including category labels for overflow, leakage, or abnormal liquid level, and construct a training dataset;

[0109] 1052. Use the training dataset to train a time-series state-based learning mechanism, input the liquid level feature time series, and optimize the internal parameters to learn the pattern changes;

[0110] 1053. Input the real-time liquid level feature time series into the trained learning mechanism, and calculate the state probability distribution at each time point through internal parameters;

[0111] 1054. Select the category with the highest probability as the output state label based on the state probability distribution.

[0112] In the above scheme, liquid level characteristics refer to the liquid surface height value and its error range measured by an acoustic sensor array. The temporal state learning mechanism refers to an artificial intelligence model that analyzes the changing patterns of liquid level over time. State labels are the system's output identifiers of abnormal content types, including overflow, leakage, and normal fluctuations. The training dataset is a collection of historical monitoring data and corresponding abnormal labels. The state probability distribution is the numerical probability of the occurrence of the three abnormal states calculated by the model.

[0113] In this embodiment, firstly, step 1051 collects historically monitored liquid level characteristic data and status labels including overflow, leakage, or abnormal liquid level categories to construct a training set. The system automatically archives liquid level records and event reports from the past six months. For example, the time series data recorded for solvent tanks in the warehouse on July 15th shows a liquid level of 0.65 meters at 10:00, rising to 0.68 meters at 11:00, and reaching 0.73 meters at 12:00, triggering an overflow warning. This data is integrated with manually verified labels to form training samples. Secondly, step 1052 uses the training set to train a time-series learning model. A long short-term memory neural network is used to process time window data. Inputting a typical sequence, such as six sets of liquid level values ​​showing an upward trend from 0.60 meters to 0.75 meters, corresponds to an overflow label; inputting a continuous three-hour downward sequence from 0.70 meters to 0.52 meters matches a leakage label. After two hundred rounds of training, the model automatically identifies feature rules, such as setting a liquid level rise exceeding 0.08 meters per hour as an overflow risk feature. Next, in step 1053, the real-time liquid level feature time series is input into the trained learning mechanism for real-time analysis. Input the current monitoring window data, for example, the afternoon liquid level sequence for a certain tank: 0.78 meters at 14:00, dropping to 0.72 meters at 15:00, dropping to 0.65 meters at 16:00, dropping to 0.55 meters at 17:00, and dropping to 0.48 meters at 18:00. The model calculates the state probability distribution at each time point using internal parameters. This sequence shows a 90% probability of leakage due to a sharp drop of 0.3 meters over three hours, a 2% probability of overflow, and an 8% probability of normal fluctuation. Finally, in step 1054, a state marker is generated. Based on the state probability distribution, the highest probability category is selected as the output state marker. For example, if the system selects a 90% leakage probability as the highest value, a leakage state marker is output with an attached confidence level value. Simultaneously, a yellow warning signal is activated and pushed to the management terminal, completing the automatic state determination process.

[0114] In practical application, chemical company A deployed this system in its raw material warehouse. The historical database contains 300 abnormal records of type B solvent tanks. Overflow events are characterized by a level rise exceeding 0.07 meters per hour, while leakage events are characterized by a level drop exceeding 0.18 meters over three hours. Real-time monitoring showed a tank level sequence of 0.75 meters at 14:00, dropping to 0.67 meters at 15:00, dropping to 0.55 meters at 16:00, and dropping to 0.40 meters at 17:00. The model identified a three-hour level drop of 0.35 meters for this sequence, calculated a 95% probability of leakage, automatically triggered a level-two leakage alarm, and pushed the alert to the warehouse dispatch center.

[0115] This solution enables intelligent trend analysis and abnormal pattern recognition of liquid level changes; automatically distinguishes between normal fluctuations and safety accident risks; pushes multi-level early warning information in real time to improve response speed; reduces the false judgment rate through a probability-based decision-making mechanism; and optimizes the model's judgment accuracy using historical data.

[0116] Figure 2 This application provides a scenario diagram illustrating a method for real-time monitoring of the contents of a plastic packaging drum integrating multiple sensors, as an embodiment of the present application. Figure 2 As shown, a complete embodiment of steps 101-105 includes:

[0117] In the solvent storage center of chemical company A, for the storage of flammable liquids in standard 200-liter Type B plastic containers, the system first continuously collects weight data through a load-bearing support: the initial total weight of a storage container, numbered C107, was recorded as 215 kg, of which the container itself weighed 15 kg and the contents weighed 200 kg. After continuous monitoring for five days, the weight suddenly dropped to 110 kg, triggering a weight anomaly warning threshold. Simultaneously, six piezoelectric sensor arrays mounted on the container wall were activated. After filtering out ambient background noise, the system focused on analyzing the target frequency band of 110 to 130 Hz. On the fifth day, at a height of 60 cm from the bottom of the container, the main peak frequency was detected to have shifted from the reference value of 125 Hz to 128 Hz, generating an acoustic feature spectrum with frequency shift markers. In the data fusion phase, the system calibrates the acoustic reference frequency to 64 Hz based on the weight change rate. After identifying an abnormal energy gradient of 4.5 dB per Hz at the 128 Hz location, the frequency at that point is corrected to 126 Hz, and the energy value is adjusted to 82 dB. After time alignment processing, a correlation matrix is ​​generated, revealing the physical law that every 50 kg decrease in weight corresponds to a positive frequency shift of 1.8 Hz. Next, liquid level feature extraction is performed: monitoring data from the middle of the tank, 60 to 100 cm from the bottom, is stacked and analyzed through sensor spatial grouping. An energy peak of 85 dB at 80 cm from the bottom is identified. After adding a 2 cm solution characteristic compensation, the current liquid level is output as 82 cm. In the final state determination phase, the time-series analysis model processes five days of continuous liquid level data: from 178 cm on the first day to 82 cm on the fifth day, the cumulative decrease over three days reaches 96 cm, far exceeding the normal evaporation rate. The model calculates a leakage probability of 98%, triggering a level one alarm. The system automatically locates the coordinates of the leaking tank and initiates the safety handling procedure, successfully replacing the sealing ring within 20 minutes to prevent a 1200 liter solvent leak.

[0118] This solution utilizes a multi-sensor collaborative monitoring mechanism to achieve comprehensive perception and intelligent early warning of the contents of plastic packaging containers. It integrates weight monitoring and acoustic analysis to construct a dual-dimensional detection system, effectively identifying abnormal consumption patterns of the contents; a dynamic calibration mechanism eliminates environmental interference and container resonance effects, improving the accuracy of liquid level inversion; a time-series analysis model automatically identifies risk states such as overflow and leakage and generates early warning signals; a non-intrusive design ensures safe warehousing operations; and multi-source data fusion technology provides real-time decision support for inventory management, significantly enhancing the safety protection capabilities of hazardous chemical storage.

[0119] Figure 3This application provides a schematic diagram of the structure of a real-time monitoring system for the contents of a plastic packaging drum that integrates multiple sensors, as shown in the embodiment of this application. Figure 3 As shown, the system includes:

[0120] The data acquisition module 31 is used to integrate an embedded weighing sensor on the load-bearing bracket at the storage location of the plastic packaging barrel, and to continuously acquire real-time weight data of the barrel's own weight and the contents based on the embedded weighing sensor.

[0121] The generation module 32 is used to collect real-time sound signals of barrel wall vibration based on the cantilever structure sensor on the outer wall of the plastic packaging barrel, select the inherent resonance frequency band of the plastic packaging barrel according to the noise distribution characteristics of the storage environment, perform frequency shift analysis on the real-time sound signal, and generate a frequency domain feature map containing the characteristics of liquid level reflected sound waves.

[0122] The generation module 32 is also used to adjust the calibration suspension reference value based on the weight data, perform resonant cavity compensation by combining the energy gradient characteristics of the frequency domain feature map, generate optimized frequency domain data, and fuse the real-time weight data and the optimized frequency domain data by time alignment to generate state analysis data describing the relationship between the liquid level fluctuation of the contents and the weight of the barrel.

[0123] The detection module 33 is used to perform neighborhood stacking detection on the state analysis data based on the spatial distribution of the piezo-acoustic resonance sensor array of the plastic packaging barrel, and extract liquid level features characterizing the liquid level height.

[0124] The output module 34 is used to input the liquid level features into a time-series state-based learning mechanism and output a status marker reflecting whether the contents of the plastic packaging barrel are overflowing, leaking, or have an abnormal liquid level, thus forming a real-time monitoring method for the contents of the plastic packaging barrel that integrates multiple sensors.

[0125] Figure 3 The aforementioned real-time monitoring system for the contents of plastic packaging drums integrating multiple sensors can perform... Figure 1 The implementation principle and technical effects of the real-time monitoring method for the contents of a plastic packaging drum integrating multiple sensors, as described in the illustrated embodiment, will not be repeated here. The specific operation methods of each module and unit in the real-time monitoring system for the contents of a plastic packaging drum integrating multiple sensors in the above embodiments have been described in detail in the embodiments related to this method, and will not be elaborated upon here.

[0126] In one possible design, Figure 3 The real-time monitoring system for the contents of a plastic packaging drum integrating multiple sensors, as shown in the embodiment, can be implemented as a computing device, such as... Figure 4 As shown, the computing device may include a storage component 41 and a processing component 42;

[0127] The storage component 41 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 42.

[0128] The processing component 42 is used for the above Figure 1 The embodiment describes a method for real-time monitoring of the contents of a plastic packaging barrel that integrates multiple sensors.

[0129] The processing component 42 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-described method. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method.

[0130] Storage component 41 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0131] Of course, computing devices may also include other components, such as input / output interfaces, display components, communication components, etc.

[0132] Input / output interfaces provide interfaces between processing components and peripheral interface modules, which can be output devices, input devices, etc.

[0133] The communication components are configured to facilitate wired or wireless communication between computing devices and other devices.

[0134] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform. In this case, the computing device can refer to a cloud server, and the aforementioned processing components, storage components, etc., can be basic server resources rented or purchased from the cloud computing platform.

[0135] This application also provides a computer storage medium storing a computer program, which, when executed by a computer, can perform the above-described functions. Figure 1 The illustrated embodiment provides a method for real-time monitoring of the contents of a plastic packaging drum that integrates multiple sensors.

[0136] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0137] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0138] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0139] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for real-time monitoring of the contents of a plastic packaging drum integrating multiple sensors, characterized in that, include: An embedded weighing sensor is integrated on the load-bearing bracket at the storage location of the plastic packaging drum. Based on the embedded weighing sensor, real-time weight data of the drum's own weight and the contents are continuously collected. Based on the cantilever structure sensor on the outer wall of the plastic packaging barrel, the real-time sound signal of the barrel wall vibration is collected. According to the noise distribution characteristics of the storage environment, the inherent resonance frequency band of the plastic packaging barrel is selected. Frequency shift analysis is performed on the real-time sound signal to generate a frequency domain feature map containing the characteristics of liquid level reflected sound waves. Based on the real-time weight data, the suspension reference value is adjusted and calibrated. The resonant cavity is compensated by combining the energy gradient characteristics of the frequency domain feature map to generate optimized frequency domain data. The real-time weight data and the optimized frequency domain data are fused by time alignment to generate state analysis data describing the relationship between the liquid level fluctuation of the contents and the weight of the barrel. Based on the spatial distribution of the piezo-acoustic resonance sensor array in the plastic packaging barrel, neighborhood stacking detection is performed on the state analysis data to extract liquid level features characterizing the liquid level height. The liquid level features are input into a time-series state-based learning mechanism, and the output is a state marker reflecting whether the contents of the plastic packaging barrel are overflowing, leaking, or have an abnormal liquid level, thus forming a real-time monitoring method for the contents of plastic packaging barrels that integrates multiple sensors. Based on the real-time weight data, the suspension reference value is adjusted and calibrated. Resonance cavity compensation is performed using the energy gradient characteristics of the frequency domain feature map, generating optimized frequency domain data, including: Extract the total weight value from the real-time weight data, calculate the adjustment factor based on the difference between the total weight value and the suspension reference value, and update the suspension reference value based on the calculated adjustment factor. Based on the energy value data in the frequency domain feature map, the rate of energy change between adjacent frequency points is calculated to generate energy gradient features; Based on the updated suspension reference value and the energy gradient characteristics, the energy anomaly region caused by barrel resonance is identified, and a compensation offset is generated. The compensation offset is applied to the corresponding frequency point of the frequency domain feature map, and the energy value of the frequency domain feature map is adjusted at the same time. The adjusted frequency domain feature map is then output as optimized frequency domain data. By fusing the real-time weight data and the optimized frequency domain data through time alignment, state analysis data describing the correlation between the liquid level fluctuation of the contents and the weight of the container is generated, including: Assign timestamps based on the absolute time point of data acquisition to real-time weight data and optimized frequency domain data, and generate time-corresponding data point pairs through timestamp matching and interpolation; The weight value and the energy value of the key frequency point are extracted from the data point pair and combined into a multi-dimensional data vector. Based on the multidimensional data vector, the covariance mode of weight value change and energy value change is analyzed, and the correlation matrix between the liquid level fluctuation of the contents and the weight of the container is generated. The rows of the correlation matrix are output as time series, and the columns of the correlation matrix represent combined parameters of weight and frequency domain features as state analysis data. Based on the spatial distribution of the piezo-acoustic resonant sensor array in the plastic packaging barrel, neighborhood stacking detection is performed on the state analysis data to extract liquid level features characterizing the liquid level height, including: Based on the physical installation location of the piezo-acoustic resonant sensor array, the sensors are divided into multiple neighborhood groups containing spatially adjacent sensors according to the sensor spacing and barrel geometry. Extract the data elements corresponding to each sensor from the state analysis data, and stack the data elements of all sensors in each neighborhood group to form a multidimensional stacked matrix. Calculate the distribution pattern of energy values ​​with height in the stacking matrix and identify the location of energy peaks; Based on the mapping between the energy peak position and the sensor height, the liquid level height value is derived as a liquid level feature.

2. The method according to claim 1, characterized in that, Based on the noise distribution characteristics of the storage environment, the inherent resonant frequency band of the plastic packaging drum is selected. Frequency shift analysis is performed on the real-time sound signal to generate a frequency domain feature map containing the characteristics of liquid level reflected sound waves, including: Collect background noise signals from the storage environment and analyze the energy distribution of the background noise signals; Based on the physical properties of plastic packaging barrels, the frequency range in which the barrel produces the maximum vibration when it is empty is determined through preliminary experiments to be the inherent resonant frequency band. Combining the inherent resonant frequency band and the energy distribution, the low-noise resonant frequency band is selected as the target frequency band. The real-time audio signal is divided into time windows, and frequency transformation is applied to each window to convert the time-domain signal into a frequency-domain representation. Within the target frequency band, the difference between the peak frequency position and the preset reference frequency in the frequency components is detected and calculated, and the difference is used as the offset relative to the reference frequency. Based on the frequency domain representation and the offset mapping, a two-dimensional coordinate system diagram is generated, and the offset abrupt change region is marked to output a frequency domain feature map.

3. The method according to claim 1, characterized in that, The liquid level features are input into a time-series state-based learning mechanism, which outputs a status marker reflecting whether the contents of the plastic packaging drum are overflowing, leaking, or have an abnormal liquid level. This constitutes a real-time monitoring method for the contents of plastic packaging drums integrating multiple sensors, including: Collect historical liquid level characteristic data and status labels including category labels such as overflow, leakage or abnormal liquid level, and construct a training dataset; The training dataset is used to train a time-series state-based learning mechanism, which is input as a liquid level feature time series and internal parameters are optimized to learn pattern changes. The real-time liquid level feature time series is input into the trained learning mechanism, and the state probability distribution at each time point is calculated through internal parameters. The highest probability category is selected as the output state label based on the state probability distribution.

4. The method according to claim 1, characterized in that, Based on the energy value data in the frequency domain feature map, the rate of energy change between adjacent frequency points is calculated to generate energy gradient features, including: Extract the energy value sequence of the frequency domain feature map in ascending order of frequency; Based on the energy value data of the energy value sequence, the absolute difference between the current point and the previous point is calculated for each frequency point, and a unit frequency change sequence is generated by combining the frequency span. A judgment threshold is set for each unit frequency change. If the value exceeds the threshold, a flag bit is set at the corresponding frequency position to form a corresponding flag sequence. The energy gradient feature is formed by combining the unit frequency change sequence and the label sequence.

5. The system for real-time monitoring of the contents of a plastic packaging drum integrating multiple sensors according to claim 1, characterized in that, include: An embedded weighing sensor is integrated on the load-bearing bracket at the storage location of the plastic packaging drum. Based on the embedded weighing sensor, real-time weight data of the drum's own weight and the contents are continuously collected. Based on the cantilever structure sensor on the outer wall of the plastic packaging barrel, the real-time sound signal of the barrel wall vibration is collected. According to the noise distribution characteristics of the storage environment, the inherent resonance frequency band of the plastic packaging barrel is selected. Frequency shift analysis is performed on the real-time sound signal to generate a frequency domain feature map containing the characteristics of liquid level reflected sound waves. Based on the weight data, the suspension reference value is adjusted and calibrated. The resonant cavity is compensated by combining the energy gradient characteristics of the frequency domain feature map to generate optimized frequency domain data. The real-time weight data and the optimized frequency domain data are fused by time alignment to generate state analysis data describing the relationship between the liquid level fluctuation of the contents and the weight of the barrel. Based on the spatial distribution of the piezo-acoustic resonance sensor array in the plastic packaging barrel, neighborhood stacking detection is performed on the state analysis data to extract liquid level features characterizing the liquid level height. The liquid level features are input into a time-series state-based learning mechanism, which outputs a status marker reflecting whether the contents of the plastic packaging barrel are overflowing, leaking, or have an abnormal liquid level, thus forming a real-time monitoring method for the contents of plastic packaging barrels that integrates multiple sensors.

6. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to realize the real-time monitoring method for the contents of a plastic packaging barrel integrating multiple sensors as described in claim 1.

7. A computer storage medium, characterized in that, The device contains a computer program that, when executed by a computer, implements a method for real-time monitoring of the contents of a plastic packaging drum integrating multiple sensors, as described in claim 1.

Images