Remote monitoring method and system for asphalt mixture mixing equipment based on internet of things

Abstract

The present application relates to the technical field of industrial internet of things monitoring, and discloses a remote monitoring method and system for asphalt mixture mixing equipment based on the internet of things.The method comprises the following steps: collecting key component operating parameters and external environmental variables, performing time domain statistical filtering to obtain an initial operating sequence; performing adaptive window length smoothing and frequency domain noise suppression to obtain a clean operating sequence; using the external environmental variables to perform environmental equivalent correction on the clean operating sequence to obtain a calibration characteristic vector; performing multi-dimensional feature correlation analysis according to the calibration characteristic vector to determine a working condition switching label; determining a dynamic reference range according to the working condition switching label; calculating a preliminary abnormality probability according to the calibration characteristic vector and the dynamic reference range; performing historical trend weighting and abnormality integration to obtain a final health assessment result.This method eliminates environmental interference and working condition misjudgment through environmental correction and dynamic reference following, and realizes accurate assessment of the health state of the equipment.

Description

Technical Field

[0001] This invention relates to the field of industrial Internet of Things (IoT) and intelligent monitoring technology for construction machinery, and in particular to a remote monitoring method and system for asphalt mixing equipment based on IoT. Background Technology

[0002] Currently, in modern road construction and maintenance projects, asphalt mixing equipment serves as a core production facility, and the stability of its operation directly determines the quality and progress of the project. Because this type of equipment operates for extended periods in complex construction site environments characterized by high dust levels, large temperature differences, and drastic load variations, real-time monitoring of the health status of its key components has become a necessary means to ensure construction safety and efficiency.

[0003] In one existing technology, a microcontroller (MCU) deployed on the device typically collects sensor data and uses preset upper and lower threshold values ​​to make simple judgments on parameters such as vibration, temperature, or current exceeding limits. While this monitoring method can provide some protection when the equipment is running stably, its core flaw lies in ignoring the dynamic nature of the equipment's operating state and its coupling with the environment. For example, when the ambient temperature drops suddenly or the equipment undergoes normal load changes, operating parameters often experience significant "normal fluctuations." These fluctuations may numerically exceed conventional static thresholds, but they are not actually faults. Conversely, some early, weak fault signals may be masked by large fluctuations in operating conditions and fail to trigger alarms.

[0004] Therefore, existing technologies have the technical problem of being unable to accurately distinguish between normal fluctuations and potential faults in complex environments, thus leading to high false alarm or false alarm rates. Summary of the Invention

[0005] This invention provides a remote monitoring method and system for asphalt mixture mixing equipment based on the Internet of Things, in order to solve the technical problems of high false alarm rate or missed alarm rate in the prior art.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a remote monitoring method for asphalt mixture mixing equipment based on the Internet of Things, comprising:

[0007] The operating parameters of key components of the asphalt mixture mixing equipment and external environmental variables are collected, and the operating parameters of the key components are subjected to time-domain statistical filtering to obtain the initial operating sequence.

[0008] Based on the initial running sequence, adaptive window length smoothing and frequency domain noise suppression are performed to obtain a clean running sequence.

[0009] Time-varying features are extracted from the cleanroom operation sequence, and environmental equivalence correction is performed using the external environmental variables to obtain a calibration feature vector.

[0010] Based on the calibration feature vector, perform multidimensional feature correlation analysis and pattern matching to determine the operating condition switching label;

[0011] Based on the operating condition switching label, retrieve the corresponding health status data from the preset historical database or maintain the current benchmark to determine the dynamic benchmark range;

[0012] Based on the calibration feature vector and the dynamic benchmark range, deviation calculation and trend direction analysis are performed to obtain the preliminary anomaly probability;

[0013] Based on the preliminary anomaly probability, historical trend weighting and anomaly integral calculations are performed to obtain the final health assessment result.

[0014] Secondly, the present invention provides a remote monitoring system for asphalt mixture mixing equipment based on the Internet of Things, comprising:

[0015] The data acquisition and preprocessing module is used to acquire the operating parameters of key components of the asphalt mixture mixing equipment and external environmental variables, and to perform time-domain statistical filtering on the operating parameters of the key components to obtain the initial operating sequence.

[0016] The smoothing and denoising module is used to perform adaptive window length smoothing and frequency domain noise suppression based on the initial running sequence to obtain a clean running sequence;

[0017] The feature extraction and correction module is used to extract time-varying features based on the clean operation sequence and perform environmental equivalence correction processing using the external environmental variables to obtain a calibration feature vector.

[0018] The operating condition identification module is used to perform multi-dimensional feature association analysis and pattern matching based on the calibration feature vector to determine the operating condition switching label;

[0019] The dynamic baseline determination module is used to switch tags according to the working conditions, retrieve corresponding health status data from a preset historical database or maintain the current baseline, and determine the dynamic baseline range.

[0020] The deviation analysis module is used to perform deviation calculation and trend direction analysis based on the calibration feature vector and the dynamic benchmark range to obtain a preliminary anomaly probability.

[0021] The health assessment module is used to calculate the final health assessment result by performing historical trend weighting and anomaly integral calculation based on the preliminary anomaly probability.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] (1) This invention uses external environmental variables to perform environmental equivalent correction on the time-varying features extracted from the clean operation sequence, and determines the dynamic reference range based on the working condition switching label. This dynamic correction and reference following mechanism can separate the normal parameter fluctuations caused by changes in ambient temperature or load adjustment of the equipment from the real fault signals, so that the monitoring system can always maintain the judgment standard that is adapted to the current working condition in the complex and ever-changing construction site environment. It solves the technical problem that the existing technology relies on static thresholds, which leads to a high false alarm rate when the environment changes suddenly or the working condition switches, and cannot accurately distinguish between normal fluctuations and potential faults.

[0024] (2) This invention performs multidimensional feature correlation analysis on the calibration feature vector, calculates the feature change synchronization rate to identify the operating condition switching label, and locks the dynamic benchmark when there is no sudden change in operating condition. This logical judgment mechanism based on multidimensional data coordination can effectively identify whether the equipment is in a normal mode switching process (multi-parameter synchronous change) or a sudden failure (single parameter outlier), and prevents the incorrect use of the health benchmark due to the system misjudging the operating condition. It solves the problem that the prior art is prone to misjudging the sudden failure as an operating condition switch, resulting in missed reports, or misjudging the normal switch as an abnormality, resulting in false reports.

[0025] (3) This invention obtains the preliminary anomaly probability by combining deviation calculation and trend direction analysis, and further performs historical trend weighting and anomaly integral calculation to obtain the final health assessment result; this multi-level assessment strategy from instantaneous deviation to continuous trend and then to time integral introduces the cumulative effect verification in the time dimension, effectively filtering out the occasional instantaneous interference that may still remain even after preprocessing; it solves the problem that the existing technology triggers alarms based on a single point exceeding the limit, lacks comprehensive consideration of the anomaly persistence and development trend, and leads to poor system robustness. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the remote monitoring method for asphalt mixture mixing equipment based on the Internet of Things provided in the first embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of the remote monitoring system for an asphalt mixing plant based on the Internet of Things, provided in the second embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Reference Figure 1 The first embodiment of the present invention provides a remote monitoring method for asphalt mixture mixing equipment based on the Internet of Things, including the following steps:

[0030] S11, Collect the operating parameters of key components of the asphalt mixture mixing equipment and external environmental variables, and perform time-domain statistical filtering on the operating parameters of the key components to obtain the initial operating sequence;

[0031] S12, based on the initial running sequence, perform adaptive window length smoothing and frequency domain noise suppression to obtain a clean running sequence;

[0032] S13, extract time-varying features based on the clean operation sequence, and perform environmental equivalence correction processing using the external environmental variables to obtain a calibration feature vector;

[0033] S14. Based on the calibration feature vector, perform multidimensional feature correlation analysis and pattern matching to determine the working condition switching label;

[0034] S15, based on the working condition switching tag, retrieve the corresponding health status data from the preset historical database or maintain the current benchmark to determine the dynamic benchmark range;

[0035] S16, Based on the calibration feature vector and the dynamic reference range, perform deviation calculation and trend direction analysis to obtain the preliminary anomaly probability;

[0036] S17. Based on the preliminary abnormality probability, perform historical trend weighting and abnormality integral calculation to obtain the final health assessment result.

[0037] In step S11, the operating parameters of key components of the asphalt mixture mixing equipment and external environmental variables are collected, and the operating parameters of the key components are subjected to time-domain statistical filtering to obtain an initial operating sequence, including:

[0038] Real-time acquisition of the operating parameters of the key components and the external environmental variables;

[0039] Based on the pre-stored historical key component operating parameters, calculate the statistical distribution characteristics, and determine the current dynamic statistical threshold based on the statistical distribution characteristics;

[0040] The initial operating sequence is obtained by removing the numerical points in the operating parameters of the key components that exceed the dynamic statistical threshold.

[0041] In one implementation, this embodiment uses a multi-dimensional sensor network deployed on the asphalt mixture mixing equipment to collect the operating parameters of the key components and the external environmental variables in real time.

[0042] Specifically, regarding the operating parameters of the key components, this embodiment deploys a high-frequency Hall current sensor with a sampling rate of no less than 1kHz at the three-phase input terminal of the main stirring motor to acquire instantaneous current data reflecting load changes; a piezoelectric vibration acceleration sensor is deployed at the bearing seat of the stirring cylinder to acquire vibration data reflecting the smoothness of mechanical operation; and an optical encoder or speed sensor is installed at the drive end of the stirring shaft to acquire the real-time speed signal of the equipment, providing an angular domain reference for subsequent frequency domain order analysis. Regarding the external environmental variables, this embodiment deploys an industrial-grade weather station integrating a thermistor and a capacitive humidity sensor outside the equipment control room to acquire ambient temperature and relative humidity data. This embodiment synchronously receives the analog signals output by the above sensors through a multi-channel data acquisition card, performs analog-to-digital conversion on them, and uniformly adds a hardware timestamp with microsecond-level precision, thereby forming the original operating data set and the original environmental data set.

[0043] It should be noted that the preset historical reference window is determined based on the autocorrelation analysis of the signal, ensuring that the data within the window is statistically stationary in the short term. During the stable operation phase of the equipment, a preset length of raw operating data sequence is collected, and the autocorrelation function of this sequence is calculated. This function characterizes the degree of correlation of the signal at different lag times. Subsequently, the decay trend of the autocorrelation coefficient with lag time is analyzed, and the first decay of the autocorrelation coefficient to a preset decorrelation threshold (e.g., ...) is selected. The time length corresponding to a value below approximately 0.368 is used as the length of the preset historical reference window. The preset decorrelation threshold is selected. Based on the correlation time definition in statistical signal processing, this indicates that the linear correlation between data points has significantly decreased to a negligible level. This determination ensures that the window contains a sufficient sample size for robust statistics while avoiding the introduction of non-stationary long-term trends due to an excessively long window, thus guaranteeing the local adaptability of subsequent threshold calculations.

[0044] In one implementation, this embodiment calculates statistical distribution characteristics based on the running data records within the preset historical reference window. To completely resolve the masking effect of the traditional mean-variance method when facing large-value fault signals—that is, the problem of extremely large fault values ​​raising the mean and variance, leading to an erroneous expansion of the threshold range—this embodiment employs a robust statistical method.

[0045] Specifically, this embodiment acquires a historical data sequence with a window length preceding the current time, calculates the median of the sequence as a location estimate, and calculates the absolute median difference as a dispersion estimate. The absolute median difference is calculated by taking the absolute value of the difference between each data point in the sequence and the median, and then taking the median of these absolute values. The median and the absolute median difference together constitute the statistical distribution characteristic, which has a high collapse point for outliers, ensuring the robustness of the statistical benchmark.

[0046] It is worth noting that this embodiment determines the current dynamic statistical threshold based on the aforementioned statistical distribution characteristics. This threshold is determined by constructing a dynamic confidence interval based on Chebyshev's inequality or the probability density function of the data distribution. Specifically, the upper threshold is equal to the median plus the product of the coefficient k and the absolute median difference, and the lower threshold is equal to the median minus the product of the coefficient k and the absolute median difference.

[0047] The method for selecting the coefficient k is as follows: A labeled validation dataset containing known normal noise and abnormal spikes is constructed. The range of values ​​for coefficient k is traversed, and for each k value, the true positive rate for noise removal and the false positive rate for normal signal deletion are calculated. Using receiver operating characteristic (ROC) curve analysis, the Youden index (the difference between the true positive rate and the false positive rate) is calculated. The k value that maximizes the Youden index is selected as the parameter to be fixed in the system. The coefficient k determined by this method statistically achieves the optimal balance between sensitivity and specificity in anomaly identification.

[0048] The labeled verification dataset generates abnormal spike samples by injecting simulated instantaneous pulse signals into known healthy operating data segments of the device; at the same time, the portion of the healthy data without injected pulses is selected as normal noise samples; together, they constitute the labeled dataset used to optimize coefficient k.

[0049] In another implementation, this embodiment removes numerical points from the original running data set that exceed the dynamic statistical threshold to obtain the initial running sequence. To maintain the continuity of the time series and prevent time gaps caused by directly deleting data points, this embodiment employs a numerical clamping strategy for the removal process.

[0050] Specifically, the current sampling point is monitored in real time. If the value of the sampling point is greater than the upper threshold, it is replaced with the upper threshold; if the value of the sampling point is less than the lower threshold, it is replaced with the lower threshold; if the value of the sampling point is between the upper and lower thresholds, it remains unchanged. The data stream output after the above processing is the initial running sequence. This sequence retains the original trend of the data while limiting all instantaneous outlier noise within a reasonable dynamic statistical range.

[0051] In step S12, based on the initial running sequence, adaptive window length smoothing and frequency domain noise suppression are performed to obtain a clean running sequence, including:

[0052] Based on the fluctuation amplitude of the initial running sequence, adaptive segmentation and local weighted mean calculation are performed on the initial running sequence to obtain a time-domain smoothed sequence;

[0053] The time-domain smoothed sequence is converted to the frequency domain by a fast Fourier transform, and signal components in the preset interference frequency band are identified and filtered out to obtain frequency domain cleaned data.

[0054] The frequency domain purification data is reconstructed by inverse Fourier transform to obtain the clean operation sequence.

[0055] In one implementation, this embodiment performs adaptive segmentation and local weighted mean calculation on the initial running sequence based on the fluctuation amplitude of the initial running sequence. Because the data fluctuation characteristics of asphalt mixing equipment vary greatly under different operating conditions (such as the violent oscillations at the moment of aggregate feeding and the slight fluctuations during the stable mixing period), a filter with a fixed window length cannot simultaneously achieve both noise reduction and signal feature preservation. Therefore, this embodiment adopts a dynamic window length strategy.

[0056] Specifically, the system first calculates the local variance of each data point in the initial running sequence within its neighborhood. This local variance characterizes the degree of signal fluctuation at the current moment. Then, in this embodiment, the filter window length at the current moment is calculated based on a preset inverse proportional mapping function. The calculation formula is set as follows:

[0057]

[0058] in, and These are the preset minimum and maximum window lengths (e.g., 5 and 50). This is the sensitivity coefficient. For local variance, This represents the rounding function. The physical meaning of this formula is that when the signal fluctuates drastically (i.e., the variance is large), the exponent term approaches zero, and the window length shrinks to near its minimum, thus preserving the transient details of the signal. When the signal fluctuates smoothly (i.e., the variance is small), the exponent term approaches one, and the window length expands to near its maximum, thus maximizing the smoothing and denoising effect. Using rounding ensures that the calculation result covers the set maximum window boundary, avoiding the boundary value loss problem that may occur when rounding down. In this embodiment, based on the calculated dynamic window length, a Gaussian weighted average is performed on the data points within the window to obtain the smoothed time-domain sequence.

[0059] It should be noted that the preset minimum and maximum window lengths are not arbitrarily set, but are calibrated in engineering based on the system's sampling frequency, the duration of the transient characteristics of the critical fault signal, and the maximum permissible response delay. Specifically, the minimum window length is set according to the transient response characteristics of the fault signal. To prevent excessive smoothing from blurring the rising edge characteristics of the signal when a real fault occurs in the equipment (such as a sudden current jump caused by mechanical jamming), this value should be set to the minimum number of sampling points required to cover the shortest effective transient pulse. For example, it is set to 5 points at a 1kHz sampling rate to retain 5ms-level transient details. The maximum window length, on the other hand, needs to strike a balance between signal-to-noise ratio gain and system hysteresis. It is set to the number of sampling points that can cover at least one complete cycle of the main background noise (such as power frequency interference or low-frequency mechanical ripple) without exceeding the maximum tolerable delay of the system's real-time monitoring (e.g., 50ms).

[0060] It should be noted that the sensitivity coefficient... It was determined through offline simulation, aiming to balance the response speed to abrupt signals with the ability to suppress steady-state noise. A standard signal containing typical transient changes (such as the current curve of a motor starting process) was acquired and superimposed with Gaussian white noise; different... The values ​​are filtered, and the mean square error (MSE) between the filtered result and the noise-free standard signal is calculated; the value that minimizes the MSE is selected. The value serves as the final system parameter, ensuring optimal performance of the algorithm in practical applications.

[0061] In one implementation, this embodiment performs a Fast Fourier Transform (FFT) on the time-domain smoothed sequence, mapping it from the time domain to the frequency domain to generate a spectrum. Based on a preset interference frequency band, this embodiment identifies the corresponding signal components in the spectrum and sets their amplitude to zero or attenuates them to the background noise level, thereby obtaining frequency-domain cleaned data.

[0062] It is worth noting that the preset interference frequency band is determined through spectral characteristic analysis of the equipment. This determination process involves two levels: first, when the equipment is powered on but not in a stirring state, sensor data is collected and fixed frequency spikes (such as power frequency interference) with amplitudes significantly higher than the base noise are identified; second, when the equipment is under load, based on the real-time speed signal obtained from the aforementioned speed sensor or the speed estimated based on the fundamental frequency of the current signal, order analysis is used to identify fixed frequency components (such as power supply harmonics or structural natural frequency resonance) that are not linearly correlated with the equipment's spindle speed. The frequency bands containing these fixed frequency spikes (e.g., center frequency) are then analyzed. The preset interference frequency band is determined. This mechanism can accurately eliminate periodic electromagnetic interference or mechanical resonance interference that is unrelated to the actual operating conditions, preventing it from misleading subsequent fault diagnosis.

[0063] In another implementation, this embodiment reconstructs the frequency domain cleaned data using Inverse Fast Fourier Transform (IFFT). This reconstruction process converts the processed complex spectral data back into a real-domain time series. To eliminate the Gibbs Phenomenon that may occur due to frequency domain truncation—that is, ringing at signal transitions—this embodiment applies a smoothing window function (such as a Hanning window) to the frequency domain data before reconstruction. The data sequence with the maximized signal-to-noise ratio generated after the dual processing of adaptive time-domain smoothing and fixed-point frequency-domain culling is the clean running sequence.

[0064] In step S13, time-varying features are extracted based on the cleanroom operation sequence, and environmental equivalence correction is performed using the external environmental variables to obtain a calibration feature vector, including:

[0065] The cleanroom operation sequence is subjected to time-series statistical analysis to calculate the local fluctuation amplitude and the rate of change of operating parameters within a preset time window, and to obtain initial feature data;

[0066] Based on the external environmental variables, the corresponding gain value is retrieved from the preset environmental impact factor curve to obtain the environmental compensation gain under the current environment.

[0067] The initial feature data is corrected by division using the environmental compensation gain to obtain the calibration feature vector.

[0068] In one implementation, this embodiment performs time-series statistical analysis on the cleanroom operation sequence to extract key indicators reflecting the equipment's operating status. Specifically, this embodiment employs a sliding window technique, setting a preset time window (e.g., 10 seconds) and calculating the statistical characteristics of the sequence within this window. For the local fluctuation amplitude, this embodiment calculates the standard deviation of the data within the window; this indicator is used to quantify the dispersion and vibration energy of the equipment under the current operating conditions. For the rate of change of the operating parameters, this embodiment performs linear regression fitting on the data within the window and uses the slope of the fitted line as the rate of change indicator; this indicator is used to quantify the evolution trend of the parameters over time. The standard deviation and the slope together constitute the initial feature data.

[0069] It should be noted that the preset environmental impact factor curve is constructed through offline correlation analysis and physical modeling. During the healthy operation phase of the equipment's entire life cycle, historical operating data and synchronized environmental variable data covering a preset operating temperature range (e.g., -10 degrees Celsius to 40 degrees Celsius) are collected.

[0070] Specifically, for each ambient temperature sampling point, this embodiment first calculates the arithmetic mean of the stable operating data of the equipment at that temperature as a baseline value, and compares it with the baseline value under standard operating conditions (i.e., the rated ambient temperature at which the equipment was calibrated at the factory, such as 25 degrees Celsius) to calculate the baseline ratio corresponding to that temperature point, thereby constructing a discrete sample set containing multiple sets of "temperature-baseline ratio" data. Subsequently, the least squares method is used to analyze the functional relationship between the key operating parameters in this sample set and the changes in external environmental variables.

[0071] Due to the rheological properties of asphalt mixtures, their viscosity increases non-linearly with decreasing temperature, leading to increased mixing resistance. This means that the functional relationship exhibits significant non-linear characteristics. Therefore, this embodiment employs a high-order polynomial fitting technique (e.g., a quadratic or cubic polynomial regression model), setting the fitting function as:

[0072]

[0073] in, For temperature, This is the gain coefficient. , ... These are the undetermined coefficients, which together form an undetermined coefficient vector. By minimizing the sum of squared residuals between the discrete sample set and the fitted function, the optimal coefficient vector is calculated, thereby solidifying this physical correlation into one or more environmental impact factor curves. This curve uses the environmental variable values ​​as the horizontal axis and the ratio of the parameter baseline values ​​under the current environment to the baseline values ​​under standard operating conditions (i.e., the fitted gain coefficients) as the vertical axis.

[0074] It is worth noting that this embodiment utilizes the environmental compensation gain to perform a division-equivalent correction on the initial feature data. Specifically, this correction strategy primarily targets load-related parameters (such as stirring motor current and drive torque) with clear physical correlations, such as changes in medium viscosity and mechanical resistance caused by temperature. For parameters that may exhibit nonlinear responses, such as vibration, the system can choose to skip the correction or use an independent nonlinear mapping curve. The specific calculation method is to divide the load-related values ​​in the initial feature data by the corresponding environmental compensation gain. The specific calculation formula is:

[0075]

[0076] in, The initial feature data, To extract from the environmental impact factor curve based on current environmental variables (such as temperature) The environmental compensation gain is retrieved from the database. Through this division correction operation, the system restores the characteristic values ​​that are naturally increased due to environmental factors (such as low temperature and high viscosity) (e.g., increased to 1.2 times the standard value) to the equivalent level under standard environmental conditions (i.e., divided by 1.2), thereby obtaining the calibration characteristic vector. This process effectively removes environmental interference, ensuring that subsequent fault diagnosis only targets changes in the physical state of the equipment itself.

[0077] In step S14, based on the calibration feature vector, multidimensional feature correlation analysis and pattern matching are performed to determine the operating condition switching label, including:

[0078] Calculate the synchronization rate of changes between features of different dimensions in the calibration feature vector to obtain the feature change synchronization rate;

[0079] The synchronization rate of the feature changes is compared with a preset coordination threshold.

[0080] If the feature change synchronization rate is greater than or equal to the preset coordination threshold, the data in the corresponding time period is marked as a potential switching segment, the similarity between the potential switching segment and each standard template in the preset historical working condition mode library is calculated, and the working condition name corresponding to the template with the highest similarity is determined as the working condition switching label.

[0081] If the synchronization rate of the feature change is less than the preset coordination threshold, then a working condition switching tag is generated that is identified as a non-working condition change.

[0082] In one implementation, this embodiment calculates the synchronization rate of changes between features of different dimensions in the calibration feature vector. Because the key physical parameters of an asphalt mixing plant typically exhibit strong physical coupling when performing normal operating condition switching, such as transitioning from no-load to mixing mode. For example, as mixing resistance increases, the motor current and mechanical vibration amplitude should show a synchronous upward trend. Therefore, this embodiment quantifies this degree of synchronization by calculating the correlation between features.

[0083] Specifically, this embodiment selects a sliding time window. The length of this window is determined based on the mechanical response delay characteristics of the device, and is typically set to cover the entire load build-up process. Within this window, this embodiment calculates feature sequences of different dimensions (e.g., current feature sequences). With vibration characteristic sequence Pearson correlation coefficient between ) This is taken as the synchronization rate of the aforementioned feature change, and the calculation formula is as follows:

[0084]

[0085] in, The number of sample points within the window. and These are the mean of the sequences, For a very small positive number (e.g.) This correlation coefficient is used to prevent division-by-zero errors caused by a standard deviation of zero, thus enhancing the robustness of the algorithm in a static state. The value of this correlation coefficient ranges from -1 to 1; the closer the value is to 1, the stronger the positive correlation, indicating more coordinated changes in the multidimensional parameters.

[0086] It should be noted that the preset coordination threshold is determined through statistical analysis of historical normal operating data. Multidimensional data from the acquisition equipment during known normal operating condition switching processes are used to calculate the synchronization rate distribution within the aforementioned sliding window. Subsequently, the lower quartile of this distribution or a lower limit determined based on the 3-sigma principle is selected as the coordination threshold. This threshold serves as a logical boundary distinguishing between systemic operating condition changes and local single-parameter mutations; values ​​below this threshold typically indicate a break in the coupling relationship between physical parameters.

[0087] In one implementation, if the feature change synchronization rate is greater than or equal to the preset coordination threshold, it indicates that the current data change has multi-parameter coordination, conforms to the normal operating logic of the physical equipment, and is highly likely caused by equipment condition adjustments. In this embodiment, the data within this time window is marked as a potential switching segment. Subsequently, this embodiment calls a preset historical operating condition pattern library for matching.

[0088] It should be explained in detail that the construction process of the preset historical working condition mode library is as follows: First, based on the equipment PLC control log, original characteristic waveform segments with clear working condition switching markers (such as "cold start command" and "aggregate delivery command") are extracted from historical data; second, for each working condition type, the dynamic time warping (DTW) algorithm is used to align the time axis of multiple extracted waveform segments to eliminate the influence of differences in operation duration; third, the K-Means clustering algorithm or K-Shape time series clustering algorithm based on DTW distance is used to perform cluster analysis on the aligned waveforms and remove abnormal samples with large deviations; finally, the centroid or arithmetic mean waveform of the retained sample clusters is calculated and solidified as the "standard characteristic waveform template" of the working condition and stored in the library.

[0089] This embodiment employs the Dynamic Time Warping (DTW) algorithm to calculate the similarity distance between the current potential switching segment and each standard feature waveform template in the library. The operating condition name corresponding to the template with the smallest distance, below a preset matching tolerance, is then identified as the operating condition switching label. The preset matching tolerance is determined based on the intra-cluster distance of each operating condition cluster during the aforementioned clustering analysis, typically set to 1.2 to 1.5 times the intra-cluster average distance, ensuring that the statistical significance of the identification results is maintained while tolerating a certain degree of waveform distortion.

[0090] It is worth noting that if the synchronization rate of the characteristic changes is less than the preset coordination threshold, it indicates that the changes in each parameter exhibit decoupling or discrete characteristics. For example, if only a sudden increase in vibration amplitude is detected while the drive current remains unchanged, this usually suggests a loose mechanical component or sensor malfunction, rather than an increase in load. In this case, this embodiment generates a condition switching tag identified as a non-conditional abrupt change. This mechanism effectively prevents abnormal fluctuations in a single parameter from being misjudged as normal condition switching, thereby avoiding the incorrect use of inapplicable health benchmarks in subsequent steps.

[0091] In step S15, the dynamic reference range is determined based on the operating condition switching tag, including:

[0092] Based on the operating condition switching label, determine the category to which the device belongs, and obtain the current operating condition mode category of the device;

[0093] If the operating condition mode category is a known operating condition mode, then the corresponding historical health feature samples are retrieved and extracted from the preset historical database, and the distribution statistical analysis of the historical health feature samples is performed to obtain the dynamic benchmark range.

[0094] If the operating condition mode category is a non-operating condition change, then the dynamic reference range determined at the previous moment is obtained and assigned as the dynamic reference range at the current moment.

[0095] In one implementation, this embodiment determines the category based on the working condition switching label. The working condition switching label is output from the previous processing step and contains specific working condition semantic information (such as "aggregate drying stage" or "asphalt spraying stage") or special status identifiers (such as "non-working condition change"). This embodiment categorizes the tag into two main types: "Known Operating Condition Mode" and "Non-Operating Condition Abrupt Change," using a lookup table mapping or state machine logic. The lookup table mapping involves pre-establishing a table that stores the mapping relationship between all possible operating condition names (such as "aggregate drying stage," "asphalt spraying stage," "standby stage," etc.) and the category "Known Operating Condition Mode," as well as mapping the special identifier "Non-Operating Condition Abrupt Change" to the category "Non-Operating Condition Abrupt Change." The category can be obtained by directly querying this table based on the input operating condition switching tag. The state machine logic maintains an internal state machine. If the input operating condition switching tag is a specific operating condition name (not "Non-Operating Condition Abrupt Change"), the state machine transitions to the "Known Operating Condition Mode" state; if the input tag is "Non-Operating Condition Abrupt Change," the state machine transitions to the "Non-Operating Condition Abrupt Change" state, potentially triggering a baseline lock timer. The state machine's transition condition is based on the judgment of the input tag string content. This identification process constitutes the divergence point for subsequent baseline determination strategies, ensuring that the system can adopt distinctly different baseline management strategies for normal operation and abnormal abrupt changes.

[0096] It should be noted that the preset historical database is a "digital fingerprint database" built based on the health operation data of the equipment throughout its entire lifecycle. The database is constructed as follows: during the equipment's commissioning or post-maintenance health operation cycle, the system automatically records key parameter feature sequences under different operating conditions; using unsupervised clustering algorithms (such as K-Means) or manual annotation based on business logic, these data fragments are classified and stored according to operating condition categories. For example, the database stores vibration feature sample sets under "stirring mode" and current feature sample sets under "standby mode." This database provides a reliable health reference system for the generation of dynamic benchmarks.

[0097] In one implementation, if a known operating condition mode is identified, this embodiment retrieves and extracts historical health characteristic samples corresponding to that mode from the preset historical database. Subsequently, this embodiment performs distribution statistical analysis on the samples to determine the dynamic benchmark range. Specifically, this embodiment calculates the mean of the sample set ( ) and standard deviation ( Based on the principles of statistical process control (SPC), this embodiment constructs numerical boundaries covering a pre-set confidence interval (e.g., 99.7%):

[0098]

[0099] in, For safety, a value of 3 is typically used. The calculated closed interval... This refers to the dynamic reference range adapted to the current operating conditions. This mechanism ensures that the judgment criteria are adaptively adjusted according to changes in the operating load, avoiding false alarms caused by using a low threshold during standby to measure a high load during stirring, or false alarms caused by using a wide threshold during high load to measure the standby state.

[0100] It should be noted that, The choice of 3 as the safety factor is based on the "3-Sigma principle" in statistics. When equipment is in a healthy state, the random fluctuations of its key operating parameters usually follow or approximately follow a normal distribution. According to the probability density function characteristics of the normal distribution, the numerical distribution is within a range of the mean plus or minus three standard deviations (where...). The probability within the given range is approximately 99.73%. This means that... Setting it to 3 can theoretically cover 99.73% of normal operating condition fluctuation data, keeping the false alarm rate under normal conditions below 0.3%. Compared to value 2 (coverage of 95.45%, but with a high false alarm rate) or value 4 (although the false alarm rate is lower, it may lead to missed detection of early minor faults), value 3 achieves the best balance between fault detection rate and false alarm rate in industrial monitoring scenarios.

[0101] It is worth noting that if a non-operational mutation is identified, it indicates that the current data change exhibits low synchronization rate or discrete characteristics, which is highly likely caused by a fault (such as sensor failure or mechanical breakage) rather than normal operation. In this case, blindly updating the benchmark may absorb faulty data as normal standard, leading to a masking effect. Therefore, this embodiment adopts a zero-order hold strategy, that is, performing dynamic benchmark range locking and maintenance processing. The system forcibly ignores the current mutation characteristics and continues to use the previous time step ( The dynamic reference range determined at time ( ) is used as the current time ( ) The standard is based on the time frame. This strategy ensures that subsequent deviation calculations can be based on the normal standard before the mutation, calculating large deviation values ​​and thus accurately capturing anomalies. At the same time, to prevent system logic deadlock, the system has a built-in timeout reset mechanism. If the duration of the non-operating condition mutation exceeds the preset safety time limit (e.g., 30 seconds), the lock is forcibly released and the system switches to the preset safety mode reference range to deal with extreme situations such as long-term sensor failure.

[0102] In step S16, based on the calibration feature vector and the dynamic reference range, deviation calculation and trend direction analysis are performed to obtain a preliminary anomaly probability, including:

[0103] Based on the calibration feature vector and the dynamic reference range, a numerical distance is calculated to obtain the deviation amplitude value;

[0104] The slope of change is calculated for the calibration feature vector, and the direction of the continuous divergence trajectory is determined based on the positive or negative sign of the slope of change.

[0105] Based on the deviation amplitude value and the continuous divergence trajectory, the preliminary anomaly probability is obtained by quantitative calculation using a preset probability scoring model.

[0106] In one implementation, this embodiment calculates the numerical distance between the calibration feature vector and the dynamic reference range. Considering the multidimensional attributes of the feature vector (such as including multiple dimensions like current and vibration), this embodiment first performs Z-score standardization (or range normalization) on each dimension of the feature to give it a unified dimension. Then, this embodiment calculates the current feature point. The Mahalanobis distance to the boundary of the dynamic reference range. Specifically, for a single-dimensional feature, the deviation magnitude value. The calculation formula is:

[0107]

[0108] in, and The upper and lower bounds of the dynamic reference determined by S15, This represents the historical standard deviation under this operating condition. This formula not only quantifies the absolute value of the deviation but also eliminates the influence of volatility differences under different operating conditions by dividing by the standard deviation, thus achieving a standardized measure of the degree of deviation.

[0109] It should be noted that this embodiment performs slope calculation and direction identification processing on the calibration feature vector. Simple deviations may be instantaneous noise, while a continuously deteriorating trend is the definitive symptom of a fault. Therefore, this embodiment selects a backtracking time window (e.g., data from the past 5 seconds) and uses the least squares method for linear regression to calculate the slope of the feature value change. The sustained divergence trajectory is defined as a Boolean value or quantization factor. The judgment logic is as follows: if the eigenvalue exceeds the upper limit and the slope is... (Positive divergence), or eigenvalues ​​below the lower limit and slope If there is negative divergence, then it is determined that there is a continuous divergence trajectory.

[0110] In one implementation, this embodiment utilizes a pre-defined probability scoring model for quantitative calculation. The probability scoring model is constructed as a logistic regression model, whose mathematical expression is in the form of a sigmoid function:

[0111]

[0112] in, The output is the preliminary anomaly probability. and These are the weighting coefficients for the deviation magnitude and the slope of change, respectively. This is a bias term.

[0113] It is worth noting that the parameters of the probability scoring model ( The weights are determined through supervised learning training. and It not only plays a role in weighting importance, but also implicitly includes a scaling transformation function for normalizing characteristic dimensions, thereby solving the problem of dimensionless bias values. With dimensionless slope The problem of physical consistency when directly adding elements.

[0114] The training process includes collecting historical running datasets labeled "normal" and "faulty" and dividing them into training and validation sets; using deviation magnitude and slope as input features and fault labels (0 or 1) as targets; constructing a log-loss function based on the maximum likelihood estimation principle and using an iterative optimization algorithm to solve for the optimal parameters.

[0115] Specifically, the system first initializes the model parameters (weight coefficients and bias terms) to random values ​​close to zero. Then, in each iteration, the training samples are substituted into the current Sigmoid function to calculate the predicted probability, and the cross-entropy loss between the predicted probability and the true fault label is calculated. Next, the partial derivative of the loss function with respect to each model parameter (i.e., the gradient) is calculated using gradient descent, and the model parameters are updated in the opposite direction of the gradient according to the preset learning rate. This iterative update process continues until the stopping condition is met.

[0116] The stopping conditions include the change in the loss function value falling below a preset convergence threshold or the number of iterations reaching a preset maximum number of iterations. The preset convergence threshold is determined through offline loss curve analysis. Specifically, a decay curve of the loss function value as a function of iterations is plotted on the validation set. The value at which the gradient of the loss value descent becomes gradual and the relative rate of change is less than a preset small amount (e.g., one ten-thousandth) is selected as the convergence threshold, balancing computational accuracy and response speed. The preset maximum number of iterations is set based on a deadlock prevention mechanism. Specifically, it is the average number of iterations required for the model to converge in historical training, and a safety multiple of this average number of steps (e.g., 1.5 to 2 times) is set as the maximum number of iterations to prevent the program from entering an infinite loop in abnormal situations such as gradient vanishing or oscillations. The parameters determined at this point are the final optimized model parameters.

[0117] Furthermore, the specific values ​​of the preset convergence threshold and the maximum number of iterations are empirically set after performing multiple pre-training operations on a typical training set and observing the decline curve of the loss function. This aims to ensure that the parameters converge to a stable optimal solution while avoiding unnecessary computational overhead.

[0118] For cold-start equipment lacking historical fault data, the system employs a statistical migration strategy based on historical data from a group of similar equipment to determine initial parameters. Specifically, this embodiment pre-collects a large amount of historical operation and maintenance data from asphalt mixing plants of the same model, constructs a general fault evolution model, and calculates the statistical mean of the weights of each item in this general model as the initial model parameters. For example, based on the general model, statistical data is obtained... The value range is usually between [0.8, 1.2]. The value range is typically between [0.3, 0.6]. This parameter distribution reflects the objective physical law in fault judgment that "the deviation magnitude (i.e., the degree of deviation that has occurred) is the main factor, and the change slope (i.e., the deterioration trend) is the secondary factor." This ensures that cold-start equipment still has basic monitoring capabilities based on the common characteristics of the group when there is no training data. Subsequently, the above parameters are individually fine-tuned through online learning during operation.

[0119] For example, assume the historical average under the current operating conditions. Historical standard deviation The dynamic benchmark range calculated based on the 3-standard-deviation principle is: The measured value at the current moment is The value exceeds the upper limit. At this point, the calculated standardized deviation magnitude value... Meanwhile, linear regression shows that the data over the past few seconds has been... The slope continues to rise. and Substituting into the above probability model (assuming...) The initial anomaly probability can be calculated. If the value is still 115 at the next moment but the slope becomes 0 (no longer deteriorating), the probability output by the model will decrease significantly.

[0120] In step S17, based on the preliminary anomaly probability, historical trend weighting and anomaly integral calculations are performed to obtain the final health assessment result, including:

[0121] Based on the historical preliminary anomaly probability sequence within a preset time period, a weighted score of the historical trend is obtained by using a preset time decay factor for weighted calculation.

[0122] Based on the current preliminary anomaly probability and the historical trend weighted score, a weighted sum is calculated to obtain a fusion index, and the fusion index is integrated over a sliding window time to obtain a cumulative anomaly score.

[0123] The cumulative abnormal score is compared and the preset dynamic alarm threshold is used to determine the level, and the final health assessment result is obtained.

[0124] In one implementation, this embodiment uses a preset time decay factor to perform weighted calculations based on a historical preliminary anomaly probability sequence within a preset time period. Since equipment failure is typically a gradual, cumulative process, while transient interference often manifests as isolated pulses, recent data has significantly higher reference value than older data. This embodiment employs an exponential decay weighted algorithm based on a sliding window.

[0125] Specifically, this embodiment obtains the past Preliminary anomaly probability sequence at each time step This embodiment utilizes the preset time decay factor. (For example ), calculate the historical trend weighted score :

[0126]

[0127] This formula ensures that anomalies closer to the current time are given a higher weight. Specifically, This represents the number of time steps that historical data is lagging behind the current time (e.g., Represents the previous sampling time. (Represents the oldest moment within the time window).

[0128] It should be noted that the preset time decay factor It is determined through autocorrelation analysis of historical interference signals. The probability waveform of the acquisition equipment affected by transient interference (such as voltage fluctuations) during normal operation is obtained; the autocorrelation function decay rate of this waveform is calculated; and the autocorrelation coefficient is selected to decay to... Using the time constant corresponding to the time as a reference, the solution is obtained that allows the instantaneous disturbance to rapidly decay to negligible levels within a preset integration period. value.

[0129] In one implementation, this embodiment performs numerical fusion and time integration based on the current preliminary anomaly probability and the historical trend weighted score. This embodiment first constructs a fused anomaly intensity index. This indicator is the current instantaneous probability. Compared with historical trends Weighted sum (e.g.) This is used to balance instantaneous impacts and long-term trends. Subsequently, this embodiment operates within a fixed integration time window. The fusion index is integrated over time to calculate the cumulative anomaly integral value. :

[0130]

[0131] in The sampling interval is consistent with the data acquisition sampling period, ensuring the continuity of the integral calculation over time. This integration process characterizes the cumulative effect of anomalous energy over time. Only when the anomalous state persists and its intensity is high will the integral value increase significantly, thereby completely filtering out high-amplitude but short-duration sporadic glitches.

[0132] It is worth noting that this embodiment compares and determines the level of abnormality based on the accumulated abnormality score and a preset dynamic alarm threshold. The preset dynamic alarm threshold is not a fixed value, but is associated with the operating condition mode category determined in S15. This embodiment obtains the score threshold corresponding to the current operating condition through a lookup table. This threshold table is constructed based on the extreme value statistical distribution of historical health data. Specifically, the 95th percentile of the maximum score calculated within the historical healthy operating cycle is set as the yellow warning threshold, and the score level before the occurrence of historical known fault cases is set as the red threshold. If the accumulated abnormality score is lower than the yellow warning threshold, it is determined to be "healthy"; if it is between the yellow and red thresholds, it is determined to be "potential fault risk"; if it is higher than the red threshold, it is determined to be "certain fault". This graded assessment mechanism constitutes the final health assessment result, which directly guides subsequent maintenance decisions.

[0133] In summary, this invention achieves adaptive and accurate assessment of equipment health status in complex and ever-changing construction site environments by constructing a closed-loop monitoring system that includes environmental physical correction, multi-dimensional operating condition decoupling, and dynamic benchmark following. Specifically, it eliminates nonlinear interference from environmental factors on monitoring data using division-equivalent correction, effectively avoids false alarms caused by operating condition switching using feature synchronization rate analysis, and prevents the masking effect of fault data through a dynamic benchmark locking mechanism. This solves the problems of monitoring blind spots and frequent false alarms caused by existing technologies relying on static thresholds, providing strong technical support for the safe and efficient operation of asphalt mixing equipment.

[0134] Reference Figure 2 The second embodiment of the present invention provides a remote monitoring system for asphalt mixture mixing equipment based on the Internet of Things, comprising:

[0135] The data acquisition and preprocessing module is used to acquire the operating parameters of key components of the asphalt mixture mixing equipment and external environmental variables, and to perform time-domain statistical filtering on the operating parameters of the key components to obtain the initial operating sequence.

[0136] The smoothing and denoising module is used to perform adaptive window length smoothing and frequency domain noise suppression based on the initial running sequence to obtain a clean running sequence;

[0137] The feature extraction and correction module is used to extract time-varying features based on the clean operation sequence and perform environmental equivalence correction processing using the external environmental variables to obtain a calibration feature vector.

[0138] The operating condition identification module is used to perform multi-dimensional feature association analysis and pattern matching based on the calibration feature vector to determine the operating condition switching label;

[0139] The dynamic baseline determination module is used to switch tags according to the working conditions, retrieve corresponding health status data from a preset historical database or maintain the current baseline, and determine the dynamic baseline range.

[0140] The deviation analysis module is used to perform deviation calculation and trend direction analysis based on the calibration feature vector and the dynamic benchmark range to obtain a preliminary anomaly probability.

[0141] The health assessment module is used to calculate the final health assessment result by performing historical trend weighting and anomaly integral calculation based on the preliminary anomaly probability.

[0142] It should be noted that the remote monitoring system for asphalt mixture mixing equipment based on the Internet of Things provided in this embodiment of the invention is used to execute all the process steps of the remote monitoring method for asphalt mixture mixing equipment based on the Internet of Things in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0143] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a remote monitoring program for an asphalt mixture mixing plant based on the Internet of Things (IoT). When the processor executes the computer program, it implements the steps in the various embodiments of the remote monitoring method for asphalt mixture mixing plants based on the IoT described above, for example... Figure 1 Steps S11 to S17 are shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above system embodiments, for example... Figure 2 The module shown includes data acquisition and preprocessing, smoothing and denoising, feature extraction and correction, working condition identification, dynamic benchmark determination, deviation analysis, and health assessment.

[0144] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0145] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0146] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0147] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0148] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0149] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and 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. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0150] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A remote monitoring method for asphalt mixing equipment based on the Internet of Things, characterized in that, include: The operating parameters of key components of the asphalt mixture mixing equipment and external environmental variables are collected, and the operating parameters of the key components are subjected to time-domain statistical filtering to obtain the initial operating sequence. Based on the initial running sequence, adaptive window length smoothing and frequency domain noise suppression are performed to obtain a clean running sequence. Time-varying features are extracted from the cleanroom operation sequence, and environmental equivalence correction is performed using the external environmental variables to obtain a calibration feature vector. Based on the calibration feature vector, perform multidimensional feature correlation analysis and pattern matching to determine the operating condition switching label; Based on the operating condition switching label, retrieve the corresponding health status data from the preset historical database or maintain the current benchmark to determine the dynamic benchmark range; Based on the calibration feature vector and the dynamic benchmark range, deviation calculation and trend direction analysis are performed to obtain the preliminary anomaly probability; Based on the preliminary anomaly probability, historical trend weighting and anomaly integral calculations are performed to obtain the final health assessment result; The step of determining the operating condition switching label by performing multidimensional feature correlation analysis and pattern matching based on the calibration feature vector includes: Calculate the synchronization rate of changes between features of different dimensions in the calibration feature vector to obtain the feature change synchronization rate; The synchronization rate of the feature changes is compared with a preset coordination threshold. If the feature change synchronization rate is greater than or equal to the preset coordination threshold, the data in the corresponding time period is marked as a potential switching segment, the similarity between the potential switching segment and each standard template in the preset historical working condition mode library is calculated, and the working condition name corresponding to the template with the highest similarity is determined as the working condition switching label. If the synchronization rate of the feature change is less than the preset coordination threshold, then a working condition switching tag is generated that identifies it as a non-working condition abrupt change; The step of retrieving corresponding health status data from a preset historical database or maintaining the current baseline based on the working condition switching tag, and determining the dynamic baseline range, includes: Based on the operating condition switching label, determine the category to which the device belongs, and obtain the current operating condition mode category of the device; If the operating condition mode category is a known operating condition mode, then the corresponding historical health feature samples are retrieved and extracted from the preset historical database, and the distribution statistical analysis of the historical health feature samples is performed to obtain the dynamic benchmark range. If the operating condition mode category is a non-operating condition change, then the dynamic reference range determined at the previous moment is obtained and assigned as the dynamic reference range at the current moment.

2. The remote monitoring method for asphalt mixing equipment based on the Internet of Things according to claim 1, characterized in that, The system collects the operating parameters of key components of the asphalt mixture mixing equipment and external environmental variables, and performs time-domain statistical filtering on the operating parameters of the key components to obtain an initial operating sequence, including: Real-time acquisition of the operating parameters of the key components and the external environmental variables; Based on the pre-stored historical key component operating parameters, calculate the statistical distribution characteristics, and determine the current dynamic statistical threshold based on the statistical distribution characteristics; The initial operating sequence is obtained by removing the numerical points in the operating parameters of the key components that exceed the dynamic statistical threshold.

3. The remote monitoring method for asphalt mixing equipment based on the Internet of Things according to claim 1, characterized in that, The step of performing adaptive window length smoothing and frequency domain noise suppression based on the initial running sequence to obtain a clean running sequence includes: Based on the fluctuation amplitude of the initial running sequence, adaptive segmentation and local weighted mean calculation are performed on the initial running sequence to obtain a time-domain smoothed sequence; The time-domain smoothed sequence is converted to the frequency domain by a fast Fourier transform, and signal components in the preset interference frequency band are identified and filtered out to obtain frequency domain cleaned data. The frequency domain purification data is reconstructed by inverse Fourier transform to obtain the clean operation sequence.

4. The remote monitoring method for asphalt mixing equipment based on the Internet of Things according to claim 1, characterized in that, The step of extracting time-varying features based on the cleanroom operation sequence and performing environmental equivalence correction using the external environmental variables to obtain a calibration feature vector includes: The cleanroom operation sequence is subjected to time-series statistical analysis to calculate the local fluctuation amplitude and the rate of change of operating parameters within a preset time window, and to obtain initial feature data; Based on the external environmental variables, the corresponding gain value is retrieved from the preset environmental impact factor curve to obtain the environmental compensation gain under the current environment. The initial feature data is corrected by division using the environmental compensation gain to obtain the calibration feature vector.

5. The remote monitoring method for asphalt mixing equipment based on the Internet of Things according to claim 1, characterized in that, The step of calculating the deviation and analyzing the trend direction based on the calibration feature vector and the dynamic reference range to obtain the preliminary anomaly probability includes: Based on the calibration feature vector and the dynamic reference range, a numerical distance is calculated to obtain the deviation amplitude value; The slope of change is calculated for the calibration feature vector, and the direction of the continuous divergence trajectory is determined based on the positive or negative sign of the slope of change. Based on the deviation amplitude value and the continuous divergence trajectory, the preliminary anomaly probability is obtained by quantitative calculation using a preset probability scoring model.

6. The remote monitoring method for asphalt mixing equipment based on the Internet of Things according to claim 1, characterized in that, The step of calculating the final health assessment result based on the preliminary anomaly probability by weighting historical trends and calculating the anomaly integral includes: Based on the historical preliminary anomaly probability sequence within a preset time period, a weighted score of the historical trend is obtained by using a preset time decay factor for weighted calculation. Based on the current preliminary anomaly probability and the historical trend weighted score, a weighted sum is calculated to obtain a fusion index, and the fusion index is integrated over a sliding window time to obtain a cumulative anomaly score. The cumulative abnormal score is compared and the preset dynamic alarm threshold is used to determine the level, and the final health assessment result is obtained.

7. A remote monitoring system for an asphalt mixture mixing plant based on the Internet of Things (IoT), used to implement the remote monitoring method for an asphalt mixture mixing plant based on the IoT as described in any one of claims 1-6, characterized in that, include: The data acquisition and preprocessing module is used to acquire the operating parameters of key components of the asphalt mixture mixing equipment and external environmental variables, and to perform time-domain statistical filtering on the operating parameters of the key components to obtain the initial operating sequence. The smoothing and denoising module is used to perform adaptive window length smoothing and frequency domain noise suppression based on the initial running sequence to obtain a clean running sequence; The feature extraction and correction module is used to extract time-varying features based on the clean operation sequence and perform environmental equivalence correction processing using the external environmental variables to obtain a calibration feature vector. The operating condition identification module is used to perform multi-dimensional feature association analysis and pattern matching based on the calibration feature vector to determine the operating condition switching label; The dynamic baseline determination module is used to switch tags according to the working conditions, retrieve corresponding health status data from a preset historical database or maintain the current baseline, and determine the dynamic baseline range. The deviation analysis module is used to perform deviation calculation and trend direction analysis based on the calibration feature vector and the dynamic benchmark range to obtain a preliminary anomaly probability. The health assessment module is used to calculate the final health assessment result by performing historical trend weighting and anomaly integral calculation based on the preliminary anomaly probability.

Images