Mes-oriented industrial equipment data acquisition edge computing processing method

Abstract

This invention discloses an edge computing processing method for industrial equipment data acquisition in MES (Manufacturing Execution System), relating to the technical field of industrial equipment data acquisition and edge computing processing. It includes: applying to an edge computing gateway, periodically capturing multi-source device data and extracting time-sensitive features, recording the numerical change trajectory of each feature dimension within the current time window, and generating a feature trajectory descriptor composed of the change rate and fluctuation amplitude; inputting the feature trajectory descriptor into a state machine matching matrix, dynamically comparing trajectory shape similarity and determining the device operating stage based on device state transition conditions; obtaining the MES business event code and compression strategy identifier based on the operating stage, encapsulating the business event code and key timestamps into an edge-side output message, and compressing the multi-source device data according to the compression strategy.

Description

Technical Field

[0001] This invention relates to the field of industrial equipment data acquisition and edge computing processing technology, specifically to an industrial equipment data acquisition and edge computing processing method for MES. Background Technology

[0002] Industrial Manufacturing Execution Systems (MES) rely on the real-time aggregation and processing of equipment-level data, with edge computing gateways playing a crucial role in data acquisition and uploading. In existing technologies, edge computing gateways typically collect multi-source equipment data at fixed intervals, uploading it directly or distributing it to the MES system after simple threshold filtering. This processing model struggles to identify dynamic changes in equipment operation phases from high-frequency, multi-source data, resulting in the MES system receiving only discrete data points and lacking equipment status information with business semantics. Furthermore, the uniform compression ratio is commonly used when uploading equipment data, failing to differentiate the business value of data at different operational stages, leading to high-value and low-value data consuming the same transmission resources. Regarding feature extraction from multi-source equipment data, existing technologies often employ single-dimensional threshold detection or simple time-frequency domain statistical calculations, lacking descriptions of the collaborative changes between different feature dimensions, making it difficult to accurately capture the transition process from stable operation to abnormal states. Furthermore, existing solutions for triggering MES business events based on device data largely rely on manually preset fixed condition codes. They cannot match the device's operational stage based on the shape similarity between the device's real-time behavior trajectory and standard state templates, nor can they combine the previous stage with state machine compliance verification during state transition judgment. One problem this invention aims to solve is how to construct a trajectory descriptor reflecting the linkage between different feature dimensions from multiple time-sensitive features in multi-source heterogeneous device data scenarios, and how to use this descriptor to perform flexible trajectory shape matching with multiple preset standard trajectory templates to accurately identify the current operational stage of the device. Another problem this invention aims to solve is how to adaptively generate MES business event codes based on the operational stage and its relationship with the previous state after determining the device's operational stage, and how to differentiate compression strategies based on multi-dimensional information such as the operational stage and vibration characteristics, thereby reducing data upload pressure while preserving high-value data details. Summary of the Invention

[0003] This invention provides an edge computing processing method for industrial equipment data acquisition in MES (Manufacturing Execution System). It extracts time-sensitive features from multi-source equipment data and constructs multi-dimensional collaboratively changing feature trajectory descriptors. A state machine matching matrix is ​​used to dynamically compare the feature trajectory descriptors with multiple standard trajectory templates to determine the trajectory shape. Combined with state transition conditions, the device's operating stage is determined. Furthermore, based on the device's operating stage and associated information, a table is looked up to obtain MES business event codes and compression strategy identifiers. While outputting business events to the MES system, differentiated compression processing is implemented for data from different operating stages, achieving deep collaboration between data acquisition, state identification, and compressed transmission.

[0004] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides an edge computing processing method for industrial equipment data acquisition in MES, applied to an edge computing gateway, the method comprising: The edge computing gateway periodically captures multi-source device data within the current time window. Based on the device type and parameter encoding of the data source, it extracts multiple time-sensitive features for each data frame. Based on these multiple time-sensitive features, it records the numerical change trajectory of each feature dimension within the current time window to generate a set of feature trajectory descriptors composed of change rate and fluctuation amplitude. The feature trajectory descriptors are input into a preset state machine matching matrix. The state machine matching matrix stores multiple standard trajectory templates and the corresponding device state transition conditions for each standard trajectory template. The state machine matching matrix dynamically compares the trajectory shape similarity between the current feature trajectory descriptor and the multiple standard trajectory templates. Based on the standard trajectory template with the highest trajectory shape similarity and that satisfies the device state transition conditions, it determines the device operation stage corresponding to the current time window. Based on the device operation stage, it obtains the corresponding MES service event code and compression strategy identifier by looking up a table in the state machine matching matrix. The MES service event code and the key timestamps within the current time window are encapsulated into an edge-side output message, and the multi-source device data within the current time window is compressed according to the compression strategy identifier.

[0005] By extracting time-sensitive features and generating feature trajectory descriptors that reflect cross-dimensional collaborative changes, the transient and transitional characteristics of equipment operation can be condensed, eliminating the impact of differences in the dimensions and numerical ranges of different equipment. By using a state machine matching matrix to dynamically match the trajectory descriptors with standard state transition templates and combining this with condition verification of the previous equipment operation stage, the current operation stage of the equipment can be accurately identified under limited sample conditions, suppressing erroneous state switching caused by local disturbances. Then, based on the identified equipment operation stage, a lookup table is used to obtain the required business event codes and appropriate compression strategies for the MES side. This ensures timely reporting of critical events in the manufacturing execution system and, by performing adaptive compression processing on multi-source data within the current window that matches the equipment state, significantly reduces the amount of data transmitted from the edge side to the upper-layer system while retaining effective feature information, alleviating network load and saving storage costs.

[0006] As a preferred embodiment of the present invention, the process of extracting multiple time-sensitive features from each data frame includes: parsing each data frame to obtain device identifier, data type identifier, and original numerical sequence; calculating local extreme points within a sliding window for the original numerical sequence, using the timestamps and amplitudes of the local extreme points as original feature points, performing spline interpolation reconstruction on multiple original feature points to obtain an equally spaced sampling sequence; calculating the absolute value of the slope between adjacent sampling points, the difference between adjacent absolute values ​​of the slope, and the higher-order difference value of the sampled values ​​in the equally spaced sampling sequence to form multiple time-sensitive features for each data frame. Preferably, the spline interpolation reconstruction uses cubic spline interpolation, and the number of sampling points in the equally spaced sampling sequence is determined by the product of the current time window length and a preset sampling frequency, wherein the preset sampling frequency is dynamically adjusted according to the data source device type. In this way, a regular feature sequence can be recovered from non-uniform original extreme points, ensuring the temporal consistency of subsequent change trajectory descriptions and providing differentiated feature granularity for devices with different response characteristics.

[0007] As a further preferred technical solution of the present invention, the process of recording the numerical change trajectory of each feature dimension within the current time window includes: taking the start time of the current time window as the origin, arranging multiple time-sensitive features of each feature dimension according to the collection timestamp to generate multiple sub-dimensional trajectory curves; performing amplitude normalization on the multiple sub-dimensional trajectory curves to a unified amplitude range, and using the median of the unified amplitude range as a benchmark, marking the upward and downward crossing states of trajectory segments of each sub-dimensional trajectory curve in multiple sub-time periods; counting the number of trajectory segments with the same upward or downward crossing state in the same sub-time period among the multiple sub-dimensional trajectory curves, using the statistical result as the degree of coordinated change of the sub-time period, and concatenating the degrees of coordinated change of each sub-time period in chronological order to form the feature trajectory descriptor. More preferably, when performing amplitude normalization on the multiple sub-dimensional trajectory curves to a unified amplitude range, linear normalization is performed with the global maximum and global minimum values ​​of each sub-dimensional trajectory curve as boundaries, and the unified amplitude range is [-1, 1]. This approach abstracts the changes in multidimensional features into a cross-channel cooperative directional description, forming a compact and discriminative expression of the waveform shape, so that subsequent trajectory matching is not affected by the scaling of the absolute amplitude of the signal.

[0008] As another preferred technical solution of the present invention, the process of dynamically comparing the trajectory shape similarity between the current feature trajectory descriptor and the multiple standard trajectory templates using the state machine matching matrix specifically includes: reading the standard cooperative change sequence and standard transition trigger threshold corresponding to each standard trajectory template from the state machine matching matrix; performing dynamic time warping on the cooperative change degree sequence in the feature trajectory descriptor and the standard cooperative change sequence, and calculating the warped path length as the initial similarity; determining whether the initial similarity is lower than the standard transition trigger threshold, and if it is lower than the standard transition trigger threshold, selecting the standard trajectory template as a candidate template, otherwise discarding the standard trajectory template. Furthermore, when determining the equipment operation stage corresponding to the current time window based on the standard trajectory template with the highest trajectory shape similarity and satisfying the equipment state transition conditions, the edge computing gateway selects the standard trajectory template with the lowest initial similarity from multiple candidate templates as the matching trajectory template, reads the equipment state transition conditions corresponding to the matching trajectory template, obtains the previous equipment operation stage corresponding to the previous time window before the current time window, compares the previous equipment operation stage with the source state set in the equipment state transition conditions, if the previous equipment operation stage belongs to the source state set, then the target state in the equipment state transition conditions is determined as the equipment operation stage corresponding to the current time window; if the previous equipment operation stage does not belong to the source state set, then the remaining candidate templates are traversed sequentially in ascending order of initial similarity until a candidate template that satisfies the source state set is found, and the target state corresponding to the candidate template is determined as the equipment operation stage corresponding to the current time window. This design utilizes dynamic time warping to tolerate nonlinear scaling on the time axis, while using state transition logic to impose secondary constraints on the matching results, effectively preventing state jumps that do not conform to process logic, ensuring the temporal continuity of stage identification and the interpretability of the industrial site. If the edge computing gateway fails to find a candidate template containing the previous device operation stage after traversing all candidate templates, it will maintain the device operation stage corresponding to the current time window as the previous device operation stage and generate a state transition anomaly alarm event code to expose potential process anomalies in a timely manner.

[0009] As another preferred technical solution of the present invention, the process of obtaining the corresponding MES service event code and compression strategy identifier by looking up the state machine matching matrix according to the equipment operation stage includes: using the equipment operation stage as the row index and the average load rate interval of the equipment within the current time window as the column index, performing a two-dimensional lookup in the mapping table of the state machine matching matrix to obtain the MES service event code that matches both the row index and the column index; simultaneously, using the equipment operation stage as the first index and whether the maximum amplitude of the vibration feature among the multiple time-sensitive features exceeds the preset amplitude boundary as the second index, performing a two-dimensional lookup in the compression mapping table of the state machine matching matrix to obtain the compression strategy identifier, which includes the compression ratio and the compression start position. Thus, the equipment operation stage is mapped to a combination of multi-dimensional operating conditions such as load status and vibration intensity, and differentiated service reporting and compression actions are mapped to achieve adaptive adjustment of the reported content and compression intensity based on operating conditions.

[0010] In the above scheme, the process of compressing multi-source device data within the current time window according to the compression strategy identifier includes: parsing the compression ratio in the compression strategy identifier; sampling data frames of multi-source device data within the current time window at equal intervals according to the compression ratio to obtain a first compressed data frame sequence; parsing the compression start position in the compression strategy identifier; discarding data frames located before the compression start position from the first compressed data frame sequence to obtain a second compressed data frame sequence; performing an XOR differential operation on each frame of data in the second compressed data frame sequence with the previous frame of data to obtain a differential coding sequence, and using the differential coding sequence as the compressed transmission data. Through the joint compression of sampling and differential coding, redundancy is further eliminated while preserving the trend of the evolving operating conditions. Furthermore, the setting of the compression start position ensures that only data segments sensitive to the current state are retained, thereby prioritizing the data fidelity required by MES under limited bandwidth.

[0011] The technical effects and advantages provided by the present invention in the above technical solution are as follows: The time-sensitive features of multiple feature dimensions within the current time window are converted into feature trajectory descriptors composed of a sequence of coordinated change degrees. This descriptor records the number of trajectory segments with consistent amplitude crossings (either upward or downward) across multiple sub-time periods for different feature dimensions, representing the linkage between multi-dimensional features as a one-dimensional coordinated change sequence. Multiple standard trajectory templates and their standard coordinated change sequences are pre-set in the state machine matching matrix. After receiving the feature trajectory descriptors, dynamic time warping is performed on the coordinated change degree sequence and the standard coordinated change sequence, calculating the warped path length as the trajectory shape similarity. Dynamic time warping can absorb non-linear offsets on the time axis, avoiding time alignment errors caused by fluctuations in equipment operating rhythm that interfere with the matching results. Based on the trajectory shape similarity, the matching process further filters candidate templates that meet the standard transition trigger thresholds and verifies whether the previous equipment operating stage belongs to the source state set in the candidate template's equipment state transition conditions, thereby ensuring that state transitions logically conform to the equipment operating rules. This matching method does not rely on instantaneous triggering of a single threshold. Instead, it integrates elastic trajectory shape matching and state transition compliance judgment to reconstruct the migration of equipment operation stages from multi-dimensional data within a continuous time window, improving the accuracy of stage determination and anti-interference capability. Based on the determined equipment operation stage, a two-dimensional lookup is performed through the mapping table in the state machine matching matrix, using the equipment operation stage and the average load rate range of the equipment within the current time window as indexes to obtain the corresponding MES business event code. This method combines equipment operation stages and load rates into fine-grained business semantics, allowing different stages combined with different load levels to correspond to different events, eliminating the need for manually writing complex event triggering rules. Simultaneously, using the equipment operation stage and whether the maximum amplitude of the vibration characteristics exceeds the preset amplitude boundary as indexes, a compression strategy identifier is obtained by looking up the compression mapping table. This identifier includes the compression ratio and compression start position. The compression strategy is selected based on the real-time requirements of the operation stage and vibration information. During periods of abnormally high vibration, a low compression ratio or compression starting from a more recent data frame can be automatically used to retain more original details; during periods of stable operation and normal vibration amplitude, a high compression ratio is used, and earlier data frames are appropriately discarded to reduce the amount of transmitted data. By performing equal-interval sampling, discarding data frames before the start position, and performing XOR differential operations according to the compression strategy, differentiated compression encoding is achieved frame by frame. This does not increase the processing load of the edge computing gateway, and the uploaded data achieves better bandwidth utilization efficiency under the uneven distribution of business value. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0013] Figure 1 This is a flowchart of an industrial equipment data acquisition edge computing processing method for MES (Manufacturing Execution System). Figure 2 This is a flowchart for determining the equipment operation stage based on trajectory similarity and state transition conditions; Figure 3 This is a schematic diagram of the original numerical sequence and its local extreme points within the sliding window; Figure 4 This is a schematic diagram of the normalized trajectory curves for each feature dimension; Figure 5 It is a comparison curve of the dynamic time warping of the characteristic trajectory descriptor and the standard co-variation sequence; Figure 6 This is a comparison chart of the degree of co-variation between the feature trajectory descriptor and the standard trajectory template; Figure 7 This is a schematic diagram of equally spaced sampling points for data from multiple sources. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0015] See Figure 1 This invention provides an edge computing processing method for industrial equipment data acquisition in MES (Manufacturing Execution System), applied to an edge computing gateway. The method involves the edge computing gateway periodically capturing multi-source device data within the current time window, extracting multiple time-sensitive features from each data frame based on the device type and parameter encoding. Based on these time-sensitive features, the numerical change trajectory of each feature dimension within the current time window is recorded, generating a set of feature trajectory descriptors composed of change rate and fluctuation amplitude. The feature trajectory descriptors are input into a preset state machine matching matrix, which stores multiple standard trajectory templates and the corresponding device state transition conditions for each template. The state machine matching matrix dynamically compares the trajectory shape similarity between the current feature trajectory descriptor and the multiple standard trajectory templates. Based on the standard trajectory template with the highest trajectory shape similarity and that satisfies the device state transition conditions, the device operating stage corresponding to the current time window is determined. Based on the equipment operation phase, the corresponding MES service event code and compression strategy identifier are obtained by looking up the state machine matching matrix. The MES service event code and the key timestamps in the current time window are encapsulated into edge-side output messages, and the multi-source device data in the current time window are compressed according to the compression strategy identifier. Example

[0016] In practical implementation, when the edge computing gateway extracts multiple time-sensitive features from each data frame, it parses each received data frame to obtain the device identifier, data type identifier, and original numerical sequence. Based on a pre-defined communication protocol corresponding to the data source device type, the edge computing gateway reads the device identifier and data type identifier fields from a fixed offset in the frame header of the data frame, and extracts the original numerical sequence from the payload segment of the data frame. The original numerical sequence consists of multiple sampled values ​​arranged in chronological order of acquisition time.

[0017] For the original numerical sequence, the edge computing gateway calculates local extrema within a sliding window. The edge computing gateway sets a sliding window containing a preset number of sampling points; the window length is determined based on the signal variation frequency corresponding to the data source device type. The edge computing gateway slides the sliding window from the start to the end of the original numerical sequence, sampling points one by one. After each slide, it calculates the maximum and minimum points within the current sliding window, extracting the timestamp and amplitude of the maximum or minimum point in the original numerical sequence as a single original feature point. For sampling points whose edges cannot form a complete window, the edge computing gateway does not calculate their local extrema. All extracted original feature points constitute the original feature point set.

[0018] The edge computing gateway performs spline interpolation reconstruction on multiple original feature points to obtain an equally spaced sampling sequence. During the spline interpolation reconstruction process, the edge computing gateway employs a cubic spline interpolation algorithm, using the timestamps of the original feature points as independent variable nodes and the amplitudes of the original feature points as function value nodes, constructing a piecewise polynomial function with three nodes, ensuring the continuity of function values, first derivative values, and second derivative values ​​at adjacent polynomial functions at the nodes. Based on the continuous function obtained from the cubic spline interpolation, the edge computing gateway resamples at equal time intervals within the time range from the start to the end of the current time window, generating an equally spaced sampling sequence. The number of sampling points in the equally spaced sampling sequence is determined by the product of the current time window length and the preset sampling frequency. The preset sampling frequency is dynamically adjusted according to the data source device type. The edge computing gateway maintains a mapping table between device type and preset sampling frequency. When the data source device type is a vibration sensor, the preset sampling frequency is mapped to the first frequency value; when the data source device type is a temperature sensor, the preset sampling frequency is mapped to the second frequency value; and when the data source device type is a pressure sensor, the preset sampling frequency is mapped to the third frequency value. The edge computing gateway reads the mapping table to obtain the corresponding preset sampling frequency and calculates the number of sampling points. ,in Indicates the duration of the current time window. Indicates the preset sampling frequency. This represents the number of sampling points in an equally spaced sampling sequence.

[0019] After obtaining the equally spaced sampling sequence, the edge computing gateway calculates the absolute value of the slope between adjacent sampling points in the equally spaced sampling sequence. Let the sampling points in the equally spaced sampling sequence be arranged in chronological order as follows: The time interval between adjacent sampling points is Then the first The absolute value of the slope between adjacent sampling points for ,in from to , In the equally spaced sampling sequence, the first... The value of each sampling point This indicates the preset resampling time interval.

[0020] The edge computing gateway calculates the difference between the absolute values ​​of adjacent slopes. For the resulting sequence of absolute slope values... The edge computing gateway calculates the absolute value of the difference between two adjacent absolute values ​​of slope. ,in from to , Indicates the first The absolute value of the slope between adjacent sampling points Indicates the first The absolute value of the slope between adjacent sampling points.

[0021] The edge computing gateway calculates the higher-order differences of the sampled values ​​in an equally spaced sampling sequence. Taking the second-order difference as an example, the calculation formula for the edge computing gateway is:

[0022] in, In the equally spaced sampling sequence, the first... The value of each sampling point In the equally spaced sampling sequence, the first... The value of each sampling point In the equally spaced sampling sequence, the first... The value of each sampling point Indicates the first The second-order difference value corresponding to each sampling point The range of values ​​is to .

[0023] The edge computing gateway will calculate the sequence of absolute slope values. The sequence of differences between the absolute values ​​of adjacent slopes and the higher-order difference sequence of the sampled values These together constitute multiple time-sensitive features for each data frame. For each data frame, the edge computing gateway parses and calculates according to the above process, outputting the corresponding time-sensitive feature set. When the data source device type changes, the edge computing gateway readjusts the preset sampling frequency and sliding window length according to the mapping table to adapt to the signal characteristics of different devices.

[0024] See Figure 3 The figure illustrates the process by which the edge computing gateway in Example 1 extracts local extrema through a sliding window and constructs an equally spaced sampling sequence from the received raw numerical sequence. The horizontal axis represents the time offset in seconds, indicating the time shift of the sampling point within the current time window relative to the start time of the time window; the vertical axis represents the amplitude in meters per second, indicating the signal amplitude of the corresponding sampling point. The gray dashed line in the figure represents the raw numerical sequence, reflecting the original signal waveform composed of multiple sampled values ​​obtained by the edge computing gateway from the data frame payload segment in chronological order of acquisition. The black solid line represents the equally spaced sampling sequence obtained by reconstructing and uniformly resampling using a cubic spline interpolation algorithm; the curve is continuous and smooth, effectively reflecting the signal's changing trend. Red triangles in the figure indicate local maxima calculated within the sliding window, and blue inverted triangles indicate local minima. As described in Example 1, the sliding window, with a preset number of sampling points as its length, slides point by point across the raw numerical sequence to calculate local extrema, extracting the timestamps and amplitudes of the maxima and minima, thus forming the original feature point set. Example

[0025] In practical implementation, when the edge computing gateway records the numerical change trajectory of each feature dimension within the current time window, it uses the start time of the current time window as the origin and arranges multiple time-sensitive features of each feature dimension according to the acquisition timestamp, generating multiple sub-dimensional trajectory curves. Each feature dimension corresponds to a physical parameter type, which includes the absolute slope value dimension, the difference between adjacent absolute slope values ​​dimension, and the higher-order difference value dimension of the sampled values. For all time-sensitive features of a dimension, the edge computing gateway reads the time information corresponding to each time-sensitive feature from the acquisition timestamp of the data frame, representing each time-sensitive feature as a two-dimensional coordinate point. The horizontal coordinate of the two-dimensional coordinate point is the time offset between the acquisition timestamp corresponding to the time-sensitive feature and the start time of the current time window, and the vertical coordinate is the value of the time-sensitive feature. The edge computing gateway connects all two-dimensional coordinate points in ascending order of time offset to form a sub-dimensional trajectory curve. The above operation is performed for all feature dimensions to generate the corresponding number of sub-dimensional trajectory curves.

[0026] When performing amplitude normalization to a unified amplitude range for multiple multi-dimensional trajectory curves, the edge computing gateway performs linear normalization using the global maximum and global minimum values ​​of each multi-dimensional trajectory curve as boundaries. For a single multi-dimensional trajectory curve, the edge computing gateway iterates through all ordinate values ​​of the entire multi-dimensional trajectory curve and extracts the global maximum value. and global minimum And according to the following linear mapping formula, the original values ​​of each ordinate on the multidimensional trajectory curve are... Mapped to normalized values This maps the multidimensional trajectory curves to a uniform amplitude range. :

[0027] in, This represents the global maximum value of the multidimensional trajectory curve. This represents the global minimum value of the multidimensional trajectory curve. This indicates the first dimension of the trajectory curve. The original ordinate values ​​of each sampling point This indicates the first dimension of the trajectory curve. The normalized ordinate values ​​of each sampling point The range of values ​​is The total number of sampling points on this multi-dimensional trajectory curve. The edge computing gateway performs a normalization operation on each multi-dimensional trajectory curve according to the above formula, so that the range of the ordinate values ​​of all multi-dimensional trajectory curves is mapped to... Within the range.

[0028] Using the median of a uniform amplitude interval as a benchmark, the upward and downward crossing states of each sub-dimensional trajectory curve in multiple sub-time periods are marked. The median of the uniform amplitude interval is the numerical value. The edge computing gateway divides the current time window into multiple sub-time periods of a preset length. For a normalized, multi-dimensional trajectory curve segment within a sub-time period, the edge computing gateway compares each numerical point of the trajectory segment with the median. Compare the data; if the value in the trajectory segment is greater than the median... If the number of numerical points exceeds a preset threshold proportion of the total number of numerical points in a trajectory segment, then the trajectory segment of that sub-dimensional trajectory curve in the sub-time period is marked as "upward crossing"; if the number of numerical points in the trajectory segment is less than the median... If the number of numerical points exceeds a preset percentage threshold for the total number of numerical points in a trajectory segment, then the trajectory segment of that sub-dimensional trajectory curve in the sub-time period is marked as an underpass state; if neither of the above two conditions is met, then the trajectory segment of that sub-dimensional trajectory curve in the sub-time period is marked as a level state. The preset percentage threshold is pre-configured as a fixed percentage value in the edge computing gateway.

[0029] The edge computing gateway counts the number of trajectory segments with consistent upward or downward states within the same sub-time period across multiple multi-dimensional trajectory curves, and uses this result as the degree of co-change for that sub-time period. For a sub-time period, the edge computing gateway aggregates the marked states of multiple multi-dimensional trajectory curves into a marked state set. It then determines the absolute value of the difference between the number of trajectory segments marked as upward and downward in the marked state set. The ratio obtained by subtracting the absolute value of the difference from the larger of these two values ​​is taken as the degree of co-change for that sub-time period. A higher degree of co-change indicates more consistency in the direction and magnitude of change of the multi-dimensional trajectory curves within that sub-time period. The edge computing gateway concatenates the degrees of co-change for each sub-time period in chronological order, forming a one-dimensional sequence composed of the values ​​of the degrees of co-change for each sub-time period. This one-dimensional sequence is the feature trajectory descriptor.

[0030] See Figure 4 The figure shows the normalized amplitude trends of multiple time-sensitive features corresponding to multiple multi-dimensional trajectory curves within the current time window. The horizontal axis represents the time offset in seconds, ranging from 0 to 0.6 seconds, reflecting the offset of the sampling timestamp at each moment within the time window relative to the start time of the time window. The vertical axis represents the normalized amplitude, uniformly mapped to... An interval describes the magnitude of numerical variation in each feature dimension.

[0031] In the figure, the orange solid line represents the normalized trajectory of the absolute slope dimension, the green dashed line represents the normalized trajectory of the difference dimension between adjacent absolute slope values, the purple dotted line represents the normalized trajectory of the higher-order difference dimension of the sampled values, and the gray dashed line is the median baseline, which is constant at 0. As described in Example 2, the absolute slope dimension, the difference dimension between adjacent slopes, and the higher-order difference dimension correspond to feature dimensions of different physical parameter types, and the trajectory curve of each dimension is formed by connecting two-dimensional coordinate points in chronological order.

[0032] As shown in the figure, the three trajectory curves exhibit frequent numerical fluctuations with evenly distributed amplitudes within the normalized amplitude range, all oscillating around the median of 0, reflecting the dynamic changes in the feature dimensions within this time window. The absolute slope dimension curve displays multiple distinct peaks and troughs, indicating the instantaneous rate of signal change; the adjacent slope difference dimension curve is relatively flat but has many low-to-medium amplitude fluctuations, reflecting subtle differences between the rates of change; the higher-order difference dimension curve of the sampled values ​​shows larger fluctuation amplitudes, indicating complex changes in signal acceleration or curvature. Example

[0033] In practical implementation, when the state machine matching matrix dynamically compares the trajectory shape similarity between the current feature trajectory descriptor and multiple standard trajectory templates, the edge computing gateway reads the standard cooperative change sequence and standard transition trigger threshold corresponding to each standard trajectory template from the state machine matching matrix. The state machine matching matrix is ​​organized by entries in the edge computing gateway's memory. Each entry corresponds to a standard trajectory template, and the fields of each entry include a template identifier, a standard cooperative change sequence, device state transition conditions, and a standard transition trigger threshold. The standard cooperative change sequence is a one-dimensional numerical sequence with a length equal to the cooperative change degree sequence length in the feature trajectory descriptor. Each value in the sequence represents the cooperative change degree of the standard trajectory within the corresponding sub-time period. The standard transition trigger threshold is a preset floating-point value, the magnitude of which is determined by the state machine matching matrix during initialization based on the discriminative power of the standard trajectory.

[0034] The edge computing gateway performs dynamic time warping on the covariance degree sequence in the feature trajectory descriptor and the standard covariance degree sequence, calculating the warped path length as the initial similarity. The covariance degree sequence in the feature trajectory descriptor is represented as a time series. ,in This indicates the length of the co-variance sequence, i.e., the total number of sub-time periods. Indicates the first The degree of coordinated change in each sub-time period. The standard coordinated change sequence read from the state machine matching matrix is ​​represented as a time series. ,in This represents the length of the standard cooperative variation sequence, and and equal, Indicating the first standard cooperative change sequence Standard co-variance of individual time periods.

[0035] The edge computing gateway builds a size of OK Cumulative distance matrix of columns , cumulative distance matrix The Line 1 Column elements Set as , will the The sum of the remaining elements in the row and column 1 All elements in the remaining rows of the column are set to infinity. The edge computing gateway iterates through the rows. from to Column-wise traversal from to Calculate the co-variance sequence The Middle element Co-variation sequence with standard The Middle element The distance between them is calculated using the absolute value of the Euclidean distance, i.e. The edge computing gateway calculates the cumulative distance matrix according to the following recursive formula. The Middle Line number The possible values ​​for column elements:

[0036] in, Represents the th element in the cumulative distance matrix Line number The element values ​​of the column, Represents the th element in the cumulative distance matrix Line number The element values ​​of the column, Represents the th element in the cumulative distance matrix Line number The element values ​​of the column, Represents the th element in the cumulative distance matrix Line number The element values ​​of the column, This represents a function that takes the minimum of three values. After the edge computing gateway completes all recursive calculations, it will accumulate the distance matrix. The Middle Line number Column element values The normalized path length is the initial similarity.

[0037] After calculating the initial similarity, the edge computing gateway determines whether the initial similarity is lower than the standard transition trigger threshold. For each entry read from the state machine matching matrix, the edge computing gateway compares the calculated initial similarity with the standard transition trigger threshold in the entry. If the initial similarity is lower than the standard transition trigger threshold, the edge computing gateway selects the standard trajectory template corresponding to the entry as a candidate template and adds the template identifier and corresponding initial similarity of the candidate template to the candidate list. If the initial similarity is greater than or equal to the standard transition trigger threshold, the edge computing gateway discards the standard trajectory template corresponding to that entry and does not include it in the candidate list. The edge computing gateway performs dynamic time warping calculations and threshold comparisons on each entry corresponding to all standard trajectory templates stored in the state machine matching matrix, generating a candidate list containing several candidate templates.

[0038] See Figure 5 In the figure, the horizontal axis represents the sub-time period number, ranging from 0 to 500, and the vertical axis represents the degree of co-variance, ranging from approximately 0.1 to 0.9. The figure contains two curves: the solid curve represents the characteristic trajectory descriptor, and the dashed curve represents the standard co-variance sequence. The characteristic trajectory descriptor curve exhibits a periodic fluctuation trend with significant ups and downs, with a maximum value close to 0.88 and a minimum value of approximately 0.12. The standard co-variance sequence curve shows relatively smaller fluctuations, with a maximum value of approximately 0.77 and a minimum value of approximately 0.2, and its overall trend maintains good synchronization with the characteristic trajectory descriptor curve. Both curves show peaks at sub-time period numbers of approximately 50 and 300, corresponding to a high level of co-variance, reflecting strong co-variance characteristics of multiple sub-dimensional trajectory curves within that time period. Example

[0039] In specific implementation, please refer to Figure 2 When determining the device operation stage corresponding to the current time window based on the standard trajectory template with the highest trajectory shape similarity and meeting the device state transition conditions, the edge computing gateway selects the standard trajectory template with the lowest initial similarity from multiple candidate templates as the matching trajectory template. The edge computing gateway reads the candidate list, where each candidate template entry contains two fields: template identifier and initial similarity. The edge computing gateway sorts the candidate list in ascending order based on the value of the initial similarity field, and selects the standard trajectory template corresponding to the template identifier in the first candidate template entry after sorting as the matching trajectory template.

[0040] The edge computing gateway reads the device state transition conditions corresponding to the matching trajectory template. Using the template identifier of the matching trajectory template as an index, the edge computing gateway searches for the entry corresponding to that template identifier in the state machine matching matrix and reads the device state transition condition field from the entry. The device state transition condition consists of two parts: a source state set and a target state. The source state set consists of one or more valid previous device operation stage identifiers, and the target state is a single device operation stage identifier. Each element in the source state set is a predefined device operation stage enumeration value, including idle stage identifier, startup stage identifier, stable operation stage identifier, deceleration stage identifier, and shutdown stage identifier.

[0041] The edge computing gateway obtains the previous device's operational stage corresponding to the time window preceding the current time window. The edge computing gateway maintains a status register in its internal memory, which stores the device operational stage identifier determined in the previous time window. The edge computing gateway reads the previous device operational stage identifier from the status register.

[0042] The edge computing gateway compares the previous device operation phase with the source state set in the device state transition conditions. The edge computing gateway performs an equality comparison on each valid previous device operation phase identifier in the source state set. If the previous device operation phase identifier is equal to any valid previous device operation phase identifier in the source state set, then the previous device operation phase belongs to the source state set, and the edge computing gateway determines the target state in the device state transition conditions as the device operation phase corresponding to the current time window. The edge computing gateway writes the target state identifier to the state register and updates the device operation phase identifier stored in the state register.

[0043] If the identifier of the previous device operation stage is not equal to any of the valid identifiers of the previous device operation stage in the source state set, then the previous device operation stage does not belong to the source state set. The edge computing gateway traverses the remaining candidate templates in ascending order of initial similarity. The remaining candidate templates refer to all candidate template entries in the candidate list that are ranked after the matching trajectory template. The edge computing gateway takes the candidate template with the highest current ranking from the remaining candidate templates, reads the source state set in the device state transition conditions of the standard trajectory template corresponding to the candidate template, and compares whether the identifier of the previous device operation stage belongs to the source state set. If the identifier of the previous device operation stage belongs to the source state set, the edge computing gateway stops traversing and determines that the target state in the device state transition conditions corresponding to the candidate template is the device operation stage corresponding to the current time window, and writes the target state identifier into the state register. If the identifier of the previous device operation stage does not belong to the source state set, the edge computing gateway continues to take the next candidate template for comparison until a candidate template that satisfies the source state set is found.

[0044] If, after traversing all remaining candidate templates, the edge computing gateway still fails to find a candidate template whose source state set contains the identifier of the previous device operation phase, the edge computing gateway maintains the device operation phase corresponding to the current time window as the previous device operation phase. The edge computing gateway does not update the device operation phase identifier in the status register; the value in the status register remains the identifier of the previous device operation phase. Simultaneously, the edge computing gateway generates a state transition anomaly alarm event code. This code is a predefined fixed code value, indicating that the transition process of the device operation phase did not match a valid state transition path. The edge computing gateway writes the state transition anomaly alarm event code to the alarm log, appending the timestamp information of the current time window and the identifier information of the previous device operation phase.

[0045] See Figure 6 In the figure, the horizontal axis represents the sub-time period number, and the vertical axis represents the value of the degree of coordinated change. The black solid line in the figure is the feature trajectory descriptor calculated within the current acquisition time window, reflecting the coordinated change trend of multi-dimensional time-sensitive features in each sub-time period. The orange dashed line, green dotted line, and purple dotted line represent three standard trajectory templates preset in the state machine matching matrix, corresponding to the startup phase, stable operation phase, and deceleration phase of the device operation, respectively.

[0046] As observed in the figure, the current characteristic trajectory descriptor curve is highly consistent with the shape of the standard trajectory template curve during the startup phase. Especially within the sub-time period from 0 to 250, both exhibit obvious periodic fluctuations, with the degree of coordination varying between 0.15 and 0.85, indicating that the equipment is in a transitional phase from idle to startup. This characteristic trajectory descriptor shows significantly greater fluctuations than the standard trajectory template (green dotted line) during the stable operation phase, exceeding the relatively stable degree of coordination range of approximately 0.45 to 0.55. Furthermore, it differs significantly from the standard trajectory template (purple dashed line) during the deceleration phase, ruling out the possibility that the equipment has entered the deceleration phase.

[0047] After the sub-time period exceeds 250, the current characteristic trajectory description sub-curve gradually deviates from the start-up phase template, and the degree of coordinated change shows a decreasing trend, approaching the trajectory shape of the standard trajectory template for the deceleration phase, but still maintaining a certain difference, indicating that the equipment operating status has changed but has not yet fully entered the deceleration phase. The degree of coordinated change in this phase gradually decreases from 0.8 to 0.2, reflecting a reduction in the amplitude of equipment performance fluctuations. Example

[0048] In practical implementation, based on the device's operating phase, the corresponding MES service event code and compression strategy identifier are obtained by looking up a table using a state machine matching matrix. The edge computing gateway uses the device operating phase identifier corresponding to the current time window as the first-dimensional index, which is used for two-dimensional table lookup operations in both the mapping table and the compression mapping table.

[0049] For obtaining MES business event codes, the edge computing gateway uses the device operation phase identifier as the row index. The edge computing gateway calculates the average device load rate range within the current time window as the column index. The edge computing gateway extracts device load parameters from all data frames within the current time window, calculates the ratio of the sum of device load parameters to the total number of data frames, and obtains the average device load rate value. The edge computing gateway compares the average device load rate value with predefined load rate range boundary values, including a first load threshold and a second load threshold, where the first load threshold is less than the second load threshold. If the average device load rate value is less than the first load threshold, the column index is marked as a low load range. If the average device load rate value is greater than or equal to the first load threshold and less than the second load threshold, the column index is marked as a medium load range. If the average device load rate value is greater than or equal to the second load threshold, the column index is marked as a high load range. The edge computing gateway performs a two-dimensional lookup in the mapping table of the state machine matching matrix using row and column indices. The mapping table is a two-dimensional array structure, where rows correspond to the enumeration identifiers of the device's operating phase, and columns correspond to the low-load, medium-load, and high-load intervals. Each cell in the mapping table stores a MES service event code. The edge computing gateway reads the value stored in the cell where the row and column indices intersect; this value is the MES service event code obtained from the lookup.

[0050] For obtaining the compression strategy identifier, the edge computing gateway uses the device operation phase identifier as the first index. The edge computing gateway extracts the maximum amplitude of the vibration feature from multiple time-sensitive features within the current time window. The vibration feature originates from the original numerical sequence of the time-sensitive features in the data frame corresponding to the vibration sensor. The edge computing gateway compares the maximum amplitude of the vibration feature with a preset amplitude boundary. The preset amplitude boundary is written during edge computing gateway initialization and is a fixed threshold set according to the vibration sensor's range. If the maximum amplitude of the vibration feature exceeds the preset amplitude boundary, the second index marks it as an out-of-bounds state. If the maximum amplitude of the vibration feature does not exceed the preset amplitude boundary, the second index marks it as a non-out-of-bounds state. The edge computing gateway performs a two-dimensional lookup in the compression mapping table of the state machine matching matrix using the first and second indices. The compression mapping table is a two-dimensional array structure, with rows corresponding to the device operation phase enumeration identifiers and columns corresponding to the out-of-bounds and non-out-of-bounds states. Each cell in the compression mapping table stores a compression strategy identifier. The edge computing gateway reads the compression strategy identifier stored in the cell where the first and second indices intersect. The compression strategy identifier consists of two fields: one field is the compression ratio, and the other field is the compression start bit.

[0051] The edge computing gateway encapsulates the MES service event code obtained from the lookup table and the key timestamp within the current time window into an edge-side output message. The key timestamp is the system timestamp corresponding to the start time of the current time window. The edge computing gateway constructs an edge-side output message, writing a predefined message type identifier into the message header, and sequentially writing the MES service event code field and the key timestamp field into the message payload. The MES service event code field occupies four bytes, and the key timestamp field occupies eight bytes. The edge computing gateway sends the encapsulated edge-side output message to the MES system through the communication interface.

[0052] When compressing multi-source device data within the current time window according to the compression strategy identifier, the edge computing gateway parses the compression ratio in the compression strategy identifier. The compression ratio field is a single-precision floating-point number, with a value greater than zero and less than or equal to one, specifically read from a cell in the compression mapping table. The edge computing gateway performs equally spaced sampling of data frames from the multi-source device data within the current time window based on the compression ratio, obtaining the first compressed data frame sequence. The edge computing gateway then arranges all data frames collected within the current time window in ascending order of their collection timestamps, forming the original data frame sequence. Let the total number of frames in the original data frame sequence be... The compression ratio is denoted as Edge computing gateway computation extraction step size , ,in Indicates to Divide by compression ratio Round the quotient down to the nearest integer. If Less than Then Set as The edge computing gateway extracts data frames from the original data frame sequence to generate the first compressed data frame sequence according to the following formula:

[0053] in, Represents the original data frame sequence. Represents the first data frame in the original data frame sequence. One data frame, For frame number, The range of values ​​is to . This represents the extraction step size calculated from the compression ratio, and its value is a positive integer. Indicates the first compressed data frame sequence. Indicates the first compressed data frame sequence One data frame, The range of values ​​is to , This represents the total number of frames in the first compressed data frame sequence. .

[0054] The edge computing gateway parses the compression start bit in the compression strategy identifier. The compression start bit field is an integer value indicating the starting frame number in the first compressed data frame sequence from which data needs to be retained. The edge computing gateway reads the value of the compression start bit and records it as... The edge computing gateway discards data frames preceding the compression start position from the first compressed data frame sequence to obtain the second compressed data frame sequence. Specifically, the edge computing gateway retains data frames from the first compressed data frame sequence starting from the compression start position. The set of frames from the beginning of the first frame to the last frame is used as the second compressed data frame sequence.

[0055] The edge computing gateway performs an XOR differential operation on each frame of data in the second compressed data frame sequence with the previous frame to obtain a differentially encoded sequence. The edge computing gateway directly uses the first frame of data in the second compressed data frame sequence as the first frame of the differentially encoded sequence without performing an XOR operation. For the second frame and each subsequent frame in the second compressed data frame sequence, the edge computing gateway performs a bitwise XOR operation on each byte of that frame's data with the byte at the corresponding offset position of the previous frame's data to generate a new data frame. This newly generated data frame is then used as the frame at the corresponding position in the differentially encoded sequence. The edge computing gateway stores the differentially encoded sequence as compressed transmission data in a transmission buffer, awaiting uploading to the MES system.

[0056] See Figure 7 In the figure, the horizontal axis represents the sampling point number, ranging from 0 to 600, and the vertical axis represents the original value, with the unit being volts (V), ranging approximately from -2 to 2. The gray solid curve represents the sequence of original data frames collected, exhibiting obvious periodic fluctuations, with high-frequency small fluctuations within each period, reflecting the changing trend of the continuous signal collected by the equipment vibration sensor. The red solid dots mark the equally spaced sampling frames, evenly distributed throughout the sampling sequence, with stable intervals. The actual number of sampling points is much less than the number of original data points, reflecting the equally spaced sampling operation based on the compression ratio in the compression strategy identifier.

[0057] This diagram corresponds to the data compression process in Example 5. The edge computing gateway first acquires the original multi-source device data frame sequence within the current time window (i.e., the signal shown by the gray curve), then calculates the extraction step size according to the parsed compression ratio, and extracts points from the original data frame sequence at equal intervals based on the step size to form the first compressed data frame sequence. The red solid dots in the diagram represent the data frames retained in the first compressed data frame sequence, reflecting the data simplification effect after extraction. This extraction method effectively reduces the amount of transmitted data while preserving the fluctuation characteristics and key change trends of the original signal as much as possible.

[0058] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. An edge computing processing method for industrial equipment data acquisition in MES, characterized in that, Applied to an edge computing gateway, the method includes: The edge computing gateway periodically captures multi-source device data within the current time window and extracts multiple time-sensitive features of each data frame based on the device type and parameter encoding of the data source. Based on the aforementioned multiple time-sensitive features, the numerical change trajectory of each feature dimension within the current time window is recorded to generate a set of feature trajectory descriptors consisting of the rate of change and the amplitude of fluctuation. The feature trajectory descriptor is input into a preset state machine matching matrix. The state machine matching matrix stores multiple standard trajectory templates and the device state transition conditions corresponding to each standard trajectory template. The state machine matching matrix dynamically compares the trajectory shape similarity between the current feature trajectory descriptor and the multiple standard trajectory templates, and determines the device operation stage corresponding to the current time window based on the standard trajectory template with the highest trajectory shape similarity and that satisfies the device state transition conditions. Based on the device operation phase, the corresponding MES service event code and compression strategy identifier are obtained by looking up the state machine matching matrix. The MES service event code and the key timestamps in the current time window are encapsulated into edge-side output messages, and the multi-source device data in the current time window are compressed according to the compression strategy identifier.

2. The edge computing processing method for industrial equipment data acquisition for MES according to claim 1, characterized in that, Extract multiple time-sensitive features from each data frame, including: Each data frame is parsed to obtain the device identifier, data type identifier, and original numerical sequence; For the original numerical sequence, calculate the local extreme points within the sliding window, use the timestamps and amplitudes of the local extreme points as the original feature points, and perform spline interpolation reconstruction on multiple original feature points to obtain an equally spaced sampling sequence. The absolute value of the slope between adjacent sampling points in the equally spaced sampling sequence, the difference between adjacent absolute values ​​of the slope, and the higher-order difference value of the sampling value are calculated to form multiple time-sensitive features of each data frame.

3. The edge computing processing method for industrial equipment data acquisition for MES according to claim 2, characterized in that, Record the numerical change trajectory of each feature dimension within the current time window, including: Using the start time of the current time window as the origin, multiple time-sensitive features of each feature dimension are arranged according to the collection timestamp to generate multiple multi-dimensional trajectory curves; The amplitude of the multiple subdimension trajectory curves is normalized to a unified amplitude range, and the upward and downward crossing states of each subdimension trajectory curve in multiple sub-time periods are marked based on the median of the unified amplitude range. The edge computing gateway counts the number of trajectory segments with consistent upward or downward states within the same sub-time period in the multiple multi-dimensional trajectory curves, uses the statistical results as the degree of coordinated change for that sub-time period, and concatenates the degrees of coordinated change of each sub-time period in chronological order to form the feature trajectory descriptor.

4. The edge computing processing method for industrial equipment data acquisition for MES according to claim 3, characterized in that, The state machine matching matrix dynamically compares the trajectory shape similarity between the current feature trajectory descriptor and the multiple standard trajectory templates, including: Read the standard cooperative change sequence and standard transition trigger threshold corresponding to each standard trajectory template from the state machine matching matrix; The edge computing gateway performs dynamic time warping on the co-variance sequence in the feature trajectory descriptor and the standard co-variance sequence, and calculates the warped path length as the initial similarity. The edge computing gateway determines whether the initial similarity is lower than the standard transition trigger threshold. If it is lower than the standard transition trigger threshold, the standard trajectory template is selected as a candidate template; otherwise, the standard trajectory template is discarded.

5. The edge computing processing method for industrial equipment data acquisition for MES according to claim 4, characterized in that, Based on the standard trajectory template with the highest trajectory shape similarity and satisfying the device state transition conditions, the device operation stage corresponding to the current time window is determined, including: The edge computing gateway selects the standard trajectory template with the lowest initial similarity from multiple candidate templates as the matching trajectory template, and reads the device state transition conditions corresponding to the matching trajectory template. The edge computing gateway obtains the previous device operation stage corresponding to the previous time window before the current time window, compares the previous device operation stage with the source state set in the device state transition condition, and if the previous device operation stage belongs to the source state set, then the target state in the device state transition condition is determined as the device operation stage corresponding to the current time window. If the previous device operation phase does not belong to the source state set, the edge computing gateway will traverse the remaining candidate templates in ascending order of initial similarity until a candidate template that satisfies the source state set is found, and determine that the target state corresponding to the candidate template is the device operation phase corresponding to the current time window.

6. The edge computing processing method for industrial equipment data acquisition for MES according to claim 5, characterized in that, Based on the device's operating phase, the corresponding MES service event code and compression strategy identifier are obtained by looking up the state machine matching matrix, including: Using the device operation phase as the row index and the average load rate range of the device within the current time window as the column index, a two-dimensional lookup is performed in the mapping table of the state machine matching matrix to obtain the MES service event code that matches both the row index and the column index. Using the device operation phase as the first index and whether the maximum amplitude of the vibration feature among the multiple time-sensitive features exceeds the preset amplitude boundary as the second index, a two-dimensional lookup is performed in the compression mapping table of the state machine matching matrix to obtain the compression strategy identifier, which includes the compression ratio and the compression start position.

7. The edge computing processing method for industrial equipment data acquisition for MES according to claim 6, characterized in that, The multi-source device data within the current time window is compressed according to the compression strategy identifier, including: The compression ratio in the compression strategy identifier is parsed, and the data frames of multi-source device data in the current time window are sampled at equal intervals according to the compression ratio to obtain the first compressed data frame sequence. Parse the compression start bit in the compression strategy identifier, discard the data frames located before the compression start bit from the first compressed data frame sequence, and obtain the second compressed data frame sequence; The edge computing gateway performs an XOR differential operation on each frame of data in the second compressed data frame sequence with the previous frame of data to obtain a differential encoding sequence, and uses the differential encoding sequence as the compressed transmission data.

8. The edge computing processing method for industrial equipment data acquisition for MES according to claim 2, characterized in that, The spline interpolation reconstruction uses cubic spline interpolation, and the number of sampling points in the equally spaced sampling sequence is determined by the product of the current time window length and the preset sampling frequency. The preset sampling frequency is dynamically adjusted according to the data source device type.

9. The edge computing processing method for industrial equipment data acquisition for MES according to claim 3, characterized in that, When performing amplitude normalization to a unified amplitude range for the multiple sub-dimensional trajectory curves, linear normalization is performed with the global maximum and global minimum values ​​of each sub-dimensional trajectory curve as boundaries, and the unified amplitude range is [-1, 1].

10. The edge computing processing method for industrial equipment data acquisition for MES according to claim 5, characterized in that, If the edge computing gateway fails to find a candidate template containing the previous device operation stage after traversing all candidate templates, it will maintain the device operation stage corresponding to the current time window as the previous device operation stage and generate a state transition abnormal alarm event code.

Images