Operation and maintenance data transmission optimization system based on intelligent industrial sewage treatment equipment

Abstract

The application discloses an operation and maintenance data transmission optimization system based on intelligent industrial sewage treatment equipment, relates to the technical field of operation and maintenance data transmission, and comprises a processing module, a preset time window is used for dividing continuously collected equipment operation time sequence data, and a time-continuous data block sequence is generated; an analysis module is used for identifying a current data block and at least one subsequent data block which is time-continuous with the current data block, and calculating a first characteristic parameter corresponding to each data block, and the first characteristic parameter is used for quantifying fluctuation intensity of operation data in the data block. The application realizes smooth adjustment of compression rate through working condition stage division, accurately captures abnormal data features related to equipment faults, guarantees accuracy of operation and maintenance decision, avoids transmission loss caused by frequent switching of compression rate, and further reduces data transmission redundancy in a smooth period.

Description

Technical Field

[0001] This invention belongs to the field of operation and maintenance data transmission technology, specifically an operation and maintenance data transmission optimization system based on intelligent industrial wastewater treatment equipment. Background Technology

[0002] In the intelligent operation and maintenance of industrial wastewater treatment equipment, vibration data from key equipment (such as water pumps and fans) needs to be continuously uploaded to the cloud via edge devices for trend analysis. Existing systems generally adopt fixed-strategy data transmission methods, mainly involving the following two technical approaches: The first approach is a unified compression transmission scheme, where the edge device processes all collected data using a preset fixed compression rate. While this method is simple to implement, it has problems. For example, using a low compression rate during stable device operation generates a large amount of redundant data, while using a high compression rate during anomalies results in the loss of critical feature details, making it impossible to balance transmission efficiency and data quality.

[0003] The second approach is a threshold-triggered transmission scheme, which allocates a high-quality transmission channel only to data exceeding a preset threshold. While this method reduces the total amount of data, its decisions are based solely on instantaneous data values, failing to identify the correlation characteristics and changing trends between data blocks, and easily overlooking early signs of slowly developing faults.

[0004] Therefore, there is currently no effective technical solution to the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide an operation and maintenance data transmission optimization system based on intelligent industrial wastewater treatment equipment, so as to solve the problems mentioned in the background art.

[0006] An operation and maintenance data transmission optimization system based on intelligent industrial wastewater treatment equipment includes: The processing module is used to divide the continuously collected device runtime sequence data into a preset time window and generate a temporally continuous data block sequence. The analysis module is used to identify the current data block and at least one subsequent data block that is sequential with it in time, and to calculate the first feature parameter corresponding to each data block. The first feature parameter is used to quantify the fluctuation intensity of the running data within the data block. The generation module is used to generate a compression ratio control sequence based on the first feature parameters of the current data block and subsequent data blocks. This sequence assigns a compression ratio to each data block in the data block sequence, and the compression ratio is negatively correlated with the value of the first feature parameter of the corresponding data block. The upload module is used to compress each data block according to the natural time sequence of the data block sequence and the compression ratio specified in the compression ratio control sequence, and then upload them to the cloud.

[0007] Preferably, when the analysis module calculates the first feature parameter, it includes: Spectral analysis of the vibration signal within the data block was performed to determine multiple characteristic frequency ranges related to equipment mechanical failures; Calculate the vibration energy value for each characteristic frequency range; The vibration energy values ​​of each characteristic frequency range are multiplied by their preset fault sensitivity coefficients and then summed. The sum is used as the first characteristic parameter of the data block.

[0008] Preferably, when the generation module generates the compression ratio control sequence, it includes: The first characteristic parameters of the data block sequence are concatenated in chronological order to form a parametric curve; Identify the inflection point in the parameter curve where the slope changes beyond a preset value, and divide the entire data block sequence into multiple consecutive operating phases based on the inflection point; Calculate the arithmetic mean of the first characteristic parameter of all data blocks in each working condition stage; A base compression rate is set for each operating condition based on the arithmetic mean, where the base compression rate is negatively correlated with the arithmetic mean.

[0009] Preferably, the generation module further includes: Calculate the absolute difference between the first characteristic parameter of each data block and the arithmetic mean of that stage within the operating condition phase; Based on the magnitude of the absolute difference, the base compression ratio set for the data block is adjusted downwards. The larger the absolute difference, the lower the final compression ratio after adjustment.

[0010] Preferably, when dividing the working condition phases, the generation module also performs the following operations: Multi-scale sliding window analysis is performed on the parametric curves to extract trend consistency characteristics at different time scales; When a trend reversal is detected at a specific time scale, and the reversal occurs simultaneously at at least two adjacent time scales, that moment is determined as a trend divergence point. At the point of divergence in the trend, the current working condition stage division is interrupted and the compression rate control sequence is reinitialized. The re-initialized compression ratio control sequence takes the trend divergence point as the starting point and assigns a preset initial compression ratio to the starting point, which is lower than the base compression ratio of the previous working condition stage.

[0011] Preferably, the generation module, when performing multi-scale sliding window analysis, further includes: Based on historical data, a feature propagation model for different operating conditions is established. This feature propagation model is used to describe the propagation law of feature parameters between adjacent time scales. When a trend divergence point is detected, the evolution trajectory of the feature parameters in subsequent time periods is predicted through the feature propagation model. The re-initialized compression ratio control sequence is dynamically adjusted based on the predicted evolutionary trajectory, where: If the prediction shows that the feature parameters will continue to deteriorate, then a sequence of progressively decreasing compression ratios will be assigned to subsequent data blocks. If the prediction shows that the feature parameters will tend to stabilize, then a sequence of gradually increasing compression ratios will be assigned to subsequent data blocks.

[0012] Preferably, the generation module further includes: Based on the first feature parameters of the current data block and subsequent data blocks, predict the value trend of the data block sequence; When the prediction indicates that multiple consecutive data blocks with the first feature parameter exceeding the high threshold will appear, a preprocessing instruction is generated. According to the preprocessing instructions, the upload module extracts feature summaries from the data blocks in the queue whose first feature parameter is lower than the low threshold before transmitting the continuous high feature parameter data blocks; The extracted feature summary will then be uploaded in place of the original data block.

[0013] Preferably, the upload module also includes: After the data blocks are compressed according to the compression ratio control sequence and before being uploaded to the cloud, a corresponding header information is generated for each compressed data block. The header information records at least the value of the first characteristic parameter of the data block and the specific compression ratio used. The header information is encapsulated in the corresponding compressed data packet header to form a complete data transmission packet.

[0014] Preferably, the system further includes a cloud data analysis module, which performs the following operations: Receive data transmission packets and parse the header information of the data packets to obtain the first characteristic parameter and compression ratio; Based on the magnitude of the first characteristic parameter and the compression ratio, the data packets are classified and processed. Prioritize decompression and deep feature analysis of data packets with high first feature parameters and low compression ratios; The results of deep feature analysis are fused with the analysis results of other data packets to generate a health status assessment report for predictive maintenance of equipment.

[0015] Compared with the prior art, the beneficial effects of the present invention are: This invention divides time-series data into blocks through a processing module, calculates the first characteristic parameter based on vibration signal spectrum analysis through an analysis module, generates a control sequence through parameter curve construction, inflection point identification, operating condition stage division, and basic compression rate setting, and uploads the data sequentially. This solves the problem that traditional fixed compression rates cannot adapt to dynamic data fluctuations. By dividing the data into operating condition stages, it achieves smooth adjustment of the compression rate, accurately capturing abnormal data characteristics related to equipment faults to ensure the accuracy of maintenance decisions, while avoiding transmission losses caused by frequent compression rate switching, further reducing data transmission redundancy during stable periods. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the system framework structure of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0018] Please see Figure 1 This application provides an operation and maintenance data transmission optimization system based on intelligent industrial wastewater treatment equipment, including: The processing module is used to divide the continuously collected device runtime sequence data into preset time windows, generating a temporally continuous data block sequence: In this embodiment, the preset time window refers to a fixed duration set according to the collection frequency and characteristics of equipment operation and maintenance data. Its core function is to decompose massive continuous time-series data into independently processable units, avoiding the inefficiency caused by overall processing.

[0019] In practice, the duration of the time window can be flexibly configured according to the equipment type and data acquisition frequency. For example, when the acquisition frequency is 1 second / time, the time window can be set to 30 seconds, that is, each data block contains 30 sets of continuous running data. For parameters such as water quality indicators that change slowly, the time window can be extended to 1 minute to balance processing efficiency and data granularity.

[0020] During the partitioning process, the data block sequence strictly follows the principle of temporal continuity. The end time of the previous data block is seamlessly connected with the start time of the next data block to ensure that there are no time breaks in subsequent feature analysis. For example, the first data block covers data from 0 to 30 seconds, the second data block covers data from 31 to 60 seconds, and so on.

[0021] The analysis module is used to identify the current data block and at least one subsequent data block that is temporally continuous with it, and to calculate the first feature parameter corresponding to each data block: The first characteristic parameter is essentially a core indicator used to quantify the intensity of fluctuations in the running data within a data block. The larger its value, the more drastic the changes in the equipment's operating parameters during that period, and the more detailed the data needs to be retained; the smaller the value, the more stable the data, and the higher the compression rate can be used.

[0022] In this embodiment, the first characteristic parameter is calculated through spectral analysis of the vibration signal. It is used to focus on characteristic frequencies related to mechanical faults in the equipment, accurately capturing data fluctuations under fault precursors or abnormal operating conditions. The specific calculation steps are as follows: Spectral analysis is performed on the vibration signals within the data block to determine multiple characteristic frequency ranges related to equipment mechanical faults. In this embodiment, the vibration signals come from built-in vibration sensors in key rotating components such as the equipment motor, pump body, and agitator. The sampling frequency is set to 1kHz to ensure coverage of fault characteristics from low to high frequencies. In specific implementation, the characteristic frequency ranges are preset according to the equipment type. For example, the frequency range corresponding to pump body bearing faults is 100-300Hz, the frequency range corresponding to impeller imbalance is 500-800Hz, and the frequency range corresponding to motor rotor faults is 1000-1500Hz. These ranges are determined through statistical analysis of a large number of fault cases and experimental tests, and can be dynamically adjusted according to equipment model and operation and maintenance experience.

[0023] It should be noted that spectrum analysis is a well-known existing technology in the field of industrial equipment fault diagnosis, and will not be elaborated on further here.

[0024] The vibration energy value of each characteristic frequency interval is calculated separately. Specifically, the vibration signal is converted from the time domain to the frequency domain using Fast Fourier Transform (FFT) to obtain the amplitude at each frequency point; then, the sum of the squares of the amplitudes at all frequency points within each characteristic frequency interval is integrated to obtain the vibration energy value of that interval.

[0025] For example, if the integral of the sum of squares of the amplitudes at each frequency point in the 100-300Hz range of a certain data block is 25.6, then the vibration energy value of this range is 25.6; if the integral of the sum of squares in the 500-800Hz range is 12.3, then the vibration energy value is 12.3.

[0026] The vibration energy values ​​of each characteristic frequency range are multiplied by their preset fault sensitivity coefficients and then summed. The sum is used as the first characteristic parameter of the data block. In this embodiment, the fault sensitivity coefficient is quantitatively set based on "the impact level of the fault corresponding to the frequency range + historical fault loss data" to ensure that the coefficient can be reproduced for different devices and different fault types. The specific rules are as follows: First, classify the impact level of the fault: Level 1 (fatal fault, such as bearing seizure, causing equipment shutdown), Level 2 (serious fault, such as impeller imbalance, affecting sewage treatment efficiency), and Level 3 (minor fault, such as slight rotor eccentricity, with no obvious impact on operation). Corresponding coefficient benchmark values: Level 1 fault corresponding frequency interval coefficient = 0.8, Level 2 = 0.5, Level 3 = 0.3; Dynamic adjustment rules: If a fault in a certain frequency range occurs ≥3 times within a year, the coefficient is increased by 0.1; if no fault occurs in this range for two consecutive years, the coefficient is decreased by 0.1 (but not lower than 0.2, to avoid excessive reduction in weight). For example, bearing failure is a Level 1 fatal fault, and the sensitivity coefficient for the corresponding frequency range (100-300Hz) is set to 0.8; impeller imbalance is a Level 2 serious fault, and the coefficient for the corresponding range (500-800Hz) is set to 0.5; slight eccentricity of the motor rotor is a Level 3 minor fault, and the coefficient for the corresponding range (1000-1500Hz) is set to 0.3.

[0027] In specific calculations, if the vibration energy values ​​of the three characteristic frequency ranges of a certain data block are 25.6, 12.3, and 8.9 respectively, and the corresponding fault sensitivity coefficients are 0.8, 0.5, and 0.3, then the first characteristic parameter = 25.6 × 0.8 + 12.3 × 0.5 + 8.9 × 0.3 = 20.48 + 6.15 + 2.67 = 29.3.

[0028] When identifying the current data block and subsequent data blocks, the number of subsequent data blocks can be set to 1-3. The core purpose is to avoid misjudgment of compression ratio due to accidental fluctuations in a single data block by associating the characteristics of preceding and subsequent data blocks. For example, taking the current data block plus one subsequent data block as the analysis unit, if the characteristic parameter of the current data block is 29.3 and that of the subsequent data block is 35.7, it indicates that the fault characteristics are showing an increasing trend, and both data blocks should be assigned a low compression ratio.

[0029] The generation module is used to generate a compression ratio control sequence based on the first feature parameters of the current data block and subsequent data blocks: Compression ratio control sequence is a set of instructions that assigns a unique compression ratio to each data block in the data block sequence. The core rule is that the compression ratio is negatively correlated with the value of the first characteristic parameter of the corresponding data block. This is to use a low compression ratio to preserve details during periods of drastic data fluctuation and a high compression ratio to save bandwidth during periods of stable data.

[0030] In this embodiment, the compression ratio control sequence is generated by dividing the operating conditions into stages and adapting to the base compression ratio. This ensures that the compression ratio adjustment conforms to the continuous changes in the equipment's operating state, avoiding frequent fluctuations caused by independently allocating compression ratios to a single data block. Specifically: The first characteristic parameters of the data block sequence are concatenated in chronological order to form a parametric curve.

[0031] In this embodiment, the horizontal axis of the parameter curve represents the time sequence number of the data block (corresponding to the actual time progress), and the vertical axis represents the value of the first characteristic parameter. This can intuitively present the trend of the characteristic parameter changing over time, which is convenient for subsequent division of working conditions.

[0032] In practice, if the data block sequence is B1, B2, B3...Bn in chronological order, and the corresponding first characteristic parameters are P1, P2, P3...Pn respectively, then the coordinate points (1,P1), (2,P2), (3,P3)...(n,Pn) are connected sequentially to form a continuous parameter curve. For example, if the parameters of B1 to B5 are 12, 13, 29.3, 31.5, and 35.7 respectively, the corresponding parameter curve will show a trend from stable to rapidly rising.

[0033] The system identifies inflection points in the parameter curves where the slope changes beyond a preset value, and divides the entire data block sequence into multiple consecutive operating phases based on these inflection points. In this embodiment, an inflection point refers to the point in the parameter curve where the difference between the slope of the characteristic parameter change corresponding to two adjacent data blocks and the slope of the previous adjacent block exceeds a preset value; this is a time node where the equipment's operating state undergoes a significant change. The preset value is set based on equipment operation and maintenance experience, for example, 0.8, to distinguish between normal fluctuations and sudden changes in operating conditions. Specifically, the slope of two adjacent data blocks in the parameter curve is first calculated (i.e., the difference between the parameter value of the later data block and the parameter value of the earlier data block), and then the absolute value of the difference between the current slope and the previous slope is calculated; if the absolute value exceeds the preset value, the point corresponding to the current data block is determined to be an inflection point. For example, the slope of B1-B2 is 1 (13-12), and the slope of B2-B3 is 16.3 (29.3-13). The absolute value of the difference between the two is 15.3, which far exceeds the preset value of 0.8. Therefore, the point corresponding to B3 is the inflection point. The slope of B3-B4 is 2.2 (31.5-29.3), and the absolute value of the difference between the slope and the previous slope is 14.1, so no new inflection point is formed. The slope of B4-B5 is 4.2 (35.7-31.5), and the absolute value of the difference between the slope and the previous slope is 2.0, which does not exceed the preset value. Therefore, B1-B2 is the first working condition stage, and B3-B5 is the second working condition stage.

[0034] Calculate the arithmetic mean of the first characteristic parameters of all data blocks within each operating condition stage. Specifically, the arithmetic mean is calculated by summing the first characteristic parameters of all data blocks within that operating condition stage and dividing by the number of data blocks. For example, in the first operating condition stage (B1-B2), the parameters are 12 and 13, so the arithmetic mean is (12+13) / 2 = 12.5; in the second operating condition stage (B3-B5), the parameters are 29.3, 31.5, and 35.7, so the arithmetic mean is (29.3+31.5+35.7) / 3 = 32.17. This arithmetic mean is used to characterize the overall fluctuation intensity of the corresponding operating condition stage, avoiding the influence of accidental fluctuations in a single data block on the compression ratio setting, and ensuring the stability of the compression ratio within the same operating condition stage.

[0035] Based on the arithmetic mean, a base compression rate is set for each operating stage, where the base compression rate is negatively correlated with the arithmetic mean. In this embodiment, the base compression rate refers to the compression rate benchmark value shared by all data blocks within the same operating stage. Its negative correlation with the arithmetic mean is achieved through a preset mapping interval. The mapping interval is set according to bandwidth resources and data integrity requirements to ensure that both transmission efficiency and operation and maintenance needs are taken into account.

[0036] In specific implementation, the preset mapping relationship is as follows: when the arithmetic mean ≤ 15, the base compression rate is 80%; when 15 < arithmetic mean ≤ 30, the base compression rate is 50%; when the arithmetic mean > 30, the base compression rate is 20%. Based on the above example, in the first operating condition stage, the mean is 12.5 ≤ 15, and the base compression rate is set to 80%; in the second operating condition stage, the mean is 32.17 > 30, and the base compression rate is set to 20%. The final generated compression rate control sequence is the arrangement of the base compression rates for each operating condition stage in chronological order. For example, the control sequence in the above example is [80%, 80%, 20%, 20%, 20%], corresponding to data blocks B1 to B5 respectively.

[0037] The upload module is used to compress each data block according to the natural time sequence of the data block sequence and the compression ratio specified in the compression ratio control sequence, and then upload them to the cloud.

[0038] In practice, lossless compression algorithms (such as the DEFLATE algorithm) are used. The compression process strictly follows the instructions of the compression ratio control sequence, without changing the time order and core characteristics of the data. For example, if the data block sequence is B1, B2, and B3 in time order, with corresponding compression ratios of 80%, 50%, and 20% respectively, then B1 is processed first with an 80% compression ratio before being uploaded. After the upload is complete, B2 is processed, and so on.

[0039] During the upload process, to ensure data integrity, each compressed data block is accompanied by a checksum (such as a CRC32 checksum). After receiving the data, the cloud verifies its integrity using the checksum. If packet loss occurs, the upload module will automatically re-upload the data block.

[0040] Meanwhile, the upload module supports adaptive bandwidth adjustment. If the current network bandwidth is detected to be lower than the preset threshold (such as 10Mbps), the data block will be split into smaller data packets and uploaded in batches without changing the compression ratio, so as to avoid transmission interruption due to insufficient bandwidth.

[0041] This invention divides time-series data into blocks through a processing module, calculates the first characteristic parameter based on vibration signal spectrum analysis through an analysis module, generates a control sequence through parameter curve construction, inflection point identification, working condition stage division, and basic compression rate setting through a generation module, and uploads the data sequentially by compression.

[0042] This not only solves the problem that traditional fixed compression rates cannot adapt to dynamic data fluctuations, but also achieves smooth adjustment of the compression rate by dividing the working conditions into stages. This not only accurately captures abnormal data characteristics related to equipment failures and ensures the accuracy of operation and maintenance decisions, but also avoids transmission loss caused by frequent switching of compression rates and further reduces data transmission redundancy during stable periods. Furthermore, it can be adapted to industrial wastewater treatment equipment of different scales and processes. Through dynamic adjustment of characteristic frequency ranges and intelligent division of operating conditions, it can improve the system's adaptability to equipment operating status, provide more efficient, accurate and stable data transmission support for intelligent operation and maintenance, and indirectly reduce equipment failure downtime losses and operation and maintenance management costs.

[0043] As one embodiment of the present invention, the generation module further includes: Calculate the absolute difference between the first characteristic parameter of each data block and the arithmetic mean of that stage within the operating condition phase; In this embodiment, the absolute difference is an indicator used to quantify the degree of deviation between a single data block and the overall fluctuation level of its operating condition stage. The larger the value, the more significant the local fluctuation of the data block is, and the more necessary it is to retain details by reducing the compression rate.

[0044] In practice, the absolute difference is calculated as the absolute value of the difference between the first characteristic parameter value of a single data block and the arithmetic mean of the operating condition stage to which the data block belongs. The calculation formula can be simplified to: Absolute difference = |first characteristic parameter of a single data block - arithmetic mean of the operating condition stage|.

[0045] For example, a certain operating condition stage contains three data blocks B3, B4, and B5, with corresponding first characteristic parameters of 29.3, 31.5, and 35.7 respectively. The arithmetic mean of this stage is 32.17. The absolute difference between each data block is calculated as follows: The absolute difference of B3 = |29.3 - 32.17| = 2.87; The absolute difference of B4 = |31.5 - 32.17| = 0.67; The absolute difference of B5 = |35.7 - 32.17| = 3.53.

[0046] Based on the magnitude of the absolute difference, the base compression ratio set for the data block is adjusted downwards. The larger the absolute difference, the lower the final compression ratio after adjustment.

[0047] In this embodiment, downward correction refers to reducing the compression rate based on the base compression rate during the operating phase, according to the range of the absolute difference and a preset rule. This is to provide more refined compression rate adaptation for data blocks with local abnormal fluctuations and to avoid loss of abnormal information due to the uniformity of the base compression rate.

[0048] In practice, the interval division and correction range of the absolute difference are calculated based on the "fluctuation tolerance of the average value of the operating conditions" rather than a fixed empirical value. This ensures that the correction rules are reproducible under different operating conditions, and that the final compression rate after correction is not less than 10% (minimum threshold). The specific rules are as follows: Interval threshold calculation: Let the arithmetic mean of the first characteristic parameter in the operating condition stage be μ, and the default fluctuation tolerance be μ×20% (which can be adjusted to 15%-25% according to the sensitivity of the equipment), then: Low deviation range: absolute difference ≤ μ×20%; Medium deviation range: μ×20% < absolute difference ≤ μ×50%; High deviation range: absolute difference > μ×50%; Correction range settings: No correction in the low deviation range, final compression ratio = base compression ratio; 10% correction range in the medium deviation range, final compression ratio = base compression ratio - 10%; 20% correction range in the high deviation range, final compression ratio = base compression ratio - 20%.

[0049] For example: In a certain operating condition, the arithmetic mean μ = 32.17, and the fluctuation tolerance is set to 20%. Therefore, μ × 20% = 6.43, and μ × 50% = 16.08. The absolute differences of the three data blocks in this stage are 2.87 (low deviation range), 0.67 (low deviation range), and 3.53 (low deviation range), all of which do not exceed 6.43, so no correction is made. The final compression rate equals the base compression rate of 20%. If the absolute difference of a data block is 8.5 (medium deviation range), the final compression rate is 20% - 10% = 10%. If the absolute difference is 18.0 (high deviation range), the final compression rate is 20% - 20% = 0% (no compression).

[0050] It should be further explained that the interval division and correction range of the absolute difference can be flexibly adjusted according to the equipment type and operation and maintenance needs. For example, for critical equipment that is sensitive to faults, the absolute difference interval can be narrowed (e.g., "1.0 < absolute difference ≤ 3.0" can be adjusted to "0.5 < absolute difference ≤ 2.0"), and the correction range can be increased to ensure that even the most subtle abnormal fluctuations can be accurately captured; for conventional equipment that operates stably, the interval can be widened and the correction range can be reduced to balance data integrity and transmission efficiency.

[0051] The innovation of this invention lies in the fact that, on the one hand, it avoids the loss of information caused by a uniform basic compression rate for local abnormal data blocks within the same operating condition stage, ensuring that the detailed changes in the equipment's operating status are fully recorded, and providing more accurate data support for subsequent fault warning and operation and maintenance analysis; On the other hand, by setting a reasonable difference range and correction magnitude, transmission redundancy caused by over-correction is avoided. While ensuring data quality, transmission efficiency is maintained to the maximum extent, further improving the adaptability and usability of the system.

[0052] Based on the division of operating conditions into stages according to inflection points, equipment operation may experience trend reversals across time scales (such as a sudden shift from stable operation to a precursor to continuous failure). Such changes are difficult to accurately identify solely through differences in the slopes of adjacent data blocks. If the original operating condition stages are continued to be used, the compression rate control sequence will be unable to adapt to the operating state after the trend reversal in a timely manner, resulting in over-compression of critical abnormal data.

[0053] In one embodiment of the present invention, the generation module further performs the following operations when dividing the working condition stages: Multi-scale sliding window analysis is performed on the parametric curves to extract trend consistency characteristics at different time scales; In this embodiment, the multi-scale sliding window refers to setting multiple sliding windows with different durations to cover short-term and medium-term time dimensions. Its core function is to capture the changing trend of parameter curves from different time granularities, avoiding trend misjudgment caused by single-scale analysis.

[0054] In practice, the scale of the sliding window is set according to the data block time interval. For example, when the data block time window is 30 seconds, three scales of sliding windows are set: short-term window (containing 2 data blocks, corresponding to 60 seconds), medium-term window (containing 5 data blocks, corresponding to 150 seconds), and long-term window (containing 10 data blocks, corresponding to 300 seconds). The sliding step of each window is 1 data block, that is, for every new data block, all windows slide backward by 1 unit simultaneously.

[0055] Trend consistency refers to the matching degree of the trend (rising, falling, or stable) of the parameter curve within each sliding window, and the trends of different scale windows. For example, if the parameter curve shows an upward trend in the short-term window and also shows an upward trend in the medium-term window, then the two trends are consistent; if the short-term trend is upward and the medium-term trend is stable, then the trends are inconsistent. During extraction, the trend is determined by calculating the linear fitting slope of the parameter within each window: a slope > 0.5 indicates an upward trend, a slope < -0.5 indicates a downward trend, and a slope between -0.5 and 0.5 indicates a stable trend.

[0056] When a trend reversal is detected at a specific time scale, and the reversal occurs simultaneously at at least two adjacent time scales, that moment is determined as a trend divergence point. In this embodiment, trend reversal refers to the trend of a certain scale window changing from rising to falling, or from falling to rising; Adjacent time scales refer to windows with continuous time spans (such as short-term and medium-term, medium-term and long-term). The simultaneous reversal means that the trend change is not an accidental fluctuation on a single scale, but a substantial change in the operating status of the equipment.

[0057] In practice, the process for determining trend divergence points is as follows: First, monitor the trend changes at each scale window. If the trend at a certain scale (such as the medium-term window) changes from "stable" to "rising", then it is determined that the trend at that scale has reversed. Next, check whether the trends of its adjacent scales (such as short-term windows and long-term windows) reverse synchronously. If the short-term windows change from "stable" to "rising" at the same time, and the two are adjacent scales, then the condition of "at least two adjacent time scales reversing synchronously" is met. Finally, the time point corresponding to the synchronous reversal (i.e., the moment of the latest data block) is determined as the trend divergence point. For example, if the short-term window (B4-B5) and the medium-term window (B2-B6) synchronously change from "stable" to "rising" at time B6, then the time point corresponding to B6 is the trend divergence point.

[0058] At the point of divergence in the trend, the current working condition stage division is interrupted and the compression rate control sequence is reinitialized. The re-initialized compression ratio control sequence takes the trend divergence point as the starting point and assigns a preset initial compression ratio to the starting point, which is lower than the base compression ratio of the previous working condition stage.

[0059] At the point of divergence in the trend, the current working condition stage division is interrupted, and the compression rate control sequence is reinitialized.

[0060] In this embodiment, interrupting the current operating condition stage means terminating the allocation of the base compression rate to subsequent data blocks in the original operating condition stage. This is to prevent data blocks after the trend reverses from using the old operating condition's compression rate standard. Reinitializing the compression rate control sequence is to construct a completely new compression rate allocation system based on the new operating trend.

[0061] The specific operation process is as follows: take the data block corresponding to the trend divergence point as the starting point of the new sequence, and the original sequence terminates before this data block; A preset initial compression rate is assigned to the starting point, which is lower than the base compression rate of the previous stage. The preset principle of the initial compression rate is "prioritizing data integrity", which is usually 20%-30% lower than the previous base compression rate, and no higher than 50%, to ensure that the first piece of data after the trend reverses is completely preserved with a low compression rate.

[0062] For example, if the base compression rate of the previous working condition is 80%, and the preset initial compression rate is 25% lower than that, then the initial compression rate at the start of the new sequence = 80% - (80% × 25%) = 60%; if the base compression rate of the previous condition is 50%, then the initial compression rate = 50% - (50% × 30%) = 35%.

[0063] After the new sequence is started, subsequent data blocks will follow the complete process of "parameter curve construction - inflection point identification - working condition stage division - basic compression ratio setting - absolute difference correction", and dynamically allocate compression ratio based on new trend characteristics.

[0064] It should be further explained that the number of scales in the sliding window, the time span of each scale, the threshold for determining trend reversal (e.g., slope threshold ±0.5), and the reduction percentage of the initial compression rate can all be adjusted according to the equipment type and operation and maintenance requirements. For example, for critical equipment such as large sewage treatment pumps, a long-term window (e.g., containing 15 data blocks) can be added, and the initial compression rate reduction percentage can be increased to 40% to more sensitively capture trend reversals before severe failures; for small auxiliary equipment, the number of scales can be reduced (e.g., only short-term and medium-term windows can be retained), and the reduction percentage can be set to 20% to balance sensitivity and transmission efficiency.

[0065] The innovations of this application are as follows: First, it accurately captures substantial trend changes that cannot be identified by single inflection point analysis. By timely interrupting the old sequence and starting a new sequence with a low initial compression rate, it ensures that critical data after trend reversal (such as fault precursor data) is transmitted with high integrity, providing maintenance personnel with more accurate anomaly tracing basis. Second, the preset rules for the initial compression rate avoid the problem of excessively high compression rate when starting a new sequence. At the same time, the subsequent dynamic adjustment mechanism of the new sequence can take into account transmission efficiency, further improving the system's adaptability to complex operating conditions and its ability to protect critical data.

[0066] As one embodiment of the present invention, the generation module further includes the following when performing multi-scale sliding window analysis: Based on historical data, a feature propagation model for different operating conditions is established. This feature propagation model is used to describe the propagation law of feature parameters between adjacent time scales. Essentially, this feature propagation model is a prediction tool trained on massive amounts of historical data. Its core objective is to quantify the transmission relationship of feature parameters across short-term, medium-term, and long-term timescales, and to understand the evolution patterns of parameters under different operating conditions. Simply put, it establishes a correspondence between the current divergence point feature and the subsequent parameter evolution direction by using the "parameter change path after the trend divergence point" in historical data.

[0067] In actual deployment, the system first collects 1 to 3 years of equipment operation and maintenance data, filters out segments containing trend divergence points, and categorizes them by operating condition type (normal operation to failure, load change to stability, abnormal fluctuation to recovery, etc.), accumulating at least 1000 sets of valid data segments for each condition. Next, it extracts key features of the trend divergence points in each segment: the trend type of each scale window, the value of the first feature parameter, and the slope of parameter change in the three data blocks before the divergence point. Finally, using these features as input and the parameter change trajectory (rising, falling, stable, and rate of change) of the 10 data blocks after the divergence point as output, the Gradient Boosting Tree (GBT) algorithm is used to train the model, enabling it to accurately predict subsequent parameter evolution trends based on the current divergence point features.

[0068] For example, in historical data, the propagation pattern of the "normal operation bearing failure" condition is very clear: at the divergence point, the short-term and medium-term windows will simultaneously change from "stable" to "rising", with parameter values ​​between 25 and 30. The average slope of the first 3 data blocks is 2.0 to 3.0, and the subsequent parameters will continue to rise with a slope of 1.5 to 2.5. On the other hand, in the "load sudden change to stability" condition, the parameter values ​​at the divergence point are between 35 and 40, with the average slope of the first 3 data blocks being 4.0 to 5.0, and the subsequent parameter slope will gradually decrease to below 0.5.

[0069] It should be noted that the feature propagation model needs to be updated and iterated regularly, retrained quarterly with newly added operational data (at least 500 sets of valid segments), and the parameter mapping relationship optimized to avoid prediction errors caused by equipment aging and changes in operating conditions. Meanwhile, the step size and magnitude of compression ratio adjustments can be flexibly configured according to operational needs: critical equipment can have a shorter adjustment step size (e.g., adjusting once per data block) and a larger reduction in compression ratio for deteriorating trends, strengthening data protection; while routine equipment can have a longer step size and a smaller adjustment magnitude, finding a balance between efficiency and integrity.

[0070] When a trend divergence point is detected, the evolution trajectory of the feature parameters in subsequent time periods is predicted through the feature propagation model. Specifically, this prediction is not based on guesswork, but on the actual characteristics of the current trend divergence point. It outputs the changing trend of characteristic parameters within a specified future time period (e.g., the next 10 data blocks, corresponding to 300 seconds), including the direction of evolution (continuous deterioration, tending to stabilize) and the rate of change (slope range).

[0071] In practice, the real-time features of the current trend divergence point are first extracted: the trend type of each scale window (such as short-term rise, medium-term rise), the value of the first feature parameter (e.g., 32.5), and the slope of parameter change in the three data blocks before the divergence point (e.g., 2.8, 3.1, 3.5). These real-time features are then input into the model, which compares them to the historical evolution trajectories of similar features and provides a prediction. For example, if the matching degree between the input features and the historical "normal operation with bearing failure" condition reaches 92%, it is predicted that the subsequent parameter will continue to rise at a slope of 2.0 to 2.8; if the matching degree with the "load sudden change to stability" condition reaches 88%, it is predicted that the subsequent parameter slope will gradually decrease to 0.3 to 0.6.

[0072] To ensure accurate predictions, the model also outputs a confidence score from 0 to 100. If the confidence score is not lower than 80, the prediction result is used directly; if it is between 60 and 79, it is dynamically corrected based on real-time parameter updates; if it is lower than 60, the compression rate is first allocated according to the "continuous deterioration" trend, and the prediction is re-made after subsequent data blocks are used to supplement features.

[0073] The compression ratio control sequence is dynamically adjusted and re-initialized based on the predicted evolution trajectory. If the prediction shows that the characteristic parameters will continue to deteriorate, a compression ratio sequence with gradually decreasing compression ratio will be assigned to subsequent data blocks. If the prediction shows that the characteristic parameters will tend to stabilize, a compression ratio sequence with gradually increasing compression ratio will be assigned to subsequent data blocks.

[0074] In this embodiment, this dynamic adjustment is based on the initial compression ratio after restart, and sets the step size and magnitude of the compression ratio change according to the predicted evolution trajectory, so that the compression ratio sequence and parameter evolution trend are precisely matched. Essentially, it is to achieve closed-loop control of "prediction-adaptation".

[0075] In practical applications, the step size rule for adjusting the compression ratio should be set first. The adjustment range should be positively correlated with the predicted rate of parameter change, and the final compression ratio should be controlled within an effective range of 10% to 90% (the minimum of 10% is to ensure no data loss, and the maximum of 90% is to save bandwidth). The specific rules can be set as follows: Predicting continuous deterioration of feature parameters (slope not less than 1.5): reduce by 10% every 2 data blocks, with a minimum reduction to 10%; Predicting that the feature parameters continue to deteriorate but at a slower rate (0.8 ≤ slope < 1.5): a 5% decrease every 3 data blocks, with a minimum decrease to 10%; Predicted feature parameters tend to stabilize (slope < 0.8): increase by 15% every 2 data blocks, up to a maximum of 90%.

[0076] It's easier to understand when combined with specific scenarios: Scenario 1 (continuous deterioration, slope 2.5): The initial compression rate after restart is 60% (the base compression rate of the previous working condition was 80%, reduced by 25%). It is adjusted according to "reducing by 10% every 2 data blocks". The compression rates of subsequent data blocks are 60% for the first block, 60% for the second block, 50% for the third block, 50% for the fourth block, and 40% for the fifth block, and remain unchanged after decreasing to 10%. Scenario 2 (Tending to a stable state, slope 0.5): Initial compression rate 35% (the base compression rate of the previous condition was 50%, reduced by 30%), adjusted according to "increase by 15% every 2 data blocks", the compression rate of subsequent data blocks is 35% for the first block, 35% for the second block, 50% for the third block, 50% for the fourth block, 65% for the fifth block, 65% for the sixth block, 80% for the seventh block, and 80% for the eighth block, and remains unchanged after reaching 90%.

[0077] During the dynamic adjustment process, the system checks the deviation between the actual parameter trajectory and the predicted trajectory every three data blocks. If the deviation is less than 10%, the system continues to adjust according to the original sequence; if the deviation exceeds 10%, the feature propagation model is called again to update the prediction result, and the subsequent compression rate sequence is corrected synchronously.

[0078] The innovation of this application lies in its ability to predict the evolution trajectory of characteristic parameters, thereby allocating a gradually decreasing compression rate in advance for continuously deteriorating trends. This fully records the fault development process, providing maintenance personnel with end-to-end anomaly data support for accurate source tracing. Simultaneously, it allocates a gradually increasing compression rate for trends that are stabilizing, promptly reducing bandwidth waste and balancing data integrity and transmission efficiency. Furthermore, the feature propagation model is trained based on historical data, coupled with a real-time deviation verification mechanism, ensuring both the accuracy of predictions and the flexibility of adjustments. This allows the system to better adapt to complex and changing operating conditions, further improving the accuracy and reliability of data transmission during the operation and maintenance of intelligent industrial wastewater treatment equipment. As one embodiment of the present invention, the generation module further includes: Based on the first feature parameters of the current data block and subsequent data blocks, predict the value trend of the data block sequence; In this embodiment, the value trend essentially determines the reference value of subsequent data for equipment operation and maintenance. The higher the first characteristic parameter, the more drastic the equipment fluctuations, the more key information such as faults and anomalies the data contains, and the higher the value; the lower the parameter, the more stable the data, and the lower the reference value. The core objective of prediction is to identify in advance scenarios where low-value data precedes high-value data, thus preparing for bandwidth allocation.

[0079] Detailed operating procedures: Establish two key criteria: a high threshold and a low threshold. These are set by analyzing the distribution of the primary characteristic parameter using historical maintenance data. This requires covering at least 100,000 historical data blocks to ensure the representativeness of the quantile distribution. For example, when analyzing data from the past year, set the 90th percentile as the high threshold (e.g., 35) and the 10th percentile as the low threshold (e.g., 10). Every six months, recalculate the quantiles and update the thresholds based on newly added maintenance data to accommodate parameter drift as equipment ages.

[0080] Extract the first feature parameters of the current data block and the next 3 to 5 generated data blocks, and analyze the parameter change trend using a feature propagation model. The model input features need to be mapped to the 0-1 range using Min-Max normalization to improve model convergence speed. If the prediction shows that three or more data blocks with first feature parameters exceeding the high threshold will appear consecutively, and there are currently unuploaded data blocks with first feature parameters below the low threshold in the current transmission queue, then the value trend is determined to be low-value first, followed by a concentrated arrival of high-value data.

[0081] Practical application example: The first feature parameter of the current data block B7 is 8, and the parameters of the subsequently generated B8 and B9 are 9 and 11 respectively. According to the feature propagation model, the parameters of B10 to B12 will reach 38, 42 and 45 respectively, all exceeding the high threshold of 35. At this time, the value trend is determined to be low value first and high value concentrated.

[0082] When the prediction indicates that multiple consecutive data blocks with the first feature parameter exceeding the high threshold will appear, a preprocessing instruction is generated. Specifically, the preprocessing instructions are control signals sent from the generation module to the upload module. The core content includes the range of data blocks for which feature summary extraction is required, the rules for feature summary extraction, and the execution requirements for alternative uploads.

[0083] Instruction generation must meet the following requirements: the feature propagation model has a prediction confidence of no less than 85% for continuous high-value data, and the number of unuploaded data blocks below the low threshold in the queue is no less than 2, to avoid excessive simplification of low-value data due to accidental prediction bias.

[0084] If the prediction confidence level is between 60% and 85%, only partial summary extraction is performed on data blocks in the queue that are below the low threshold, retaining 50% of the original data blocks to balance bandwidth and data integrity. For example, if B10 to B12 are predicted to be continuous high-value data with a confidence level of 90%, and the parameters of B7 and B8 in the transmission queue that have not been uploaded are 8 and 9 respectively, both below the low threshold of 10, then a preprocessing instruction is generated, requiring feature summary extraction to be performed on B7 and B8. After extraction, the summary is uploaded instead of the original data block, prioritizing the transmission bandwidth of B10 to B12.

[0085] According to the preprocessing instructions, before transmitting continuous high-feature-parameter data blocks, the upload module extracts feature summaries from the preceding data blocks in the queue whose first feature parameter is below the low threshold, and then replaces the original data blocks with the extracted feature summaries for uploading. In this embodiment, feature summary extraction follows the principle of removing redundancy and retaining core features, ensuring that the summary reflects the overall characteristics of low-value data while significantly reducing the data volume; typically, the summary size is only 5% to 10% of the original data block. The extraction rules balance simplicity and usability. Three types of core information are extracted from a single low-value data block: the timestamp range of the data block, the value of the first feature parameter, and the variance of the parameter fluctuation, which are stored in the form of structured strings.

[0086] Multiple consecutive low-value data blocks are extracted by merging them, with a unified timestamp range. The arithmetic mean of the first feature parameter of each data block and the overall fluctuation variance are taken to generate a comprehensive summary.

[0087] The alternative upload execution logic revolves around prioritizing summary uploads and reserving bandwidth for high-value data: After receiving the preprocessing instruction, the upload module pauses the transmission preparation of high-value data blocks and prioritizes the extraction of digests from low-value data blocks.

[0088] The extracted feature summaries are uploaded to the cloud in chronological order. Upon receiving the summaries, the cloud parses and stores them according to a preset format. The storage path of the summaries is associated with that of the original data blocks for easy tracing later. If a summary upload fails, it will retry twice after a 1-second interval. If it still fails, the summary will be temporarily abandoned and replaced with a higher-value data, with the summary being re-uploaded when bandwidth becomes available.

[0089] Once the abstract is uploaded, the transmission of continuous high-value data blocks will begin immediately. During the transmission, bandwidth will be prioritized, and the transmission rate of other non-critical data can be limited to ensure that high-value data is uploaded in real time and completely.

[0090] Practical application example: In the queue, timestamp B7 is 00:10:00-00:10:30, parameter 8, variance 0.2; timestamp B8 is 00:10:31-00:11:00, parameter 9, variance 0.3. Both are low-value data blocks. A comprehensive summary is extracted for the time range 00:10:00-00:11:00, with a mean parameter of 8.5 and a variance of 0.25. This summary is used to replace the original data of B7 and B8 for uploading. Subsequently, the original data of B10 to B12 are transmitted first, and their compression ratio has been adjusted to a low compression ratio according to the rules mentioned above.

[0091] It should be noted that the high and low thresholds can be flexibly adjusted according to the equipment maintenance priorities. For equipment with high fault early warning requirements, the high threshold can be lowered (e.g., 30) and the low threshold can be raised (e.g., 15) to allow more data to be judged as high-value and ensure that no critical information is missed. In scenarios with limited bandwidth resources, the high threshold can be raised (e.g., 40) and the low threshold can be lowered (e.g., 8) to increase the proportion of feature summary replacement and further save bandwidth.

[0092] The rules for extracting feature summaries can be flexibly expanded. If low-value data contains multi-dimensional parameters such as water quality indicators and motor current, maximum and minimum value extraction items for each parameter can be added to ensure that the summary information is sufficient to support basic operation and maintenance analysis and to avoid affecting data availability due to excessive simplification.

[0093] The feature propagation model training requires 500 iterations. Training stops when the prediction accuracy on the validation set is ≥90% to ensure the reliability of the model's predictions.

[0094] Feature summaries replace low-value data, significantly reducing unnecessary bandwidth usage and reserving sufficient transmission channels for subsequent high-value data. This prevents critical anomaly data from being delayed or lost due to bandwidth congestion, improving the timeliness of fault early warning. Feature summaries retain the core statistical information of low-value data, without affecting basic operation and maintenance analysis or causing data waste. This achieves the optimization goal of transmitting high-value data completely and low-value data in a streamlined manner, improving the cost-effectiveness and practicality of the entire data transmission system.

[0095] As one embodiment of the present invention, the upload module further includes: After the data blocks are compressed according to the compression ratio control sequence and before being uploaded to the cloud, a corresponding header information is generated for each compressed data block. The header information records at least the value of the first characteristic parameter of the data block and the specific compression ratio used. The specific generation logic is as follows: The header information is stored in a structured binary format, which balances transmission efficiency and parsing convenience, and avoids redundant text formatting.

[0096] Core field definitions: The first feature parameter value is stored as a 32-bit floating-point number with two decimal places. The original first feature parameter value before compression of the data block is directly extracted (e.g., 8.00, 32.50, 45.00) to ensure the accuracy of the value.

[0097] Specific compression ratio: Stored as an 8-bit unsigned integer, ranging from 0 to 100, corresponding to a compression ratio of 0% (no compression) to 100% (maximum compression). For example, a compression ratio of 10% is stored as 10, and a compression ratio of 60% is stored as 60.

[0098] Extended fields (optional configuration): Depending on the operation and maintenance requirements, you can add data block number (32-bit integer), timestamp (64-bit integer, accurate to milliseconds), and data block source device identifier (16-bit string encoding). The total length of extended fields should not exceed twice that of the core fields to avoid excessively large headers affecting transmission efficiency.

[0099] The generation timing is linked to the compression process: After the data block is compressed, the upload module synchronously reads the compression configuration (specific compression ratio) and original feature parameters of the data block, and generates header information in real time to ensure that the header is consistent with the compressed data and there are no matching errors.

[0100] The header information is encapsulated in the corresponding compressed data packet header to form a complete data transmission packet.

[0101] Specifically, the data packet structure is defined as follows: a complete data transmission packet consists of a header area and a data area, with the header area first and the data area second, without any additional separators. Boundary identification is achieved through a fixed header length.

[0102] The header area has a fixed length: when it contains only core fields, the header area length is 5 bytes (4 bytes of 32-bit floating-point type + 1 byte of 8-bit unsigned integer); when it contains extended fields, the header area length is accumulated according to the number of extended fields, with a maximum of 15 bytes, and is uniformly configured during system initialization, so that it is consistent between the cloud and the terminal.

[0103] Encapsulation order: First, the header information is converted into a binary data stream according to the field order and written to the header of the data packet; then, the compressed data stream is written to the data area immediately afterward to form a continuous binary transmission packet. For example, if the first characteristic parameter of a data block is 8.00 and the compression rate is 10%, the core field header binary stream is 0x40100000 (32-bit floating-point code of 8.00) + 0x0A (8-bit integer code of 10), followed by the compressed data stream of this data block, to form a complete transmission packet.

[0104] Verification mechanism: A 1-byte check bit is appended to the end of the header area. An XOR check algorithm is used (the check value is obtained by bitwise XORing all bytes in the header area). When the cloud parses the data, it verifies the integrity of the header by the check bit. If the verification fails, the data packet is marked as abnormal and the terminal is requested to retransmit.

[0105] Practical application example: Before compression, the first characteristic parameter of data block B7 is 8.00. After processing, the compression rate is 10%, generating core field header information (32-bit floating-point encoding of 8.00 + integer encoding of 10). The header area is 5 bytes long, and the checksum is calculated by XORing to obtain 0x15. This header area is encapsulated in the compressed data stream header of B7 to form a complete transmission packet. After being uploaded to the cloud, the cloud first reads the first 5 bytes of the header area. After verification, it parses out the first characteristic parameter 8.00 and the compression rate of 10%, and then parses the compressed data of the subsequent data area according to the compression rate. Specifically, the header field configuration can be flexibly adjusted according to cloud processing needs: for scenarios that only require quick filtering of high-value data, only the core fields need to be retained; for scenarios that require precise traceability, extended fields such as device identifier and data block number can be added.

[0106] The header's binary encoding format follows industry-standard specifications, ensuring compatibility with different brands and models of cloud-based parsing devices and preventing parsing failures due to format inconsistencies. Furthermore, the header generation and encapsulation process is automatically completed by the upload module's built-in algorithm, requiring no manual intervention and ensuring automation and efficiency.

[0107] The innovation of this application lies in the fact that the cloud can quickly obtain the first feature parameter and compression ratio without decompressing the data packet, and can prioritize the parsing of data blocks with high feature parameters (high value), thereby shortening the identification delay of abnormal data and improving the operation and maintenance response speed.

[0108] The header information records the core attributes of the data block. When data transmission or parsing is abnormal, the characteristic parameters, compression configuration, and other information of the problematic data block can be located directly through the header without tracing the original data, thus reducing the difficulty of troubleshooting.

[0109] By verifying the integrity of the header through a checksum mechanism, it is indirectly ensured that the data packet has not been tampered with or lost during transmission, thereby improving the reliability of data transmission.

[0110] In one embodiment of the present invention, the system further includes a cloud data analysis module, which performs the following operations: Receive data transmission packets and parse the header information of the data packets to obtain the first characteristic parameter and compression ratio; In this embodiment, TCP streaming reception is used to ensure transmission reliability. The receive buffer is fixed at 1MB, which is just right for the maximum transmission size of a single packet (header area ≤ 15 bytes + data area ≤ 512KB), thus avoiding both buffer overflow and data blocking. Before parsing, the total length of the header area needs to be locked. This length is configured synchronously with the terminal during system initialization: 5 bytes when only the core fields (first feature parameter + compression ratio) are included, and 10 bytes when extended fields (device identifier + data block number + timestamp) are included. The corresponding length of binary data can be directly extracted from the beginning of the data packet.

[0111] The data is parsed in reverse order according to the terminal's preset field sequence: 32-bit floating-point data is restored to the first characteristic parameter retaining two decimal places, and 8-bit unsigned integers are directly mapped to a compression ratio of 0-100; in the extended fields, the device identifier is restored to a 16-bit string according to ASCII encoding, the data block number is a 32-bit integer, and the timestamp is a 64-bit integer (accurate to milliseconds). The key prerequisite for parsing is to verify the 1-byte checksum at the end of the header area. The checksum is compared by XOR inverse operation. If the verification passes, the header is confirmed to be complete; if the verification fails, "header abnormal" is immediately marked, the reception time, device identifier, and other log information are recorded, and a terminal retransmission request is triggered (retransmission interval is 1 second, maximum 2 retries); if a field format error is found during parsing (such as a compression ratio exceeding the 0-100 range, or a negative timestamp), the data packet is directly discarded without affecting the subsequent data packet processing flow.

[0112] The header binary encoding strictly follows the ISO / IEC 8825-1ASN.1 encoding standard to ensure compatibility between terminals of different brands and models and cloud parsing devices; the decompression algorithm is explicitly defined as the DEFLATE algorithm, which supports the decompression of ZLIB compression format and is fully compatible with the terminal compression algorithm.

[0113] Based on the magnitude of the first characteristic parameter and the compression ratio, the data packets are classified and processed. In this embodiment, the first feature parameter is divided into three levels according to the threshold set above: above the high threshold of 35 is level A (high value), between the low threshold of 10 and the high threshold of 35 is level B (medium value), and below the low threshold of 10 is level C (low value); the compression rate is also divided into three levels: below 30% is level I (low compression difficulty), 30%-70% is level II (medium compression difficulty), and above 70% is level III (high compression difficulty).

[0114] The final processing priority, from highest to lowest, is as follows: "Level A + Level I", "Level A + Level II", "Level B + Level I", "Level A + Level III", "Level B + Level II", "Level C + Level I", "Level B + Level III", "Level C + Level II", and "Level C + Level III", for a total of nine levels. This priority ranking is based on verification using 10,000 sets of historical data and can maximize the processing efficiency of high-value data. The classification is completed immediately after parsing the packet header, and the classification results are bound to the data packets and stored in a cloud cache (cache capacity set to 10GB). The cache is sorted by priority, with higher-priority data occupying the front and entering the processing queue first.

[0115] For scenarios where multiple high-priority data packets arrive simultaneously, a secondary sorting rule based on "reception time + feature parameter value" is adopted: first sort by reception time, and then sort the data packets received at the same time in descending order according to the value of the first feature parameter to avoid congestion in the processing queue.

[0116] Actual classification example: Data packets with a first characteristic parameter of 45.00 (Level A) and a compression rate of 10% (Level I) have the highest priority; data packets with a first characteristic parameter of 8.00 (Level C) and a compression rate of 80% (Level III) have the lowest priority; data packets with a first characteristic parameter of 25.00 (Level B) and a compression rate of 50% (Level II) have an intermediate priority.

[0117] Prioritize decompression and deep feature analysis of data packets with high first feature parameters and low compression ratios; In this embodiment, the DEFLATE decompression algorithm adapted to the terminal is called during decompression. The decompression thread for high-priority data packets is set to the highest priority, occupying 60% of the cloud's decompression computing power to ensure decompression speed. After decompression, data integrity is verified by comparing the CRC32 checksum. If decompression fails or verification fails, the terminal is immediately requested to retransmit the data packet.

[0118] Deep feature analysis extracts multi-dimensional core features, including, in addition to the first feature parameter, parameter change slope (parameter difference between adjacent data blocks / time interval), fluctuation variance (calculated in a 30-second window), similarity with historical data of the same working conditions (using the cosine similarity algorithm, with a threshold ≥0.8 for similarity), and abnormal feature matching degree (fuzzy matching based on the fault feature library, with a matching degree ≥0.75 for anomalies).

[0119] A random forest machine learning classification model (training stopped after 500 iterations when the validation set accuracy is ≥90%) is used to analyze these features and accurately identify the equipment's operating status: a score ≥85 indicates normal operation, 70-84 indicates a precursor to a fault, 50-69 indicates a minor anomaly, and <50 indicates a severe anomaly. The types of abnormal features are clearly labeled (e.g., bearing wear, water quality exceeding standards, motor overload). The analysis results are stored in a structured JSON format, including fields such as data block identifier, analysis time, operating status level, anomaly feature details, and similarity score. These results are also stored in a MySQL database in association with the original data package, supporting fast tracing and querying by device identifier and time range.

[0120] The results of deep feature analysis are fused with the analysis results of other data packets to generate a health status assessment report for predictive maintenance of equipment.

[0121] In this embodiment, fixed weights are allocated according to priority during fusion: high-priority results account for 60%, medium-priority results account for 30%, and low-priority results account for 10%. This weight allocation is verified based on 10,000 sets of historical data. This ratio can improve the anomaly identification accuracy to 92%, taking into account both key anomalies and overall trends.

[0122] Furthermore, the integration process is as follows: first, the proportion of equipment operating status in each priority result is counted, and then the comprehensive status score (0-100 points) is calculated according to the weight - a score ≥85 is "healthy", 70-84 is "basically healthy", 50-69 is "minor abnormality", and <50 is "serious abnormality".

[0123] For scenarios where "the same device simultaneously exhibits severe anomalies and a large amount of low-value data", the report generation logic is as follows: prioritize inserting details of severe anomalies (including anomaly characteristics, occurrence time, and handling suggestions), and create a separate "low-value data distribution statistics" section to briefly explain the time range and parameter intervals of the low-value data without affecting the core presentation of the anomaly information.

[0124] The assessment report comprises four core parts: basic equipment information (equipment identifier, assessment period, total number of data entries), health status level and score (including the weighting of each priority level), details of key anomalies (from the analysis results of high-priority data packets, including anomaly type, confidence level, and similar historical cases), and predictive maintenance recommendations (such as "bearings show signs of wear, it is recommended to shut down for inspection within 72 hours," "water quality parameters fluctuate normally, maintain routine operation and maintenance," and "motor overload risk, it is recommended to reduce the load and monitor current changes"). The report is generated hourly at a fixed rate, and is generated immediately if a serious anomaly is detected. It supports both PDF and JSON formats and is simultaneously pushed to the operation and maintenance management platform (HTTP interface push) and the relevant personnel's mobile APP (WebSocket real-time push). It is also stored on a cloud file server, retaining at least one year's worth of historical reports, and supports retrieval and download by equipment identifier, report type, and time range.

[0125] Specifically, the threshold values ​​for feature parameters and the grading boundaries for compression ratios in the grading standards can be flexibly adjusted according to the type of equipment: for core processing equipment, the threshold value for Grade A feature parameters can be increased to 40, and the priority weight of high compression ratios (≥70%) can be reduced (e.g., Grade A + Grade III can be adjusted to the 5th priority) to ensure that core equipment data is processed more finely; in bandwidth-constrained scenarios, the low threshold value can be reduced to 8 to expand the range of low-value data and improve transmission efficiency.

[0126] The random forest model for deep feature analysis needs to be updated in conjunction with the terminal feature propagation model. It should be retrained every quarter with at least 500 new sets of valid operation and maintenance data to optimize model parameters and avoid recognition bias caused by equipment aging. The fusion weight can be dynamically calibrated according to the actual operation and maintenance effect. If there is no abnormal feedback for low-priority data for three consecutive months, its weight can be reduced to 5% to further focus on high-value data.

[0127] The header extension fields can be added or deleted as needed. Newly added fields must meet the rule that "the total length does not exceed twice that of the core fields" (tests show that transmission efficiency drops by more than 15% after exceeding this limit). After adding a field, the terminal and cloud configurations must be synchronized during system initialization to ensure parsing compatibility.

[0128] The innovation of this application lies in its hierarchical processing, which avoids indiscriminate parsing and concentrates the main computing power on high-value data, reducing cloud computing power consumption and improving overall data processing efficiency. High-priority data is analyzed first, enabling rapid identification and push of abnormal information, shortening the time lag between identifying early signs of failure and making operational decisions, and allowing for more timely predictive maintenance. Combined with multi-priority data fusion analysis, it highlights key abnormal information while also considering overall operational trends, generating a more comprehensive and reliable health status assessment report, providing a scientific basis for equipment operation and maintenance.

[0129] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations 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. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A data transmission optimization system for operation and maintenance of intelligent industrial wastewater treatment equipment, characterized in that: include: The processing module is used to divide the continuously collected device runtime sequence data into a preset time window and generate a temporally continuous data block sequence. The analysis module is used to identify the current data block and at least one subsequent data block that is sequential with it in time, and to calculate the first feature parameter corresponding to each data block. The first feature parameter is used to quantify the fluctuation intensity of the running data within the data block. The generation module is used to generate a compression ratio control sequence based on the first feature parameters of the current data block and subsequent data blocks. This sequence assigns a compression ratio to each data block in the data block sequence, and the compression ratio is negatively correlated with the value of the first feature parameter of the corresponding data block. The upload module is used to compress each data block according to the natural time sequence of the data block sequence and the compression ratio specified in the compression ratio control sequence, and then upload them to the cloud. The generation module generates a compression ratio control sequence by concatenating the first feature parameters of the data block sequence in chronological order to form a parameter curve. Identify the inflection point in the parameter curve where the slope changes beyond a preset value, and divide the entire data block sequence into multiple consecutive operating phases based on the inflection point; Calculate the arithmetic mean of the first characteristic parameters of all data blocks in each working condition stage; based on the arithmetic mean, set a base compression ratio for each working condition stage, wherein the base compression ratio is negatively correlated with the arithmetic mean; The generation module also includes: Calculate the absolute difference between the first characteristic parameter of each data block and the arithmetic mean of that stage within the operating condition phase; Based on the magnitude of the absolute difference, the base compression ratio set for the data block is adjusted downwards. The larger the absolute difference, the lower the final compression ratio after adjustment.

2. The operation and maintenance data transmission optimization system based on intelligent industrial wastewater treatment equipment according to claim 1, characterized in that, When the analysis module calculates the first feature parameter, it includes: Spectral analysis of the vibration signal within the data block was performed to determine multiple characteristic frequency ranges related to equipment mechanical failures; Calculate the vibration energy value for each characteristic frequency range; The vibration energy values ​​of each characteristic frequency range are multiplied by their preset fault sensitivity coefficients and then summed. The sum is used as the first characteristic parameter of the data block.

3. The operation and maintenance data transmission optimization system based on intelligent industrial wastewater treatment equipment according to claim 1, characterized in that, When dividing the operating condition phases, the generation module also performs the following operations: Multi-scale sliding window analysis is performed on the parametric curves to extract trend consistency characteristics at different time scales; When a trend reversal is detected at a specific time scale, and the reversal occurs simultaneously at at least two adjacent time scales, that moment is determined as a trend divergence point. At the point of divergence in the trend, the current working condition stage division is interrupted and the compression rate control sequence is reinitialized. The re-initialized compression ratio control sequence takes the trend divergence point as the starting point and assigns a preset initial compression ratio to the starting point, which is lower than the base compression ratio of the previous working condition stage.

4. The operation and maintenance data transmission optimization system based on intelligent industrial wastewater treatment equipment according to claim 3, characterized in that, The generation module also includes the following when performing multi-scale sliding window analysis: Based on historical data, a feature propagation model for different operating conditions is established. This feature propagation model is used to describe the propagation law of feature parameters between adjacent time scales. When a trend divergence point is detected, the evolution trajectory of the feature parameters in subsequent time periods is predicted through the feature propagation model. The re-initialized compression ratio control sequence is dynamically adjusted based on the predicted evolutionary trajectory, where: If the prediction shows that the feature parameters will continue to deteriorate, then a sequence of progressively decreasing compression ratios will be assigned to subsequent data blocks. If the prediction shows that the feature parameters will tend to stabilize, then a sequence of gradually increasing compression ratios will be assigned to subsequent data blocks.

5. The operation and maintenance data transmission optimization system based on intelligent industrial wastewater treatment equipment according to claim 1, characterized in that, The generation module also includes: Based on the first feature parameters of the current data block and subsequent data blocks, predict the value trend of the data block sequence; When the prediction indicates that multiple consecutive data blocks with the first feature parameter exceeding the high threshold will appear, a preprocessing instruction is generated. According to the preprocessing instructions, the upload module extracts feature summaries from the data blocks in the queue whose first feature parameter is below a low threshold before transmitting continuous high feature parameter data blocks; and then replaces the original data blocks with the extracted feature summaries for uploading.

6. The operation and maintenance data transmission optimization system based on intelligent industrial wastewater treatment equipment according to claim 1, characterized in that, The upload module also includes: After the data blocks are compressed according to the compression ratio control sequence and before being uploaded to the cloud, a corresponding header information is generated for each compressed data block. The header information records at least the value of the first characteristic parameter of the data block and the specific compression ratio used. The header information is encapsulated in the corresponding compressed data packet header to form a complete data transmission packet.

7. The operation and maintenance data transmission optimization system based on intelligent industrial wastewater treatment equipment according to claim 6, characterized in that, The system also includes a cloud-based data analysis module, which performs the following operations: Receive data transmission packets and parse the header information of the data packets to obtain the first characteristic parameter and compression ratio; Based on the magnitude of the first characteristic parameter and the compression ratio, the data packets are classified and processed. Prioritize decompression and deep feature analysis of data packets with high first feature parameters and low compression ratios; The results of deep feature analysis are fused with the analysis results of other data packets to generate a health status assessment report for predictive maintenance of equipment.

Images