Metal powdering process optimization system based on data analysis

Abstract

This invention relates to the field of data processing technology, specifically to a metal pulverizing process optimization system based on data analysis. The system includes a pressure abrupt change identification module, a multi-source data sequence registration module, a particle size trend highlighting module, a velocity-pressure offset decomposition module, and a trend time period classification module. In this invention, by extracting pressure abrupt change points to construct a unified time index, precise alignment of multi-source data such as energy consumption, temperature, and particle size is achieved in the time dimension. Combined with directional consistency analysis, linked time periods are screened. Furthermore, stable trend segments are identified based on changes in the average particle size. The relationship between velocity and pressure difference is decomposed and directionally matched to quantify, constructing a trend response index system. Finally, trend levels are classified according to the degree of matching, and time and directional information are integrated. This achieves the aggregated expression of multi-variable linkage features and the hierarchical presentation of particle size change trends, improving the accuracy of process identification and control response capabilities under complex operating conditions.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a metal powdering process optimization system based on data analysis. Background Technology

[0002] The field of data processing technology primarily involves the collection, organization, transformation, storage, and analysis of raw data to achieve efficient information utilization and management. This technology encompasses core aspects such as data acquisition, data preprocessing, data modeling, and data mining and analysis, and is widely applied in various scenarios including industrial manufacturing, intelligent control, financial management, and healthcare. Data processing not only relies on basic statistical analysis and logical reasoning methods but also integrates data analysis algorithms and knowledge reasoning techniques tailored to specific business objectives. Its key lies in effectively organizing and analyzing massive amounts of data to extract valuable information, providing decision support for the optimization of various systems. Among these, the traditional metal pulverizing process optimization system refers to addressing the parameter control and process selection issues in the physical pulverizing process of metal materials. It involves adjusting physical parameters such as pulverizing temperature, speed, and pressure through manual experience or based on single-variable experimental methods to achieve process optimization. This type of optimization method usually uses orthogonal experimental design or response surface methodology in experimental statistics to explore the relationship between process factors, and then combines physical property analysis data to perform parameter fitting and optimization. However, these methods often rely on expert experience, are highly dependent on data and lack flexibility. They are difficult to handle the relationship between large-scale multivariate process data and cannot meet the needs for precise optimization of the powdering process in complex manufacturing scenarios.

[0003] Existing technologies adjust process parameters using experience-based modeling, primarily focusing on the analysis of process factor ratios under static conditions. When faced with real-time changes in operating conditions, they lack the ability to characterize the synchronicity and trend linkage between data. When multiple variables fluctuate simultaneously, it is difficult to identify the driving relationship of key variables, which can easily lead to problems such as ambiguous response to particle size changes and a high proportion of redundant data. Especially in continuous production scenarios, the difficulty in timely extraction of trend signals will affect the accuracy and timeliness of process intervention, resulting in potential process fluctuation risks such as unstable particle size distribution and difficulty in controlling energy consumption levels. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a data analysis-based metal powdering process optimization system.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a metal powdering process optimization system based on data analysis includes:

[0006] The pressure change identification module collects the continuous pressure sequence in the pressure regulation cavity, divides it into equal time periods, calculates the difference between the beginning and end of each segment, compares it with the average of adjacent segments, and if the current difference exceeds twice the average of adjacent segments, it records the time index, summarizes all segments that meet the conditions, and generates a pressure change time marker column.

[0007] The multi-source data sequence registration module extracts the corresponding energy consumption and temperature data based on the pressure change time marker column, aligns the time axis and calculates adjacent differences, marks the directional changes, filters time periods with the same direction and numbers them, forming a set of time periods with the same direction.

[0008] The particle size trend highlighting module uses the set of time periods with consistent direction to obtain particle size data, calculates the mean by segment, compares the direction of two consecutive sets of mean values, extracts the time periods with consistent direction and sufficient duration, and generates a sequence of stable particle size change segments.

[0009] The velocity-pressure offset decomposition module extracts the rotational velocity and pressure data for the corresponding time period from the sequence of stable particle size changes, calculates the difference between adjacent values, calculates the difference between the two to generate a new column, determines whether the direction of change is consistent with the particle size, calculates the consistency ratio, and generates a rotational pressure offset matching ratio column.

[0010] As a further embodiment of the present invention, the pressure change time marker column includes a change time point index, a corresponding time period number, and a change intensity level; the set of directional consistent time periods includes an energy consumption change direction, a temperature change direction, and a unified number identifier; the particle size stable change segment sequence includes a particle size mean sequence, a change direction label, and a duration length; and the rotational pressure offset matching ratio column includes a velocity difference sequence, a pressure difference sequence, and a directional consistency ratio.

[0011] As a further aspect of the present invention, the pressure change recognition module includes:

[0012] The pressure acquisition submodule acquires continuous pressure records in the pressure regulation cavity, divides the records into multiple time periods of equal duration, extracts the starting and ending pressures of each time period, calculates the pressure difference of the corresponding time period based on the starting and ending pressures, and obtains the segmented pressure change.

[0013] The difference calculation submodule, based on the segmented pressure change, selects the pressure change of adjacent time periods for each time period, calculates the average change level of the two periods as the benchmark data, compares the pressure change of the current time period with the benchmark data, determines whether it is greater than twice the benchmark data, and obtains the abrupt change trend comparison result.

[0014] The mutation identification submodule filters all time period indices that are determined to be mutations based on the mutation trend comparison results, collects and organizes the time positions in order, establishes a unified identifier list, and obtains the pressure mutation time marker column.

[0015] As a further aspect of the present invention, the formula for calculating the pressure change is specifically as follows:

[0016] ;

[0017] in, This represents the pressure change during the i-th time period. This represents the pressure value at the end of the i-th time interval. This represents the pressure value at the start of the i-th time interval. This represents the pressure change at the j-th sub-time point within the i-th time period. This represents the pressure change weighting coefficient corresponding to the j-th sub-time point within the i-th time period. This represents the number of sub-time points within the i-th time period. This represents a small positive correction value within the i-th time period.

[0018] As a further aspect of the present invention, the multi-source data sequence registration module includes:

[0019] The data extraction submodule extracts energy consumption data and temperature data within the corresponding time period based on all time indices of the pressure change time marker column, arranges the energy consumption data and temperature data point by point according to the time index order, adjusts the start and end positions of the data to ensure consistency of time nodes, and obtains the synchronous data arrangement result.

[0020] The difference determination submodule calculates the energy consumption change difference and temperature change difference at continuous time points based on the synchronous data arrangement result, determines the positive and negative directions of each pair of differences, marks the directional status of the two data channels in the corresponding time period, establishes a directional correspondence table, and obtains the directional change corresponding identifier.

[0021] The consistent screening submodule filters time periods with consistent directional states based on the corresponding identifiers of the directional changes, uniformly numbers consecutive data segments with consistent directions, collects the identifier values ​​in order of numbering, removes segments with different directions, and establishes a set of time periods with consistent directions.

[0022] As a further aspect of the present invention, the particle size trend highlighting module includes:

[0023] The particle size extraction submodule extracts the particle size record value corresponding to the time period number provided by the set of time periods with consistent direction, divides the time period into multiple particle size data groups according to the time period number, calculates the mean of the particle size record value of each group, establishes a list of particle size mean values ​​corresponding to the time period, and obtains a segmented particle size mean value sequence.

[0024] The direction determination submodule performs a difference calculation on each pair of adjacent average particle size values ​​based on the segmented particle size mean sequence, determines the difference result, marks the change direction between adjacent groups segment by segment, generates a direction change sequence in sequence, and obtains the particle size change direction label.

[0025] The continuous screening submodule calls the particle size change direction label, retrieves time periods with consistent direction labels, records the start and end numbers and calculates the corresponding duration, filters and retains time periods with durations not less than a preset reference value, and establishes a sequence of stable particle size change segments.

[0026] As a further aspect of the present invention, the formula for the weighted equilibrium value index is as follows:

[0027] ;

[0028] in, Representing the The equilibrium value of particle size variation over a time period. Representing the In the time period of the th time period Individual particle size observations Representing the In the time period of the th time period The time-weighted coefficients corresponding to each particle size observation. Representing the In the time period of the th time period The morphological factor of each particle size record Representing the The total number of particle size observations in each time period Representing the The average of all particle size observations over a given time period.

[0029] As a further aspect of the present invention, the velocity-pressure offset decomposition module includes:

[0030] The data extraction submodule calls each time range in the particle size stable change segment sequence, extracts the rotational speed record value and pressure record value in the corresponding time period, arranges the data in chronological order, calculates the numerical difference between the time point and the previous time point, establishes the change results of the two types of records, and obtains a list of velocity and pressure change differences.

[0031] The difference generation submodule, based on the velocity-pressure change difference list, takes the rotational speed change result and pressure change result as parameters, calculates the difference item by item in time order, constructs a new numerical column of the difference between the two types of data changes, and marks the direction of the difference to obtain the velocity-pressure difference direction column.

[0032] The consistency matching submodule determines whether the direction is consistent with the particle size change direction in the corresponding time period of the particle size stable change segment sequence based on the velocity-pressure difference direction column. It records the consistency of time points and counts the proportion of consistency points. It constructs records according to the time period number and generates a rotation pressure offset matching proportion column.

[0033] As a further aspect of the present invention, the system also includes a trend time period level division module:

[0034] The trend period classification module divides and sorts the proportion of the rotational pressure offset matching into intervals, and combines the particle size change direction, interval label and time range to generate the metal powdering process optimization results.

[0035] The optimization results of the metal grinding process include particle size change trend level, corresponding time period, and trend direction classification.

[0036] As a further aspect of the present invention, the trend time period classification module includes:

[0037] The interval division submodule calls all the proportional values ​​in the rotation pressure offset matching ratio column, sorts them by value size, divides them into three non-overlapping intervals, marks each proportional value according to its interval, records the interval label number of the corresponding time period, and generates the interval label distribution result.

[0038] The sequential sorting submodule extracts the marked time period numbers under the intervals based on the interval label distribution results, arranges the record points in each type of interval in chronological order, records the start and end time points and summarizes them by number to obtain a time interval sorting list.

[0039] The structure combination submodule calls the time interval sorting list, extracts the particle size change direction and interval label number corresponding to each time period, merges the three pieces of information row by row and establishes a data mapping structure, outputs the combination records in sequence, and establishes the metal powdering process optimization results.

[0040] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0041] In this invention, a unified time index is constructed by extracting pressure abrupt change points to achieve precise alignment of multi-source data such as energy consumption, temperature, and particle size in the time dimension. Combined with directional consistency analysis, linkage time periods are screened. Furthermore, stable trend segments are identified based on the change in average particle size. The relationship between velocity and pressure difference is decomposed by offset and directional matching is quantified to construct a trend response index system. Finally, trend levels are divided according to the degree of matching and time and directional information are integrated to achieve the aggregated expression of multi-variable linkage features and the hierarchical presentation of particle size change trends, thereby improving the process identification accuracy and control response capability under complex working conditions. Attached Figure Description

[0042] Figure 1 This is a system flowchart of the present invention;

[0043] Figure 2 This is a system block diagram of the present invention;

[0044] Figure 3 This is a flowchart of the pressure change recognition module of the present invention;

[0045] Figure 4 This is a flowchart of the multi-source data sequence registration module of the present invention;

[0046] Figure 5 This is a flowchart of the particle size trend highlighting module of the present invention;

[0047] Figure 6 This is a flowchart of the velocity-pressure offset decomposition module of the present invention;

[0048] Figure 7 This is a flowchart of the trend time period level classification module of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0050] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0051] Please see Figure 1 and Figure 2 The data analysis-based metal powdering process optimization system includes:

[0052] The pressure change identification module collects a continuous pressure record sequence in the pressure regulation cavity, divides it into equal-length time periods, calculates the difference between the first and last values ​​in each segment and compares it with the difference between adjacent segments. If the current difference is greater than twice the average difference between adjacent segments, the time index is recorded and all segments that meet the conditions are summarized to generate a pressure change time marker column.

[0053] The multi-source data sequence registration module extracts energy consumption data and temperature data for the corresponding time period based on all time indices of the pressure change time marker column, aligns the data by time and calculates the difference between adjacent points, marks the directional change results, filters three types of time periods with consistent directions and numbers them, and generates a set of time periods with consistent directions.

[0054] The particle size trend highlighting module obtains particle size record values ​​and calculates the mean of each group based on the time period number provided by the set of time periods with consistent direction. After performing the difference operation on two consecutive groups of mean values, the change direction is determined, and the time periods with consistent direction and duration not less than the preset reference value are extracted to generate a sequence of stable particle size change segments.

[0055] The velocity-pressure offset decomposition module calls each time range in the stable particle size change segment sequence, extracts the rotational velocity and pressure records, calculates the difference between adjacent points, calculates the difference between the two types of differences in turn and generates a new column, determines whether the direction of difference change is consistent with the direction of particle size change, calculates the proportion of consistent records, and generates a rotational pressure offset matching proportion column.

[0056] The trend period level division module calls all the proportion values ​​in the rotating pressure offset matching percentage column, divides them into three non-overlapping value intervals, sorts the recorded points in each interval by time, and combines and organizes the particle size direction, interval label and time range to generate the metal powdering process optimization results.

[0057] The pressure mutation time marker column includes the mutation time point index, corresponding time period number, and mutation intensity level. The set of time periods with consistent direction includes the energy consumption change direction, temperature change direction, and unified number identifier. The particle size stable change segment sequence includes the particle size mean sequence, change direction label, and duration. The rotation pressure offset matching ratio column includes the velocity difference sequence, pressure difference sequence, and direction consistency ratio. The metal grinding process optimization results include the particle size change trend level, corresponding time segment, and trend direction classification.

[0058] Please see Figure 3 and Figure 2 The pressure change recognition module includes:

[0059] The pressure acquisition submodule acquires continuous pressure records in the pressure regulation cavity, divides the records into multiple time periods of equal duration, extracts the starting and ending pressures of each time period, calculates the pressure difference of the corresponding time period based on the starting and ending pressures, and obtains the segmented pressure change.

[0060] The pressure acquisition submodule collects continuous pressure records from the pressure regulation cavity. At the start of acquisition, the system first sets the sampling period, for example, to collect pressure values ​​every 100 milliseconds. The recorded results form a continuous pressure curve. The pressure data, evenly distributed on the time axis, is grouped by time intervals. For example, every 10 sampling points constitute a time interval, corresponding to a time interval length of 1000 milliseconds. Then, the first and last pressure values ​​are extracted from each interval, serving as the starting and ending pressures for that interval, respectively. For example, the pressure values ​​collected in the first interval might be 101.2, 101.5, 101.6, 102.0, and 10... Given values ​​of 2.1, 102.4, 102.6, 102.9, 103.1, and 103.3 kPa, the initial pressure of this segment is 101.2 kPa, and the final pressure is 103.3 kPa. Therefore, the pressure change of this segment is 103.3 minus 101.2, resulting in 2.1 kPa. This process is repeated for all time periods. Assuming there are 60 seconds of records, the total number of samples is 600, divided into 60 segments. The initial and final values ​​of each segment are extracted to obtain the pressure change for each segment. The data for each segment is recorded as a list of pressure differences for subsequent trend recognition tasks, ultimately forming a complete data sequence containing the pressure changes of each segment.

[0061] The difference calculation submodule is based on the segmented pressure change. For each time period, it calculates the pressure change, selects the average change level of the two segments as the benchmark data, compares the pressure change of the current time period with the benchmark data, determines whether it is greater than twice the benchmark data, and obtains the result of the sudden change trend comparison.

[0062] The specific formula for calculating the change in pressure is as follows:

[0063] ;

[0064] in, This represents the pressure change during the i-th time period. This represents the pressure value at the end of the i-th time interval. This represents the pressure value at the start of the i-th time interval. This represents the pressure change at the j-th sub-time point within the i-th time period. represents the pressure change weighting coefficient corresponding to the j-th sub-time point within the i-th time period, and nᵢ represents the number of sub-time points within the i-th time period. This represents a small positive correction value within the i-th time period;

[0065] Explanation of parameter acquisition and assignment, and calculation examples:

[0066] Set monitoring point A, and the pressure at the beginning of a certain time period is... (Measured air pressure, normal atmospheric pressure), the pressure at the end was (Measured value, acquired through a pressure sensor).

[0067] If data is collected at 5 sub-time points within this time period, then (Time interval every 1 minute).

[0068] The instantaneous pressure changes at each sub-time point are automatically collected by the sensor, and are as follows:

[0069]

[0070] Sub-time weights It is obtained through normalized calculation of the pressure change amplitude. The normalization method is as follows:

[0071] ;

[0072] The calculation yields:

[0073] Total absolute change: ;

[0074] The weights are as follows:

[0075] ;

[0076] Correction value Take the upper limit of the pressure measurement error, referencing the accuracy of a high-precision sensor of ±0.005 kPa, and take... .

[0077] Actual calculation process:

[0078] Pressure difference in main section:

[0079] ;

[0080] Sub-time weighted values:

[0081] ;

[0082] molecular:

[0083] ;

[0084] Denominator:

[0085] ;

[0086] Overall calculation:

[0087] ;

[0088] Benchmark value setting and determination of mutation trend:

[0089] The baseline value is the average ΔP of the previous two time periods. Assume the baseline ΔP value is:

[0090] ;

[0091] The mutation threshold is set at twice the baseline:

[0092] ;

[0093] Due to the current calculation results <2.30 kPa, therefore no mutation trend is judged.

[0094] Table data support:

[0095] Table 1. Pressure Change Data During Monitoring Period

[0096] Time period number (kPa) (kPa) Sub-time change (kPa) Weighting coefficient Correction value (kPa) 1 101.3 102.8 0.3，-0.2，0.5，0.1，-0.4 0.2，0.133，0.333，0.067，0.267 0.005

[0097] Table 1 lists the monitoring data and calculation components of the embodiments.

[0098] Benefit enhancement explanation:

[0099] The advantage of the formula lies in the introduction of sub-time-time changes. Coupled with the main segment difference calculation, the sensitivity to short-term high-frequency fluctuations is improved. Simultaneously, the correction value is combined with denominator square root processing. It effectively controls the distortion caused by abnormal fluctuations, thereby achieving higher accuracy in identifying sudden changes in complex and ever-changing pressure monitoring scenarios.

[0100] This result indicates that the calculated pressure change... If the pressure is below the baseline threshold of 2.30 kPa, it indicates that the pressure fluctuation within this period is within the normal range and there is no need to trigger an abnormal response. This value directly determines the result of the step as "not judged as a sudden trend". This judgment result can be used as the basis for subsequent data screening and equipment control.

[0101] The mutation identification submodule filters all time period indices that are identified as mutations based on the mutation trend comparison results, collects and organizes the time positions in order, establishes a unified identifier list, and obtains the pressure mutation time marker column.

[0102] The mutation identification submodule summarizes the time periods identified as mutations based on the mutation judgment results of all time periods. First, it scans the mutation flag bits of each segment. For example, segments 2, 5, and 12 are marked as mutation segments, corresponding to pressure changes of 5.2, 6.5, and 7.0 kPa respectively. The changes in adjacent segments are all below 2.5 kPa, meeting the mutation judgment criteria. The system extracts the segment numbers to form a mutation index list, such as {2, 5, 12}. Then, these indices are mapped to actual time points according to segment order. If each segment corresponds to 1 second, then the time points corresponding to these mutations are the 2nd second, 5th second, and 12th second, respectively. These time points are then organized into a time series for pressure mutations. The final output is a time stamp column for mutations arranged in chronological order, such as [2.0, 5.0, 12.0] seconds, for subsequent use by the system, such as event logging and exception archiving. If the mutation times are dense, for example, the mutation interval is less than 2 seconds, time merging processing can be considered for the mutation results. That is, only the initial time point of the mutation is retained in the stamp column. For example, if the continuous mutations occur at the 10th, 11th, and 12th seconds, only the 10th second is retained as the identifier. The merging rule can be set to record only one mutation time within 5 seconds to reduce redundant judgments and output the final version of the pressure mutation time stamp column.

[0103] Please see Figure 4 and Figure 2 The multi-source data sequence registration module includes:

[0104] The data extraction submodule extracts energy consumption data and temperature data within the corresponding time period based on all time indices of the pressure change time marker column. It then arranges the energy consumption data and temperature data point by point according to the time index, adjusts the start and end positions of the data to ensure consistency of time nodes, and obtains the synchronized data arrangement results.

[0105] Based on the complete time index of the pressure change time marker column, the energy consumption and temperature data for each change period are extracted one by one. First, the start and end positions of the specific data extraction are determined according to the change time index. For example, if the change time points are the 12th, 25th, and 38th seconds, the 120th, 250th, and 380th data points in the original energy consumption and temperature data sequences are located respectively. Then, a certain range is extended before and after each change point, such as 5 seconds before and after. The start and end points of each extraction segment are the 50 sampling points before and after the change point, respectively. Assuming the data sampling period is 100 milliseconds, each segment has 100 points, so the data index of the 12th second change segment is [70, 170]. The energy consumption data sequence E(t) is extracted for all change time periods in this way. The temperature data sequence T(t) is then used as the starting point. The extracted data segments are then uniformly arranged, and the sampling start and end points of each set of energy consumption and temperature data on the time axis are compared. If the start time of the two channels is found to be inconsistent in some segments, it is adjusted by padding or truncation. For example, if the starting point of energy consumption data is 70 and the starting point of temperature data is 72, two data points can be inserted forward using the previous value expansion method. If the starting point is 66 but data is missing, the temperature data is truncated from point 70 to align with the energy consumption data. The consistency adjustment of the ending position is also handled in the same way. After all segments are aligned, each pair of data is arranged into a binary tuple in chronological order, such as (E1, T1), (E2, T2), ..., (En, Tn), finally completing the synchronized arrangement data structure of energy consumption and temperature.

[0106] The difference determination submodule calculates the difference in energy consumption and temperature at continuous time points based on the synchronous data arrangement results, determines the positive and negative directions of each pair of differences, marks the directional status of the two data channels in the corresponding time period, establishes a directional correspondence table, and obtains the corresponding identifier of directional change.

[0107] Based on the above-mentioned synchronized data arrangement results, the changes in energy consumption and temperature between consecutive time points are calculated. Specifically, the data of two adjacent time points are subtracted to obtain the difference for each time period. For example, if the energy consumption values ​​for time points t1 and t2 are 2.8 kWh and 3.1 kWh, and the temperature values ​​are 45.3℃ and 46.2℃, then the change in energy consumption is 0.3 kWh and the change in temperature is 0.9℃. After performing the same operation on all data pairs, two difference sequences ΔE and ΔT are generated. Then, the positive and negative directions of each pair of differences are determined. If ΔE > 0 and ΔT > 0, then both are positive changes and are marked as "++"; if ΔE < 0 and ΔT < 0, then they are marked as "−−"; if the negative direction is not specified, then the positive direction is not specified. If the direction is opposite, it is marked as "+-" or "-+", thus encoding the directional relationship of each time period into four states. The determination of the directional state must clearly define the positive and negative range. For energy consumption changes, ±0.05kWh is set as the boundary. If the difference is below this value, it is considered a small fluctuation. For temperature changes, ±0.2℃ is set as the boundary. If a difference is 0.02kWh or 0.1℃, it is marked as "00", indicating that the direction is unknown. All determination results are organized in chronological order to form a directional correspondence table. For example, the first three segments are "++", "+-", and "00". The system will classify, label, and collect the directional information of each pair of differences into the directional identifier sequence to obtain the corresponding identifier for directional changes.

[0108] The consistent screening submodule filters time periods with consistent directional status based on the corresponding identifier of directional change, uniformly numbers the data segments with consistent directional status, collects the identifier values ​​in the order of numbering, removes segments with different directions, and establishes a set of time periods with consistent directional status.

[0109] Based on the corresponding identifiers for directional changes, the directional state of each segment is analyzed one by one, and time periods in which energy consumption and temperature change directions are consistent are selected. The judgment logic is to traverse the sequence of directional identifiers. Any segment marked with "++" or "−−" is determined to be a segment with consistent direction. Other identifiers such as "+−", "−+" and "00" are excluded. For example, if the directional identifier sequence is "++、++、−−、−+、00、−−", then the first two segments, the third segment, and the last segment are retained and numbered as 1, 2, 3, and 4, respectively. Segments with non-contiguous numbers are considered as the start of a new sequence. For example, segments 1 and 2 are continuous, segments after segment 3 are segments with inconsistent directions, and segment 4 is a new starting segment. After numbering, they are classified into sequences A=[1,2] and B=[4], which are the segments with consistent directions in segment A and segment with consistent directions in segment B, respectively. The corresponding original data positions are reorganized after being mapped by number. The time points, energy consumption, and temperature data in each segment will be labeled with the same number. The final output only contains a set of data segments with consistent directions, which can be used for subsequent behavioral trend statistics or further data summarization and analysis.

[0110] Please see Figure 5 and Figure 2The particle size trend highlighting module includes:

[0111] The particle size extraction submodule extracts the particle size record value corresponding to the time period number provided by the set of time periods with consistent direction, divides the data into multiple particle size data groups according to the time period number, calculates the weighted variation equilibrium value index, establishes a list of particle size mean values ​​corresponding to the time period, and obtains the segmented particle size mean value sequence.

[0112] Weighted equilibrium value index formula:

[0113] ;

[0114] in, Representing the The equilibrium value of particle size variation over a time period. Representing the In the time period of the th time period Individual particle size observations Representing the In the time period of the th time period The time-weighted coefficients corresponding to each particle size observation. Representing the In the time period of the th time period The morphological factor of each particle size record Representing the The total number of particle size observations in each time period Representing the The average of all particle size observations over a given time period;

[0115] In this formula:

[0116] The numerator is the weighted sum of the product of particle size, weighting factor, and morphology factor, and the difference between it and the square root term consisting of a set of squares. This is intended to retain the core equilibrium fluctuation trend after deducting non-equilibrium terms from the total amount.

[0117] The denominator is composed of the sum of particle size deviations and the square root of the sample size, thus avoiding outliers from having a disproportionate impact on the overall trend.

[0118] Using absolute values ​​to process the overall fractional results ensures that the final index does not result in a negative value, thus enhancing the physical interpretability of the results.

[0119] Parameter Acquisition and Numerical Examples

[0120] Taking actual monitoring as an example, a time period is selected. Total sampling The particle size values ​​were sampled at 9:00, 10:00, 11:00, 12:00, and 13:00, with sampling every 10 minutes. The particle size data were read by the device as follows:

[0121] Sampling time (h) Particle size value (μm) Weight morphological factors 9:00 2.0 0.15 0.105 10:00 2.5 0.25 0.263 11:00 1.8 0.10 0.211 12:00 3.2 0.30 0.421 13:00 2.7 0.20 0.158

[0122] Table 2. Particle Size Data Sampling and Weighting Table

[0123] Among them, morphological factors All are based on the aforementioned formula The average particle size is calculated to be:

[0124] ;

[0125] Substitute calculation (Taking the first item as an example):

[0126] ;

[0127] And so on, all morphological factors are derived.

[0128] Substitute all the values ​​into the formula and calculate each part:

[0129] First term of the molecule:

[0130] ;

[0131] The square root of the second term in the numerator:

[0132] ;

[0133] Denominator:

[0134] ;

[0135] final:

[0136] ;

[0137] Interpretation and Application of Results

[0138] The results show that the particle size weighted variation equilibrium value in the first time period is 2.606. This value deviates significantly from the locally measured stable range [0.8, 1.5], indicating that the particle size distribution fluctuates significantly during this time period and has not reached a stable state. This needs to be given special attention in subsequent calculations.

[0139] The advantage of the formula lies in the introduction of morphological factors. Quantify the deviation of each particle size point from its time-limited average particle size, and correlate it with time weights. Simultaneous superposition allows the formula to not only consider the size of the particle itself, but also effectively identify and enhance the impact of data points with large fluctuations or high sampling weights on the overall trend assessment. This helps to improve the identification accuracy of particle size change trend analysis and enhance the overall particle size monitoring system's ability to identify abnormal periods.

[0140] The direction determination submodule performs a difference calculation on the average particle size of each pair of adjacent groups based on the segmented particle size mean sequence, determines the difference result, marks the change direction between adjacent groups segment by segment, generates a direction change sequence in sequence, and obtains the particle size change direction label.

[0141] Based on the established segmented average particle size sequence, the difference between the average particle size of adjacent time periods is calculated to determine whether the difference is positive or negative, thereby determining the direction of particle size change. For example, if the average particle size sequence is [12.15, 12.10, 12.12, 12.35, 12.50], then the difference between the 4th and 3rd segments is 12.10 − 12.15 = −0.05 micrometers, recorded as a decrease; the difference between the 5th and 4th segments is 12.12 − 12.10 = +0.02 micrometers, recorded as an increase; the difference between the 9th and 5th segments is 12.35 − 12.12 = +0.23 micrometers, recorded as an increase; and the difference between the 10th and 9th segments is 12.50 − 12.35 = +0.02 micrometers. 0.15 micrometers is recorded as an increase. The system summarizes the above judgment results into a direction label sequence in chronological order, with corresponding labels of [decline, increase, increase, increase]. All judgment processes must clearly distinguish the magnitude of the difference. The case where the absolute value of the difference is less than 0.01 micrometers is considered stable and unchanged. If the difference is 0.00, +0.005, −0.008, etc., it is marked as "flat". The critical value of 0.01 micrometers for this stability judgment comes from the minimum resolution unit of the detection instrument and the industry's allowable fluctuation range. The sensor accuracy is generally ±0.005 micrometers. Therefore, taking twice the accuracy as the stability boundary is a reasonable setting. After segment-by-segment difference comparison and judgment, the particle size change direction labels are output in sequence.

[0142] The continuous screening submodule calls the particle size change direction label, retrieves the time period with the same continuous direction label, records the start and end numbers and calculates the corresponding duration, filters and retains the time period number with a duration not less than the preset reference value, and establishes a sequence of stable particle size change segments.

[0143] The system invokes particle size change direction labels to identify time period groups with consistent directions. It iterates through the label sequence, counting the number of consecutive "rising" or "falling" segments and recording their start and end numbers. For example, if the direction labels are [falling, rising, rising, rising], the system records the "rising" state starting from the second item. It finds that segments 2-4 are consecutively rising, with a start number of 4 and an end number of 10, corresponding to time period numbers [4, 5, 9, 10]. The number of continuous segments is 4. Combining this with the fact that each segment lasts 1 second, the duration of this consecutive rising segment is 4 seconds. The system then determines whether this exceeds a preset reference value. The threshold is set to 3 seconds. If the system presets a stable direction reference threshold of 3 seconds, the current segment duration meets the condition and is retained; otherwise, it is discarded. The 3-second threshold is determined based on the sampling frequency and the minimum response period for particle size control in engineering applications. Typically, in industrial applications, particle size changes lasting more than 3 seconds are meaningful for identification. If it is set to 1 second, misjudgment may occur; if it is set to 5 seconds, valid segments may be missed. Therefore, setting the value to 3 seconds can cover most scenarios. Through this process, all segments with continuous and consistent directions are screened one by one, and finally, the time periods that meet the duration condition are numbered and merged to form a sequence of stable particle size change segments.

[0144] Please see Figure 6 and Figure 2 The velocity-pressure offset decomposition module includes:

[0145] The data extraction submodule calls each time range in the stable particle size change sequence, extracts the rotational velocity and pressure records within the corresponding time period, arranges the data in chronological order, calculates the numerical difference between the time point and the previous time point, establishes the change results of the two types of records, and obtains a list of velocity and pressure change differences.

[0146] The sequence of stable particle size variation segments is used to locate the start and end times of each segment. Within this time range, the rotational speed and pressure values ​​are recorded. The data for each segment are arranged in ascending order of time. For example, the fifth segment, the stable particle size interval, corresponds to 12.0 to 16.0 seconds, with a sampling period of 0.5 seconds. Therefore, the data points to be extracted are [12.0, 12.5, 13.0, ..., 16.0], a total of 9 time points. If the rotational speed values ​​are 1300, 1320, 1350, 1370, 1385, 1400, 1410, 1420, and 1430 rpm, and the pressure records are 105.5, 106.0, 106.8, 107.3, 107.6, 107.9, and 108.1 rpm, respectively. For values ​​of 108.2 and 108.4 kPa, time difference calculations are performed on each type of data, calculating the change between two adjacent sampling points. For example, the difference between the speed at the second time point and the speed at the first time point is 1320-1300=20 rpm, and the difference at the third time point is 1350-1320=30 rpm. This process is repeated for all points, resulting in pressure differences of 0.5, 0.8, 0.5, 0.3, 0.3, 0.2, 0.1, and 0.2 kPa. All changes in rotational speed and pressure are arranged side-by-side in chronological order to generate one-to-one difference data. For example, the first difference pair is (20, 0.5), and the second is (30, 0.8). All difference data are finally aggregated to form a speed change difference sequence and a pressure change difference sequence, constructing a speed-pressure change difference list.

[0147] The difference generation submodule is based on the list of velocity and pressure change differences. It takes the results of rotational velocity change and pressure change as parameters, calculates the difference item by item in time order, constructs a new numerical column of the difference between the two types of data changes, and marks the direction of the difference to obtain the velocity and pressure difference direction column.

[0148] Based on the velocity-pressure change difference list constructed above, the rotational velocity change results and pressure change results at the same time point are matched one-to-one, and the difference is calculated item by item to construct a difference value column between the two types of data. That is, the rotational velocity change value in each data set is subtracted from the corresponding pressure change value to form the difference value. For example, the data pairs (20, 0.5), (30, 0.8), (20, 0.5), and (15, 0.3) have differences of 19.5, 29.2, 19.5, and 14.7, respectively. The direction of the difference is determined by the sign of the difference value. If the result is positive, it means that the increase in rotational velocity is greater than the increase in pressure, and the direction is marked as "velocity-dominated". If it is negative, it means that the pressure change is more drastic. The direction is labeled "pressure-dominant". If the absolute value of the difference is less than the preset direction threshold, such as 1.0, it means that the two changes are basically the same and the direction is labeled "synchronous". The threshold is set to 1.0, which is derived from the discrete standard deviation calculation of the experimental data. The standard deviation of velocity change in the sample is 7.5 rpm and the standard deviation of pressure change is 0.4 kPa. After unit conversion, the equivalent comparison constant ratio of the two is 10. The difference within 1.0 within the standard deviation range is considered to be consistent. The final direction label sequence may be "velocity-dominant, velocity-dominant, velocity-dominant, velocity-dominant". If there is a violent fluctuation between adjacent values, it is also marked by setting a direction mutation screening mechanism. Finally, the velocity and pressure difference direction column is output.

[0149] The consistency matching submodule determines whether the direction of the velocity-pressure difference direction is consistent with the particle size change direction of the corresponding time period in the sequence of stable particle size change segments. It records the consistency of time points and counts the proportion of consistency points. It constructs records by time period number and generates a rotational pressure offset matching proportion column.

[0150] Based on the velocity-pressure difference direction column, the direction label corresponding to each time point is compared line by line to determine whether it is consistent with the particle size change direction recorded in the stable particle size change segment. For example, if the particle size direction is "rising", the system will determine the difference direction of each point in that time period one by one. If it is "velocity-dominated" and the particle size is also in the growth direction, it is marked as a consistent time point. If it is "pressure-dominated" or "synchronous", it is considered inconsistent. Consistent time points are recorded one by one, and their total percentage in the current segment is counted. For example, if a segment has 8 time points, 6 of them are consistent. If the particle size direction is consistent with the particle size distribution, the matching percentage for that segment is 6 / 8 = 0.75, rounded to two decimal places. The criterion for judgment is a one-to-one comparison and a sign consistency check. If the particle size is decreasing, it is only considered consistent if the difference direction column is "velocity-dominated" and the direction is decreasing. Otherwise, it is considered inconsistent. After the comparison is completed, a corresponding matching percentage record row is generated using the time period number as the index. The content of each record is in the form of segment number 5, percentage 0.75, segment number 6, percentage 0.33, etc. After being summarized in order, they form the rotational pressure offset matching percentage column.

[0151] Please see Figure 7 and Figure 2 The trend time period classification module includes:

[0152] The interval division submodule calls all the proportional values ​​in the rotating pressure offset matching percentage column, sorts them by value size, divides them into three non-overlapping intervals, marks each proportional value according to its interval, records the interval label number of the corresponding time period, and generates the interval label distribution result.

[0153] The algorithm retrieves all proportional values ​​from the rotational pressure offset matching column, extracts all values, and sorts them in ascending order. Given 10 data points: 0.25, 0.33, 0.38, 0.45, 0.50, 0.57, 0.63, 0.70, 0.75, and 0.82, the sorted results are divided into three non-overlapping intervals: the first third is the "low percentage interval," the middle third is the "medium percentage interval," and the last third is the "high percentage interval." The boundaries are determined by the number and position of the sorted data: the first three data points form the first group, the middle four and six the second group, and the last seven and ten the third group. Therefore, the low interval is [0.2]. The original ratio values ​​are 0.33, 0.38, [0.45, 0.50, 0.57], [0.63, 0.70, 0.75, 0.82], and the middle interval is [0.45, 0.50, 0.57]. The high interval is [0.63, 0.70, 0.75, 0.82]. Each original ratio value will be labeled according to its interval, with the three interval numbers represented by numbers 1, 2, and 3 respectively. For example, the ratio value of the second segment is 0.33, which corresponds to the low interval and is recorded as label 1; the ratio value of the fifth segment is 0.50, which falls within the middle interval and is recorded as label 2; and the ratio value of the tenth segment is 0.82, which belongs to the high interval and is recorded as label 3. Finally, a one-to-one correspondence is established between all time interval numbers and their labels, and the records are recorded in a structure such as {segment 2: 1, segment 5: 2, segment 10: 3}, generating a complete interval label distribution result.

[0154] The sequential sorting submodule extracts the time period numbers marked under the intervals based on the interval label distribution results, arranges the record points in each type of interval in chronological order, records the start and end time points and summarizes them by number, and obtains a sorted list of time intervals.

[0155] Based on the above interval label distribution results, all time period numbers with labels are grouped according to interval category. Within each group, the time period numbers are then sorted in ascending order of chronological sequence. For example, under label 1, the segment numbers are [1, 3, 7], under label 2 they are [2, 5, 8], and under label 3 they are [4, 6, 9, 10]. The sorting result within each group represents the sequential structure of the corresponding time periods within that interval. The starting and ending segments of each group are extracted, and their corresponding time point information is used to construct an interval time sequence structure table. For example, if label 1's segments are segments 1 to 7, then... The time points range from 12 seconds to 24 seconds. Label 2 corresponds to segments 2 to 8, with times ranging from 16 seconds to 32 seconds. Label 3 corresponds to segments 4 to 10, with times ranging from 24 seconds to 40 seconds. The system summarizes the time period number, start and end number, and start and end time points in each interval, forming a format as follows: label number, time period start number, time period end number, start time, end time. For example, the record item is {label 2: segment 2 segment 8, time 16s 32s}. By sequentially completing the sorting and structuring of all interval data, a sorted list of time intervals is obtained.

[0156] The structural combination submodule calls the time interval sorting list, extracts the particle size change direction and interval label number corresponding to each time period, merges the three pieces of information by row and establishes a data mapping structure, outputs the combination records in sequence, and establishes the metal powdering process optimization results.

[0157] The system calls the time interval sorting list and reads the segment information within each time interval number range one by one. It extracts the particle size change direction and the marked interval label number from the original data structure, and integrates the three items by row: segment number, particle size change direction, and interval label number. For example, the particle size direction of segment 5 is "rising", the interval label is 2, and the record item is {segment 5, rising, 2}. The system processes all time interval numbers in sequence, keeping the time order unchanged, and outputs all structural information as a combined data record structure. Each record contains three attributes: time interval number, direction status, and percentage interval. The combined structure can be used for subsequent process optimization reference. Finally, the system outputs a complete set of data rows in the record order to establish the metal powdering process optimization results.

[0158] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A data analysis-based metal powdering process optimization system, characterized in that, The system includes: The pressure change identification module collects the continuous pressure sequence in the pressure regulation cavity, divides it into equal time periods, calculates the difference between the beginning and end of each segment, compares it with the average of adjacent segments, and if the current difference exceeds twice the average of adjacent segments, it records the time index, summarizes all segments that meet the conditions, and generates a pressure change time marker column. The multi-source data sequence registration module extracts the corresponding energy consumption and temperature data based on the pressure change time marker column, aligns the time axis and calculates adjacent differences, marks the directional changes, filters time periods with the same direction and numbers them, forming a set of time periods with the same direction. The particle size trend highlighting module uses the set of time periods with consistent direction to obtain particle size data, calculates the mean by segment, compares the direction of two consecutive sets of mean values, extracts the time periods with consistent direction and sufficient duration, and generates a sequence of stable particle size change segments. The velocity-pressure offset decomposition module extracts the rotational velocity and pressure data for the corresponding time period from the sequence of stable particle size changes, calculates the difference between adjacent values, calculates the difference between the two to generate a new column, determines whether the direction of change is consistent with the particle size, calculates the consistency ratio, and generates a rotational pressure offset matching ratio column.

2. The metal powdering process optimization system based on data analysis according to claim 1, characterized in that: The pressure change time marker column includes the change time point index, the corresponding time period number, and the change intensity level. The set of directional consistent time periods includes the energy consumption change direction, the temperature change direction, and a unified number identifier. The particle size stable change segment sequence includes the particle size mean sequence, the change direction label, and the duration length. The rotational pressure offset matching ratio column includes the velocity difference sequence, the pressure difference sequence, and the directional consistency ratio.

3. The data analysis-based metal powdering process optimization system according to claim 1, characterized in that, The pressure surge detection module includes: The pressure acquisition submodule acquires continuous pressure records in the pressure regulation cavity, divides the records into multiple time periods of equal duration, extracts the starting and ending pressures of each time period, calculates the pressure difference of the corresponding time period based on the starting and ending pressures, and obtains the segmented pressure change. The difference calculation submodule, based on the segmented pressure change, selects the pressure change of adjacent time periods for each time period, calculates the average change level of the two periods as the benchmark data, compares the pressure change of the current time period with the benchmark data, determines whether it is greater than twice the benchmark data, and obtains the abrupt change trend comparison result. The mutation identification submodule filters all time period indices that are determined to be mutations based on the mutation trend comparison results, collects and organizes the time positions in order, establishes a unified identifier list, and obtains the pressure mutation time marker column.

4. The data analysis-based metal powdering process optimization system according to claim 3, characterized in that, The specific formula for calculating the pressure change is as follows: ； in, This represents the pressure change during the i-th time period. This represents the pressure value at the end of the i-th time interval. This represents the pressure value at the start of the i-th time interval. This represents the pressure change at the j-th sub-time point within the i-th time period. This represents the pressure change weighting coefficient corresponding to the j-th sub-time point within the i-th time period. This represents the number of sub-time points within the i-th time period. This represents the positive correction value within the i-th time period.

5. The data analysis-based metal powdering process optimization system according to claim 1, characterized in that, The multi-source data sequence registration module includes: The data extraction submodule extracts energy consumption data and temperature data within the corresponding time period based on all time indices of the pressure change time marker column, arranges the energy consumption data and temperature data point by point according to the time index order, adjusts the start and end positions of the data to ensure consistency of time nodes, and obtains the synchronous data arrangement result. The difference determination submodule calculates the energy consumption change difference and temperature change difference at continuous time points based on the synchronous data arrangement result, determines the positive and negative directions of each pair of differences, marks the directional status of the two data channels in the corresponding time period, establishes a directional correspondence table, and obtains the directional change corresponding identifier. The consistent screening submodule filters time periods with consistent directional states based on the corresponding identifiers of the directional changes, uniformly numbers consecutive data segments with consistent directions, collects the identifier values ​​in order of numbering, removes segments with different directions, and establishes a set of time periods with consistent directions.

6. The data analysis-based metal powdering process optimization system according to claim 1, characterized in that, The particle size trend highlighting module includes: The particle size extraction submodule extracts the particle size record value corresponding to the time period number provided by the set of time periods with consistent direction, divides the data into particle size data groups according to the time period number, calculates the mean of the particle size record value of each group, establishes a list of particle size mean values ​​corresponding to the time period, and obtains a segmented particle size mean value sequence. The direction determination submodule performs a difference calculation on each pair of adjacent average particle size values ​​based on the segmented particle size mean sequence, determines the difference result, marks the change direction between adjacent groups segment by segment, generates a direction change sequence in sequence, and obtains the particle size change direction label. The continuous screening submodule calls the particle size change direction label, retrieves time periods with consistent direction labels, records the start and end numbers and calculates the corresponding duration, filters and retains time periods with durations not less than a preset reference value, and establishes a sequence of stable particle size change segments.

7. The data analysis-based metal powdering process optimization system according to claim 6, characterized in that, The formula for the weighted equilibrium value index is as follows: ； in, Representing the The equilibrium value of particle size variation over a time period. Representing the In the time period of the th time period Individual particle size observations Representing the In the time period of the th time period The time-weighted coefficients corresponding to each particle size observation. Representing the In the time period of the th time period The morphological factor of each particle size record Representing the The total number of particle size observations in each time period Representing the The average of all particle size observations over a given time period.

8. The data analysis-based metal powdering process optimization system according to claim 1, characterized in that, The velocity-pressure offset decomposition module includes: The data extraction submodule calls each time range in the particle size stable change segment sequence, extracts the rotational speed record value and pressure record value in the corresponding time period, arranges the data in chronological order, calculates the numerical difference between the time point and the previous time point, establishes the change results of the two types of records, and obtains a list of velocity and pressure change differences. The difference generation submodule, based on the velocity-pressure change difference list, takes the rotational speed change result and pressure change result as parameters, calculates the difference item by item in time order, constructs a new numerical column of the difference between the two types of data changes, and marks the direction of the difference to obtain the velocity-pressure difference direction column. The consistency matching submodule determines whether the direction is consistent with the particle size change direction in the corresponding time period of the particle size stable change segment sequence based on the velocity-pressure difference direction column. It records the consistency of time points and counts the proportion of consistency points. It constructs records according to the time period number and generates a rotation pressure offset matching proportion column.

9. The data analysis-based metal powdering process optimization system according to claim 1, characterized in that, The system also includes a trend time period classification module: The trend period classification module divides and sorts the proportion of the rotational pressure offset matching into intervals, and combines the particle size change direction, interval label and time range to generate the metal powdering process optimization results. The optimization results of the metal grinding process include particle size change trend level, corresponding time period, and trend direction classification.

10. The metal powdering process optimization system based on data analysis according to claim 1, characterized in that, The trend time period classification module includes: The interval division submodule calls all the proportional values ​​in the rotation pressure offset matching ratio column, sorts them by value size, divides them into three non-overlapping intervals, marks each proportional value according to its interval, records the interval label number of the corresponding time period, and generates the interval label distribution result. The sequential sorting submodule extracts the marked time period numbers under the intervals based on the interval label distribution results, arranges the record points in each type of interval in chronological order, records the start and end time points and summarizes them by number to obtain a time interval sorting list. The structure combination submodule calls the time interval sorting list, extracts the particle size change direction and interval label number corresponding to each time period, merges the three pieces of information row by row and establishes a data mapping structure, outputs the combination records in sequence, and establishes the metal powdering process optimization results.

Images