Intelligent analysis and early warning methods, systems, equipment and media for production management data

Abstract

This invention provides a method, system, device, and medium for intelligent analysis and early warning of production management data, belonging to the field of production process monitoring technology. The method includes: acquiring indicator data for each production management indicator within the current time window and constructing a normalized production management state vector; calculating correlation coefficients and time phase offsets based on the normalized production management state vector to construct a coupling analysis matrix, and then correcting it to obtain a corrected coupling analysis matrix; calculating the production state balance degree and calculating the lower bound of the dynamic early warning threshold; comparing the production state balance degree with the lower bound of the dynamic early warning threshold, and when the production state balance degree is less than the lower bound of the dynamic early warning threshold, calculating a sensitivity value based on the normalized production management state vector and the corrected coupling analysis matrix, determining the early warning indicator, and outputting the corresponding early warning result. This invention achieves quantitative identification of the overall operational imbalance state of production management and precise location of key abnormal indicators.

Description

Technical Field

[0001] This invention relates to the field of production process monitoring technology, specifically to a method, system, equipment, and medium for intelligent analysis and early warning of production management data. Background Technology

[0002] In existing production management systems, anomaly detection typically involves statistical analysis of individual production management indicators such as output achievement rate, quality pass rate, inventory level, and equipment utilization rate, combined with fixed thresholds or empirical rules. When an indicator exceeds a set range, the system triggers an alert, requiring manual analysis and handling by management personnel. This method is simple to implement and had some practical value in the early stages of information technology development.

[0003] However, as production organization models gradually shift towards multi-variety, small-batch, and dynamic scheduling, the correlation and coupling between various production management indicators have significantly increased. An abnormality in a single indicator does not necessarily indicate an imbalance in production, while slight synchronous shifts in multiple indicators may reflect structural risks at the system level. Most existing methods are based on univariate threshold judgments or simple weighted scoring models, making it difficult to characterize the dynamic relationships between indicators and to identify false anomalies caused by factors such as changes in statistical methods and differences in data collection timing, resulting in high false alarm rates or delayed warnings.

[0004] Furthermore, in actual operation, the data update times of various business systems are inconsistent, and there are time misalignments between different indicators, making direct correlation analysis prone to bias. Simultaneously, when the external environment changes or production rhythms are adjusted, some indicators may exhibit trend drift, which traditional models often misinterpret as abnormal fluctuations, lacking a mechanism to identify structural changes. Therefore, how to construct an analytical method that can comprehensively reflect the overall stability of production status under multi-indicator coupling, and on this basis, achieve more accurate early warning judgments, has become a pressing technical problem to be solved in the field of production management technology. Summary of the Invention

[0005] The purpose of this invention is to provide a method, system, device and medium for intelligent analysis and early warning of production management data, so as to at least solve the problem that the prior art is difficult to reflect the dynamic coupling relationship of multiple production management indicators and is prone to misjudgment in the presence of time misalignment and structural drift.

[0006] To achieve the above objectives, the first aspect of the present invention provides a method for intelligent analysis and early warning of production management data. The method includes: acquiring indicator data for each production management indicator within the current time window, and constructing a normalized production management state vector based on the indicator data; calculating a correlation coefficient and a time phase offset based on the normalized production management state vector to construct a coupling analysis matrix, and correcting the coupling analysis matrix based on a structure drift index to obtain a corrected coupling analysis matrix; calculating a production state balance degree based on the normalized production management state vector and the corrected coupling analysis matrix, and calculating a lower bound of a dynamic early warning threshold based on the production state balance degree within a historical time window; comparing the production state balance degree with the lower bound of the dynamic early warning threshold, and when the production state balance degree is less than the lower bound of the dynamic early warning threshold, calculating a sensitivity value based on the normalized production management state vector and the corrected coupling analysis matrix, determining an early warning indicator, and outputting the corresponding early warning result.

[0007] Optionally, constructing a normalized production management state vector based on the indicator data includes: statistically analyzing historical indicator data corresponding to each production management indicator within a preset normalized historical time interval, and determining the historical lower bound and historical upper bound of each production management indicator; acquiring indicator data of each production management indicator within the current time window, and performing interval mapping processing on the indicator data and the corresponding historical lower bound and historical upper bound to generate corresponding normalized indicator values; and combining the normalized indicator values ​​according to a predetermined arrangement order of production management indicators to construct the normalized production management state vector.

[0008] Optionally, the process of calculating correlation coefficients and time phase offsets based on the normalized production management state vectors to construct a coupling analysis matrix includes: within a preset analysis time interval, obtaining the normalized production management state vectors corresponding to each time window, and calculating the correlation coefficient between any two production management indicators based on the normalized index values ​​at corresponding positions within each time window; for any two production management indicators, performing time offset processing on one of the production management indicators within a preset time offset interval, and recalculating the correlation coefficient between the two production management indicators under each time offset; determining the time offset with the largest absolute value of the correlation coefficient within the time offset interval as the time phase offset between the two production management indicators, and determining the corresponding correlation coefficient as the corrected correlation coefficient; and constructing the coupling analysis matrix based on the corrected correlation coefficients between each production management indicator and the corresponding time phase offsets.

[0009] Optionally, the coupling analysis matrix is ​​modified based on the structural drift index to obtain a modified coupling analysis matrix, including: within a preset long-term statistical time interval, calculating the historical mean and historical fluctuation values ​​corresponding to each production management indicator, and determining the long-term statistical characteristics of each production management indicator based on the historical mean and historical fluctuation values; within the current analysis time interval, calculating the current mean corresponding to each production management indicator, and calculating the structural drift index of each production management indicator based on the degree of difference between the current mean and the corresponding historical mean and the historical fluctuation value; determining whether the structural drift index corresponding to each production management indicator exceeds a preset drift judgment threshold, and performing attenuation processing on the correlation coefficients corresponding to production management indicators whose structural drift index exceeds the drift judgment threshold; and reconstructing the coupling analysis matrix based on the attenuated correlation coefficients to obtain the modified coupling analysis matrix.

[0010] Optionally, calculating the production state balance degree based on the normalized production management state vector and the modified coupling analysis matrix, and calculating the lower bound of the dynamic early warning threshold based on the production state balance degree within the historical time window, includes: performing matrix mapping processing on the normalized production management state vector based on the modified coupling analysis matrix to obtain the corresponding coordinated state vector; calculating the difference between the normalized production management state vector and the coordinated state vector to obtain the production state misalignment measure, and determining the production state balance degree based on the production state misalignment measure; obtaining the production state balance degree corresponding to each time window within a preset balance degree statistical time interval, and performing statistical processing on the production state balance degree corresponding to each time window to obtain the historical balance degree mean and historical balance degree fluctuation value; and determining the lower bound of the dynamic early warning threshold based on the historical balance degree mean and the historical balance degree fluctuation value.

[0011] Optionally, the production status balance is compared with the lower bound of the dynamic early warning threshold. When the production status balance is less than the lower bound of the dynamic early warning threshold, a sensitivity value is calculated based on the normalized production management status vector and the modified coupling analysis matrix, and an early warning index is determined, and the corresponding early warning result is output. This includes: when the production status balance is less than the lower bound of the dynamic early warning threshold, calculating the sensitivity value corresponding to each production management index based on the normalized production management status vector and the modified coupling analysis matrix; comparing the absolute values ​​of each sensitivity value, and determining the production management index with the largest absolute value as the early warning index; generating corresponding early warning level information based on the early warning index, and outputting the early warning result.

[0012] Optionally, calculating the sensitivity value corresponding to each production management indicator based on the normalized production management state vector and the modified coupling analysis matrix includes: performing matrix mapping processing on the normalized production management state vector based on the modified coupling analysis matrix to obtain a coordinated state vector; calculating the difference between the normalized production management state vector and the coordinated state vector to obtain an imbalance propagation vector; performing backpropagation processing on the imbalance propagation vector based on the modified coupling analysis matrix to obtain an indicator propagation contribution vector; and determining the sensitivity value corresponding to each production management indicator based on the numerical value corresponding to each position in the indicator propagation contribution vector.

[0013] A second aspect of this invention provides an intelligent analysis and early warning system for production management data. The system includes: a first module for acquiring indicator data of various production management indicators within the current time window, and constructing a normalized production management state vector based on the indicator data; a second module for calculating correlation coefficients and time phase offsets based on the normalized production management state vector to construct a coupling analysis matrix, and correcting the coupling analysis matrix based on a structure drift index to obtain a corrected coupling analysis matrix; a third module for calculating the production state balance degree based on the normalized production management state vector and the corrected coupling analysis matrix, and calculating a lower bound of a dynamic early warning threshold based on the production state balance degree within a historical time window; and a fourth module for comparing the production state balance degree with the lower bound of the dynamic early warning threshold, and when the production state balance degree is less than the lower bound of the dynamic early warning threshold, calculating a sensitivity value based on the normalized production management state vector and the corrected coupling analysis matrix, determining an early warning indicator, and outputting the corresponding early warning result.

[0014] A third aspect of the present invention provides an electronic device, comprising: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the above-described intelligent analysis and early warning method for production management data.

[0015] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the above-described intelligent analysis and early warning method for production management data.

[0016] Through the above technical solution, the present invention constructs a normalized production management state vector, uniformly mapping multiple production management indicators to the same analysis space. Based on this, it introduces time phase offset and structural drift index to correct the coupling analysis matrix, enabling the depiction of the true coupling relationship between indicators while eliminating the influence of data time misalignment and trend drift. Furthermore, it reflects the overall operating status through production state balance and makes judgments based on the lower bound of the dynamic early warning threshold, achieving system-level identification of imbalance risks. Simultaneously, it determines early warning indicators based on sensitivity values, giving the early warning results clear directionality, thereby improving the accuracy and stability of anomaly identification in multi-indicator linkage scenarios. Attached Figure Description

[0017] Figure 1 This is a flowchart of the steps of an intelligent analysis and early warning method for production management data provided in one embodiment of the present invention;

[0018] Figure 2 This is a system structure diagram of a production management data intelligent analysis and early warning system provided in one embodiment of the present invention;

[0019] Figure 3 This is a schematic diagram of the structure of a computer system suitable for implementing terminal devices or servers according to one embodiment of the present invention. Detailed Implementation

[0020] like Figure 1 As shown, embodiments of the present invention provide a method for intelligent analysis and early warning of production management data, the method comprising:

[0021] Step S100: Obtain the indicator data of each production management indicator within the current time window, and construct a normalized production management state vector based on the indicator data.

[0022] Specifically, constructing a normalized production management state vector based on the indicator data includes: statistically analyzing the historical indicator data corresponding to each production management indicator within a preset normalized historical time interval, and determining the historical lower bound and historical upper bound of each production management indicator; acquiring the indicator data of each production management indicator within the current time window, and performing interval mapping processing on the indicator data and the corresponding historical lower bound and historical upper bound to generate the corresponding normalized indicator value; and combining the normalized indicator values ​​according to the predetermined arrangement order of the production management indicators to construct the normalized production management state vector.

[0023] In this embodiment of the invention, the production management system typically involves multi-dimensional data, including output indicators, quality indicators, inventory indicators, equipment operation indicators, and plan execution indicators. These production management indicators are generated in different business subsystems, and their data granularity, statistical scope, and units differ. For example, output indicators may be represented by the quantity completed, quality indicators by the pass rate, and inventory indicators by the stock size. If correlation analysis or matrix operations are directly performed on these indicators, the differences in numerical scale will have an unbalanced impact on the construction of the subsequent coupling analysis matrix. Therefore, before proceeding with the coupling analysis, it is necessary to construct a normalized production management state vector with a unified numerical expression.

[0024] In practice, within a pre-defined normalized historical time interval, the set of production management indicators is as follows:

[0025]

[0026] Analyze historical data separately. For any production management indicator... The system acquires the historical indicator data sequence formed within the preset normalized historical time interval, and determines the historical lower bound value based on the historical indicator data sequence. Historical threshold The historical lower bound and historical upper bound are used to define the effective fluctuation range of the production management indicator in a historical statistical sense. They can be determined based on the statistical boundaries or stable intervals of historical samples to ensure that the normalized interval reflects the true operating range.

[0027] In the current time window Within this process, obtain the indicator data for each production management indicator and record it as follows: And compare the indicator data with the corresponding historical lower bound value. and historical threshold Perform interval mapping to convert it into a normalized index value within a uniform numerical range. The interval mapping process is based on the relative position of the current indicator data within the historical statistical interval, performing a proportional mapping to ensure comparability of production management indicators with different dimensions in the numerical space. When the current indicator data exceeds the historical statistical interval, it can be restricted according to boundary constraints to ensure the numerical stability of the normalized production management state vector.

[0028] Obtain the normalized index values ​​corresponding to each production management indicator. Then, the production management indicators are combined according to a predefined order to construct a normalized production management state vector:

[0029]

[0030] The normalized production management state vector serves as a unified input object for subsequent coupling analysis matrix construction, structural drift index correction, and production state balance calculation, enabling multi-dimensional production management indicators to participate in calculations within the same analysis space, thereby providing a consistent data foundation for multi-indicator coupling analysis and early warning judgment.

[0031] Step S200: Calculate the correlation coefficient and time phase offset based on the normalized production management state vector to construct a coupling analysis matrix, and correct the coupling analysis matrix based on the structure drift index to obtain a corrected coupling analysis matrix.

[0032] Specifically, the process of calculating correlation coefficients and time phase offsets based on the normalized production management state vectors to construct a coupling analysis matrix includes: within a preset analysis time interval, obtaining the normalized production management state vectors corresponding to each time window, and calculating the correlation coefficient between any two production management indicators based on the normalized index values ​​at corresponding positions within each time window; for any two production management indicators, performing time offset processing on one of the production management indicators within a preset time offset interval, and recalculating the correlation coefficient between the two production management indicators under each time offset; determining the time offset with the largest absolute value of the correlation coefficient within the time offset interval as the time phase offset between the two production management indicators, and determining the corresponding correlation coefficient as the corrected correlation coefficient; and constructing the coupling analysis matrix based on the corrected correlation coefficients between each production management indicator and the corresponding time phase offsets.

[0033] Furthermore, the coupling analysis matrix is ​​modified based on the structural drift index to obtain the modified coupling analysis matrix, including: within a preset long-term statistical time interval, calculating the historical mean and historical fluctuation values ​​corresponding to each production management indicator, and determining the long-term statistical characteristics of each production management indicator based on the historical mean and historical fluctuation values; within the current analysis time interval, calculating the current mean corresponding to each production management indicator, and calculating the structural drift index of each production management indicator based on the degree of difference between the current mean and the corresponding historical mean and the historical fluctuation value; determining whether the structural drift index corresponding to each production management indicator exceeds a preset drift judgment threshold, and performing attenuation processing on the correlation coefficients corresponding to production management indicators whose structural drift index exceeds the drift judgment threshold; and reconstructing the coupling analysis matrix based on the attenuated correlation coefficients to obtain the modified coupling analysis matrix.

[0034] In this embodiment of the invention, the normalized production management state vector is constructed. Subsequently, this implementation method further models the correlation structure between multiple production management indicators to form a coupling analysis matrix that reflects the dynamic coupling relationship between the indicators, and then corrects it based on structural drift factors. This process maintains continuity with the aforementioned normalization step at the data object level, and the normalized production management state vector serves as the sole input object in correlation calculation and matrix construction, ensuring the consistency of the overall analysis path.

[0035] Within a preset analysis time interval, the system continuously acquires normalized production management status vectors corresponding to multiple time windows, forming a time series. Each vector is:

[0036]

[0037] The Indicates the first Production management indicators within the time window The normalized index values ​​within the time series. Based on the above time series, for any two production management indicators... and Extract the index value sequence at the corresponding position to obtain:

[0038]

[0039] Within the preset analysis time interval, the correlation coefficient is calculated by statistically analyzing the degree of co-variance between the two sequences. The correlation coefficient is used to characterize the statistical consistency of changes in two production management indicators, and its numerical range is usually limited to [range missing]. Within the interval, the closer the value is to 1 or -1, the stronger the correlation. To avoid extreme samples having an excessive impact on the calculation results, the sequence can be mean-centered before calculation, and boundary restrictions can be applied to outlier samples.

[0040] Considering that production management data originates from different business systems and that there are statistical time differences between different indicators (e.g., some indicators have been updated within the current time window, while others have a confirmation delay), a time phase offset is introduced when constructing the coupling relationship. Specifically, within a preset time offset interval... Internally, regarding indicators The normalized index value sequence is time-shifted, i.e., a time-shifting process is constructed. and at each time offset Recalculate its correlation with the index Correlation coefficient between After traversing the time offset intervals, select those that satisfy... The time offset is the time phase offset between two production management indicators. And the corresponding correlation coefficient is denoted as the corrected correlation coefficient. Through the above processing, the coupling relationship is no longer limited to the statistical results at the synchronization time point, but takes into account the possible lag response structure between indicators.

[0041] Based on the adjusted correlation coefficients among all indicator pairs Constructing the coupling analysis matrix , where matrix elements Indicates the first The production management indicators and the first The modified correlation strength among the various production management indicators is described in this coupling analysis matrix, which characterizes the overall network structure of the relationships among the multiple production management indicators and serves as a key input for subsequent calculations of production state balance.

[0042] After obtaining the coupling analysis matrix, it is also necessary to consider the structural changes that may occur during the operation of the indicators. In the production operation environment, some production management indicators may show a continuous shift trend due to product structure adjustments, changes in business strategies, or changes in statistical methods. If this trend change is not identified, the correlation strength in the coupling analysis matrix may be amplified by long-term shifts, thereby affecting subsequent status judgments.

[0043] Therefore, within a pre-set long-term statistical period, the historical average values ​​of various production management indicators were calculated. Compared with historical fluctuation values Historical averages are used to represent long-term stable operating levels, while historical fluctuations represent the normal fluctuation range of the indicator. Within the current analysis time interval, calculate the current average of each production management indicator. And based on:

[0044]

[0045] Calculate the structural drift index When the structural drift index exceeds the preset drift threshold... At that time, it was determined that the production management indicator showed a trend of deviation at the current stage.

[0046] For production management indicators whose structural drift index exceeds the drift judgment threshold, the corresponding correlation coefficients in the coupling analysis matrix are attenuated. This attenuation can be performed using exponential attenuation, for example, by defining an attenuation coefficient. ,in These are the attenuation control parameters. The attenuated matrix elements can be represented as:

[0047]

[0048] Or, in the form of two-sided attenuation, it can be expressed as:

[0049]

[0050] This yields the modified coupling analysis matrix. This modified coupling analysis matrix retains the multi-index coupling structure while suppressing the correlation amplification problem caused by structural shifts.

[0051] A stable corrected coupling analysis matrix is ​​formed through two steps: time phase shift identification and structural drift correction. This matrix and the normalized production management state vector These are all used as inputs for subsequent production status balance calculations. The time phase offset is used to correct for temporal differences between different production management indicators, and the structural drift index is used to identify the impact of long-term trend changes on the coupling relationship. Together, they ensure that the coupling relationship maintains consistency and interpretability in a dynamic operating environment. This processing flow guarantees that multiple production management indicators have completed temporal alignment and structural stability correction before entering the production status balance calculation stage, providing a reliable data foundation for subsequent dynamic early warning judgments.

[0052] In another implementation, based on the construction of the modified coupling analysis matrix, a coupling stability decomposition mechanism is introduced to perform feature structure analysis on the modified coupling analysis matrix in order to identify abnormal propagation channels in the coupled network.

[0053] Specifically, for the modified coupling analysis matrix Perform eigenvalue decomposition to obtain the eigenvalue set. and the corresponding eigenvectors. When there exist eigenvalues ​​that satisfy... At that time, it is determined that a high-gain propagation path exists in the coupling structure; further, based on the magnitude of each component in the corresponding eigenvector, the set of production management indicators with amplification effect in the coupling network is determined. For the set of production management indicators, a stability constraint coefficient can be applied to its correlation coefficient in the coupling analysis matrix. The stabilization matrix is ​​obtained as follows:

[0054]

[0055] Alternatively, targeted attenuation processing can be applied to specific rows and columns. In this way, without changing the original related structure identification logic, structural suppression can be performed on coupling channels that may form abnormal cascading amplification, thereby enhancing the stability of coupling modeling in complex operating environments.

[0056] Step S300: Calculate the production state balance degree based on the normalized production management state vector and the modified coupling analysis matrix, and calculate the lower bound of the dynamic early warning threshold based on the production state balance degree within the historical time window.

[0057] Specifically, the normalized production management state vector is matrix mapped based on the modified coupling analysis matrix to obtain the corresponding coordinated state vector; the difference between the normalized production management state vector and the coordinated state vector is calculated to obtain the production state misalignment measure, and the production state balance is determined based on the production state misalignment measure; the production state balance corresponding to each time window is obtained within a preset balance statistical time interval, and the production state balance corresponding to each time window is statistically processed to obtain the historical balance mean and historical balance fluctuation value; the lower bound of the dynamic early warning threshold is determined based on the historical balance mean and the historical balance fluctuation value.

[0058] In this embodiment of the invention, after obtaining the normalized production management state vector... and the modified coupling analysis matrix Subsequently, this implementation further calculates the production state balance based on the two, and determines the lower bound of the dynamic early warning threshold based on this. This step follows the aforementioned data object, with the normalized production management state vector and the modified coupling analysis matrix serving as the sole inputs, without introducing a new analysis system, thereby ensuring the continuity of the overall technical path.

[0059] Specifically, the normalized production management state vector is subjected to matrix mapping processing based on the modified coupling analysis matrix to obtain the coordinated state vector. Let the current time window be... Then the coordinated state vector can be expressed as The coordination state vector is used to characterize the coordinated state that should exist among various production management indicators under the constraints of the coupled structure. When the system is running stably, the difference between the normalized production management state vector and the coordination state vector is small; when the coordinated relationship between multiple indicators is disrupted, the deviation between the two will increase.

[0060] Based on this, the difference between the normalized production management state vector and the coordinated state vector is calculated. The difference can be calculated using the vector norm method; for example, the production state discrepancy measure can be defined as:

[0061]

[0062] in This represents the L2 norm, used to comprehensively characterize the overall degree of deviation among multidimensional indicators. (Production status measurement) The larger the value, the greater the deviation between the current production operation status and the coordination status reflected by the coupling structure.

[0063] The production state balance is determined based on the aforementioned production state imbalance. The production state balance can be defined as:

[0064]

[0065] in The maximum deviation observed within the statistical interval or a preset normalization constant is used to ensure the balance of production status. fall into Within the interval. This method maps the imbalance of multiple coupled indicators into a single scalar for subsequent early warning judgment. Within the preset balance statistical time interval, the production status balance sequence corresponding to each time window is continuously acquired. The production state balance sequence is statistically processed to obtain the historical balance mean. and historical balance fluctuation value The historical equilibrium mean reflects the overall stability level of the system within the statistical interval, while the historical equilibrium fluctuation value reflects the normal fluctuation range of the equilibrium.

[0066] The lower bound of the dynamic early warning threshold is determined based on the historical average balance value and the historical balance fluctuation value. The lower bound of the dynamic early warning threshold can be expressed as:

[0067]

[0068] in, This is a threshold sensitivity parameter used to adjust the strictness of the early warning judgment. By constructing a dynamic lower threshold using historical statistical features, the early warning judgment no longer relies on a fixed value, but rather adaptively adjusts based on the system's historical operating status. When the production status balance is lower than the dynamic early warning threshold lower bound, it indicates that the current operating status deviates from the historical stable range, providing a basis for subsequent early warning indicator identification.

[0069] Through the above steps, we have realized the process of mapping high-dimensional production management status to a single balance index from a multi-index coupling structure, and constructing a dynamic early warning threshold lower bound by combining historical statistical characteristics, thus providing a unified quantitative basis for subsequent early warning triggering and index positioning.

[0070] Step S400: Compare the production status balance with the lower bound of the dynamic early warning threshold. When the production status balance is less than the lower bound of the dynamic early warning threshold, calculate the sensitivity value and determine the early warning index based on the normalized production management status vector and the modified coupling analysis matrix, and output the corresponding early warning result.

[0071] Specifically, when the production state balance is less than the lower bound of the dynamic early warning threshold, the sensitivity value corresponding to each production management indicator is calculated based on the normalized production management state vector and the modified coupling analysis matrix; the absolute values ​​of each sensitivity value are compared, and the production management indicator with the largest absolute value is determined as the early warning indicator; the corresponding early warning level information is generated based on the early warning indicator, and the early warning result is output.

[0072] Furthermore, calculating the sensitivity values ​​corresponding to each production management indicator based on the normalized production management state vector and the modified coupling analysis matrix includes: performing matrix mapping processing on the normalized production management state vector based on the modified coupling analysis matrix to obtain a coordinated state vector; calculating the difference between the normalized production management state vector and the coordinated state vector to obtain an imbalance propagation vector; performing backpropagation processing on the imbalance propagation vector based on the modified coupling analysis matrix to obtain an indicator propagation contribution vector; and determining the sensitivity value corresponding to each production management indicator based on the numerical values ​​corresponding to each position in the indicator propagation contribution vector.

[0073] In this embodiment of the invention, after achieving production state balance... and the lower bound of the dynamic early warning threshold After the calculation, this implementation method enters the early warning judgment and location stage. The production status balance degree serves as a quantitative representation of the overall system operating status. When it falls below the lower boundary of the dynamic early warning threshold, it indicates that the operating status of multiple indicators within the current time window deviates from the historical stable range, and it is necessary to further identify the key production management indicators that cause the imbalance.

[0074] When the judgment condition is met At that time, based on the normalized production management state vector With the modified coupling analysis matrix Calculate the sensitivity values ​​corresponding to each production management indicator. The sensitivity values ​​are used to characterize the contribution intensity of each production management indicator to the degree of imbalance in the current production state. The calculation process is consistent with the aforementioned balance calculation and does not introduce new analytical structures.

[0075] Specifically, based on the modified coupling analysis matrix, the normalized production management state vector is subjected to matrix mapping processing to obtain the coordinated state vector:

[0076]

[0077] Subsequently, the difference between the normalized production management state vector and the coordination state vector is calculated to obtain the imbalance propagation vector:

[0078]

[0079] The imbalance propagation vector This represents the degree of deviation of each production management indicator from the constraints of the coupled structure within the current time window. Each component in the vector... Reflecting the Local offset of a production management indicator.

[0080] To further identify the role paths of each indicator in the overall imbalance propagation, this implementation method performs backpropagation processing on the imbalance propagation vector based on the modified coupling analysis matrix. The backpropagation processing can be expressed as:

[0081]

[0082] in To correct the transpose of the coupling analysis matrix, the index propagation contribution vector Each component in Indicates the first The degree of contribution of each production management indicator to the propagation of the coupled network is determined by introducing a transpose matrix to match the propagation direction with the original coupled structure, thereby reflecting the influence path of each indicator in the network.

[0083] The sensitivity value for each production management indicator is determined based on the numerical value corresponding to each position in the indicator propagation contribution vector. The sensitivity value can be defined as follows: ,in This represents absolute value operations, used to standardize the degree of influence of offsets in different directions. By comparing the absolute values ​​of each sensitivity value, a value that satisfies the condition is selected. Production management indicators As an early warning indicator, the aforementioned early warning indicator represents the production management indicator that has the greatest impact on the system's balance in the current imbalance propagation structure.

[0084] After determining the warning indicator, based on the sensitivity value corresponding to the indicator and the difference between the production status balance and the lower bound of the dynamic warning threshold, corresponding warning level information is generated. The warning level information can be divided according to a preset level range, for example, based on... The numerical range determines the different levels of warning indicators. The final warning result is output, which includes the warning indicator indicator and the corresponding warning level information.

[0085] Through the above steps, the transformation from overall balance anomaly to specific early warning indicator location is achieved. This transformation process is based on the joint calculation of the normalized production management state vector and the modified coupling analysis matrix, ensuring consistency between the early warning judgment result and the multi-indicator coupling structure, and providing the production management system with clear anomaly indication information.

[0086] In another implementation, after obtaining the sensitivity values ​​corresponding to each production management indicator, all sensitivity values ​​are first sorted, and the degree of difference between the highest and second-highest sensitivity values ​​is statistically analyzed. When the highest sensitivity value is significantly higher than the other indicators, it is determined that the current imbalance is mainly dominated by a single production management indicator; when the sensitivity values ​​of multiple production management indicators are at similar levels and the difference does not reach a preset proportional threshold, it is determined that the current imbalance belongs to a multi-indicator collaborative offset state.

[0087] When an imbalance is determined to be a single dominant imbalance, the production management indicator with the highest sensitivity value is still identified as the early warning indicator according to the original rules, and the corresponding early warning level information is generated. When an imbalance is determined to be a multi-indicator coordinated imbalance, multiple production management indicators with sensitivity values ​​exceeding a preset percentage threshold are selected to form a joint early warning indicator set, and the comprehensive early warning level is determined based on the weighted average of the sensitivity values ​​of each indicator.

[0088] By implementing the above rules, the early warning output can not only identify single-point anomalies, but also identify multi-indicator coordinated disturbances, thereby improving the interpretability and applicability of the early warning results in complex production management environments, while maintaining consistency with the original normalized production management state vector and the calculation path of the modified coupling analysis matrix.

[0089] like Figure 2 As shown, this invention provides an intelligent analysis and early warning system for production management data. The system includes: a first module for acquiring indicator data of each production management indicator within the current time window and constructing a normalized production management state vector based on the indicator data; a second module for calculating correlation coefficients and time phase offsets based on the normalized production management state vector to construct a coupling analysis matrix, and correcting the coupling analysis matrix based on the structure drift index to obtain a corrected coupling analysis matrix; a third module for calculating the production state balance degree based on the normalized production management state vector and the corrected coupling analysis matrix, and calculating a lower bound of the dynamic early warning threshold based on the production state balance degree within a historical time window; and a fourth module for comparing the production state balance degree with the lower bound of the dynamic early warning threshold. When the production state balance degree is less than the lower bound of the dynamic early warning threshold, a sensitivity value is calculated based on the normalized production management state vector and the corrected coupling analysis matrix, and an early warning indicator is determined, and the corresponding early warning result is output.

[0090] A third aspect of the present invention provides an electronic device, comprising: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the above-described intelligent analysis and early warning method for production management data.

[0091] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the above-described intelligent analysis and early warning method for production management data.

[0092] The following is for reference. Figure 3 It shows a schematic diagram of the structure of a computer system 400 suitable for implementing a terminal device of the present invention. Figure 3 The terminal device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0093] like Figure 3 As shown, the computer system 400 includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 402 or programs loaded from storage section 408 into random access memory (RAM) 403. The RAM 403 also stores various programs and data required for the operation of the system 400. The CPU 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0094] The following components are connected to I / O interface 405: an input section 406 including a keyboard, mouse, etc.; an output section 407 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN card, modem, etc. The communication section 409 performs communication processing via a network such as the Internet. Drive 410 is also connected to I / O interface 405 as needed. Removable media 411, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 410 as needed so that computer programs read from them can be installed into storage section 408 as needed.

Claims

1. A method for intelligent analysis and early warning of production management data, characterized in that, The method includes: Obtain the indicator data of each production management indicator within the current time window, and construct a normalized production management state vector based on the indicator data; Based on the normalized production management state vector, correlation coefficients and time phase offsets are calculated to construct a coupling analysis matrix. The coupling analysis matrix is ​​then corrected based on the structure drift index to obtain a corrected coupling analysis matrix. The rules for determining the structural drift index are as follows: in, The structural drift index, This represents the current average of each production management indicator. The historical averages of each production management indicator. These are the historical fluctuation values ​​of each production management indicator; The process of constructing a coupling analysis matrix based on the normalized production management state vector includes: obtaining the normalized production management state vector corresponding to each time window within a preset analysis time interval, and calculating the correlation coefficient between any two production management indicators based on the normalized index values ​​at corresponding positions within each time window; for any two production management indicators, performing time offset processing on one of the production management indicators within a preset time offset interval, and recalculating the correlation coefficient between the two production management indicators under each time offset; determining the time offset with the largest absolute value of the correlation coefficient within the time offset interval as the time phase offset between the two production management indicators, and determining the corresponding correlation coefficient as the corrected correlation coefficient; and constructing the coupling analysis matrix based on the corrected correlation coefficient between each production management indicator and the corresponding time phase offset. The modified coupling analysis matrix is ​​obtained by correcting the coupling analysis matrix based on the structural drift index. This includes: calculating the historical mean and historical fluctuation values ​​of each production management indicator within a preset long-term statistical time interval, and determining the long-term statistical characteristics of each production management indicator based on the historical mean and historical fluctuation values; calculating the current mean of each production management indicator within the current analysis time interval, and calculating the structural drift index of each production management indicator based on the degree of difference between the current mean and the corresponding historical mean and the historical fluctuation values; determining whether the structural drift index of each production management indicator exceeds a preset drift judgment threshold, and attenuating the correlation coefficients of production management indicators whose structural drift indices exceed the drift judgment threshold; and reconstructing the coupling analysis matrix based on the attenuated correlation coefficients to obtain the modified coupling analysis matrix. The production state balance is calculated based on the normalized production management state vector and the modified coupling analysis matrix, and the lower bound of the dynamic early warning threshold is calculated based on the production state balance within the historical time window. The production status balance is compared with the lower bound of the dynamic early warning threshold. When the production status balance is less than the lower bound of the dynamic early warning threshold, a sensitivity value is calculated based on the normalized production management status vector and the modified coupling analysis matrix, and an early warning index is determined. The corresponding early warning result is then output. This includes: When the production status balance is less than the lower bound of the dynamic early warning threshold, the sensitivity value corresponding to each production management indicator is calculated based on the normalized production management status vector and the modified coupling analysis matrix; the absolute values ​​of each sensitivity value are compared, and the production management indicator with the largest absolute value is determined as the early warning indicator; the corresponding early warning level information is generated based on the early warning indicator, and the early warning result is output. The sensitivity values ​​for each production management indicator are calculated based on the normalized production management state vector and the modified coupling analysis matrix, including: performing matrix mapping processing on the normalized production management state vector based on the modified coupling analysis matrix to obtain a coordinated state vector; calculating the difference between the normalized production management state vector and the coordinated state vector to obtain an imbalance propagation vector; performing backpropagation processing on the imbalance propagation vector based on the modified coupling analysis matrix to obtain an indicator propagation contribution vector; and determining the sensitivity value for each production management indicator based on the numerical values ​​corresponding to each position in the indicator propagation contribution vector.

2. The intelligent analysis and early warning method for production management data according to claim 1, characterized in that, Based on the aforementioned indicator data, a normalized production management state vector is constructed, including: Within a preset normalized historical time interval, statistical analysis is performed on the historical indicator data corresponding to each production management indicator, and the historical lower bound and historical upper bound values ​​of each production management indicator are determined respectively. Obtain the indicator data of each production management indicator within the current time window, and perform interval mapping processing on the indicator data with the corresponding historical lower bound value and historical upper bound value to generate the corresponding normalized indicator value; The normalized index values ​​are combined according to a predetermined order of production management indicators to construct the normalized production management state vector.

3. The intelligent analysis and early warning method for production management data according to claim 1, characterized in that, The production state balance is calculated based on the normalized production management state vector and the modified coupling analysis matrix, and the lower bound of the dynamic early warning threshold is calculated based on the production state balance within the historical time window, including: Based on the modified coupling analysis matrix, the normalized production management state vector is subjected to matrix mapping processing to obtain the corresponding coordination state vector. The difference between the normalized production management state vector and the coordination state vector is calculated to obtain the production state mismatch measure, and the production state balance is determined based on the production state mismatch measure. Within a preset balance degree statistical time interval, the production status balance degree corresponding to each time window is obtained, and the production status balance degree corresponding to each time window is statistically processed to obtain the historical balance degree mean and historical balance degree fluctuation value. The lower bound of the dynamic early warning threshold is determined based on the historical average balance value and the historical balance fluctuation value.

4. A production management data intelligent analysis and early warning system, characterized in that, The system is used to execute the intelligent analysis and early warning method for production management data as described in any one of claims 1-3, and the system includes: The first module is used to obtain the indicator data of each production management indicator within the current time window, and to construct a normalized production management state vector based on the indicator data. The second module is used to calculate the correlation coefficient and time phase offset based on the normalized production management state vector to construct a coupling analysis matrix, and to correct the coupling analysis matrix based on the structure drift index to obtain a corrected coupling analysis matrix. The third module is used to calculate the production state balance degree based on the normalized production management state vector and the modified coupling analysis matrix, and to calculate the lower bound of the dynamic early warning threshold based on the production state balance degree within the historical time window. The fourth module is used to compare the production status balance with the lower bound of the dynamic early warning threshold. When the production status balance is less than the lower bound of the dynamic early warning threshold, the module calculates the sensitivity value and determines the early warning index based on the normalized production management status vector and the modified coupling analysis matrix, and outputs the corresponding early warning result.

5. An electronic device, characterized in that, include: One or more processors; A storage device having stored one or more programs thereon, which, when executed by one or more processors, cause the one or more processors to implement the intelligent analysis and early warning method for production management data as described in any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the intelligent analysis and early warning method for production management data as described in any one of claims 1-3.

Images