Auxiliary decision system based on small and micro enterprise data analysis

Abstract

This invention relates to the field of industrial data decision-making technology and discloses an auxiliary decision-making system based on data analysis of micro and small enterprises. The system includes: firstly, extracting change event data based on product structure and time windows, and then determining the main supply cycle through spectrum analysis; subsequently, calculating layered phasors and interlayer complex gains using hierarchical partitioning, constructing a transfer matrix, and extracting the maximum singular value as a proxy for change propagation intensity; finally, automatically determining the safe release depth and adjusting the change effective time based on the comparison results between the proxy and the safety line, thereby achieving prediction and control of concentrated fluctuations in engineering changes and improving the intelligent decision-making level and production stability of micro and small enterprises.

Description

Technical Field

[0001] This invention relates to the field of industrial data decision-making technology, and more specifically, to an auxiliary decision-making system based on data analysis of micro and small enterprises. Background Technology

[0002] In small and micro-sized manufacturing enterprises, Product Data Management (PDM) systems are the core carriers of design, process, and production information. With increasing customization demands, supply chain instability, and frequent changes in regulations and materials, the number of design change (ECR / ECO) events within PDM systems has risen sharply. Engineering changes do not occur in isolation but rather tend to occur in concentrated periods and propagate continuously across multiple levels of the product structure. Due to the limited scale of production systems and low integration of information systems in small and micro-sized enterprises, these changes often cause problems such as resource consumption, process conflicts, and scheduling chaos within local time windows, leading to delayed or distorted production adaptation decisions. Traditional change management is based solely on single-event analysis, lacking a holistic description of the time series and structural propagation effects, making it difficult to explain why the same type of change can cause drastically different production disruptions at different times.

[0003] In reality, the supply rhythm of micro and small enterprises is constrained by external factors such as raw material procurement cycles, outsourced processing volumes, and customer order review cycles, resulting in a clear cyclical pattern in production and change behavior over time. Simultaneously, change events recorded by the PDM system exhibit a trigger-response propagation chain within its internal hierarchical structure. Any design adjustment to a critical component can trigger continuous modifications at multiple subsequent assembly levels. When the cyclical fluctuations of external supply overlap with the internal multi-level change propagation, change events in the PDM system can erupt in a concentrated burst, generating short-term, high-intensity change waves. Such fluctuations not only cause backlogs in engineering approval and implementation processes but can also lead to simultaneous congestion in material preparation, process switching, and work scheduling, thereby disrupting the original production rhythm. Because these events superficially appear as concentrated periods of regular ECOs, traditional statistical analysis often struggles to extract their causes. Summary of the Invention

[0004] This invention provides an auxiliary decision-making system based on data analysis of micro and small enterprises, which solves the technical problem of how to identify and quantify whether the system is in a critical state of change fluctuation driven by both external supply cycles and internal multi-layer propagation, based solely on the time distribution and hierarchical relationship of change events in the PDM system, thereby avoiding concentrated instability in advance in production adaptation decisions.

[0005] This invention provides an auxiliary decision-making system based on data analysis of micro and small enterprises, comprising: The data start-up module, based on the central component determined by the product structure information and a preset time window, acquires change event data and product structure information in product data management. The supply master cycle determination module generates a counting sequence based on change event data according to a preset statistical step size, and determines the supply master cycle from the spectral peaks of the counting sequence through spectral analysis. The proxy feature construction module calculates the layered phasor of each layer according to the product structure information at the frequency point corresponding to the main supply cycle. Based on the layered phasor of adjacent layers, it obtains the interlayer complex gain, constructs the interlayer transfer matrix containing the interlayer complex gain, and takes the maximum singular value of the interlayer transfer matrix as the proxy quantity of the periodic phase-locked layered gain spectrum radius. The decision execution module compares the proxy quantity with the preset safety line, determines the minimum cutting depth based on the comparison result, and only allows engineering changes that do not exceed this depth. At the same time, it determines the layered phase stagger based on the phase information of the interlayer complex gain and converts it into the effective time of the preset granularity.

[0006] The beneficial effects of this invention are as follows: By combining time series analysis of change events in a PDM data management system with product structure hierarchical propagation characteristic modeling, an auxiliary decision-making mechanism is proposed that can quantify the propagation risk of engineering changes and automatically adjust the change activation rhythm. This invention utilizes spectrum analysis to identify the main cycle of external supply rhythms and, through constructing inter-layer transfer matrices and singular value analysis, achieves a quantitative characterization of the coupling strength of internal multi-level changes, thereby identifying the critical state of change fluctuations in advance. Compared to traditional single-event change management, this invention can proactively adjust the release depth and activation time before concentrated change risks occur, effectively reducing the probability of resource conflicts and scheduling instability, and significantly improving the production adaptability and decision-making intelligence level of small and micro manufacturing enterprises in complex supply environments. Attached Figure Description

[0007] Figure 1 This is a block diagram of the present invention. Detailed Implementation

[0008] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0009] like Figure 1 As shown, the decision support system based on data analysis of micro and small enterprises includes: The data start-up module, based on the central component determined by the product structure information and a preset time window, acquires change event data and product structure information in product data management. The supply master cycle determination module generates a counting sequence based on change event data according to a preset statistical step size, and determines the supply master cycle from the spectral peaks of the counting sequence through spectral analysis. The proxy feature construction module calculates the layered phasor of each layer according to the product structure information at the frequency point corresponding to the main supply cycle. Based on the layered phasor of adjacent layers, it obtains the interlayer complex gain, constructs the interlayer transfer matrix containing the interlayer complex gain, and takes the maximum singular value of the interlayer transfer matrix as the proxy quantity of the periodic phase-locked layered gain spectrum radius. The decision execution module compares the proxy quantity with the preset safety line, determines the minimum cutting depth based on the comparison result, and only allows engineering changes that do not exceed this depth. At the same time, it determines the layered phase stagger based on the phase information of the interlayer complex gain and converts it into the effective time of the preset granularity.

[0010] In one embodiment of the present invention, based on the central component determined by product structure information and a preset time window as the starting point, change event data and product structure information in product data management are acquired, including: Determine the central component and a preset time window, which has a start time and an end time; Read all change event data from product data management to form a complete change event data set. Each change event in the complete change event data set contains the change object and the change time. Filter out change events whose change time falls between the start and end times of a preset time window from the entire change event data set, forming a change event data set within the time window; Read product structure information from product data management. This product structure information includes a set of all components and a set of connections between the components. Based on the set of connection relationships in the product structure information, identify the components that have path connections with the central component, forming a set of components related to the central component. The starting point of this path is the central component, and the ending point is the related component. From the change event data set within the time window, filter out change events where the change object belongs to the component set related to the central component or the change object is the central component, and form a change event data set related to the central component.

[0011] Specifically, by using time windows to eliminate expired or non-occurring invalid changes, and by using central component path association to eliminate redundant data unrelated to core changes, this dual filtering can accurately identify change events that are critical to production and have a timely impact, solving the problem of complex and low-relevance change data in PDM systems for small and micro enterprises. For example, first set the time window (including start and end times) and the central component; then read all changes and product structure information from the PDM; and then complete the time filtering → spatial association filtering in sequence to finally obtain the target change set.

[0012] Criteria for determining central components: Based on the product's functional architecture, select components that play a decisive role in the core performance of the product and occupy a pivotal position in the assembly relationship (such as the stator of motor products and the spindle of machine tool products). These components are pre-designed by the company's technical personnel in conjunction with product design documents and production experience.

[0013] The specific method for determining path connections is based on the assembly subordination relationship in the product structure. That is, starting from the central component, the components that can be traced along the assembly link from the parent component to the child component (e.g., if the central component is the mobile phone motherboard, the path connection components include directly assembled components such as chips and interfaces on the motherboard, as well as indirectly assembled components such as capacitors connected to the chips) are automatically identified by traversing the BOM assembly relationship table in the PDM.

[0014] The entities and tools used to perform the filtering operation are as follows: The PDM system automatically executes the filtering process by using the system's built-in time range filtering component association filtering function module. After inputting the central component number and time window parameters, the system automatically calls the change event library and product structure library data to complete the filtering and output the results.

[0015] In one embodiment of the present invention, a counting sequence is generated based on change event data according to a preset statistical step size, and the main supply period is determined from the spectral peaks of the counting sequence through spectral analysis, including: The number of time slices is calculated based on the start time, end time, and preset statistical step size of the preset time window. This number is the integer ratio of the length of the time window to the preset statistical step size. The start time of each time slice is determined. The start time of the first time slice is the start time of the preset time window, and the start time of each subsequent time slice is the sum of the start time of the previous time slice and the preset statistical step size. The number of change events related to the central component within each time slice is counted to form a counting sequence. The time range of each time slice is the sum of the start time of the time slice and the preset statistical step size. Determine the set of angular frequencies, which is based on the number of time slices and does not include zero angular frequencies; Based on the counting sequence and the set of angular frequencies, calculate the spectral power corresponding to each angular frequency: the spectral power is obtained by taking the square of the absolute value of each value in the counting sequence and the corresponding angular frequency component. Select the angular frequency with the largest spectral power from the set of angular frequencies as the angular frequency corresponding to the spectral peak; calculate the main supply period based on the angular frequency corresponding to the spectral peak, and the main supply period is the ratio of 2π to the angular frequency corresponding to the spectral peak.

[0016] The main supply cycle influences demand changes through factors such as raw material procurement rhythm and outsourcing delivery frequency (e.g., changes to adaptable raw materials tend to occur more frequently during the procurement cycle), causing the temporal distribution of change events to follow a pattern consistent with the supply cycle. By extracting strong periodic signals from the change count sequence through spectral analysis, the main supply cycle can be deduced. This eliminates the need for direct supply chain data; cycle identification can be achieved using only PDM change data.

[0017] The preset statistical step size is a time interval set according to the frequency of changes of micro and small enterprises (such as 1 day or 3 days). It must meet the requirement that each time segment has a certain number of change events (to avoid meaningless statistics) and does not exceed 1 / 5 of the supply cycle (to avoid masking the cycle pattern). It is preset by technical personnel in combination with the enterprise's historical change data.

[0018] Angular frequency is a physical quantity that describes the speed of a periodic signal. Zero angular frequency corresponds to a constant signal without period (which has no practical supply period significance), so it is excluded. The angular frequency in the set needs to be determined based on the number of time slices (e.g., if the number of time slices is 30, the angular frequency is 2π×1 / (total time length) 2π×2 / (total time length)...2π×15 / (total time length) to ensure coverage of the possible supply period range).

[0019] The change count for each time slice is combined with the periodic signal component of the corresponding angular frequency (e.g., if a certain angular frequency corresponds to a weekly cycle, this component is the signal that repeats weekly). The square of the absolute value is used to amplify the power value of the strong periodic signal, making the peak value corresponding to the cycle more significant and easier to identify.

[0020] The angular frequency with the highest spectral power corresponds to the most significant period in the change counting sequence (i.e., the period that coincides with the main supply period); by dividing 2π by this angular frequency, the angular frequency (unit: radians / time) can be converted into the actual time period (e.g., if the angular frequency corresponds to 2π / 7 radians / day, then the main supply period is 7 days).

[0021] Specifically, taking a 30-day time window and a 1-day statistical step as an example: 1. The calculation involves 30 time slices, with each time slice corresponding to 1 day; 2. Count the number of changes related to the central components each day, forming a counting sequence of [5,3,6,2,5,3,6,…]. 3. Construct an angular frequency set (remove zero angular frequency, take the angular frequencies corresponding to 1 to 15, covering a period range of 1 to 30 days). 4. Calculate the spectral power of each angular frequency. The power corresponding to 2π / 7 radians / day is the largest. 5. The main supply cycle is calculated to be 7 days, meaning that enterprises supply once a week.

[0022] It should be noted that if there are multiple peak values ​​with similar power, the period corresponding to each peak value is compared with the period of the enterprise's known supply chain links (such as procurement and outsourcing), and the angular frequency corresponding to the peak value that matches the period of the supply chain link is selected to deduce the main supply cycle; if there is no known supply chain cycle, the longest period is taken (the supply cycle is mostly a medium-to-long cycle, and the short cycle is mostly random fluctuation).

[0023] In one embodiment of the present invention, at the frequency point corresponding to the main supply cycle, the hierarchical phasor of each level is calculated according to the hierarchy divided by product structure information, including: Determine a function to represent the hierarchical depth of a component relative to the central component, where the hierarchical depth is the number of edges in the shortest path from the central component to the component in the product structure information; Based on the hierarchy depth function, a set of components with a hierarchy of level d is divided from the component set of product structure information. The hierarchy depth of all components in this set is d. From the set of change events related to the central component, filter out the change events whose change objects belong to the component set at level d, and form a set of change event data at level d; For each change event in the change event data set of level d, perform calculations based on the change time of the change event and the angular frequency corresponding to the master supply cycle, sum all the calculation results, and obtain the layered phasor of level d.

[0024] The shortest path edge count defines the hierarchical depth, which can pinpoint the direct / indirect propagation range of changes in central components (avoiding cross-layer interference). Combined with the supply master cycle angular frequency calculation, it can convert change time into a quantifiable value indicating whether it is synchronized with the external cycle. The summed hierarchical phasors can simultaneously condense the overall response intensity of the change at that level (the amplitude of the summed result) and the degree of synchronization with the supply cycle (the phase of the summed result). Specifically, the component levels are first divided according to the shortest path edge count, then change events at each level are filtered, and finally, the results are summed using time-angular frequency calculation. These steps can be automatically executed using the product structure library and change event library of the PDM system.

[0025] Hierarchical phasors are used to characterize the temporal correlation between change events at each level and the supply master cycle. That is, the supply master cycle causes change events to occur in a concentrated manner at specific time nodes, and the change responses at different levels are different. Through phasors, this temporal synchronicity and hierarchical response intensity can be condensed into a characterization quantity.

[0026] The hierarchy depth function uniquely determines the hierarchy of a component relative to the central component. The shortest path edge count refers to the minimum number of links from the central component to the target component along the product structure connection relationship (such as assembly, subordinate relationship) (e.g., the central component is level 0, directly connected components are level 1, and components connected through one intermediate component are level 2, avoiding ambiguity in multi-level division).

[0027] The hierarchy depth function uses a breadth-first traversal algorithm, starting from the central component and traversing the product structure connection relationships (such as assembly subordination relationships) layer by layer, recording the minimum number of links (edges) to reach each component. This number is the hierarchy depth of that component.

[0028] The hierarchical change event filtering is based on the hierarchical depth. From the set of related changes of the central component, the event groups that only contain changes of the components at that level are split according to the level to which the change object belongs (e.g., change events of only the components at level 1 form the level 1 change set).

[0029] The angular frequency corresponding to the main supply cycle reflects the time pattern of the supply rhythm. The purpose of the calculation is to convert the change time into a quantity that can reflect whether it is synchronized with the supply cycle (for example, if a change time happens to be at the peak node of the supply cycle, the change will contribute more significantly to the hierarchical phase after the calculation).

[0030] The summation of the calculation results of all change events at the same level yields a hierarchical phasor that comprehensively reflects the overall response intensity of the change at that level (the amplitude of the summation result) and its synchronization with the supply cycle (the phase of the summation result).

[0031] Specifically, taking the engine block (layer 0) as the central component, and the supply main cycle angular frequency corresponding to the weekly rhythm as an example: 1. Determine the layer depth: Cylinder block (0 layers), piston (direct connection, 1 layer), piston rings (connected through piston, 2 layers). 2. Filter change events: From the changes related to the central components, separate the change sets of layer 0 (cylinder block), layer 1 (piston), and layer 2 (piston ring); 3. Calculation and summation: For the time of each piston change in layer 1, combine it with the weekly angular frequency calculation, and then add all the results to obtain the layer phasor of layer 1 (if most piston changes are concentrated in the same period of the week, the phasor amplitude is large, indicating that the changes in layer 1 are highly synchronized with the supply cycle).

[0032] In one embodiment of the present invention, the interlayer complex gain is obtained based on the layered phasors of adjacent layers, and an interlayer transfer matrix containing the interlayer complex gain is constructed, including: Determine the maximum level depth, which is the maximum level depth of all components in the product structure information; For each level with a level depth from 0 to the maximum level depth minus one, calculate the inter-level complex gain, which is the ratio of the layer phasor with a level depth of the current level plus one to the layer phasor with a level depth of the current level. Construct an inter-layer transfer matrix, the elements of which are determined according to the following rules: When the column position of the matrix is ​​equal to the row position plus one, the element is the interlayer complex gain corresponding to the row position; when the column position of the matrix is ​​not equal to the row position plus one, the element is a complex number with both real and imaginary parts being zero.

[0033] Interlayer complex gain is used to quantify the intensity and time synchronization of changes transmitted from the current layer to the next layer. That is, the phasor amplitude ratio reflects the amplification / attenuation degree of change transmission (if the ratio amplitude is greater than 1, it means that the change response of the next layer is stronger), and the phase difference reflects the time offset of change transmission (if the phase difference is positive, it means that the change of the next layer lags behind the current layer). The upper double diagonal matrix only retains the transmission relationship between adjacent layers because when changes are transmitted along the product structure, they can only be transmitted directly from the current layer to the next layer (e.g., changes to components in layer 1 can only trigger layer 2, and cannot directly trigger layer 3), eliminating cross-layer interference and ensuring that the matrix is ​​consistent with the actual transmission path.

[0034] First, iterate through the layer depths of all relevant components and take the largest value as the maximum layer depth. Then, for each combination of the current layer and the next layer, divide the phasor of the next layer by the phasor of the current layer to obtain the complex gain. Finally, fill in the complex gain with column = row + 1 and fill in the rest with zero complex numbers to complete the construction of the inter-layer transfer matrix.

[0035] The complex ratio is calculated according to the conventional complex division rules: the magnitude of the next layer phasor (numerator) is divided by the magnitude of the current layer phasor (denominator), and this is used as the magnitude of the complex gain; the phase of the complex gain is obtained by subtracting the phase of the current layer phasor from the phase of the next layer phasor. If the current layer phasor is zero (there is no change event in this layer), the corresponding inter-layer complex gain is set to zero (indicating no change propagation).

[0036] Maximum level depth boundary determination: The hierarchy depth of related components that have a path connection to the central component is counted, while the hierarchy depth of unrelated components (without a path connection) is not included to avoid irrelevant data interfering with the determination of the maximum depth (e.g., if the maximum hierarchy depth of related components of the central component is 3, then the maximum hierarchy depth is 3, and the hierarchy depth of unrelated components is 5, which is not included).

[0037] In one embodiment of the present invention, the maximum singular value of the interlayer transfer matrix is ​​taken as the surrogate quantity of the periodic phase-locked loop layer gain spectral radius, including: Determine the total number of layers, which is the maximum layer depth plus one; The inter-layer transfer matrix is ​​transposed to obtain the transpose of the inter-layer transfer matrix. The transpose process involves keeping the real part of each element in the original matrix unchanged, inverting the imaginary part, and then swapping the rows and columns of the matrix. Multiply the conjugate transpose of the interlayer transfer matrix with the interlayer transfer matrix to obtain the product matrix; Obtain all eigenvalues ​​of the product matrix, and select the eigenvalue with the largest value from all eigenvalues ​​as the largest eigenvalue of the product matrix. Calculate the square root of the largest eigenvalue of the product matrix, and take this square root as the largest singular value of the interlayer transfer matrix; The maximum singular value of the interlayer transfer matrix is ​​determined as the surrogate quantity of the periodic phase-locked layer gain spectrum radius.

[0038] The radius of the periodic phase-locked layered gain spectrum is used to determine whether the change propagates explosively along the hierarchy: When the spectral radius is less than 1, the propagation of the change will attenuate (stabilize). When the spectral radius approaches 1, the change will be significantly amplified (critical state).

[0039] However, direct calculation of the spectral radius is complex, while the maximum singular value of the interlayer transfer matrix is ​​approximately equal to the spectral radius (especially when the matrix has an upper double diagonal structure). Therefore, the maximum singular value can be obtained by the process of conjugate transpose → matrix multiplication → square root of the maximum eigenvalue, which can indirectly quantify the spectral radius and realize a simplified calculation of changing the criticality.

[0040] Specifically, first determine the total number of layers by the maximum layer depth + 1 (e.g., if the maximum layer depth is 3, the total number of layers is 4, and the matrix is ​​4×4); then, complete the conjugate transpose by keeping the real part unchanged, inverting the imaginary part, and exchanging rows and columns.

[0041] The acquisition of eigenvalues ​​of the product matrix includes: using conventional complex matrix eigenvalue calculation algorithms in this field (such as QR decomposition and Jacobi method), and automatically calculating them using computer tools (such as MATLAB and Python's numpy library). That is, inputting the product matrix (complex matrix), the algorithm will output all eigenvalues ​​(including the values ​​of the real and imaginary parts).

[0042] The spectral radius of the periodic phase-locked layer gain is the maximum amplification factor of change propagation in the interlayer transfer matrix (the larger the spectral radius, the easier it is for the change to spread explosively); while the maximum singular value of the matrix is ​​the maximum stretch factor of the matrix to the vector. For an upper double diagonal interlayer transfer matrix with non-zero elements only in adjacent layers, its maximum singular value is approximately equal to the spectral radius and can replace the spectral radius. Therefore, it can be used as its proxy quantity to quantify the critical amplification capability of change propagation.

[0043] In one embodiment of the present invention, the agent quantity is compared with a preset safety line, and a minimum cutting depth is determined based on the comparison result, and only engineering changes not exceeding this depth are allowed, including: Compare the proxy amount of the periodic phase-locked layered gain spectrum radius with the preset safety line; The range of candidate cut depths is set based on the maximum level depth, which is all integers from 0 to the maximum level depth; For each candidate cut-off depth, extract all rows and columns from layer 0 to the layer corresponding to the candidate cut-off depth in the interlayer transfer matrix to form a submatrix corresponding to the candidate cut-off depth. Calculate the maximum singular value of the submatrix corresponding to each candidate cut-off depth; The minimum candidate cutting depth is selected from the candidate cutting depths, and the maximum singular value of the submatrix corresponding to the minimum cutting depth does not exceed the preset safety line. From the set of change events related to the central component, change events whose hierarchical depth does not exceed the minimum cutting depth are selected to form a set of engineering changes that can be released.

[0044] The cut-off depth reflects the risk range of control changes propagating along the hierarchy. That is, the singular value of the submatrix reflects the propagation risk of the change within the corresponding depth (the smaller the value, the lower the risk). Selecting the minimum depth where the singular value of the submatrix does not exceed the safety line can effectively control the risk while preserving the range of changes to the maximum extent (avoiding excessive cut-off that leads to production adaptation delays).

[0045] Specifically, first, set all candidate cut-off depths from 0 to the maximum layer depth; for each candidate depth, extract the rows and columns corresponding to layer 0 to that candidate depth in the interlayer transfer matrix, form a submatrix and calculate its maximum singular value; compare the singular value with the safety line, and select the minimum candidate depth that meets the conditions, which is the minimum cut-off depth.

[0046] The preset safety line is set based on the company's historical change data: when past changes did not cause production instability, the maximum singular value of the inter-layer transfer matrix submatrix is ​​statistically analyzed, and 90% to 95% of this maximum value is taken as the safety line (e.g., if the historical maximum singular value is 0.8, the safety line is set to 0.75); if there is no historical data, the safety line benchmark of similar products of small and micro enterprises in the same industry is referenced (e.g., 0.8).

[0047] The row and column indices of the inter-layer transfer matrix correspond one-to-one with the layer depth (row index d corresponds to the transfer relationship starting from layer d, and column index d corresponds to the relationship transferred to layer d); when the candidate cut-off depth is L, the row index of the submatrix is ​​taken from 0 to L and the column index is taken from 0 to L, ensuring that the submatrix only contains the change transfer relationship within layers 0 to L.

[0048] It should be noted that if the maximum singular value of the submatrix of all candidate cut-off depths exceeds the safety line, the maximum level depth is taken as the cut-off depth, and layered phase staggering is superimposed to further reduce the risk; if it still cannot be satisfied, the current change is suspended, and the coupling reasons between the supply master cycle and the central component change are investigated before execution.

[0049] In one embodiment of the present invention, determining the layered phase misalignment based on the phase information of the interlayer complex gain and converting it into an effective time with a preset granularity includes: Based on the interlayer complex gain, obtain the phase of the interlayer complex gain for each layer; Using the angular frequency and pi corresponding to the main supply period, for each level, subtract the phase of the inter-level complex gain of that level from pi, and then divide the result by the angular frequency corresponding to the main supply period to obtain the peak shifting time increment between that level and the next level. Set the cumulative off-peak time of level zero to zero. For each level after level zero, add the cumulative off-peak time of the previous level to the off-peak time increment corresponding to the previous level to obtain the cumulative off-peak time of the current level. Using a preset time window end time, preset granularity, and minimum cutting depth, for each level with a level depth not greater than the minimum cutting depth, calculate the ratio of cumulative staggered time to preset granularity, round the ratio to the nearest integer, and then multiply the resulting integer by the preset granularity to obtain the granular time adjustment amount; add the granular time adjustment amount to the preset time window end time to obtain the effective time of that level.

[0050] The phase of the inter-layer complex gain reflects the temporal synchronicity of changes at adjacent levels. If the phases are close, the changes will take effect simultaneously, causing resource conflicts. By subtracting the phase from the pi, the phases of changes at adjacent levels are reversed (out of phase), and then converted into staggered time increments, which can offset the risk of overlapping synchronization. Accumulated staggered time ensures continuous staggered time between multiple levels, while granular conversion allows the effective time to adapt to the actual production rhythm of the enterprise (such as shifts). Specifically, the phase is first extracted from the inter-layer complex gain, the phase is subtracted from the pi → the staggered time increment is calculated by dividing by the angular frequency, then the staggered time is accumulated starting from level 0, and finally, the effective time of each level is obtained by combining the window end time and the preset granularity.

[0051] Interlayer complex gain is a complex number (containing real and imaginary parts). The phase is obtained through the arctangent operation of the complex number. The arctangent value is calculated with the imaginary part of the complex gain as the numerator and the real part as the denominator. The resulting angle is the phase (in radians). It can be automatically calculated using computer tools (such as Excel or Python's math library).

[0052] The preset granularity is set based on the enterprise's production organization model or change approval cycle. If the enterprise implements an 8-hour production shift, the preset granularity is 8 hours (to ensure that the effective time is consistent with the shift). If the change requires daily approval, the granularity is 24 hours (to facilitate unified daily processing), so that the granularity is consistent with the actual operating rhythm of the enterprise.

[0053] The handling of cumulative staggered peak time being negative is as follows: If the cumulative staggered time is negative, the duration of the main supply cycle is added to the cumulative staggered time to convert it into a positive number (e.g., if the main supply cycle is 7 days, -2 days is converted into 5 days), ensuring that the direction of time adjustment is in line with the production logic (only postponed, not brought forward to the period that has already passed within the window).

[0054] The change events within the time window have been statistically analyzed (for the initial layered phasor and complex gain calculations). The effective time is adjusted based on the end time of the window to avoid affecting the changes that have already occurred within the window, while ensuring that the effective time of subsequent changes is within a reasonable period after statistical analysis and does not conflict with the previous data.

[0055] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. A decision support system based on data analysis of micro and small enterprises, characterized in that, include: The data start-up module, based on the central component determined by the product structure information and a preset time window, acquires change event data and product structure information in product data management. The supply master cycle determination module generates a counting sequence based on change event data according to a preset statistical step size, and determines the supply master cycle from the spectral peaks of the counting sequence through spectral analysis. The proxy feature construction module calculates the layered phasor of each layer according to the product structure information at the frequency point corresponding to the main supply cycle. Based on the layered phasor of adjacent layers, it obtains the interlayer complex gain, constructs the interlayer transfer matrix containing the interlayer complex gain, and takes the maximum singular value of the interlayer transfer matrix as the proxy quantity of the periodic phase-locked layered gain spectrum radius. The decision execution module compares the proxy quantity with the preset safety line, determines the minimum cutting depth based on the comparison result, and only allows engineering changes that do not exceed this depth. At the same time, it determines the layered phase stagger based on the phase information of the interlayer complex gain and converts it into the effective time of the preset granularity.

2. The auxiliary decision-making system based on data analysis of micro and small enterprises according to claim 1, characterized in that, Starting with the core components identified by the product structure information and a preset time window, the system acquires change event data and product structure information from product data management, including: Determine the central component and a preset time window, which has a start time and an end time; Read all change event data from product data management to form a complete change event data set. Each change event in the complete change event data set contains the change object and the change time. Filter out change events whose change time falls between the start and end times of a preset time window from the entire change event data set, forming a change event data set within the time window; Read product structure information from product data management. This product structure information includes a set of all components and a set of connections between the components. Based on the set of connection relationships in the product structure information, identify the components that have path connections with the central component, forming a set of components related to the central component. The starting point of this path is the central component, and the ending point is the related component. From the change event data set within the time window, filter out change events where the change object belongs to the component set related to the central component or the change object is the central component, and form a change event data set related to the central component.

3. The auxiliary decision-making system based on data analysis of micro and small enterprises according to claim 2, characterized in that, Based on change event data, a counting sequence is generated according to a preset statistical step size. The main supply cycle is determined from the spectral peaks of the counting sequence through spectral analysis, including: The number of time slices is calculated based on the start time, end time, and preset statistical step size of the preset time window. This number is the integer ratio of the length of the time window to the preset statistical step size. The start time of each time slice is determined. The start time of the first time slice is the start time of the preset time window, and the start time of each subsequent time slice is the sum of the start time of the previous time slice and the preset statistical step size. The number of change events related to the central component within each time slice is counted to form a counting sequence. The time range of each time slice is the sum of the start time of the time slice and the preset statistical step size.

4. The auxiliary decision-making system based on data analysis of micro and small enterprises according to claim 3, characterized in that, Based on change event data, a counting sequence is generated according to a preset statistical step size. The main supply period is determined from the spectral peaks of the counting sequence through spectral analysis. The process also includes: determining the set of angular frequencies, where the angular frequencies are determined based on the number of time slices and do not include zero angular frequencies. Based on the counting sequence and the set of angular frequencies, calculate the spectral power corresponding to each angular frequency: the spectral power is obtained by taking the square of the absolute value of each value in the counting sequence and the corresponding angular frequency component. Select the angular frequency with the largest spectral power from the set of angular frequencies as the angular frequency corresponding to the spectral peak; calculate the main supply period based on the angular frequency corresponding to the spectral peak, and the main supply period is the ratio of 2π to the angular frequency corresponding to the spectral peak.

5. The auxiliary decision-making system based on data analysis of micro and small enterprises according to claim 4, characterized in that, At the frequency points corresponding to the main supply cycle, based on the hierarchical division of product structure information, the stratified phasors of each level are calculated, including: Determine a function to represent the hierarchical depth of a component relative to the central component, where the hierarchical depth is the number of edges in the shortest path from the central component to the component in the product structure information; Based on the hierarchy depth function, a set of components with a hierarchy of level d is divided from the component set of product structure information. The hierarchy depth of all components in this set is d. From the set of change events related to the central component, filter out the change events whose change objects belong to the component set at level d, and form a set of change event data at level d; For each change event in the change event data set of level d, perform calculations based on the change time of the change event and the angular frequency corresponding to the master supply cycle, sum all the calculation results, and obtain the layered phasor of level d.

6. The auxiliary decision-making system based on data analysis of micro and small enterprises according to claim 5, characterized in that, Interlayer complex gain is obtained based on the layered phasors of adjacent layers, and an interlayer transfer matrix containing the interlayer complex gain is constructed, including: Determine the maximum level depth, which is the maximum level depth of all components in the product structure information; For each level with a level depth from 0 to the maximum level depth minus one, calculate the inter-level complex gain, which is the ratio of the layer phasor with a level depth of the current level plus one to the layer phasor with a level depth of the current level. Construct an inter-layer transfer matrix, the elements of which are determined according to the following rules: When the column position of the matrix is ​​equal to the row position plus one, the element is the interlayer complex gain corresponding to the row position; when the column position of the matrix is ​​not equal to the row position plus one, the element is a complex number with both real and imaginary parts being zero.

7. The auxiliary decision-making system based on data analysis of micro and small enterprises according to claim 6, characterized in that, The maximum singular value of the interlayer transfer matrix is ​​taken as the surrogate quantity of the periodic phase-locked loop layer gain spectral radius, including: Determine the total number of layers, which is the maximum layer depth plus one; The inter-layer transfer matrix is ​​transposed to obtain the transpose of the inter-layer transfer matrix. The transpose process involves keeping the real part of each element in the original matrix unchanged, inverting the imaginary part, and then swapping the rows and columns of the matrix. Multiply the conjugate transpose of the interlayer transfer matrix with the interlayer transfer matrix to obtain the product matrix; Obtain all eigenvalues ​​of the product matrix, and select the eigenvalue with the largest value from all eigenvalues ​​as the largest eigenvalue of the product matrix. Calculate the square root of the largest eigenvalue of the product matrix, and take this square root as the largest singular value of the interlayer transfer matrix; The maximum singular value of the interlayer transfer matrix is ​​determined as the surrogate quantity of the periodic phase-locked layer gain spectrum radius.

8. The auxiliary decision-making system based on data analysis of micro and small enterprises according to claim 7, characterized in that, Compare the agency quantity with the preset safety line, determine the minimum cutting depth based on the comparison result, and only allow engineering changes not exceeding this depth, including: Compare the proxy amount of the periodic phase-locked layered gain spectrum radius with the preset safety line; The range of candidate cut depths is set based on the maximum level depth, which is all integers from 0 to the maximum level depth; For each candidate cut-off depth, extract all rows and columns from layer 0 to the layer corresponding to the candidate cut-off depth in the interlayer transfer matrix to form a submatrix corresponding to the candidate cut-off depth. Calculate the maximum singular value of the submatrix corresponding to each candidate cut-off depth; The minimum candidate cutting depth is selected from the candidate cutting depths, and the maximum singular value of the submatrix corresponding to the minimum cutting depth does not exceed the preset safety line. From the set of change events related to the central component, change events whose hierarchical depth does not exceed the minimum cutting depth are selected to form a set of engineering changes that can be released.

9. The auxiliary decision-making system based on data analysis of micro and small enterprises according to claim 8, characterized in that, Based on the phase information of interlayer complex gain, the effective time of layered phase misalignment is determined and converted into a preset granularity, including: Based on the interlayer complex gain, obtain the phase of the interlayer complex gain for each layer; Using the angular frequency and pi corresponding to the main supply period, for each level, subtract the phase of the inter-level complex gain of that level from pi, and then divide the result by the angular frequency corresponding to the main supply period to obtain the peak shifting time increment between that level and the next level. Set the cumulative off-peak time for level zero to zero. For each level after level zero, add the cumulative off-peak time of the previous level to the off-peak time increment corresponding to that previous level to obtain the cumulative off-peak time of the current level.

10. The auxiliary decision-making system based on data analysis of micro and small enterprises according to claim 9, characterized in that, The method for determining the phase shift of layered phases based on the phase information of interlayer complex gain and converting it into an effective time with a preset granularity also includes: Using a preset time window end time, preset granularity, and minimum cutting depth, for each level with a level depth not greater than the minimum cutting depth, calculate the ratio of cumulative staggered time to preset granularity, round the ratio to the nearest integer, and then multiply the resulting integer by the preset granularity to obtain the granular time adjustment amount; add the granular time adjustment amount to the preset time window end time to obtain the effective time of that level.

Images