Industrial data processing method for cloth dyeing and finishing parameters

Abstract

The present application relates to the field of industrial data processing, and particularly to a method for processing industrial data of cloth dyeing and finishing parameters. According to the dyeing and finishing kinetics mechanism, the continuous production cycle is divided into the heating, holding and cooling mechanism stages, and the corresponding time sequence data boundary features and process constraint features are extracted; a multi-dimensional correlation graph of time sequence segments, formula parameters and cloth attributes is constructed based on the above features, the physical storage offset of the time sequence data and the stage identifier are embedded in the time sequence segment nodes to form the bidirectional mapping pointers of the multi-dimensional correlation graph and the underlying time sequence storage interval; a search request containing the cloth physical property index and the target colorimetric parameter is received, the matching node is located in the graph, and the bidirectional mapping pointers are redirected to the physical address corresponding to the stage identifier in the underlying time sequence storage area. The present application eliminates the timestamp matching operation between heterogeneous databases, shortens the cross-dimension data search path, and guarantees the time sequence alignment and consistency of the formula parameters and the production running state data.

Description

Technical Field

[0001] This invention relates to the field of industrial data processing, specifically to an industrial data processing method for fabric dyeing and finishing parameters. Background Technology

[0002] Current data processing in the fabric dyeing and finishing industry typically employs a heterogeneous database architecture with separate storage. Relational databases store discrete formula parameters and fabric attribute data, while time-series databases store continuous temperature and humidity data collected by sensors in the dyeing and finishing equipment. When cross-dimensional correlation retrieval of historical dyeing and finishing parameters is required, the conventional approach first searches the relational database for matching formula records based on fabric properties and target color parameters, obtaining the corresponding timestamp intervals. These timestamp intervals are then sent to the time-series database, which iterates through the timestamp intervals to extract the corresponding production status data.

[0003] Existing technologies suffer from low efficiency in cross-dimensional association retrieval of heterogeneous databases and temporal misalignment of associated data. Since relational databases and time-series databases are independent, cross-database joins require transferring timestamp intervals and performing data alignment operations between the two heterogeneous storage systems. The continuity of timestamps in time-series databases and the discreteness of recipe records in relational databases increase the retrieval path length and computational overhead. Furthermore, industrial network transmission jitter can easily cause timestamp offsets, resulting in temporal misalignment when the extracted recipe parameters and production operation status data are associated. Summary of the Invention

[0004] The purpose of this invention is to provide an industrial data processing method for fabric dyeing and finishing parameters, which can effectively solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: Industrial data processing methods for fabric dyeing and finishing parameters include: Based on the dyeing and finishing kinetics mechanism, the continuous production cycle of fabric dyeing and finishing is divided into the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage. The temporal data boundary features and process constraint features corresponding to the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage are extracted respectively. Based on the time-series data boundary features and the process constraint features, a multi-dimensional correlation graph is constructed between time-series segment nodes, formula parameter nodes, and fabric attribute nodes. The physical storage offset and stage identifier of the time-series data are embedded in the time-series segment nodes of the multi-dimensional correlation graph to form a bidirectional mapping pointer between the multi-dimensional correlation graph and the underlying time-series storage area. The system receives a retrieval request containing fabric property indicators and target color parameters, locates a matching node in the formula parameter node and fabric attribute node of the multidimensional correlation graph, redirects the request to the physical address corresponding to the stage identifier in the underlying time-series storage area through the bidirectional mapping pointer, and extracts the time-series operation parameters corresponding to the heating mechanism stage, the heat preservation mechanism stage, or the cooling mechanism stage from the physical address.

[0006] Preferably, the step of extracting the time-series data boundary features and process constraint features corresponding to the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage respectively includes: Calculate the first and second derivative characteristics of temperature and humidity time series data within a sliding window during the continuous production cycle of fabric dyeing and finishing. Compare the first derivative characteristics with a preset dyeing and finishing kinetic stage transition threshold to determine the stage switching point between the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage. Based on the stage switching point, the time series data boundary features of the temperature and humidity time series data are extracted, and the process constraint features characterizing the chemical reaction rate of each mechanism stage are extracted in combination with the second derivative features.

[0007] Preferably, the step of constructing a multidimensional correlation graph between time-series nodes, recipe parameter nodes, and fabric attribute nodes includes: Using the fabric attribute node as the root node, the formula parameter node as the intermediate node, and the time sequence node as the leaf node, a directed edge is established between the root node, the intermediate node, and the leaf node according to the physical flow direction of the fabric dyeing and finishing process. The underlying time-series storage area is divided into data partitions corresponding to the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage, respectively. The starting physical storage offset of the data partition is embedded into the leaf node as the stage identifier to construct the node topology structure of the multidimensional association graph.

[0008] Preferably, the step of forming a bidirectional mapping pointer between the multidimensional correlation graph and the underlying temporal storage area includes: A logical pointer pointing to the logical address space is written into the time segment node of the multidimensional association graph, and a graph pointer pointing to the time segment node is written into the data block header of the underlying time storage area, thereby establishing a bidirectional mapping relationship between the logical pointer and the graph pointer. A logical-physical mapping table is constructed between the logical address space and the physical storage offset. When the underlying time-series storage area undergoes data reconstruction or location migration, the physical storage offset in the logical-physical mapping table is updated while keeping the logical pointer and the graph pointer fixed.

[0009] Preferably, the step of receiving a retrieval request containing fabric property indicators and target color parameters, and locating a matching node in the formula parameter node and the fabric attribute node of the multidimensional correlation graph, includes: The fabric physical property index and the target color parameter are converted into a multi-dimensional feature vector. The first similarity distance between the multi-dimensional feature vector and the fabric attribute node is calculated in the multi-dimensional association map. Target fabric attribute nodes whose first similarity distance meets the first preset threshold are selected. In the multidimensional association graph, traverse the recipe parameter nodes associated with the target fabric attribute node along the directed edge, calculate the second similarity distance between the target chromaticity coordinates in the retrieval request and the historical chromaticity coordinates recorded in the recipe parameter nodes, and determine the historical recipe parameter nodes whose second similarity distance satisfies the second preset threshold as matching nodes.

[0010] Preferably, the step of redirecting the bidirectional mapping pointer to the physical address corresponding to the stage identifier in the underlying timing storage area, and extracting the timing operation parameters corresponding to the heating mechanism stage, the heat preservation mechanism stage, or the cooling mechanism stage from the physical address includes: The corresponding logical-physical mapping table entries are obtained by using the bidirectional mapping pointer associated with the matching node, and the physical address ranges of the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage in the underlying time-series storage area are parsed out. According to the chronological order of the fabric dyeing and finishing process, the time-series operation parameter subsets of each mechanism stage are read from the physical address range. The timestamps of adjacent time-series operation parameter subsets at the switching point are compared, the data records corresponding to overlapping timestamps are removed, and the time-series operation parameter subsets after removing overlaps are spliced ​​into a continuous time-series operation parameter set.

[0011] Preferably, the step of calculating the first and second derivative characteristics of the temperature and humidity time-series data within a sliding window during the continuous production cycle of fabric dyeing and finishing includes: Perform multi-scale Gaussian smoothing on the temperature and humidity time series data to filter out high-frequency burrs in the temperature and humidity time series data that have a frequency higher than the preset dyeing and finishing physical change frequency; On the temperature and humidity time series data after multi-scale Gaussian smoothing, calculate the first and second derivative features of the center data point within the sliding window; The step of comparing the first derivative feature with a preset dyeing and finishing kinetics stage transition threshold to determine the stage switching point between the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage includes: When the first derivative feature crosses the stage transition threshold of the dyeing and finishing kinetics and the second derivative feature is in the extreme value range, the time point corresponding to the center data point of the sliding window is marked as the stage switching point.

[0012] Preferably, the step of establishing directed edges between the root node, the intermediate node, and the leaf node, using the formula parameter node as the intermediate node and the time sequence node as the leaf node, according to the physical flow direction of the fabric dyeing and finishing process, includes: The formula parameter node is split into a dye type node and a dye concentration node. The dye type node is used as a first-level intermediate node to connect to the fabric attribute node, and the dye concentration node is used as a second-level intermediate node to connect to the dye type node. Based on the dye chemical properties represented by the dye type nodes, the time segment nodes corresponding to dyes with different chemical properties are classified and connected to the dye concentration nodes, so that the time segment nodes of the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage form a heterogeneous branch topology in the multidimensional correlation graph according to the differences in dye chemical properties.

[0013] Preferably, when the underlying time-series storage area undergoes data reconstruction or location migration, the step of updating the physical storage offset in the logical-physical mapping table while keeping the logical pointer and the map pointer fixed includes: Set up a front-end logical-physical mapping table and a back-end logical-physical mapping table. The front-end logical-physical mapping table is used to respond to the retrieval request of the bidirectional mapping pointer. The back-end logical-physical mapping table is used to write the updated physical storage offset when data reconstruction or location migration occurs in the underlying time-series storage area. After the backend logical-physical mapping table completes all update operations, the frontend logical-physical mapping table and the backend logical-physical mapping table are atomically swapped, so that the retrieval request can obtain the original physical storage offset based on the unupdated frontend logical-physical mapping table during data reconstruction.

[0014] Preferably, the step of converting the fabric physical property index and the target chromaticity parameter into a multidimensional feature vector, and calculating the first similarity distance between the multidimensional feature vector and the fabric attribute node in the multidimensional association graph, includes: The physical weights of the fabric's physical properties are determined based on the fabric fiber color absorption dynamics, and the chromaticity weights of the target chromaticity parameters are determined based on the human eye perception characteristics of the color space. Based on the physical weights and the chromaticity weights, a normalized mapping is performed on the fabric property indicators and the target chromaticity parameters, and the normalized fabric property indicators and the target chromaticity parameters are combined into the multidimensional feature vector. In the multidimensional association graph, the weighted Mahalanobis distance between the multidimensional feature vector and the attribute vector of the fabric attribute node is calculated, and the weighted Mahalanobis distance is used as the first similarity distance.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention constructs a multi-dimensional correlation graph of "time series segment - formula parameters - fabric properties" and embeds the physical storage offset of time series data and stage identifiers into the graph nodes to form a bidirectional mapping pointer. This integrates discrete formula retrieval and continuous time series data extraction into a single graph traversal and pointer redirection operation. When a retrieval request is received and a matching node is located in the graph, the time series operation parameters are directly extracted by redirecting to the physical address of the corresponding stage identifier in the underlying time series storage area via the bidirectional mapping pointer. This eliminates timestamp matching and alignment operations between heterogeneous databases, shortens the cross-dimensional data retrieval path, and ensures the time series alignment and consistency between formula parameters and production operation status data.

[0016] 2. This invention establishes a front-end logical-physical mapping table and a back-end logical-physical mapping table, and performs atomic swaps after updating the back-end. When data reconstruction or location migration occurs in the underlying time-series storage area, it ensures that the retrieval request obtains the original physical storage offset based on the unupdated front-end mapping table, avoiding retrieval interruption and pointer invalidation during data migration. By comparing the timestamps of adjacent time-series runtime parameter subsets at the switching point and removing data records corresponding to overlapping timestamps, it eliminates data redundancy at the stage switching point and improves the continuity of the spliced ​​time-series runtime parameter set. Attached Figure Description

[0017] Figure 1 This is a flowchart of the overall method for processing industrial data of fabric dyeing and finishing parameters according to the present invention. Figure 2 This is a flowchart of the dyeing and finishing kinetics stage division and temporal feature extraction of the present invention; Figure 3 This is a flowchart of the multidimensional association graph construction and heterogeneous topology generation process of the present invention; Figure 4 This is a flowchart of the retrieval request feature conversion and map matching node localization process of the present invention; Figure 5 This is a flowchart of the bidirectional pointer redirection and timing parameter extraction and splicing process of the present invention; Figure 6 This is a flowchart of the front-end and back-end dual mapping table data migration fault tolerance processing of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please refer to Figure 1 This embodiment provides an industrial data processing method for fabric dyeing and finishing parameters. Based on the dyeing and finishing kinetics mechanism, the continuous production cycle of fabric dyeing and finishing is divided into a heating mechanism stage, a heat preservation mechanism stage, and a cooling mechanism stage. The method extracts the temporal data boundary features and process constraint features corresponding to each of the three stages. The dyeing and finishing kinetics mechanism describes the diffusion, adsorption, and fixation process of dye molecules within the fabric fibers. This process is strongly coupled with the dye liquor temperature, heat preservation time, and the physicochemical properties of the fabric fibers. The heating mechanism stage corresponds to the migration of dye molecules from the dye liquor to the fiber surface. In this stage, the temperature increases monotonically with time, and the diffusion rate of dye molecules increases exponentially with increasing temperature. The heat preservation mechanism stage corresponds to the diffusion of dye molecules from the fiber surface to the fiber interior and the completion of fixation. In this stage, the temperature remains near a preset constant value, and the diffusion rate of dye molecules reaches its maximum value and remains stable. The cooling mechanism stage corresponds to the process of the dye bath temperature dropping to room temperature. During this stage, the diffusion rate of dye molecules decreases exponentially with decreasing temperature until the distribution of dye molecules inside the fiber reaches equilibrium.

[0020] Specifically, refer to Figure 2 The first and second derivative features of temperature and humidity time-series data within a sliding window are calculated during a continuous fabric dyeing and finishing production cycle. The first derivative features are compared with preset dyeing and finishing kinetic stage transition thresholds to determine the stage switching points between the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage. Temperature and humidity time-series data are collected by temperature and humidity sensors installed inside the dyeing and finishing equipment at a fixed sampling frequency of 1Hz. Each sampling point includes three fields: timestamp, temperature value, and humidity value. The sliding window length is set to 11 sampling points, and the sliding step size is set to 1 sampling point to ensure that each data point is fully calculated. The first derivative feature characterizes the rate of change of temperature and humidity over time, while the second derivative feature characterizes the acceleration of the rate of change of temperature and humidity. The combination of these two features can accurately identify the switching moments of the dyeing and finishing kinetic stages.

[0021] Based on the stage switching points, the temporal data boundary features of the temperature and humidity time-series data are extracted, and combined with the second derivative features, process constraint features characterizing the chemical reaction rates of each mechanistic stage are extracted. The temporal data boundary features include the start time stamp, end time stamp, and duration of each mechanistic stage, while the process constraint features include the maximum temperature change rate, average temperature change rate, and temperature fluctuation range of each mechanistic stage. These features collectively constitute the kinetic characteristic set of the dyeing and finishing production process, providing fundamental data support for the subsequent construction of multidimensional correlation maps.

[0022] Based on the temporal data boundary features and the process constraint features, a multidimensional association graph is constructed between temporal segment nodes, formula parameter nodes, and fabric attribute nodes. The physical storage offset and stage identifier of the temporal data are embedded in the temporal segment nodes of the multidimensional association graph, forming a bidirectional mapping pointer between the multidimensional association graph and the underlying temporal storage area. The multidimensional association graph is represented using a directed graph structure, where each node represents an entity object, and each directed edge represents the association relationship between entity objects. The fabric attribute nodes store the physicochemical properties of the fabric, including fiber type, fiber fineness, fabric weight, fabric thickness, and fabric weave type. The formula parameter nodes store the formula information used in the dyeing and finishing process, including dye type, dye concentration, auxiliary agent type, auxiliary agent concentration, and dye liquor pH value. The temporal segment nodes store the temporal data metadata of each mechanistic stage in the dyeing and finishing process, including stage identifier, temporal data length, data sampling frequency, and physical storage offset.

[0023] Specifically, using the fabric attribute node as the root node, the formula parameter node as the intermediate node, and the time segment node as the leaf node, directed edges are established between the root node, the intermediate node, and the leaf node according to the physical flow direction of the fabric dyeing and finishing process. The physical flow direction of the fabric dyeing and finishing process is as follows: first, the physicochemical properties of the fabric are determined; then, a suitable dye formula is selected based on the fabric properties; finally, the dyeing and finishing production process is executed according to the dye formula, generating corresponding time-series operation data. Therefore, the direction of the directed edges is set from the fabric attribute node to the formula parameter node, and then from the formula parameter node to the time segment node, ensuring that the topology of the multidimensional correlation graph is consistent with the actual production process.

[0024] The underlying time-series storage area is divided into data partitions corresponding to the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage, respectively. The starting physical storage offset of each data partition is embedded as a stage identifier into the leaf node to construct the node topology of the multidimensional association graph. The underlying time-series storage area is organized using contiguous disk storage space, and each data partition stores all time-series operational data for the corresponding mechanism stage. The heating mechanism stage data partition stores the temperature and humidity time-series data collected during the heating process, the heat preservation mechanism stage data partition stores the temperature and humidity time-series data collected during the heat preservation process, and the cooling mechanism stage data partition stores the temperature and humidity time-series data collected during the cooling process. The starting physical storage offset of each data partition uniquely identifies the position of that data partition in the underlying time-series storage area. After embedding it into the time-series segment node, direct location from the graph node to the underlying storage can be achieved.

[0025] Logical pointers pointing to the logical address space are written into the time-series nodes of the multi-dimensional association graph, and graph pointers pointing to the time-series nodes are written into the header of the data blocks in the underlying time-series storage area, establishing a bidirectional mapping relationship between the logical pointers and the graph pointers. The logical pointer is a 64-bit unsigned integer that uniquely identifies the position of the time-series node in the logical address space. The graph pointer is also a 64-bit unsigned integer that uniquely identifies the position of the time-series node in the multi-dimensional association graph. This bidirectional mapping relationship ensures that data in the underlying time-series storage area can be located from nodes in the multi-dimensional association graph, and vice versa.

[0026] A logical-physical mapping table is constructed between the logical address space and the physical storage offset. When data reconstruction or location migration occurs in the underlying time-series storage area, the physical storage offset in the logical-physical mapping table is updated while keeping the logical pointer and the graph pointer unchanged. The logical-physical mapping table is a key-value hash table, where the key is the logical pointer and the value is the corresponding physical storage offset. When a data block in the underlying time-series storage area is migrated, only the physical storage offset value of the corresponding logical pointer in the logical-physical mapping table needs to be updated, without modifying the logical pointer in the multidimensional association graph or the graph pointer in the header of the underlying data block, thus achieving transparency of data migration to upper-layer retrieval operations.

[0027] The system receives a retrieval request containing fabric properties and target color parameters. It locates a matching node in the formula parameter node and fabric attribute node of the multidimensional correlation graph, redirects it to the physical address corresponding to the stage identifier in the underlying time-series storage area via the bidirectional mapping pointer, and extracts the time-series operating parameters corresponding to the heating mechanism stage, the heat preservation mechanism stage, or the cooling mechanism stage from the physical address. The retrieval request is initiated by a dyeing and finishing process engineer through an industrial control terminal, and includes the fabric properties of the fabric to be produced and the desired target color parameters. After receiving the retrieval request, the system first searches for the fabric attribute node and formula parameter node matching the requested parameters in the multidimensional correlation graph, then directly locates the corresponding physical address in the underlying time-series storage area via the bidirectional mapping pointer, and finally reads the required time-series operating parameters from that physical address.

[0028] Specifically, the fabric properties and the target chromaticity parameters are converted into multi-dimensional feature vectors. A first similarity distance between the multi-dimensional feature vectors and the fabric attribute nodes is calculated in the multi-dimensional correlation graph. Target fabric attribute nodes whose first similarity distance meets a first preset threshold are then selected. Fabric properties include fiber type, fiber fineness, fabric weight, fabric thickness, and fabric weave type. The target chromaticity parameters are represented using the CIELab color space, including three components: L value, a value, and b value. After converting these parameters into multi-dimensional feature vectors, the fabric attribute nodes most similar to the requested parameters can be quickly selected by calculating the similarity distance between the feature vectors.

[0029] In the multidimensional association graph, the formula parameter nodes associated with the target fabric attribute node are traversed along the directed edges. A second similarity distance is calculated between the target chromaticity coordinates in the search request and the historical chromaticity coordinates recorded in the formula parameter nodes. Historical formula parameter nodes whose second similarity distance satisfies a second preset threshold are identified as matching nodes. Each fabric attribute node is associated with multiple formula parameter nodes, and each formula parameter node corresponds to a specific dye formula and the actual chromaticity parameters of the fabric produced by that formula. By calculating the similarity distance between the target chromaticity coordinates and the historical chromaticity coordinates, historical formula parameter nodes that can produce fabrics closest to the target chromaticity can be selected.

[0030] By obtaining the corresponding logical-physical mapping table entries through the bidirectional mapping pointers associated with the matching nodes, the physical address ranges of the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage in the underlying time-series storage area are parsed out. Each recipe parameter node is associated with three time-series segment nodes, corresponding to the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage, respectively. Through the logical pointers embedded in these three time-series segment nodes, the corresponding physical storage offsets can be found in the logical-physical mapping table, thereby determining the physical address ranges of the time-series data for each mechanism stage in the underlying time-series storage area.

[0031] Following the chronological order of the fabric dyeing and finishing process, subsets of time-series operating parameters for each mechanistic stage are read from the physical address range. The timestamps of adjacent time-series operating parameter subsets at the switching points are compared, and data records corresponding to overlapping timestamps are removed. The time-series operating parameter subsets after removing overlaps are then concatenated into a continuous set of time-series operating parameters. Because sensors in the dyeing and finishing equipment may repeatedly collect data at stage switching points, timestamp overlap may occur between adjacent time-series operating parameter subsets at these points. By comparing the timestamps of adjacent subsets at the switching points and removing overlapping records, the concatenated set of time-series operating parameters can be ensured to have good continuity and consistency. A comparison of the characteristic parameters of the dyeing and finishing kinetic stages is shown in Table 1.

[0032] Table 1 Comparison of characteristic parameters in the dyeing and finishing kinetic stages warming phase Monotonic rise (0,2] [-0.5,0.5] 30-60 minutes Adsorption on the surface of dye molecules Insulation stage basically constant [-0.1,0.1] [-0.05,0.05] 60-120 minutes Internal diffusion of dye molecules cooling phase monotonically decreasing [-2,0) [-0.5,0.5] 20-40 minutes dye molecule fixation equilibrium Table 1 shows a comparison of characteristic parameters for the three kinetic stages in the dyeing and finishing process. During the heating stage, the temperature shows a monotonically increasing trend, with a positive first derivative ranging from (0,2) and a second derivative fluctuating between [-0.5, 0.5]. The typical duration is 30-60 minutes, primarily involving the adsorption reaction of dye molecules on the fabric fiber surface. During the heat preservation stage, the temperature remains relatively constant, with a first derivative close to zero ranging from [-0.1, 0.1] and a second derivative fluctuating between [-0.05, 0.05]. The typical duration is 60-120 minutes, primarily involving the diffusion reaction of dye molecules from the fiber surface to the fiber interior. During the cooling stage, the temperature shows a monotonically decreasing trend, with a negative first derivative ranging from [-2, 0) and a second derivative fluctuating between [-0.5, 0.5]. The typical duration is 20-40 minutes, primarily involving the fixation and equilibrium reaction of dye molecules within the fiber. These characteristic parameters provide a quantitative basis for the accurate division of dyeing and finishing kinetic stages.

[0033] In this embodiment, multi-scale Gaussian smoothing is performed on the temperature and humidity time-series data to filter out high-frequency noises in the data that are higher than the preset physical change frequency of dyeing and finishing. The physical change process of dyeing and finishing is relatively slow, with characteristic frequencies typically below 0.1Hz. However, the temperature and humidity time-series data collected by sensors often contains high-frequency noises caused by factors such as electrical noise and mechanical vibration. These high-frequency noises can interfere with subsequent derivative calculations and stage switching point identification. Multi-scale Gaussian smoothing, by performing Gaussian convolution on the time-series data at different scales, can effectively filter out high-frequency noise while retaining the low-frequency characteristics of the time-series data.

[0034] The mathematical expression for multi-scale Gaussian smoothing is: in, The scale is represented as The Gaussian kernel function, where t represents the time variable. This represents the scaling parameter of the Gaussian kernel. For the input temperature and humidity time series data... The output after multi-scale Gaussian smoothing is: in, This represents the convolution operation. Indicates the scale as The time series data is Gaussian smoothed. In this embodiment, three Gaussian kernels of different scales are used to smooth the time series data, with the scale parameters set as follows: The results of smoothing at three different scales are then weighted and averaged to obtain the final smoothed time series data. in, These are the weighting coefficients of the Gaussian smoothing results at three scales, and they satisfy... In this embodiment, the weighting coefficient is set to To balance the smoothing effect at different scales.

[0035] On the temperature and humidity time-series data after multi-scale Gaussian smoothing, the first and second derivative features of the center data point within the sliding window are calculated. The length of the sliding window is set to 2N+1 sampling points, where N is the half-length of the window. In this embodiment, N=5, so the sliding window length is 11 sampling points. For the i-th data point within the sliding window... The first and second derivatives are calculated using the central difference method.

[0036] The formula for calculating the first derivative is: in, In this embodiment, the sampling time interval is... Second The formula for calculating the second derivative is:

[0037] When the first derivative feature crosses the dyeing and finishing kinetics stage transition threshold and the second derivative feature is in the extreme value range, the time point corresponding to the center data point of the sliding window is marked as the stage switching point. The dyeing and finishing kinetics stage transition threshold includes a heating-holding transition threshold and a holding-cooling transition threshold. The heating-holding transition threshold is set to 0.1℃ / s. When the first derivative decreases from greater than 0.1℃ / s to less than 0.1℃ / s and the second derivative reaches a local minimum, it is marked as the switching point from the heating stage to the holding stage. The holding-cooling transition threshold is set to -0.1℃ / s. When the first derivative decreases from greater than -0.1℃ / s to less than -0.1℃ / s and the second derivative reaches a local minimum, it is marked as the switching point from the holding stage to the cooling stage.

[0038] The formulation parameter node is split into a dye type node and a dye concentration node. The dye type node serves as the first-level intermediate node connecting to the fabric attribute node, and the dye concentration node serves as the second-level intermediate node connecting to the dye type node. Dye type determines the chemical structure of dye molecules and the dyeing mechanism; different types of dyes have different optimal dyeing temperatures and holding times. Dye concentration determines the number of dye molecules in the dye bath, directly affecting the final color of the fabric. Splitting the formulation parameter node into dye type and dye concentration nodes allows for a more precise description of the composition and structure of the dyeing and finishing formulation, improving the retrieval accuracy of the multidimensional correlation map.

[0039] Based on the dye chemical properties characterized by the dye type nodes, the time-series nodes corresponding to dyes with different chemical properties are classified and connected to the dye concentration nodes. This allows the time-series nodes of the heating mechanism stage, the holding mechanism stage, and the cooling mechanism stage to form a heterogeneous branch topology in the multidimensional correlation graph based on the differences in dye chemical properties. Dyes with different chemical properties have different dyeing kinetics. For example, reactive dyes require higher holding temperatures and longer holding times, while direct dyes can be dyed at lower temperatures. Therefore, the time-series nodes corresponding to different dye types have different characteristic parameters. Classifying and connecting them to the corresponding dye concentration nodes allows the topology of the multidimensional correlation graph to better reflect the inherent laws of the dyeing and finishing process. The attributes of the time-series nodes corresponding to different dye types are shown in Table 2.

[0040] Table 2. Attributes of Time Segments Corresponding to Different Dye Types reactive dyes 1.0-1.5 60-80 60-90 0.5-1.0 Direct dyes 0.5-1.0 80-100 30-60 1.0-1.5 Disperse dyes 1.5-2.0 120-130 45-75 1.5-2.0 Acid dyes 1.0-1.5 80-90 45-60 1.0-1.5 Table 2 shows the time-series node attributes for four common dye types. For reactive dyes, the optimal heating rate is 1.0-1.5℃ / min, the optimal holding temperature is 60-80℃, the optimal holding time is 60-90 minutes, and the optimal cooling rate is 0.5-1.0℃ / min. For direct dyes, the optimal heating rate is 0.5-1.0℃ / min, the optimal holding temperature is 80-100℃, the optimal holding time is 30-60 minutes, and the optimal cooling rate is 1.0-1.5℃ / min. For disperse dyes, the optimal heating rate is 1.5-2.0℃ / min, the optimal holding temperature is 120-130℃, the optimal holding time is 45-75 minutes, and the optimal cooling rate is 1.5-2.0℃ / min. The optimal heating rate for acid dyes is 1.0-1.5℃ / min, the optimal holding temperature is 80-90℃, the optimal holding time is 45-60 minutes, and the optimal cooling rate is 1.0-1.5℃ / min. These properties provide a basis for constructing time-series nodes corresponding to different dye types and lay the foundation for the formation of heterogeneous branch topologies in multidimensional correlation maps.

[0041] In this embodiment, a front-end logical-physical mapping table and a back-end logical-physical mapping table are set up. The front-end logical-physical mapping table is used to respond to retrieval requests from the bidirectional mapping pointers, and the back-end logical-physical mapping table is used to write updated physical storage offsets when data reconstruction or location migration occurs in the underlying time-series storage area. The front-end and back-end logical-physical mapping tables have identical structures, both being key-value hash tables, where the key is a logical pointer and the value is the corresponding physical storage offset. Under normal operating conditions, the contents of the front-end and back-end logical-physical mapping tables are completely identical, and the system uses the front-end logical-physical mapping table to respond to all retrieval requests.

[0042] When the underlying time-series storage area needs to undergo data reconstruction or relocation, the system first creates a new backend logical-physical mapping table, and then copies the contents of the original backend logical-physical mapping table to the new one. Next, according to the data reconstruction or relocation plan, the system moves the data blocks in the underlying time-series storage area one by one, updating the physical storage offset values ​​of the corresponding logical pointers in the new backend logical-physical mapping table. Throughout the entire data migration process, the system still uses the frontend logical-physical mapping table to respond to retrieval requests; therefore, retrieval operations are not affected by the data migration.

[0043] After all updates to the backend logical-physical mapping table are completed, the frontend logical-physical mapping table is atomically swapped with the backend logical-physical mapping table. This allows the retrieval request to obtain the original physical storage offset based on the unupdated frontend logical-physical mapping table during data reconstruction. The atomic swap operation is indivisible, ensuring that no retrieval request accesses both the old and new mapping tables simultaneously during the swap. After the swap, the system begins using the new frontend logical-physical mapping table to respond to retrieval requests, while the original frontend logical-physical mapping table becomes the new backend logical-physical mapping table, awaiting the next data reconstruction or location migration operation.

[0044] The physical weights of the fabric properties are determined based on the color absorption kinetics of the fabric fibers, and the chromaticity weights of the target chromaticity parameters are determined based on the human eye's perception characteristics in the color space. The color absorption kinetics of the fabric fibers determine the degree of influence of different fabric properties on the dyeing effect. For example, fiber type has the greatest impact on the dyeing effect, followed by fiber fineness and fabric weight, while fabric thickness and fabric weave type have relatively smaller effects. Therefore, different physical weights are assigned to different fabric properties based on these differences in influence. The human eye's perception characteristics in the color space determine the sensitivity of the human eye to changes in different chromaticity parameters. For example, the human eye is most sensitive to changes in luminance (L*), followed by changes in red-green hue (a*), and relatively insensitive to changes in yellow-blue hue (b*). Therefore, different chromaticity weights are assigned to different chromaticity parameters based on the differences in human eye perception characteristics.

[0045] Based on the physical weights and the chromaticity weights, a normalization mapping is performed on the fabric property indices and the target chromaticity parameters. The normalized fabric property indices and the target chromaticity parameters are then combined into the multidimensional feature vector. The purpose of the normalization mapping is to convert parameters with different dimensions and value ranges into a unified numerical range for subsequent similarity calculations. For the fabric property indices, the min-max normalization method is used to map them to the [0,1] interval; for the target chromaticity parameters, the z-score normalization method is used to map them to the standard normal distribution interval.

[0046] The formula for calculating min-max normalization is: in, These are the original parameter values. This is the minimum value of the parameter. This is the maximum value of the parameter. These are the normalized parameter values.

[0047] The formula for calculating z-score normalization is: in, These are the original parameter values. This is the mean of the parameter. This is the standard deviation of the parameter. These are the normalized parameter values.

[0048] In the multidimensional association graph, the weighted Mahalanobis distance between the multidimensional feature vector and the attribute vector of the fabric attribute node is calculated, and the weighted Mahalanobis distance is used as the first similarity distance. The weighted Mahalanobis distance takes into account the correlation between the components of the feature vector and the difference in importance between the components, and can more accurately measure the similarity between two feature vectors.

[0049] The formula for calculating the weighted Mahalanobis distance is: in, To retrieve the requested multidimensional feature vector, This is the attribute vector for the cloth attribute node. This is the weight matrix. Let covariance matrix be the variance matrix. This is the inverse of the covariance matrix. The weight matrix. It is a diagonal matrix, where the elements on the diagonal are the weight values ​​of each feature component, and the elements off-diagonal are all 0. The expression for the weight matrix is: in, Let be the weight values ​​of each feature component, and satisfy . .

[0050] covariance matrix It describes the correlation between the components of the feature vector, and its elements Indicates the first The eigencomponent and the ... The covariance between the eigencomponents. The expression for the covariance matrix is: in, , Represents the mathematical expectation. For the first The mean of each feature component, For the first The mean of each characteristic component. The weighting of fabric properties and color parameters is shown in Table 3.

[0051] Table 3. Weighting of Fabric Properties and Color Parameters Fabric properties Fiber type 0.30 min-max Fabric properties Fiber fineness 0.20 min-max Fabric properties Fabric weight 0.15 min-max Fabric properties Fabric thickness 0.10 min-max Fabric properties Fabric weave type 0.05 min-max chromaticity parameters L* value 0.12 z-score chromaticity parameters a* value 0.05 z-score chromaticity parameters b* value 0.03 z-score Table 3 shows the weighting of fabric properties and colorimetric parameters. The total weight of the fabric properties is 0.80, with fiber type having the highest weight at 0.30, followed by fiber fineness (0.20), fabric weight (0.15), fabric thickness (0.10), and fabric weave type (0.05). The total weight of the colorimetric parameters is 0.20, with the L value having the highest weight at 0.12, followed by a* value (0.05) and b* value (0.03). The sum of the weights of all parameters is 1.00, ensuring the rationality of the weighted Mahalanobis distance calculation. The fabric properties were normalized using the min-max method, and the colorimetric parameters were normalized using the z-score method, respectively adapting to the distribution characteristics of different types of parameters.

[0052] In this embodiment, the weighted Mahalanobis distance is also used to calculate the second similarity distance between the target chromaticity coordinates in the retrieval request and the historical chromaticity coordinates recorded in the recipe parameter node. Since there is a certain correlation between the three components L*, a*, and b* of the chromaticity parameter, and the human eye has different sensitivities to different components, using the weighted Mahalanobis distance can more accurately measure the difference between two chromaticity coordinates.

[0053] For target chromaticity coordinates and historical chromaticity coordinates The weighted Mahalanobis distance between them is: in, This is the weight matrix for the chromaticity parameters. This is the covariance matrix of the chromaticity parameters. The weight matrix of the chromaticity parameters. for: in, The weight of the value is 0.6. The weight of the value is 0.25. The weight of each value is 0.15, and the sum of the weights is 1.0.

[0054] Covariance matrix of chromaticity parameters The typical value is obtained through statistical analysis of a large amount of historical chromaticity data:

[0055] The elements on the diagonal are the variances of the three components L*, a*, and b*, respectively, while the elements off-diagonal are the covariances among the components.

[0056] When the calculated weighted Mahalanobis distance is less than or equal to a second preset threshold, the historical formula parameter node is considered to match the target chromaticity parameter in the search request. The second preset threshold is set according to actual production needs, and in this embodiment, it is set to 2.0. When the weighted Mahalanobis distance is greater than 2.0, the fabric chromaticity produced by the historical formula parameter node is considered to have too large a difference from the target chromaticity and does not meet production requirements.

[0057] In this embodiment, when obtaining the corresponding logical-physical mapping table entry through the bidirectional mapping pointer associated with the matching node, the logical pointer value is first read from the timing segment node associated with the matching node. Then, the logical pointer value is used as the key to search in the logical-physical mapping table to obtain the corresponding physical storage offset value. The physical storage offset value represents the starting position of the timing data corresponding to the timing segment in the underlying timing storage area. Combined with the timing data length information stored in the timing segment node, the physical address range of the timing data in the underlying timing storage area can be determined.

[0058] For the time-series nodes in the heating mechanism stage, their physical address range is: ,in This represents the initial physical storage offset for the data during the heating phase. This represents the length of the data during the heating phase. For the time-series nodes in the heat preservation mechanism phase, their physical address range is... ,in This represents the initial physical storage offset for the data during the cooling phase. This represents the length of the data during the cooling phase.

[0059] Following the sequential order of the fabric dyeing and finishing process, subsets of time-series operating parameters for the heating, holding, and cooling stages are read from the three physical address ranges mentioned above. Because the sensors in the dyeing and finishing equipment may continuously collect multiple data points at stage switching points, timestamps may overlap between adjacent subsets of time-series operating parameters at those points. For example, the timestamp of the last data point in the heating stage may be the same as the timestamp of the first data point in the holding stage, or vice versa.

[0060] To eliminate data redundancy caused by timestamp overlap, it is necessary to compare the timestamps of adjacent time series operation parameter subsets at the switching point. Specifically, the timestamp of the last data point of the time series operation parameter subset during the heating phase needs to be obtained. The timestamp of the first data point in the subset of timing parameters for the heat preservation stage. .if Then the timestamp in the subset of time sequence operation parameters for the heat preservation stage will be less than or equal to All data records were removed. Similarly, the timestamp of the last data point of the subset of time-series operating parameters for the insulation stage was obtained. The timestamp of the first data point in the subset of timing parameters during the cooling phase. .if If the timestamp in the subset of timing parameters for the cooling phase is less than or equal to... All data records were removed.

[0061] The time-series operating parameters subsets for the heating, holding, and cooling stages, after removing overlapping data records, are concatenated in chronological order to form a continuous set of time-series operating parameters. This set of time-series operating parameters contains all temperature and humidity time-series data throughout the entire dyeing and finishing production cycle, and there are no timestamp overlaps or data redundancy, providing an accurate data foundation for subsequent dyeing and finishing process analysis and optimization.

[0062] In a preferred embodiment, when data reconstruction or location migration occurs in the underlying time-series storage area, the system generates a data migration log, recording the original physical storage offset and the new physical storage offset of each data block. The data migration log is stored in an append-only manner to ensure the traceability of the data migration process. If a system failure occurs during the data migration process, the system can be restored to the state before the failure based on the data migration log after restarting and continue to complete the remaining data migration operations.

[0063] Furthermore, the system periodically defragments the underlying time-series storage area, merging scattered time-series data blocks into contiguous storage blocks to improve data read and write performance. The defragmentation operation also employs a dual-mapping table mechanism (front-end and back-end) to ensure that normal retrieval requests are not affected during defragmentation. After defragmentation is complete, the system updates the back-end logical-physical mapping table and performs atomic swap operations, causing the front-end logical-physical mapping table to point to the new contiguous storage blocks.

[0064] In a preferred embodiment, the multidimensional association graph is stored using a distributed graph database, supporting parallel processing and querying of large-scale data. The distributed graph database distributes the nodes and edges of the multidimensional association graph across multiple physical nodes, ensuring that associated nodes and edges are stored on the same physical node through a graph partitioning algorithm, thus improving query efficiency. When a retrieval request is received, the distributed graph database distributes the query task to each physical node for parallel execution, then merges the query results from each physical node and returns the final matching node.

[0065] Furthermore, the system periodically performs incremental updates to the multidimensional correlation graph, adding newly generated dyeing and finishing production data to the graph. The incremental update operation includes creating new fabric attribute nodes, formula parameter nodes, and time series nodes, as well as establishing the relationships between these nodes. After the incremental update is complete, the system updates the bidirectional mapping pointers and the logical-physical mapping table to ensure that the newly added data can be retrieved correctly.

[0066] In a preferred embodiment, the system preprocesses the extracted time-series operating parameter set, including outlier detection and missing value imputation. Outlier detection employs the 3σ principle, marking data points in the time-series data that deviate from the mean by more than three standard deviations as outliers. Missing value imputation uses linear interpolation, calculating estimated values ​​for missing values ​​based on normal data points before and after the outliers. The preprocessed time-series operating parameter set has higher data quality, providing more reliable data support for subsequent dyeing and finishing process analysis and optimization.

[0067] This embodiment divides the continuous production cycle of fabric dyeing and finishing into three mechanistic stages based on the dyeing and finishing kinetics mechanism, and extracts the temporal data boundary features and process constraint features of each stage, providing accurate basic data for the subsequent construction of a multidimensional correlation graph. By constructing a multidimensional correlation graph containing fabric attribute nodes, formula parameter nodes, and temporal segment nodes, and embedding physical storage offsets and stage identifiers into the temporal segment nodes to form bidirectional mapping pointers, the organic integration of discrete formula data and continuous temporal data is achieved. By receiving retrieval requests and locating matching nodes in the multidimensional correlation graph, and then directly redirecting to the underlying temporal storage area via bidirectional mapping pointers to extract temporal operation parameters, the timestamp matching and alignment operations between heterogeneous databases are eliminated, shortening the cross-dimensional data retrieval path. By setting up dual logical-physical mapping tables at the front end and back end and performing atomic swap operations, the continuity and stability of retrieval requests are ensured when data reconstruction or location migration occurs in the underlying temporal storage area. By comparing the timestamps of adjacent time series operation parameter subsets at the switching point and removing overlapping records, data redundancy at the stage switching point is eliminated, improving the continuity and consistency of the spliced ​​time series operation parameter set.

[0068] In one embodiment, the collected temperature and humidity time-series data during the continuous production cycle of fabric dyeing and finishing is preprocessed to remove invalid data and outliers. Invalid data includes null values ​​caused by sensor malfunctions, values ​​exceeding physically reasonable ranges, and data with incorrect timestamps. Outliers include data points that significantly deviate from the normal trend due to factors such as electrical interference and mechanical vibration. The preprocessed temperature and humidity time-series data has higher data quality and can provide a reliable data foundation for subsequent derivative calculations and stage switching point identification.

[0069] Multi-scale Gaussian smoothing is performed on the preprocessed temperature and humidity time-series data to filter out high-frequency noise higher than the preset dyeing and finishing physical change frequency. The preset dyeing and finishing physical change frequency is set to 0.1Hz because the dyeing and finishing physical change process is relatively slow, and its characteristic frequency is usually lower than 0.1Hz. Multi-scale Gaussian smoothing effectively filters out high-frequency noise while retaining the low-frequency features of the time-series data by performing Gaussian convolution on the time-series data at different scales. In this embodiment, four Gaussian kernels of different scales are used to smooth the time-series data, and the scale parameters are set to... The results of smoothing at four different scales are then weighted and averaged to obtain the final smoothed time series data.

[0070] The weighting coefficients of the four-scale Gaussian smoothing results are set to and The sum of the weighting coefficients is 1.0. Smaller scale parameters can preserve the detailed features of time series data, while larger scale parameters can better filter out high-frequency noise. By weighted averaging the smoothing results at different scales, it is possible to effectively filter out high-frequency noise interference while preserving the important features of time series data.

[0071] On the multi-scale Gaussian smoothed temperature and humidity time-series data, the first and second derivative features of the center data point within the sliding window are calculated. The length of the sliding window is set to 15 sampling points, and the sliding step size is set to 1 sampling point. A longer sliding window can improve the stability of derivative calculation and reduce the influence of noise on the derivative calculation results. For the first derivative feature within the sliding window... Data points The first and second derivatives are calculated using the five-point central difference method to improve the accuracy of derivative calculation. The formula for calculating the first derivative using the five-point central difference method is:

[0072] The formula for calculating the second derivative using the five-point central difference method is:

[0073] Compared to the three-point central difference method, the five-point central difference method has higher computational accuracy and can more accurately capture the changing trends and accelerations of temperature and humidity time-series data. This is of great significance for accurately identifying the switching points in the dyeing and finishing kinetics stages.

[0074] When the first derivative feature crosses the dyeing and finishing kinetics stage transition threshold and the second derivative feature is in the extreme value range, the time point corresponding to the center data point of the sliding window is marked as the stage switching point. In this embodiment, the dyeing and finishing kinetics stage transition threshold is an adaptive threshold, rather than a fixed threshold. The adaptive threshold is dynamically adjusted according to the statistical characteristics of historical dyeing and finishing production data, which can better adapt to the dyeing kinetics characteristics of different fabric types and different dye formulations.

[0075] The adaptive threshold is calculated as follows: First, extract the first derivative values ​​at all switching points from the heating stage to the holding stage from historical dyeing and finishing production data, and then calculate the mean of these first derivative values. and standard deviation The heating-holding transfer threshold is then set to... in To adjust the coefficient, it is set to 1.5 in this embodiment. Similarly, the first derivative values ​​of all points where the temperature changes from the heat preservation stage to the cooling stage are extracted from historical dyeing and finishing production data, and the mean of these first derivative values ​​is calculated. and standard deviation The insulation-cooling transfer threshold is then set to... .

[0076] The extreme value range of the second derivative characteristic is also determined using an adaptive method. Second derivative values ​​at all stage switching points are extracted from historical dyeing and finishing production data, and the mean of these second derivative values ​​is calculated. and standard deviation Then the extreme value interval is set to When the second derivative characteristic is within this extreme value range, this point is considered to be a stage switching point.

[0077] By employing adaptive thresholds and adaptive extreme value ranges, the accuracy and robustness of stage switching point identification can be improved, reducing false positives and false negatives. This is of great significance for accurately dividing the dyeing and finishing kinetics stages and extracting the temporal data boundary features and process constraint features of each stage.

[0078] The temporal data boundary features of the temperature and humidity time-series data are extracted based on the stage switching points, and process constraint features characterizing the chemical reaction rates of each mechanistic stage are extracted in conjunction with the second derivative features. The temporal data boundary features include the start time stamp, end time stamp, duration, and stage switching point timestamp for each mechanistic stage. The process constraint features include the maximum temperature change rate, average temperature change rate, temperature fluctuation range, maximum humidity change rate, average humidity change rate, and humidity fluctuation range for each mechanistic stage.

[0079] Furthermore, combining second-derivative features can extract more refined process constraint features, such as the temperature rise acceleration during the heating stage, the temperature stability index during the holding stage, and the temperature fall acceleration during the cooling stage. These process constraint features can more comprehensively describe the kinetic characteristics of the dyeing and finishing production process, providing richer data support for the subsequent construction of multidimensional correlation maps and the optimization of dyeing and finishing processes.

[0080] refer to Figure 3 Using the fabric attribute node as the root node, the formula parameter node as the intermediate node, and the time sequence node as the leaf node, directed edges are established between the root node, the intermediate node, and the leaf node according to the physical flow direction of the fabric dyeing and finishing process. In this embodiment, the fabric attribute node not only stores the basic physicochemical properties of the fabric but also stores the fabric's pretreatment information, such as parameter information for pretreatment processes like singeing, desizing, scouring, and bleaching. This pretreatment information has a significant impact on the subsequent dyeing process; storing it in the fabric attribute node can improve the retrieval accuracy of the multidimensional correlation graph.

[0081] The formulation parameter node is split into a dye type node and a dye concentration node. The dye type node is used as a first-level intermediate node to connect to the fabric attribute node, and the dye concentration node is used as a second-level intermediate node to connect to the dye type node. In this embodiment, the dye type node not only stores the chemical name and CAS number of the dye, but also stores detailed chemical attribute information such as the dye's molecular structure, molecular weight, solubility, and optimal dyeing pH value. This chemical attribute information determines the dyeing kinetics of the dye. Storing it in the dye type node can provide a more accurate basis for classifying the time-series nodes corresponding to different dye types.

[0082] The dye concentration node stores the concentration information of various dyes in the dye bath, including the concentration of the primary dye and the concentration of auxiliary dyes. For mixed dye formulations, the dye concentration node stores the specific concentration value of each dye and the proportional relationship between them. This information is crucial for accurately reproducing dyeing and finishing formulations and producing fabrics that meet the target color requirements.

[0083] Based on the dye chemical properties characterized by the dye type nodes, the time-series nodes corresponding to dyes with different chemical properties are classified and connected to the dye concentration nodes. This allows the time-series nodes of the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage to form a heterogeneous branch topology in the multidimensional correlation graph based on the differences in dye chemical properties. In this embodiment, dyes are classified into six major categories according to their application: reactive dyes, direct dyes, disperse dyes, acid dyes, vat dyes, and sulfur dyes. Each major category of dyes can be further subdivided into multiple subcategories based on their chemical structure. For example, reactive dyes can be divided into triazine reactive dyes, vinyl sulfone reactive dyes, and mixed reactive dyes.

[0084] Dyes with different chemical structures exhibit different dyeing kinetics, thus their corresponding time-series nodes possess different characteristic parameters. For example, triazine reactive dyes have high reactivity, lower optimal holding temperatures, and shorter holding times; while vinyl sulfone reactive dyes have lower reactivity, higher optimal holding temperatures, and longer holding times. Classifying and connecting the time-series nodes corresponding to dyes with different chemical structures to their corresponding dye concentration nodes allows the topological structure of the multidimensional correlation map to more precisely reflect the inherent laws of the dyeing and finishing process, improving the accuracy of the retrieval results.

[0085] In this embodiment, each time segment node stores not only basic information such as stage identifier, time data length, data sampling frequency, and physical storage offset, but also the corresponding process parameter information, such as the starting temperature, ending temperature, and heating rate of the heating stage; the holding temperature and holding time of the holding stage; and the starting temperature, ending temperature, and cooling rate of the cooling stage. These process parameters are the core control parameters of the dyeing and finishing production process. Storing them in the time segment nodes allows for direct retrieval of complete dyeing and finishing process parameters after finding a matching time segment node, eliminating the need for further queries from other data sources.

[0086] The underlying time-series storage area is divided into data partitions corresponding to the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage, respectively. The starting physical storage offset of each data partition is embedded as a stage identifier into the leaf node to construct the node topology of the multidimensional correlation graph. In this embodiment, the underlying time-series storage area is organized using a columnar storage format, with each data partition storing temperature and humidity time-series data column-wise. The columnar storage format has a higher data compression ratio and faster query speed, making it particularly suitable for the storage and analysis of time-series data.

[0087] Each data partition is further divided into multiple data blocks, each with a size of 64MB. The data block is the basic read / write unit of the underlying time-series storage area. Each data block contains a data block header and a data block body. The data block header stores the data block's metadata, including the checksum, creation time, data block length, and a graph pointer to the time-series node. The data block body stores the actual temperature and humidity time-series data, compressed using a compression algorithm to save storage space.

[0088] Logical pointers pointing to logical address spaces are written into the time-series nodes of the multi-dimensional correlation graph, and graph pointers pointing to the time-series nodes are written into the header of the data blocks in the underlying time-series storage area, establishing a bidirectional mapping relationship between the logical pointers and the graph pointers. In this embodiment, the logical pointers are designed with a hierarchical structure, including three parts: partition number, block number, and intra-block offset. The partition number identifies the mechanism stage data partition to which the data belongs, the block number identifies the data block to which the data belongs, and the intra-block offset identifies the starting position of the data within the data block. This hierarchical logical pointer structure can more efficiently locate data in the underlying time-series storage area.

[0089] The graph pointer also employs a hierarchical structure, comprising three parts: node type, node number, and version number. The node type identifies the node's type, such as a fabric attribute node, dye type node, dye concentration node, or time series node. The node number uniquely identifies the node within that type. The version number identifies the node's version, incrementing as the node's content is updated. This hierarchical graph pointer structure allows for more efficient location of nodes within a multidimensional relational graph.

[0090] A logical-physical mapping table is constructed between the logical address space and the physical storage offset. When data reconstruction or location migration occurs in the underlying time-series storage area, the physical storage offset in the logical-physical mapping table is updated while keeping the logical pointer and the graph pointer unchanged. In this embodiment, the logical-physical mapping table is organized using a B+ tree structure instead of a hash table structure. The B+ tree structure has better range query performance and higher storage efficiency, making it particularly suitable for storing and managing large-scale logical-physical mapping relationships. Each leaf node of the B+ tree stores a mapping relationship from a logical pointer to a physical storage offset, and the leaf nodes are connected by a doubly linked list to facilitate range queries. The non-leaf nodes of the B+ tree store index information for quickly locating the corresponding leaf node. When the data block in the underlying time-series storage area is migrated, only the physical storage offset value of the corresponding logical pointer in the B+ tree needs to be updated, without modifying the logical pointer in the multidimensional association graph or the graph pointer at the head of the underlying data block.

[0091] refer to Figure 5The system receives a retrieval request containing fabric property indicators and target color parameters. It locates a matching node among the formula parameter nodes and fabric attribute nodes in the multidimensional correlation graph. The system redirects the matching node to the physical address corresponding to the stage identifier in the underlying time-series storage area via the bidirectional mapping pointer. From the physical address, it extracts the time-series operating parameters corresponding to the heating mechanism stage, the heat preservation mechanism stage, or the cooling mechanism stage. In this embodiment, the retrieval request may also include preferred process constraints, such as maximum heating rate, minimum heat preservation temperature, and maximum heat preservation time. When locating a matching node, the system considers these process constraints simultaneously and filters out historical formula parameter nodes that satisfy all constraints.

[0092] refer to Figure 4 The fabric physical properties and the target color parameters are converted into multi-dimensional feature vectors. A first similarity distance between the multi-dimensional feature vectors and the fabric attribute nodes is calculated in the multi-dimensional correlation graph. Target fabric attribute nodes whose first similarity distance satisfies a first preset threshold are then selected. In this embodiment, the fabric physical properties also include fabric pretreatment process parameters, such as singeing temperature, desizing agent concentration, scouring temperature, and bleaching agent concentration. These pretreatment process parameters have a significant impact on the dyeing effect; incorporating them into the multi-dimensional feature vector can improve the accuracy of fabric attribute node matching.

[0093] The physical weights of the fabric's physical properties are determined based on the fabric fiber's color absorption kinetics, and the chromaticity weights of the target chromaticity parameters are determined based on the human eye's perception characteristics of the color space. In this embodiment, the physical and chromaticity weights employ an adaptive weight adjustment mechanism, dynamically adjusting the weight values ​​based on feedback information from historical search results. When the matching result of a certain parameter has a high correlation with actual production results, the weight value of that parameter is increased; when the matching result of a certain parameter has a low correlation with actual production results, the weight value of that parameter is decreased. This adaptive weight adjustment mechanism can continuously optimize the accuracy of similarity calculation and improve the quality of search results.

[0094] Based on the physical weights and the chromaticity weights, a normalized mapping is performed on the fabric property indicators and the target chromaticity parameters. The normalized fabric property indicators and the target chromaticity parameters are then combined into the multidimensional feature vector. In this embodiment, for parameters of classification types, such as fiber type, fabric weave type, and dye type, one-hot encoding is used to convert them into numerical vectors before normalized mapping. One-hot encoding can convert categorical variables into numerical forms that can be processed by machine learning algorithms, while avoiding the influence of ordinal relationships between categorical variables on similarity calculation.

[0095] In the multidimensional association graph, the weighted Mahalanobis distance between the multidimensional feature vector and the attribute vector of the fabric attribute node is calculated, and this weighted Mahalanobis distance is used as the first similarity distance. In this embodiment, the covariance matrix is ​​maintained using an incremental update method. When new fabric attribute data is added, the system automatically updates the element values ​​of the covariance matrix without recalculating the entire covariance matrix. This incremental update method can improve the update efficiency of the covariance matrix and ensure that the covariance matrix can reflect changes in data distribution in a timely manner.

[0096] In the multidimensional association graph, the system traverses the formula parameter nodes associated with the target fabric attribute node along the directed edges, calculates the second similarity distance between the target chromaticity coordinates in the retrieval request and the historical chromaticity coordinates recorded in the formula parameter nodes, and determines the historical formula parameter nodes whose second similarity distance satisfies a second preset threshold as matching nodes. In this embodiment, the system sorts the matched historical formula parameter nodes and returns the top N matching results in ascending order of the second similarity distance, where N is the number of results specified by the user. The user can select the most suitable historical formula parameter node from these matching results to view its corresponding time-series operating parameters and process parameters.

[0097] The corresponding logical-physical mapping entries are obtained through the bidirectional mapping pointer associated with the matching node, and the physical address ranges of the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage in the underlying time-series storage area are parsed out. In this embodiment, the system performs a prefetch operation on the physical address range, loading potentially accessed data blocks into the memory cache in advance to improve data read speed. The prefetch operation is based on the principle of locality of reference; when accessing a data block, the system prefetches several adjacent data blocks, reducing the number of disk I / O operations.

[0098] Following the chronological order of the fabric dyeing and finishing process, subsets of time-series operational parameters for each mechanistic stage are read from the physical address range. The timestamps of adjacent time-series operational parameter subsets at the switching points are compared, and data records corresponding to overlapping timestamps are removed. The time-series operational parameter subsets after removing overlaps are then concatenated into a continuous set of time-series operational parameters. In this embodiment, the system also performs interpolation processing on the concatenated time-series operational parameter set to fill in any missing data points. The interpolation processing uses cubic spline interpolation, which generates smooth interpolation curves, maintaining the continuity and integrity of the time-series data.

[0099] This embodiment improves the accuracy of derivative calculation by employing the five-point central difference method to calculate the first and second derivative features; it enhances the accuracy and robustness of stage division by using adaptive thresholds and adaptive extreme value intervals to identify stage switching points; it improves the retrieval accuracy of the multidimensional correlation map by incorporating fabric preprocessing information and detailed chemical property information of dyes; it improves data storage efficiency and query speed by using a columnar storage format and a logical-physical mapping table with a B+ tree structure; and it continuously optimizes the accuracy of similarity calculation by employing an adaptive weight adjustment mechanism and an incrementally updated covariance matrix. The comprehensive application of these techniques further enhances the efficiency and accuracy of industrial data processing for fabric dyeing and finishing parameters.

[0100] In one embodiment, multi-source sensor data is collected during the continuous production cycle of fabric dyeing and finishing. In addition to time-series data on temperature and humidity, this includes time-series data on dye liquor pH, dye liquor conductivity, dye concentration, stirring speed, and steam pressure. This multi-source sensor data can more comprehensively describe the state of the dyeing and finishing production process, providing richer data support for the division of dyeing and finishing kinetic stages and the extraction of process constraint characteristics.

[0101] The acquired multi-source sensor data undergoes synchronization processing to ensure that the timestamps of all sensor data remain consistent. Since different sensors may have different sampling frequencies, data from sensors with lower sampling frequencies need to be interpolated and resampled to the same sampling frequency as the temperature and humidity sensors. The synchronized multi-source sensor data has a unified time reference, accurately reflecting the state of each process parameter at the same moment.

[0102] Multi-scale Gaussian smoothing is performed on the synchronized multi-source sensor data to filter out high-frequency noise. Different scale parameters and weighting coefficients are used for smoothing different types of sensor data to adapt to the noise characteristics of different sensor data. For example, pH sensor data has relatively low noise and can use a smaller scale parameter, while conductivity sensor data has relatively high noise and requires a larger scale parameter.

[0103] On multi-source sensor data after multi-scale Gaussian smoothing, the first and second derivative features of each sensor data point within a sliding window are calculated. These features are then combined into a multi-dimensional derivative feature vector to identify the switching points in the dyeing and finishing kinetics. Compared to using only the derivative features of temperature and humidity data, using the multi-dimensional derivative feature vector from multi-source sensor data more accurately identifies these switching points and reduces the likelihood of misjudgments.

[0104] The multidimensional derivative feature vector is compared with the preset dyeing and finishing kinetics stage transition feature vector, and the cosine similarity between them is calculated. When the cosine similarity is greater than or equal to the preset similarity threshold, the time point corresponding to the center data point of the sliding window is marked as the stage switching point. The dyeing and finishing kinetics stage transition feature vector is obtained through statistical analysis of a large amount of historical dyeing and finishing production data, and contains the typical derivative features of all sensor data at each stage switching point.

[0105] Based on the phase switching points, time-series data boundary features of multi-source sensor data are extracted, and process constraint features characterizing the chemical reaction rates of each mechanistic stage are extracted by combining multidimensional derivative feature vectors. These process constraint features include the rates of temperature change, humidity change, pH change, conductivity change, and dye concentration change at each mechanistic stage. These process constraint features can more comprehensively describe the kinetic characteristics of the dyeing and finishing production process, providing a richer data foundation for the subsequent construction of multidimensional correlation maps.

[0106] Using the fabric attribute node as the root node, the formula parameter node as the intermediate node, and the time sequence node as the leaf node, directed edges are established between the root node, the intermediate node, and the leaf node according to the physical flow direction of the fabric dyeing and finishing process. In this embodiment, the formula parameter node includes not only dye type and dye concentration nodes, but also auxiliary agent type and auxiliary agent concentration nodes. Auxiliaries play an important role in the dyeing and finishing production process. For example, dyeing accelerators can improve the dyeing rate, dyeing retarders can make the dyeing uniform, and fixing agents can improve the color fastness of the dye. Including auxiliary agent type and auxiliary agent concentration in the formula parameter node can more completely describe the composition of the dyeing and finishing formula and improve the retrieval accuracy of the multidimensional correlation graph.

[0107] The formulation parameter node is broken down into dye type node, dye concentration node, auxiliary agent type node, and auxiliary agent concentration node. The dye type node and auxiliary agent type node are connected to the fabric property node as first-level intermediate nodes. The dye concentration node and auxiliary agent concentration node are connected to the auxiliary agent type node as second-level intermediate nodes. Then, the time sequence node is connected to both the dye concentration node and auxiliary agent concentration node as leaf nodes. This more refined formulation parameter node structure better reflects the synergistic effect between dyes and auxiliaries, improving the accuracy of dyeing and finishing formulation reproduction.

[0108] Based on the chemical properties represented by the dye type nodes and auxiliary agent type nodes, the time-series nodes corresponding to dye and auxiliary agent combinations with different chemical properties are classified and connected to the dye concentration nodes and auxiliary agent concentration nodes. This allows the time-series nodes of the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage to form a more complex heterogeneous branch topology in the multidimensional correlation graph based on the differences in the chemical properties of the dyes and auxiliaries. Different dye and auxiliary agent combinations have different dyeing kinetic characteristics, and therefore the corresponding time-series nodes have different characteristic parameters. Classifying and connecting the time-series nodes corresponding to different combinations enables the topology of the multidimensional correlation graph to more accurately reflect the inherent laws of the dyeing and finishing process.

[0109] The underlying time-series storage area is divided into data partitions corresponding to the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage, respectively. Each data partition is further divided into multiple sub-partitions, and each sub-partition stores one type of sensor data. For example, the data partition for the heating mechanism stage includes temperature sub-partitions, humidity sub-partitions, pH value sub-partitions, conductivity sub-partitions, etc. This partitioned storage method can improve the query efficiency of specific types of sensor data. When only one type of time-series data needs to be queried, the corresponding sub-partition can be accessed directly without reading other types of data.

[0110] Logical pointers pointing to logical address spaces are written into the time-series nodes of the multi-dimensional correlation graph, and graph pointers pointing to the time-series nodes are written into the header of the data blocks in the underlying time-series storage area, establishing a bidirectional mapping relationship between the logical pointers and the graph pointers. In this embodiment, each time-series node contains multiple logical pointers, each pointing to the logical address space of different types of sensor data. For example, a time-series node may contain logical pointers for temperature, humidity, pH, and conductivity. Through these logical pointers, the physical addresses of different types of sensor data in the underlying time-series storage area can be located respectively.

[0111] A logical-physical mapping table is constructed between the logical address space and the physical storage offset. When data reconstruction or location migration occurs in the underlying time-series storage area, the physical storage offset in the logical-physical mapping table is updated while keeping the logical pointer and the map pointer fixed. In this embodiment, a separate logical-physical mapping table is established for each type of sensor data, such as a temperature logical-physical mapping table, a humidity logical-physical mapping table, and a pH value logical-physical mapping table. This typed logical-physical mapping table improves the query efficiency of the mapping table and also facilitates independent management and maintenance of different types of data.

[0112] refer to Figure 6A front-end logical-physical mapping table and a back-end logical-physical mapping table are set up. The front-end logical-physical mapping table is used to respond to retrieval requests from the bidirectional mapping pointer, and the back-end logical-physical mapping table is used to write updated physical storage offsets when data reconstruction or location migration occurs in the underlying time-series storage area. In this embodiment, a front-end and back-end dual logical-physical mapping table are set up separately for each type of sensor data. When a certain type of sensor data needs to be reconstructed or migrated, only the corresponding type of back-end logical-physical mapping table needs to be updated, without affecting the retrieval operations of other types of data. This type-specific dual mapping table mechanism can improve the availability and reliability of the system.

[0113] After all updates to the backend logical-physical mapping table are completed, the frontend logical-physical mapping table is atomically swapped with the backend logical-physical mapping table. This allows the retrieval request to obtain the original physical storage offset based on the unupdated frontend logical-physical mapping table during data reconstruction. The atomic swap operation is implemented using database transactions to ensure the atomicity, consistency, isolation, and durability of the swap process. If a system failure occurs during the swap, the system can automatically roll back to the state before the swap after restarting, preventing data inconsistencies.

[0114] The system receives a retrieval request containing fabric property indicators and target color parameters. It locates a matching node in the formula parameter node and fabric attribute node of the multidimensional correlation graph, and redirects it to the physical address corresponding to the stage identifier in the underlying time-series storage area via the bidirectional mapping pointer. From the physical address, it extracts the time-series operating parameters corresponding to the heating mechanism stage, the heat preservation mechanism stage, or the cooling mechanism stage. In this embodiment, the retrieval request can also specify the sensor data type to be extracted, such as extracting only temperature and humidity time-series data, or extracting all types of sensor data. The system will locate the physical address in the underlying time-series storage area via the corresponding logical pointer based on the sensor data type specified in the request, and extract the corresponding time-series operating parameters.

[0115] The fabric physical properties and the target color parameters are converted into multi-dimensional feature vectors. A first similarity distance between the multi-dimensional feature vectors and the fabric attribute nodes is calculated in the multi-dimensional correlation graph. Target fabric attribute nodes whose first similarity distance meets a first preset threshold are then selected. In this embodiment, the fabric physical properties also include color fastness requirements, such as wash fastness, rubbing fastness, and sunlight fastness. These color fastness requirements are important quality indicators in dyeing and finishing production. Incorporating them into the multi-dimensional feature vector allows for the selection of fabric attribute nodes and corresponding formula parameter nodes that meet the color fastness requirements.

[0116] The physical weights of the fabric's physical properties are determined based on the fabric fiber's color absorption kinetics, and the chromaticity weights of the target chromaticity parameters are determined based on the human eye's perception characteristics of the color space. In this embodiment, the physical weights also include the weights of colorfastness requirements, with different weight values ​​assigned according to the importance of different colorfastness requirements. For example, for clothing fabrics, wash fastness and rubbing fastness have higher weights; while for curtain fabrics, sunlight fastness has a higher weight.

[0117] Based on the physical weights and the chromaticity weights, a normalized mapping is performed on the fabric property indicators and the target chromaticity parameters. The normalized fabric property indicators and the target chromaticity parameters are then combined into the multidimensional feature vector. In this embodiment, for graded parameters such as color fastness requirements, ordinal encoding is used to convert them into numerical vectors before normalized mapping. Ordinal encoding can preserve the ordinal relationships between graded parameters, making similarity calculation more reasonable.

[0118] In the multidimensional association graph, the weighted Mahalanobis distance between the multidimensional feature vector and the attribute vector of the fabric attribute node is calculated, and the weighted Mahalanobis distance is used as the first similarity distance. In this embodiment, the calculation of the weighted Mahalanobis distance also considers the nonlinear correlation between the components of the feature vector. By introducing a kernel function, the original feature space is mapped to a high-dimensional feature space, and a linear Mahalanobis distance is calculated in the high-dimensional feature space, thereby capturing the nonlinear correlation in the original feature space.

[0119] The kernel function used is the Gaussian kernel function, whose expression is: in, Let X be the Euclidean distance between the feature vectors X and Y. This represents the bandwidth parameter of the Gaussian kernel.

[0120] After mapping the original feature vectors to a high-dimensional feature space using a kernel function, the weighted Mahalanobis distance is calculated in the high-dimensional feature space: in, and They are the feature vectors and Mapping in high-dimensional feature space Let be the covariance matrix in the high-dimensional feature space.

[0121] This kernel-weighted Mahalanobis distance can more accurately measure the similarity between feature vectors with non-linear correlation, improving the accuracy of fabric attribute node matching.

[0122] In the multidimensional association graph, the system traverses the formula parameter nodes associated with the target fabric attribute node along the directed edges, calculates the second similarity distance between the target chromaticity coordinates in the search request and the historical chromaticity coordinates recorded in the formula parameter nodes, and determines the historical formula parameter nodes whose second similarity distance meets a second preset threshold as matching nodes. In this embodiment, the system also calculates the similarity between the color fastness index corresponding to the matched historical formula parameter node and the color fastness requirement in the search request, and uses the color fastness similarity as the third similarity distance. Only when both the second and third similarity distances meet their respective preset thresholds is the historical formula parameter node determined as the final matching node.

[0123] The system obtains the corresponding logical-physical mapping entries through the bidirectional mapping pointer associated with the matching node, and parses the physical address ranges of the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage in the underlying time-series storage area. In this embodiment, the system obtains the physical address ranges of the corresponding sensor data type according to the sensor data type specified in the request. For example, if the request specifies the extraction of temperature and humidity time-series data, the system obtains the physical address ranges of the temperature data and humidity data through the temperature logical pointer and humidity logical pointer, respectively.

[0124] Following the chronological order of the fabric dyeing and finishing process, subsets of time-series operating parameters for each mechanistic stage are read from the physical address range. The timestamps of adjacent time-series operating parameter subsets at the switching points are compared, and data records corresponding to overlapping timestamps are removed. The time-series operating parameter subsets after removing overlaps are then concatenated into a continuous set of time-series operating parameters. In this embodiment, for each type of sensor data, timestamp comparison, overlapping data removal, and time-series concatenation operations are performed to generate their respective continuous sets of time-series operating parameters. Then, all types of time-series operating parameter sets are aligned according to their timestamps to form a multi-source time-series operating parameter matrix. Each row of the multi-source time-series operating parameter matrix corresponds to a timestamp, and each column corresponds to a type of sensor data, which can intuitively display the changes of various process parameters over time and their interrelationships.

[0125] This embodiment comprehensively describes the state of the dyeing and finishing production process by collecting and processing multi-source sensor data; it improves the accuracy of stage division by using multi-dimensional derivative feature vectors and cosine similarity to identify stage switching points; it describes the composition of the dyeing and finishing formula more completely by incorporating auxiliary agent type and concentration into the formula parameter node; it improves data management efficiency and query speed by using categorized data partitioning and logical-physical mapping tables; it improves the accuracy of similarity calculation by introducing kernel-weighted Mahalanobis distance to capture nonlinear correlations between feature vectors; and it provides richer data support for the analysis and optimization of the dyeing and finishing process by generating a multi-source time-series operating parameter matrix. The comprehensive application of these technologies further improves the industrial data processing method for fabric dyeing and finishing parameters, enabling it to better meet the needs of complex dyeing and finishing production processes.

Claims

1. A method for industrial data processing of dyeing and finishing parameters of cloth, characterized in that, include: Based on the dyeing and finishing kinetics mechanism, the continuous production cycle of fabric dyeing and finishing is divided into the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage. The temporal data boundary features and process constraint features corresponding to the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage are extracted respectively. Based on the time-series data boundary features and the process constraint features, a multi-dimensional correlation graph is constructed between time-series segment nodes, formula parameter nodes, and fabric attribute nodes. The physical storage offset and stage identifier of the time-series data are embedded in the time-series segment nodes of the multi-dimensional correlation graph to form a bidirectional mapping pointer between the multi-dimensional correlation graph and the underlying time-series storage area. The system receives a retrieval request containing fabric property indicators and target color parameters, locates a matching node in the formula parameter node and fabric attribute node of the multidimensional correlation graph, redirects the request to the physical address corresponding to the stage identifier in the underlying time-series storage area through the bidirectional mapping pointer, and extracts the time-series operation parameters corresponding to the heating mechanism stage, the heat preservation mechanism stage, or the cooling mechanism stage from the physical address.

2. The industrial data processing method for fabric dyeing and finishing parameters according to claim 1, characterized in that, The steps for extracting the time-series data boundary features and process constraint features corresponding to the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage respectively include: Calculate the first and second derivative characteristics of temperature and humidity time series data within a sliding window during the continuous production cycle of fabric dyeing and finishing. Compare the first derivative characteristics with a preset dyeing and finishing kinetic stage transition threshold to determine the stage switching point between the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage. Based on the stage switching point, the time series data boundary features of the temperature and humidity time series data are extracted, and the process constraint features characterizing the chemical reaction rate of each mechanism stage are extracted in combination with the second derivative features.

3. The industrial data processing method for fabric dyeing and finishing parameters according to claim 2, characterized in that, The steps for constructing a multidimensional correlation graph between time-series nodes, recipe parameter nodes, and fabric attribute nodes include: Using the fabric attribute node as the root node, the formula parameter node as the intermediate node, and the time sequence node as the leaf node, a directed edge is established between the root node, the intermediate node, and the leaf node according to the physical flow direction of the fabric dyeing and finishing process. The underlying time-series storage area is divided into data partitions corresponding to the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage, respectively. The starting physical storage offset of the data partition is embedded into the leaf node as the stage identifier to construct the node topology structure of the multidimensional association graph.

4. The industrial data processing method for fabric dyeing and finishing parameters according to claim 3, characterized in that, The step of forming a bidirectional mapping pointer between the multidimensional correlation graph and the underlying temporal storage area includes: A logical pointer pointing to the logical address space is written into the time segment node of the multidimensional association graph, and a graph pointer pointing to the time segment node is written into the data block header of the underlying time storage area, thereby establishing a bidirectional mapping relationship between the logical pointer and the graph pointer. A logical-physical mapping table is constructed between the logical address space and the physical storage offset. When the underlying time-series storage area undergoes data reconstruction or location migration, the physical storage offset in the logical-physical mapping table is updated while keeping the logical pointer and the graph pointer fixed.

5. The industrial data processing method for cloth dyeing and finishing parameters according to claim 4, characterized in that, The step of receiving a retrieval request containing fabric physical property indicators and target color parameters, and locating a matching node in the formula parameter node and the fabric attribute node of the multidimensional correlation graph, includes: The fabric physical property index and the target color parameter are converted into a multi-dimensional feature vector. The first similarity distance between the multi-dimensional feature vector and the fabric attribute node is calculated in the multi-dimensional association map. Target fabric attribute nodes whose first similarity distance meets the first preset threshold are selected. In the multidimensional association graph, traverse the recipe parameter nodes associated with the target fabric attribute node along the directed edge, calculate the second similarity distance between the target chromaticity coordinates in the retrieval request and the historical chromaticity coordinates recorded in the recipe parameter nodes, and determine the historical recipe parameter nodes whose second similarity distance satisfies the second preset threshold as matching nodes.

6. The industrial data processing method for fabric dyeing and finishing parameters according to claim 5, characterized in that, The steps of redirecting the bidirectional mapping pointer to the physical address corresponding to the stage identifier in the underlying timing storage area, and extracting the timing operation parameters corresponding to the heating mechanism stage, the heat preservation mechanism stage, or the cooling mechanism stage from the physical address include: The corresponding logical-physical mapping table entries are obtained by using the bidirectional mapping pointer associated with the matching node, and the physical address ranges of the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage in the underlying time-series storage area are parsed out. According to the chronological order of the fabric dyeing and finishing process, the time-series operation parameter subsets of each mechanism stage are read from the physical address range. The timestamps of adjacent time-series operation parameter subsets at the switching point are compared, the data records corresponding to overlapping timestamps are removed, and the time-series operation parameter subsets after removing overlaps are spliced ​​into a continuous time-series operation parameter set.

7. The industrial data processing method for fabric dyeing and finishing parameters according to claim 6, characterized in that, The steps for calculating the first and second derivative characteristics of temperature and humidity time-series data within a sliding window during a continuous fabric dyeing and finishing production cycle include: Perform multi-scale Gaussian smoothing on the temperature and humidity time series data to filter out high-frequency burrs in the temperature and humidity time series data that have a frequency higher than the preset dyeing and finishing physical change frequency; On the temperature and humidity time series data after multi-scale Gaussian smoothing, calculate the first and second derivative features of the center data point within the sliding window; The step of comparing the first derivative feature with a preset dyeing and finishing kinetics stage transition threshold to determine the stage switching point between the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage includes: When the first derivative feature crosses the stage transition threshold of the dyeing and finishing kinetics and the second derivative feature is in the extreme value range, the time point corresponding to the center data point of the sliding window is marked as the stage switching point.

8. The industrial data processing method for fabric dyeing and finishing parameters according to claim 7, characterized in that, The steps of establishing directed edges between the root node, the intermediate node, and the leaf node, using the formula parameter node as the intermediate node and the time sequence node as the leaf node, according to the physical flow direction of the fabric dyeing and finishing process, include: The formula parameter node is split into a dye type node and a dye concentration node. The dye type node is used as a first-level intermediate node to connect to the fabric attribute node, and the dye concentration node is used as a second-level intermediate node to connect to the dye type node. Based on the dye chemical properties represented by the dye type nodes, the time segment nodes corresponding to dyes with different chemical properties are classified and connected to the dye concentration nodes, so that the time segment nodes of the heating mechanism stage, the heat preservation mechanism stage, and the cooling mechanism stage form a heterogeneous branch topology in the multidimensional correlation graph according to the differences in dye chemical properties.

9. The industrial data processing method for fabric dyeing and finishing parameters according to claim 8, characterized in that, When the underlying time-series storage area undergoes data reconstruction or location migration, the step of updating the physical storage offset in the logical-physical mapping table while keeping the logical pointer and the graph pointer fixed includes: Set up a front-end logical-physical mapping table and a back-end logical-physical mapping table. The front-end logical-physical mapping table is used to respond to the retrieval request of the bidirectional mapping pointer. The back-end logical-physical mapping table is used to write the updated physical storage offset when data reconstruction or location migration occurs in the underlying time-series storage area. After the backend logical-physical mapping table completes all update operations, the frontend logical-physical mapping table and the backend logical-physical mapping table are atomically swapped, so that the retrieval request can obtain the original physical storage offset based on the unupdated frontend logical-physical mapping table during data reconstruction.

10. The industrial data processing method for fabric dyeing and finishing parameters according to claim 9, characterized in that, The step of converting the fabric physical property index and the target chromaticity parameter into a multidimensional feature vector, and calculating the first similarity distance between the multidimensional feature vector and the fabric attribute node in the multidimensional association graph includes: The physical weights of the fabric's physical properties are determined based on the fabric fiber color absorption dynamics, and the chromaticity weights of the target chromaticity parameters are determined based on the human eye perception characteristics of the color space. Based on the physical weights and the chromaticity weights, a normalized mapping is performed on the fabric property indicators and the target chromaticity parameters, and the normalized fabric property indicators and the target chromaticity parameters are combined into the multidimensional feature vector. In the multidimensional association graph, the weighted Mahalanobis distance between the multidimensional feature vector and the attribute vector of the fabric attribute node is calculated, and the weighted Mahalanobis distance is used as the first similarity distance.

Images