A method and system for modeling complex electronic product instrument and equipment resources

Abstract

The present application discloses a method and system for modeling the resources of complex electronic product instruments and equipment, which includes: using a Bayesian optimization method to weight the multi-dimensional indicators of instruments and equipment to construct a process weighted feature matrix; using an improved hybrid hierarchical clustering algorithm to generate a classification of instrument and equipment resources, and combining process expert evaluation to establish an instrument and equipment classification standard; based on the instrument and equipment classification standard, on the demand side, based on the demand for instrument and equipment resources of each process in the process flow, an instrument and equipment resource demand model is established, and each process is bound to a standardized resource classification. On the supply side, a production line resource capacity model library that conforms to the standardized resource classification is established; and resource scheduling is performed using a scheduling strategy that integrates the probability of resource start-up demand and physical distance. The present application improves the scheduling feasibility and accuracy of instruments and equipment, and improves the efficiency of instrument and equipment use.

Description

Technical Field

[0001] The present application relates to the field of production and manufacturing technology, and in particular to a method and system for modeling complex electronic product instrument and equipment resources. Background Art

[0002] In the field of production and manufacturing, resource modeling is an important technical means to realize resource scheduling activities and improve resource utilization.

[0003] In the existing technology, especially for discrete production lines, instrument and equipment resources are usually modeled based on the production line. The instrument and equipment resources are bound to the process, and tasks are arranged at different workstations through production scheduling. Resources generally do not move or move very little, and the resource model usually uses the basic information of the instance.

[0004] However, for complex electronic product production lines, there are multiple process routes, many unexpected situations in the process, material movement is difficult, and the movement of instruments and equipment is relatively simple. Summary of the Invention

[0005] In the production process of complex electronic products, instruments and equipment are of high value. Improving the utilization efficiency of instruments and equipment is of great value for reducing costs and increasing efficiency. This application proposes a resource modeling method and system for complex electronic product instruments and equipment to reduce the solution time of resource scheduling schemes and improve resource utilization.

[0006] This application discloses a method for modeling complex electronic product instrument equipment resources, which includes:

[0007] Step 1: Use the Bayesian optimization method to weight the multi-dimensional indicators of instruments and equipment to construct a process weighted feature matrix;

[0008] Step 2: Use the improved hybrid hierarchical clustering algorithm to generate a classification of instrument and equipment resources, and combine it with process expert evaluation to establish an instrument and equipment classification standard;

[0009] Step 3: Based on the instrument and equipment classification standards, on the demand side, establish an instrument and equipment resource demand model based on the instrument and equipment resource requirements of each process in the process flow. Bind each process to a standardized resource classification. On the supply side, establish a production line resource capability model library that conforms to the standardized resource classification.

[0010] Step 4: Use a scheduling strategy that integrates resource start-up demand probability and physical distance to schedule resources.

[0011] Furthermore, the step 1 includes:

[0012] By obtaining the original data of multi-dimensional technical parameters of instruments and equipment, filtering out invalid data in the original data and unifying the units, constructing composite characteristic parameters, and forming an initial process characteristic matrix through robust normalization processing, then setting the weight range of parameters in the initial process characteristic matrix, setting the weight search range of different parameters based on process priority, and taking the maximization of silhouette coefficient as the goal, determining the optimal weight combination through Bayesian search, and using the optimal weight combination to weight the initial process characteristic matrix to generate a process weighted feature matrix.

[0013] Furthermore, the constructing of composite feature parameters includes:

[0014] For the implicit features formed by the combination of multiple indicators, a composite feature parameter is constructed and numerically processed. For the frequency range parameter, the frequency upper limit - frequency lower limit is constructed and logarithmically valued to establish a composite feature parameter of log_freq_range = log10 (frequency upper limit - frequency lower limit); log_freq_range is the logarithm of the frequency range.

[0015] Furthermore, the step 2 includes:

[0016] The ward algorithm is used to generate a clustering tree diagram and the dynamic threshold t is used to cut the clustering tree diagram to generate a coarse classification. The instrument and equipment samples are divided into multiple clusters, and each cluster is recorded as a coarse classification cluster. Each coarse classification cluster corresponds to a coarse-grained instrument and equipment category. When the number of samples in the coarse classification cluster exceeds the preset threshold, the K-means fine classification process is triggered to divide each coarse classification cluster into multiple fine classification clusters. The number of fine classifications is 100%. Adaptively determined according to the number of coarse classification samples n:

[0017] The obtained fine classification clusters and coarse classification clusters are merged to form the instrument and equipment cluster analysis results. The high-dimensional features in the instrument and equipment cluster analysis results are projected into 2D space through the UMAP algorithm to generate an interactive scatter plot to display the cluster distribution and outliers. The instrument and equipment cluster analysis results are integrated with the process business rules to form the instrument and equipment classification standard; the process business rules are the instrument and equipment classification rules extracted by process experts based on experience.

[0018] Furthermore, the number of subcategories is determined by the following formula :

[0019]

[0020] Among them, max is the maximum value function.

[0021] Furthermore, the step 3 includes:

[0022] On the process demand side, use the instrument and equipment classification standards to establish a production line equipment demand model: Based on the instrument and equipment classification standards and the characteristics of complex electronic product production lines, when establishing an instrument and equipment resource demand model based on the process, describe the capability level and category requirements of the instruments and equipment under the minimum capability requirements of the process requirements. In other words, if the requirements can be met with the broad category, do not describe the detailed category. If the detailed category meets the process requirements, do not specify the model. If the model meets the requirements, do not specify the number.

[0023] Use the instrument classification standards to establish a production line instrument and equipment resource model library on the supply side: Based on the instrument and equipment classification standards, classify all instruments and equipment on the production line, and establish a production line resource capability model library. When classifying, classify according to the highest capability of the instruments and equipment.

[0024] Furthermore, the step 4 includes:

[0025] Step 41: When matching the instrument and equipment resource demand model with the production line resource capability model library, a hierarchical and progressive resource matching rule is followed, that is, the instrument and equipment resource demand model is matched hierarchically and progressively in accordance with the coarse classification, fine classification, model, and serial number. If the instrument and equipment resource demand model only describes the coarse classification, the production line resource capability model library is searched only through the coarse classification; if the instrument and equipment resource demand model describes the fine classification, the production line resource capability model library is matched using both the coarse classification and the fine classification; if the instrument and equipment resource demand model describes the model, the production line resource capability model library is matched using both the coarse classification, the fine classification, and the model; if the instrument and equipment resource demand model describes the serial number, the production line resource capability model library is matched using both the coarse classification, the fine classification, the model, and the serial number;

[0026] Step 42: According to step 41, use the instrument and equipment classification standard to match the idle resources of the production line to form a resource set to be scheduled {R1, R2…R i …R N}, R i represents the i-th resource that meets the resource demand model of the instrument equipment, R N Indicates the Nth resource that meets the instrument and equipment resource demand model;

[0027] Step 43: For each resource in the set of resources to be scheduled, dynamically identify all processes that are eligible to be started but have not yet been started based on the dependencies between the process tasks that may be started in the future at the workstation where the resource is located, and form a queue to be started Q = {J1, J 2... J N}, the number of processes is N, J N is the Nth process to be started;

[0028] Step 43: Calculate R separately i Match R in the queue of waiting work at the workstation iThe number of processes M is calculated according to the random sampling probability R i The probability of being used by the next process at the workstation, that is, the resource start demand probability;

[0029] Step 44: Calculate R in the set of resources to be scheduled i The physical distance between the workstation and the current resource demand, that is, the Euclidean distance (Distance) is calculated using the workstation coordinates. i ), and add the obstacle coefficient β according to the production line layout;

[0030] Step 45: According to R i The comprehensive priority of all resources in the resource set to be scheduled is used to sort them, and resources with high priority are scheduled first. Resources that are less likely to be occupied by the current workstation in the future are more likely to be scheduled, and resources that are closer to the target workstation are more likely to be scheduled. The scheduling time unit is refined to the duration of a single process.

[0031] Furthermore, in step 43, R is calculated by the following formulas: i The probability of being used by the next process at the station P(R i ):

[0032] P(R i )=M / N 100%;

[0033] If N=0, then P(R i )=0, resource R i The probability of being scheduled later is 1-M / N 100%.

[0034] Furthermore, in step 44, R is calculated according to the following formula: i The comprehensive scheduling priority S(R i ):

[0035] S(R i )=α (1-P(R i ))+β (1 / Distance(R i ))

[0036] α+β=1

[0037] Among them, α is the weight of the probability of being scheduled by the process task of the workstation in the future, and β is the weight of the workstation distance barrier set according to the production line layout.

[0038] This application also discloses a complex electronic product instrument equipment resource modeling system, which implements the complex electronic product instrument equipment resource modeling method described above, and includes:

[0039] The matrix construction module is used to construct a process weighted feature matrix by using the Bayesian optimization method to weight the multi-dimensional indicators of instruments and equipment;

[0040] The classification standard establishment module is used to generate the classification of instrument and equipment resources using the improved hybrid hierarchical clustering algorithm and establish the classification standard of instrument and equipment in combination with the evaluation of process experts;

[0041] The model library establishment module is used to establish an instrument and equipment resource demand model based on the instrument and equipment classification standards. On the demand side, based on the process resource requirements of the process, each process is bound to a standardized resource classification. On the supply side, a production line resource model library that conforms to the standardized resource classification is established.

[0042] The resource scheduling module is used to schedule resources using a dynamic scheduling strategy that integrates resource start-up demand probability and physical distance.

[0043] Due to the adoption of the above technical solution, this application has the following advantages:

[0044] 1. This application forms an instrument and equipment resource model library, which can improve the feasibility and accuracy of instrument and equipment scheduling and improve the efficiency of instrument and equipment use.

[0045] 2. This application mainly addresses the problem that complex electronic product production lines have many types of instrument and equipment resources, many manufacturers, high complexity of indicator parameters, large differences in instrument and equipment costs, and are difficult to model in a unified manner. By using an improved hierarchical clustering algorithm, an abstract classification standard for instrument and equipment resources is implemented, and a two-layer resource model is established on the demand side and the resource side based on the abstract classification standard. The dynamic scheduling strategy that integrates demand probability and distance is used to optimize production line resource scheduling, which greatly reduces resource dimensions, reduces resource types, enhances the dynamic scheduling efficiency of resources, and improves the speed of resource scheduling solution.

[0046] 3. The method proposed in this application is based on the classification of resource capability standards provided by instruments and equipment, and is used for discrete complex electronic product production lines. There is no clear binding relationship between the debugging process and the instruments and equipment in the production line, and there is no clear inheritance relationship between the requirements of different processes for instruments and equipment. This is regarded as a characteristic of complex electronic product production lines. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments recorded in the embodiments of the present application. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0048] Figure 1 A schematic diagram of a process for establishing the instrument and equipment classification standard according to an embodiment of the present application;

[0049] Figure 2 This is a clustering dendrogram of an embodiment of the present application;

[0050] Figure 3 This is a UMAP interactive scatter diagram of an embodiment of the present application;

[0051] Figure 4 Schematic diagram of a two-layer model of the art demand side and resource supply side of an embodiment of the present application;

[0052] Figure 5 This is a schematic diagram of the process flow of an embodiment of the present application;

[0053] Figure 6 This is a flow chart of a method for modeling complex electronic product instrument and equipment resources according to an embodiment of the present application. DETAILED DESCRIPTION

[0054] The present application is further described with reference to the accompanying drawings and embodiments. The embodiments described are only a part of the embodiments of the present application, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field should fall within the scope of protection of the embodiments of the present application.

[0055] Mainly aimed at the problems of complex electronic product production lines, such as a wide variety of instrument and equipment resources, multiple manufacturers, high complexity of indicator parameters, high value of instruments and equipment, and short-term and high-frequency usage, this application can ensure the accurate description of instrument and equipment capabilities by constructing a production line resource capability model library, and support the planning and scheduling system to carry out large-scale, process-level, accurate and efficient task scheduling and dispatching of instrument and equipment resources.

[0056] See also Figure 6 The present application provides an embodiment of a method for modeling complex electronic product instrument and equipment resources, which includes:

[0057] Step 1: Use the Bayesian optimization method to weight the multi-dimensional indicators of instruments and equipment, construct a process weighted feature matrix, and establish a process for resource classification standards.

[0058] Step 1-1: Collect multi-dimensional parameter data of instruments and equipment to form raw data of samples, as shown in Table 1, which includes instrument samples of 25 signal generators.

[0059] Table 1 Sample data example table

[0060]

[0061] Step 1-2: Conduct dynamic feature engineering.

[0062] Step 1-2-1: Unify parameter units.

[0063] Process technical parameters and convert the same type of technical parameters into the same units. For example, convert the signal frequency units from GHz, MHz, and Hz to Hz. It is necessary to classify and unify the units such as GHz, MHz, ns, and dB in Table 1.

[0064] Step 1-2-2: Filter invalid data.

[0065] Filter out invalid data in the original data samples, such as samples with missing data values ​​or incorrect data values.

[0066] Step 1-2-3: Construct composite feature parameters.

[0067] For implicit features formed by the combination of multiple indicators, composite feature parameters are constructed and numerically processed to avoid magnitude differences. For example, the frequency range parameter is constructed by (upper frequency limit - lower frequency limit) and logarithmically taking values, establishing a composite feature parameter of log_freq_range = log10(upper frequency limit - lower frequency limit); log_freq_range is the logarithm of the frequency range.

[0068] Optionally, the functional parameters are binned and binarized.

[0069] For functional indicators or differences in process requirements, binary processing is performed, for example Figure 2 In the samples, the narrowest pulse is segmented at 100ns, pulse_wide = 1 (the narrowest pulse width ≥ 100ns, marking a high-precision device), the vector function is binarized, vec_flag = 1 (vector function device); pulse_wide is the pulse width, and vec_flag is the vector function flag.

[0070] Step 1-3: Perform robust normalization preprocessing on the composite feature parameters to establish the initial process feature matrix.

[0071] The composite feature parameters are robustly standardized according to the following formula, where Median represents the median to prevent outliers from affecting the center position, and IQR represents the interquartile range.

[0072]

[0073] Among them, X is the composite characteristic parameter, The eigenvalues ​​after standardization eliminate dimensional differences, unify different dimensional features to the same scale, and reduce the interference of abnormal values ​​on the data range. The standardized data is more consistent with the clustering algorithm. After data processing, the initial process feature matrix Y is formed, as shown in Table 2:

[0074] Table 2 Example of initial process feature matrix

[0075]

[0076] Steps 1-4: Bayesian weight optimization to generate a process weighted feature matrix.

[0077] By assigning weights to different feature parameters and using Bayesian search to optimize the weight combination, a multi-dimensional process weighted feature matrix is ​​formed.

[0078] Step 1-4-1: Set the process weight range of the process feature.

[0079] The weight search ranges of different features are set based on process priority, such as frequency range weight w_range∈[0.3,0.6], frequency upper limit weight w_freq∈[0.2,0.3], pulse width weight w_pulse∈[0.1,0.3], vector function weight w_vec∈[0.05,0.1], and power weight w_power∈[0.05,0.1].

[0080] Step 1-4-1 takes the maximization of the silhouette coefficient as the goal, and through Bayesian search, determines the optimal weight group w_range=0.42352941, w_freq=0.28235294, w_pulse=0.11764706, w_vec=

[0081] 0.11764706, w_power=0.05882353.

[0082] Step 1-4-2: Based on the optimal weight combination obtained in step 1-4-1, weight the initial process feature matrix obtained in step 1-3 to generate a process weighted feature matrix, as shown in Table 3.

[0083] Table 3 Example of process weighted feature matrix

[0084]

[0085] Step 2: Use the improved hybrid hierarchical clustering algorithm to generate a classification of instrument and equipment resources, and combine it with process expert evaluation to establish instrument and equipment classification standards, such as Figure 1 shown.

[0086] Step 2-1: Use the ward algorithm to generate a cluster dendrogram.

[0087] The Ward's minimum variance method is used to calculate the distance between clusters, and its objective function is:

[0088]

[0089] Among them, |A| and |B| are the number of samples in cluster A and cluster B, 、 The center points of clusters A and B are recursively merged to generate a cluster dendrogram, and the merging order and distance values ​​are recorded. Figure 2 shown. is the square of the Euclidean distance between the centers of two clusters.

[0090] Step 2-2: Generate coarse classification clusters using a dynamic threshold selection mechanism.

[0091] Based on the distance distribution characteristics of the clustering dendrogram, the percentile method is used to dynamically determine the cutting threshold t. The first 90% of the distance values ​​of all merging steps in the link distance matrix generated by hierarchical clustering are taken as the cutting threshold to ensure that the classification results adapt to the data distribution characteristics. The threshold adaptability is achieved through data-driven, avoiding the traditional fixed threshold method that relies entirely on manual adjustment. The clustering dendrogram is cut according to the dynamic threshold t, and the samples are divided into initial coarse classification clusters. Each coarse classification cluster corresponds to a coarse-grained device category, as shown in the ward clustering result column in Table 4.

[0092] Step 2-3: Use K-means clustering algorithm for local refinement.

[0093] When the number of samples in the coarse classification cluster exceeds the preset threshold (such as n threshold =10,n threshold When is the coarse classification sample number threshold, the K-means fine classification process is triggered. Clusters with too few samples usually indicate highly homogeneous instrument performance and do not need further subdivision. Clusters with a large number of samples can further improve the classification granularity through fine classification. The number of fine classifications k is adaptively determined according to the number of parent class samples n:

[0094]

[0095] By running random initializations independently multiple times to avoid local optimal solutions, the obtained fine classifications and coarse classifications are merged to form the instrument and equipment cluster analysis results, such as the sample, clustering results, and process evaluation classification mapping table in Table 4.

[0096] Table 4 Sample, clustering results, and process evaluation classification mapping table

[0097]

[0098] Table 4 Sample, clustering results, and process evaluation classification mapping table (continued)

[0099]

[0100] Step 2-4: Visualization and rule post-processing: Use the UMAP algorithm to project the dimensional features in the instrument cluster analysis results into 2D space, generate an interactive scatter plot, and intuitively display the cluster distribution and abnormal points, such as Figure 3 As shown in Figure 4, the instrument and equipment cluster analysis results obtained in steps 2-3 are evaluated by process experts. The instrument and equipment cluster analysis results are integrated with process business rules to form instrument and equipment classification standards, as shown in the two columns of Classification Standard Definition (Major Category) and Classification Standard Definition (Sub-category) in Table 4. Process business rules are instrument and equipment classification rules extracted by process experts based on experience. For example, based on process requirements, the classification of some independent clusters is added, such as the mandatory classification of the vector function of the signal generator to avoid mixing with other equipment and ensure process specificity, thus obtaining preliminary classification standards, such as Figure 5 It is divided into two major categories with a total of six subcategories. Based on the major categories and subcategories, business rules are combined with additional model and numbering levels to support precise scheduling needs and form classification standards.

[0101] Among them, the UMAP algorithm is a nonlinear dimensionality reduction algorithm used to project high-dimensional data into a low-dimensional space (usually 2D or 3D) to achieve visualization of high-dimensional data while retaining the data topology and clustering characteristics.

[0102] Step 3: Based on the instrument and equipment classification standards, on the demand side, establish an instrument and equipment resource demand model based on the process resource requirements of the process flow, bind each process to a standardized resource classification, and on the supply side, establish a production line resource capability model library that conforms to the standardized resource classification.

[0103] Based on the instrument and equipment classification standard obtained in step 2, on the demand side, based on the process resource requirements of the process, an instrument and equipment resource demand model is established, and each process is bound to a standardized resource classification. On the supply side, a production line resource model library that conforms to the standardized resource classification is established, such as Figure 4 shown.

[0104] Step 3-1: Use classification standards to establish a process instrument and equipment resource demand model on the demand side.

[0105] According to the instrument and equipment capability classification standards and the characteristics of complex electronic product production lines, there are multiple process routes and the test sequence is relatively flexible. Different test items have different requirements for instruments and equipment. When establishing an instrument and equipment process requirement model based on the process, the capability level and category requirements of the instrument and equipment are described under the minimum capability requirements that meet the process requirements. That is, when the requirements can be met in the general category, the detailed classification should not be described. When the detailed classification meets the process requirements, the model should not be specified. When the model meets the requirements, the number should not be specified. As shown in Table 5, an instrument and equipment resource requirement model is established based on each process. The difference in the accuracy of resource requirements for each process is explained as follows.

[0106] Table 5 Example of instrument and equipment resource requirement model

[0107]

[0108] a) For processes where the requirements can be met by a major category, only the major category is filled in. Resource requirements at other levels, such as process 2, indicate that all categories of Class A instruments can meet the usage requirements.

[0109] b) For processes that require precise sub-categories, both the major category and the sub-category must be filled in, such as process 3, which represents instruments of the C1 sub-category of Category A instruments.

[0110] c) For requirements that require a specific model, you need to fill in the major category, subcategory, and model. For example, process 5 indicates that a Class A instrument of subcategory C2 is required, and a T1 model is specified.

[0111] d) For requirements that require a specific number, you need to fill in the major category, subcategory, model, and number. For example, process 6 requires a C1 subcategory of a Class A instrument, and specifies the T2 model and number A0004.

[0112] e) If multiple instruments are required at the same time, fill in multiple rows. For example, as shown in process 7, one C1 and one C2 are required.

[0113] Step 3-2: Use classification standards to establish a production line resource model library on the supply side.

[0114] Based on the instrument and equipment capability classification standard established in step 3, all instruments and equipment on the production line are classified one by one, and a production line resource model library is established, as shown in Table 6. When classifying, the instruments and equipment are classified according to their highest capabilities, and each of the seven instruments on the production line is modeled one by one.

[0115] Table 6 Production line resource model library example

[0116]

[0117] Step 4: Use a dynamic scheduling strategy that integrates resource start-up demand probability and physical distance to schedule resources.

[0118] Step 4-1: When matching the instrument and equipment resource demand model with the production line resource capability model library, the hierarchical and progressive resource matching rules should be followed, and the matching should be carried out hierarchically and progressively according to the major categories, detailed categories, models, and numbers of the instrument and equipment resource demand model. If the instrument and equipment resource demand model only describes the coarse category, only the coarse category is used to search the production line resource capability model library. If it describes the detailed category, both the coarse category and the detailed category need to be used to match the production line resource capability model library. If it describes the model, the coarse category, detailed category, and model are used to match the production line resource capability model library. If it describes the number, the coarse category, detailed category, model, and number are used to match the production line resource capability model library.

[0119] Step 4-2: According to step 4-1, use the instrument and equipment capability classification standard to match the idle resources of the production line and form a set of resources to be scheduled {R1, R2…R i ...R N}, R i represents the i-th resource that meets the resource demand model of the instrument equipment, R N Indicates the Nth resource instance that meets the resource requirement model.

[0120] Step 4-3: For each resource in the resource set to be scheduled, the workstation where the resource is located is obtained according to the topological relationship of the directed acyclic graph of the process flow and the process dependency relationship. All processes that have met the preconditions but have not been started are dynamically identified to form a queue to be started Q = {J1, J2...J N}, the number of processes is N, J N is the Nth process to be started. Figure 5 As shown in the process flow diagram, if the predecessor process 1 of processes 2, 3, 4, and 5 is completed, then 2, 3, 4, and 5 will enter the queue to be started; if only 1 is not completed, 2, 3, 4, and 5 will not be included in the queue.

[0121] Step 4-3: Calculate each resource R i Matching resource R in the waiting queue of the workstation i The number of processes M is used to calculate the resource start-up demand probability, that is, the probability that the resource will continue to be used by the next process at the workstation is:

[0122] P i =M / N 100%.

[0123] If N=0, then P i =0, so the future resource R i The probability of being scheduled is 1-M / N 100%.

[0124] Step 4-4: Calculate each resource R in the set of resources to be scheduled i The physical distance from the current resource demand station is calculated using the station coordinates to calculate the Euclidean distance (Distance(R i ), and add the obstacle coefficient according to the production line layout β .

[0125] Step 4-5: Calculate the comprehensive scheduling priority of resources according to the following formula.

[0126] S(R i )=α (1-P(R i ))+β (1 / Distance(R i ))(α+β=1)

[0127] Among them, α is the weight of the probability of being scheduled by the process task of the workstation in the future, and β is the weight of the workstation distance barrier set according to the production line layout.

[0128] All resources in the resource set to be scheduled are sorted according to their comprehensive priority, and resources with high priority are scheduled first. Resources that are less occupied by the current workstation in the future are more likely to be scheduled, and resources that are closer to the target location are more likely to be scheduled. The scheduling time unit is refined to a single process. The standard classification of resources ensures that bottleneck resources can be scheduled at a higher rate, thereby improving the efficiency of instrument utilization.

[0129] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit them. Although the present application has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present application can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present application should be included in the scope of protection of the claims of the present application.

Claims

1. A complex electronic product instrument and equipment resource modeling method, characterized in that: include: Step 1: Use the Bayesian optimization method to weight the multi-dimensional indicators of instruments and equipment to construct a process weighted feature matrix; Step 2: Use the improved hybrid hierarchical clustering algorithm to generate a classification of instrument and equipment resources, and combine it with process expert evaluation to establish an instrument and equipment classification standard; Step 3: Based on the instrument and equipment classification standards, on the demand side, establish an instrument and equipment resource demand model based on the instrument and equipment resource requirements of each process in the process flow. Bind each process to a standardized resource classification. On the supply side, establish a production line resource capability model library that conforms to the standardized resource classification. Step 4: Use a scheduling strategy that integrates resource start-up demand probability and physical distance to schedule resources; The step 2 includes: The ward algorithm is used to generate a clustering tree diagram and the dynamic threshold t is used to cut the clustering tree diagram to generate a coarse classification. The instrument and equipment samples are divided into multiple clusters, and each cluster is recorded as a coarse classification cluster. Each coarse classification cluster corresponds to a coarse-grained instrument and equipment category. When the number of samples in the coarse classification cluster exceeds the preset threshold, the K-means fine classification process is triggered to divide each coarse classification cluster into multiple fine classification clusters. The number of fine classifications is 100%. Adaptively determined according to the number of coarse classification samples n: The obtained fine classification clusters are combined with the coarse classification clusters to form the instrument and equipment cluster analysis results. The high-dimensional features in the instrument and equipment cluster analysis results are projected into 2D space using the UMAP algorithm to generate an interactive scatter plot to display the cluster distribution and outliers. The instrument and equipment cluster analysis results are integrated with process business rules to form the instrument and equipment classification standard. The process business rules are the instrument and equipment classification rules extracted by process experts based on their experience. The number of subcategories is determined by the following formula : Among them, max is the maximum value function.

2. The complex electronic product instrument and equipment resource modeling method according to claim 1, characterized in that: The step 1 comprises: By obtaining the original data of multi-dimensional technical parameters of instruments and equipment, filtering out invalid data in the original data and unifying the units, constructing composite characteristic parameters, and forming an initial process characteristic matrix through robust normalization processing, then setting the weight range of the parameters in the initial process characteristic matrix, setting the weight search range of different parameters based on the process priority, and taking the maximization of the silhouette coefficient as the goal, through Bayesian search, determining the optimal weight combination, and using the optimal weight combination to weight the initial process characteristic matrix to generate a process weighted feature matrix.

3. The complex electronic product instrument and equipment resource modeling method according to claim 2, characterized in that: The constructing of composite feature parameters includes: For the implicit features formed by the combination of multiple indicators, a composite feature parameter is constructed and numerically processed. For the frequency range parameter, the frequency upper limit - frequency lower limit is constructed and logarithmically valued to establish a composite feature parameter of log_freq_range = log10(frequency upper limit - frequency lower limit); log_freq_range is the logarithm of the frequency range.

4. The complex electronic product instrument and equipment resource modeling method according to claim 1, characterized in that: The step 3 includes: On the process demand side, use the instrument and equipment classification standards to establish a production line equipment demand model: Based on the instrument and equipment classification standards and the characteristics of complex electronic product production lines, when establishing an instrument and equipment resource demand model based on the process, describe the capability level and category requirements of the instruments and equipment under the minimum capability requirements of the process requirements. In other words, if the requirements can be met with the broad category, do not describe the detailed category. If the detailed category meets the process requirements, do not specify the model. If the model meets the requirements, do not specify the number. Use the instrument classification standards to establish a production line instrument and equipment resource model library on the supply side: Based on the instrument and equipment classification standards, classify all instruments and equipment on the production line, and establish a production line resource capability model library. When classifying, classify according to the highest capability of the instruments and equipment.

5. The complex electronic product instrument and equipment resource modeling method according to claim 1, characterized in that: The step 4 comprises: Step 41: When matching the instrument and equipment resource demand model with the production line resource capability model library, a hierarchical and progressive resource matching rule is followed, that is, the instrument and equipment resource demand model is matched hierarchically and progressively in accordance with the coarse classification, fine classification, model, and serial number. If the instrument and equipment resource demand model only describes the coarse classification, the production line resource capability model library is searched only through the coarse classification; if the instrument and equipment resource demand model describes the fine classification, the production line resource capability model library is matched using both the coarse classification and the fine classification; if the instrument and equipment resource demand model describes the model, the production line resource capability model library is matched using both the coarse classification, the fine classification, and the model; if the instrument and equipment resource demand model describes the serial number, the production line resource capability model library is matched using both the coarse classification, the fine classification, the model, and the serial number; Step 42: According to step 41, use the instrument and equipment classification standard to match the idle resources of the production line to form a resource set to be scheduled {R1, R2…R i …R N }, R i represents the i-th resource that meets the resource demand model of the instrument equipment, R N Indicates the Nth resource that meets the instrument and equipment resource demand model; Step 43: For each resource in the set of resources to be scheduled, dynamically identify all processes that are eligible to be started but have not yet been started based on the dependencies between the process tasks that may be started in the future at the workstation where the resource is located, and form a queue to be started Q = {J1, J 2... J N }, the number of processes is N, J N is the Nth process to be started; Step 43: Calculate R separately i Match R in the queue of waiting work at the workstation i The number of processes M is calculated according to the random sampling probability R i The probability of being used by the next process at the workstation, that is, the resource start demand probability; Step 44: Calculate R in the set of resources to be scheduled i The physical distance between the workstation and the current resource, that is, the Euclidean distance (Distance(R)) is calculated using the workstation coordinates. i ), and add the obstacle coefficient β according to the production line layout; Step 45: According to R i The comprehensive priority of all resources in the resource set to be scheduled is used to sort them, and resources with high priority are scheduled first. Resources that are less likely to be occupied by the current workstation in the future are more likely to be scheduled, and resources that are closer to the target workstation are more likely to be scheduled. The scheduling time unit is refined to the duration of a single process.

6. The complex electronic product instrument and equipment resource modeling method according to claim 5, characterized in that: In step 43, R is calculated by the following formula: i The probability of being used by the next process at the station P(R i ): P(R i )=M / N 100%; If N=0, then P(R i )=0,R i The probability of being scheduled later is 1-M / N 100%.

7. The complex electronic product instrument and equipment resource modeling method according to claim 6, characterized in that: In step 44, R is calculated according to the following formula: i The comprehensive scheduling priority S(R i ): S(R i )=a (1-P(R i ))+b (1 / Distance(R i )) α+β=1 Among them, α is the weight of the probability of being scheduled by the process task of the workstation in the future, and β is the weight of the workstation distance barrier set according to the production line layout.

8. A complex electronic product instrument and equipment resource modeling system, which implements the complex electronic product instrument and equipment resource modeling method according to any one of claims 1 to 7, characterized in that: include: The matrix construction module is used to construct a process weighted feature matrix by using the Bayesian optimization method to weight the multi-dimensional indicators of instruments and equipment; The classification standard establishment module is used to generate the classification of instrument and equipment resources using the improved hybrid hierarchical clustering algorithm and establish the classification standard of instrument and equipment in combination with the evaluation of process experts; The model library establishment module is used to establish an instrument and equipment resource demand model based on the instrument and equipment classification standards. On the demand side, based on the process resource requirements of the process, each process is bound to a standardized resource classification. On the supply side, a production line resource model library that conforms to the standardized resource classification is established. The resource scheduling module is used to schedule resources using a dynamic scheduling strategy that integrates resource start-up demand probability and physical distance.

Images