A method for predicting machine tool operation status and workshop scheduling based on product processing quality

Abstract

The present invention discloses a machine tool operation state prediction and workshop scheduling method based on product processing quality, including: step 1: constructing a machine tool, processing procedure and processing quality characteristic mapping model for the operation state of the machine tool; step 2: constructing preliminary clustered quality data for each machine tool equipment in the production process; step 3: according to the preliminary clustered quality data, using a data label sample equalization algorithm to generate sufficient processing quality data and equalize the quality data of qualified products and unqualified products; step 4: according to the equalized quality data, analyzing and predicting the machine tool operation state; step 5: according to the predicted machine tool operation state, constructing a production process scheduling model, and reallocating production and processing tasks to normal machine tools. The present invention can avoid the problem that a complex production environment interferes with the analysis of machine tool data, and at the same time, the obtained scheduling scheme can effectively shorten the production scheduling cycle.

Description

Technical Field

[0001] The invention belongs to the technical field of workshop scheduling, and in particular relates to a machine tool operation state prediction and workshop scheduling method based on product processing quality. Background Art

[0002] In recent years, companies have increasingly tended to produce small batches and multiple varieties of products to meet customers' increasingly personalized and customized product needs. Production workshops are required to have extremely high flexibility and robustness in the process of performing mixed tasks. Machine tools respond to production tasks, and the processing quality depends on the operating status of the machine tools. Problems with the operating status of the machine tools will directly affect the processing quality of the products. The machines bear a heavy burden of handling these tasks, and the operating status of the machines has a great impact on the production process, which may cause serious property losses. Therefore, it is necessary to predict the operating status of machine tools and schedule the production process in advance to avoid potential production losses.

[0003] There are two main types of prediction methods for machine tool operation status: simple model construction and big data analysis. However, these two methods rely on data collected from the machine tool operation process. This type of data is affected by the machine tool operation environment, has a large amount of interference data and complex types, resulting in low accuracy of the prediction conclusion. In addition, a large amount of computing resources is required for data preprocessing to obtain reliable machine tool operation process data for monitoring and predicting the machine tool operation status.

[0004] Production scheduling methods can also be divided into two categories: static scheduling and dynamic scheduling. In the static scheduling method, the existing technology proposes to establish a scheduling model of heuristic algorithm from the perspective of total energy consumption and delay of machine tools; or to establish a mathematical programming model that considers production cycle, maximum power consumption and carbon emissions to obtain the optimal production plan; or to use the Pareto optimal front end combined with genetic algorithm to optimize the energy consumption and total weighted delay time of the non-processing stage to obtain the optimal scheduling solution. However, these static scheduling methods are not suitable for the dynamic scheduling problems of dynamic and ever-changing flexible production workshops.

[0005] Dynamic scheduling methods can be divided into three categories: fully reactive scheduling, robust proactive scheduling, and predictive reactive scheduling. Predictive reactive scheduling is the most widely used. The existing technology proposes a random scheduling method to extract scheduling rules from the optimal schedule to minimize the total weighted delay; or build a group of agents responsible for implementing different control and monitoring tasks to enhance the flexibility and agility of the system; or propose a two-level multi-task scheduling model to manage tasks more effectively in a cloud manufacturing environment.

[0006] From the current research technology, the following problems can be summarized:

[0007] 1. Currently, collecting machine tool operation data for machine tool operation status analysis is the mainstream method. However, the complex and noisy production environment will interfere with the collection and analysis of machine tool operation data, and consume a lot of computing resources to eliminate useless and interfering data.

[0008] 2. The operating status of machine tools has a significant impact on the production process, but there is insufficient research on the production process scheduling caused by abnormal machine tool operating status. In addition, few studies have classified the machine tool operating status into levels, making it impossible to accurately schedule the production process.

[0009] Third, high efficiency is rarely considered in the selection of scheduling algorithms, resulting in a long time spent in solving scheduling solutions. Summary of the invention

[0010] The technical problem to be solved by the present invention is to provide a machine tool operation status prediction and workshop scheduling method based on product processing quality in view of the deficiencies of the above-mentioned prior art, using an improved K-means algorithm and a proposed data label sample equalization algorithm to preliminarily analyze the processing quality data, and using an improved particle swarm algorithm to predict the processing quality data clustering center, and then predict the machine tool operation status. From the perspective of the machine tool operation status, the impact of the machine tool operation status on the production process is evaluated and production tasks are dynamically scheduled, thereby reducing the negative impact of abnormal machine tool operation status on the production process.

[0011] In order to achieve the above technical objectives, the technical solution adopted by the present invention is:

[0012] A method for predicting machine tool operation status and workshop scheduling based on product processing quality comprises the following steps:

[0013] Step 1: According to the operation status of the machine tool, a machine tool, processing procedure and processing quality characteristic mapping model is constructed;

[0014] Step 2: Construct preliminary clustered quality data for each machine tool in the production process;

[0015] Step 3: Based on the quality data of the preliminary clustering, a data label sample balancing algorithm is used to generate sufficient processing quality data and balance the quality data of qualified products and unqualified products;

[0016] Step 4: Analyze and predict the machine tool operation status based on the equalized quality data;

[0017] Step 5: Based on the predicted machine tool operating status, build a production process scheduling model and reallocate production processing tasks to normal machine tools.

[0018] To optimize the above technical solutions, the specific measures taken also include:

[0019] In the above step 1, one machine tool can be mapped to multiple parts processing qualities, and the processing quality characteristics include size characteristics, shape and position characteristics, and roughness characteristics.

[0020] The above step 2 includes:

[0021] Step 2.1: For each machine tool in the production process, collect the processing quality data label samples and define the qualified product quality data and unqualified product quality data labels;

[0022] Step 2.2: Use the improved K-means algorithm to evaluate the cluster centers of the quality data and the corresponding data sets to obtain preliminary clustered quality data.

[0023] The quality data labels are defined as described in step 2.1 above, as follows:

[0024] QL i,j =[D i,j ,S i,j ,RR i,j ] (1)

[0025] In the formula, QL i,j represents the quality data label of the jth process of the part processed by the i-th machine tool; D i,j , S i,j , RR i,j Indicates the ratio of the deviation of the dimensional characteristics, shape and position characteristics, and roughness characteristics from the corresponding tolerance range;

[0026] When the value of at least one element of a quality data label is within the unqualified product quality characteristic range, it is defined as an unqualified product quality data label;

[0027] When the values ​​of the quality label elements are all within the qualified product quality characteristic range, it is defined as a qualified product quality data label.

[0028] In the above step 2.2, the processing quality data labels are regarded as point coordinates in the Cartesian coordinate system, and the improved K-means algorithm is used to obtain the preliminary clustered quality data, which specifically includes:

[0029] Step 2.2.1: Each processing quality data label has three elements, each element has two conditions: qualified and unqualified. After combination, eight types of processing quality data labels can be obtained, and random values ​​are taken within the qualified and unqualified product quality characteristic interval as the initial eight cluster centers;

[0030] Step 2.2.2: Cluster all quality data labels according to step 2.2.1: evaluate the class to which each quality data label belongs, i.e., calculate the Cartesian coordinate distance between the quality data label and the eight cluster centers, and classify the quality data label into the class to which the cluster center has the shortest distance;

[0031] Step 2.2.3: Update cluster centers: Re-evaluate and update cluster centers based on the quality data set after clustering. The formula is as follows:

[0032]

[0033] Where CCX i represents the updated cluster center label element value; i represents the i-th quality data label, j represents the j-th element value; x ij represents the jth element value of the quality data label belonging to the i-th category; Y i Represents the element value set of the i-th quality data label; n s Indicates the number of quality data labels belonging to the same category.

[0034] Step 2.2.4: Terminate the preliminary clustering process: Iterate and loop through steps 2.2.2 and 2.2.3 until the distance between two adjacent updated cluster center Cartesian coordinates is less than a predetermined threshold, and output the preliminary clustering quality data.

[0035] The above step 3 specifically includes:

[0036] Step 3.1: Eliminate abnormal data: Calculate the average Cartesian coordinate system distance between the quality data labels of the same category and the cluster center, and compare it with the distance between each quality data label and the cluster center. The quality data labels with a distance greater than the average distance are eliminated as abnormal data, and the quality data labels with a distance less than the average distance are retained for generating data samples;

[0037] Step 3.2: Balance quality data labels: For the classes whose quality data labels are less than 50% of the class with the largest number, use the following formula to generate quality data label elements:

[0038]

[0039] Where GX represents the generated quality data label element value; XR i,j Indicates the difference between the quality data label element value and the cluster center element value; Represents the mean of the difference between the element value of all quality data labels and the element value of the cluster center in the same class;

[0040] When the number of samples after generating quality data labels reaches the number of classes with the most data labels, the process of generating quality data labels and balancing quality data labels for that class is terminated.

[0041] Step 4 above includes:

[0042] Step 4.1: Quality data change trend analysis: Use the improved particle swarm algorithm to evaluate the same cluster center of all quality data. The rules are as follows:

[0043] The sum of the distances between all quality data labels and the unique cluster center Cartesian coordinate system is defined as the optimization target of the improved particle swarm algorithm. The formula is as follows:

[0044]

[0045] Where, OB is the sum of the distances between all quality data labels and the unique cluster center in the Cartesian coordinate system; QL p Indicates quality data label samples; AC k Represents the optimal cluster center; CL represents the quality data label class;

[0046] According to the optimization goal, the clustering centers of all quality data labels are continuously updated by improving the position and speed rules of the particle swarm algorithm to obtain the optimal clustering center;

[0047] Step 4.2: Predict the machine tool operating status based on the mapping model constructed in step 1 as follows:

[0048] Evaluate the processing quality data and obtain the following four situations according to the quality data label element type:

[0049] Scenario 1: The dimensional characteristic values, shape and position characteristic values, and roughness characteristic values ​​are all within the qualified product quality characteristic range;

[0050] Scenario 2: Among the dimensional characteristic value, shape and position characteristic value, and roughness characteristic value, one value is within the unqualified product quality characteristic range;

[0051] Scenario 3: Two of the dimensional characteristic values, shape and position characteristic values, and roughness characteristic values ​​are within the unqualified product quality characteristic range;

[0052] Scenario 4: Three of the dimensional characteristic values, shape and position characteristic values, and roughness characteristic values ​​are within the unqualified product quality characteristic range;

[0053] According to the above four situations, the conclusions of predicting the machine tool operation status are as follows:

[0054] When situation 1 occurs, the machine tool is operating normally;

[0055] When the second scenario occurs, the machine tool is in a sub-healthy state, and the probability of the machine tool being selected as an available machine tool in subsequent scheduling is reduced by 50%;

[0056] When situation three occurs, the machine tool is in a sub-healthy state, and the probability of the machine tool being selected as an available machine tool in subsequent scheduling is reduced by 75%;

[0057] When situation 4 occurs, the machine tool is in an abnormal operating state and is eliminated from subsequent scheduling.

[0058] The above step 5 includes:

[0059] Step 5.1: Construct the production process code and assign multiple tasks to the matrix code of the corresponding machine tools. Specifically:

[0060] The production process coding is defined in two-dimensional matrix form, and the formula is as follows:

[0061]

[0062] In the formula, each column of the matrix represents the machine tool code vector corresponding to the machining process operation of a task; the number of rows represents the number of tasks; p(n,s) represents the machine tool code used for the nth process operation of the sth task;

[0063] Step 5.2: Based on the machine tools whose operating status is abnormal as predicted in step 4, the improved genetic algorithm is used to reallocate production and processing tasks to available machine tools. Specifically, the method of reallocating production and processing tasks to available machine tools is as follows:

[0064] Taking the construction period of the production scheduling plan as the optimization target, the improved genetic algorithm is used for evaluation to calculate the scheduling plan with the shortest construction period;

[0065] The calculation formula for construction period is as follows:

[0066]

[0067] Where, d n Indicates the production duration after scheduling; Respectively indicate the last end time and the earliest start time of the production task; Indicates the last completion time and the earliest start time of the process operation of the nth task;

[0068] The optimization objective formula is as follows:

[0069] Fit(solu)=1 / d n (7)

[0070] Where Fit(solu) represents the fitness function; dn It indicates the duration of the plan after production scheduling;

[0071] According to the above optimization objective formula, the optimal production scheduling plan is obtained by improving the genetic algorithm to optimize the scheduling plan.

[0072] The present invention has the following beneficial effects:

[0073] The present invention defines product processing quality data labels, establishes a machine tool, processing procedure and processing quality characteristic mapping model; uses a three-dimensional space point coordinate type K-means algorithm to preliminarily cluster and analyze product quality data; constructs a data label sample equalization algorithm based on the preliminarily clustered product quality data, calculates and generates balanced data label samples; uses a three-dimensional space point coordinate type particle swarm optimization algorithm to solve the balanced quality data change trend, and predicts the machine tool operation status based on the machine tool, processing procedure and processing quality characteristic mapping model; uses a three-dimensional space point coordinate type ant colony algorithm to solve the production period target model, and obtains a set of optimal production scheduling solutions. The present invention can avoid the problem of complex production environment interfering with machine tool data analysis, and at the same time, the obtained scheduling solution can effectively shorten the production scheduling cycle.

[0074] (1) Compared with the direct prediction method of the machine tool operation status based on the machine tool operation data, the method of the present invention uses the product processing quality data to analyze the quality data change trend and then predicts the machine tool operation status. It has higher reliability and saves a lot of computing resources used for data preprocessing.

[0075] (2) The present invention classifies the machine operating status and uses it to optimize the dynamic scheduling scheme of the production process. The resulting scheduling scheme is more reasonable and more operational.

[0076] (3) The present invention proposes a preliminary clustering method for quality data, which solves multiple clustering centers of quality data based on qualified quality and unqualified quality, and generates sufficient and balanced quality data label samples for quality data analysis, significantly improving the efficiency of data analysis.

[0077] (4) The present invention can cope with dynamically changing production workshop scheduling problems, dynamically predict the operating status of machine tools and quickly generate effective dynamic scheduling solutions, thereby helping engineers to formulate highly feasible production plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 It is a framework for machine tool operation status prediction and workshop scheduling methods based on product processing quality.

[0079] Figure 2 It is the quality characteristic range of qualified products and the quality characteristic range of unqualified products. DETAILED DESCRIPTION

[0080] The embodiments of the present invention are further described in detail below with reference to the accompanying drawings.

[0081] See also Figure 1 , a machine tool operation status prediction and workshop scheduling method based on product processing quality, the specific implementation steps are as follows:

[0082] Step 1: In order to express the relationship between the machine tool and its processing quality, a machine tool, processing procedure and processing quality characteristic mapping model is constructed according to the operation status of the machine tool;

[0083] One machine tool can be mapped to the processing qualities of multiple parts. The processing quality characteristics include size characteristics, shape and position characteristics, and roughness characteristics. The machine tool processing procedure determines the processing quality of the part product.

[0084] Figure 1 The upper left part is the machine tool and processing quality characteristic mapping model, and the left part is the machine tool code, using M1, M m The middle part is the process operation code, which indicates milling, drilling and other operations. The t-th operation on the m-th machine and bed is represented by Op m,t Therefore, the process operation on the machine tool can be expressed as Op 1,1 、Op 1,s 、Op m,1 、Op m,t ; On the right are the processing quality characteristics of the process operation, representing the dimensional characteristics, shape and position characteristics, and roughness characteristics, which are represented by DC, SC, and RC respectively. The subscripts of DC, SC, and RC in the figure represent the corresponding process operations.

[0085] It can be seen that the model has the following characteristics:

[0086] The model includes machine tool layer, process operation layer and quality attribute layer. The process operation layer serves as the middle layer to realize the one-to-many mapping between machine tool and quality characteristics.

[0087] The quality of each process operation is evaluated by a set of quality characteristics, and then the operating status of each machine tool can be evaluated by multiple sets of quality characteristics.

[0088] There is a big difference between the amount of qualified quality data and the amount of unqualified quality data corresponding to each machine tool.

[0089] Step 2: Construct preliminary clustered quality data for each machine tool in the production process;

[0090] Step 2.1: For each machine tool in the production process, collect the processing quality data label samples and define the qualified product quality data and unqualified product quality data labels;

[0091] The purpose of the preliminary clustering method of quality data is to classify the quality data into qualified product quality data class and unqualified product quality data class.

[0092] The data label sample refers to a vector composed of the proportion of the size characteristic value, shape and position characteristic value, and roughness characteristic value deviating from the corresponding tolerance range. That is, in order to analyze the trend of quality data changes, the quality data label is defined as:

[0093] QL i,j =[D i,j ,S i,j ,RR i,j ],

[0094] In the formula, QL i,j represents the quality data label of the jth process of the part processed by the i-th machine tool; D i,j , S i,j , RR i,j Indicates the ratio of the deviation of the dimensional characteristics, shape and position characteristics, and roughness characteristic values ​​from the corresponding tolerance range.

[0095] When the value of at least one element of a quality data label is within the unqualified product quality characteristic range, it is defined as an unqualified product quality data label;

[0096] When the values ​​of the quality label elements are all within the qualified product quality characteristic range, it is defined as a qualified product quality data label.

[0097] Each element value of the quality data label defines two value ranges, such as Figure 2 As shown, it is divided into the qualified product quality characteristic interval and the unqualified product quality characteristic interval (i.e. the σ ratio part). IV i,j They respectively represent the upper and lower limits and the median value of the quality characteristic value tolerance.

[0098] Step 2.2 preliminarily clusters the quality data, i.e. uses the improved K-means algorithm to evaluate the cluster centers of the quality data and the corresponding data sets.

[0099] Step 2.2.1: Select eight cluster centers.

[0100] Each processing quality data label has three elements, each element has two conditions: qualified and unqualified. After combination, eight types of processing quality data labels can be obtained, and random values ​​in the qualified and unqualified product quality characteristic interval are used as the initial eight cluster centers;

[0101] In order to obtain the initial cluster center, a value is randomly selected in the qualified product quality characteristic interval and the unqualified product quality characteristic interval as the initial cluster center data label element value. Therefore, as shown in Table 1, eight initial cluster centers can be obtained.

[0102] Table 1 Initial cluster centers

[0103]

[0104]

[0105] Step 2.2.2: Cluster all quality data labels according to step 2.2.1.

[0106] The class to which each quality data label belongs is evaluated, that is, the Cartesian coordinate distance between the quality data label and the eight cluster centers is calculated, and the quality data label is classified into the class to which the cluster center has the shortest distance.

[0107] Step 2.2.3: Update cluster centers.

[0108] According to the quality data set after clustering, re-evaluate and update the cluster center. The formula is as follows:

[0109]

[0110] Where CCX i represents the updated cluster center label element value; i represents the i-th quality data label, j represents the j-th element value; x ij represents the jth element value of the quality data label belonging to the i-th category; Y i Represents the element value set of the i-th quality data label; n s Indicates the number of quality data labels belonging to the same category.

[0111] Step 2.2.4: Abort the preliminary clustering process.

[0112] Steps 2.2.2 and 2.2.3 are iterated and executed in a loop until the distance between the Cartesian coordinate systems of two adjacent updated cluster centers is less than a predetermined threshold.

[0113] Step 3: Based on the quality data initially clustered in step 2, construct a data label sample balancing algorithm to generate sufficient processing quality data and balance the quality data of qualified products and unqualified products.

[0114] Step 3.1: Eliminate abnormal data:

[0115] Step 3.1.1: Calculate the average Cartesian coordinate distance between the quality data labels of the same category and the cluster center. The formula is as follows:

[0116]

[0117] In the formula, AV(QL p ) represents the average distance in Cartesian coordinates. Indicates the same type of CL m Select the quality data label QL p The Cartesian distance to all quality data labels in this class. n m Represents class CL m The number of medium-quality data labels.

[0118] Step 3.1.2: Eliminate abnormal quality data labels. Calculate the distance of each quality data label from the cluster center. Quality data labels with a distance greater than the average are eliminated as abnormal data, and quality data labels with a distance less than the average are retained for generating data samples.

[0119] Step 3.2: Generate quality data labels and balance the quality data labels. After the normal quality data labels are retained, more quality data label samples are generated based on the quality data labels. The quality data label element generation formula is as follows:

[0120]

[0121] Where GX represents the generated quality data label element value; XR i,j Indicates the difference between the quality data label element value and the cluster center element value; Represents the mean of the difference between the element value of all quality data labels and the element value of the cluster center in the same class;

[0122] When the number of samples after generating quality data labels reaches the number of classes with the most data labels, the process of generating quality data labels and balancing quality data labels for that class is terminated.

[0123] Step 4: Based on the quality data after equalization in step 3, analyze and predict the machine tool operation status.

[0124] Step 4.1 analyzes the trend of quality data changes, that is, uses the improved particle swarm algorithm to evaluate the same cluster center of all quality data, and its rules are as follows:

[0125] The sum of the distances between all quality data labels and the unique cluster center Cartesian coordinate system is defined as the optimization target of the improved particle swarm algorithm. The formula is as follows:

[0126]

[0127] In the formula, represents the update speed of the bth element of the quality data label of the jth operation on the i-th machine in the (k+1)th update cycle. ω is the inertia factor, which is used to adjust the search range of the solution space. C1 and C2 are constant accelerations. represents the local optimal position of the bth element of the quality data label for the jth operation on the i-th machine within k cycles. Indicates the position of the bth element of the quality data label for the jth operation on the i-th machine in the k-th update cycle. represents the global optimal position of the bth element of the quality data label for the jth operation on the i-th machine within k cycles.

[0128] The fitness function of the improved particle swarm optimization algorithm is the sum of the Cartesian coordinate distances from all quality data labels to the optimal clustering center, as shown in the following formula:

[0129]

[0130] In the formula, OB represents the sum of the Cartesian coordinate distances from all quality data labels to the cluster center, which is required to be as small as possible. p Indicates a quality data tag. AC k Represents the optimal cluster center. CL represents the quality data label class.

[0131] Step 4.2 predicts the machine tool operating status based on the mapping model in step 1.

[0132] Evaluate the processing quality data and obtain the following four situations according to the quality data label element type:

[0133] Scenario 1: The dimensional characteristic values, shape and position characteristic values, and roughness characteristic values ​​are all within the qualified product quality characteristic range;

[0134] Scenario 2: Among the dimensional characteristic value, shape and position characteristic value, and roughness characteristic value, one value is within the unqualified product quality characteristic range;

[0135] Scenario 3: Two of the dimensional characteristic values, shape and position characteristic values, and roughness characteristic values ​​are within the unqualified product quality characteristic range;

[0136] Scenario 4: Three of the dimensional characteristic values, shape and position characteristic values, and roughness characteristic values ​​are within the unqualified product quality characteristic range;

[0137] According to the above four situations, the conclusions of predicting the machine tool operation status are as follows:

[0138] When situation 1 occurs, the machine tool is operating normally;

[0139] When the second scenario occurs, the machine tool is in a sub-healthy state, and the probability of the machine tool being selected as an available machine tool in subsequent scheduling is reduced by 50%;

[0140] When situation three occurs, the machine tool is in a sub-healthy state, and the probability of the machine tool being selected as an available machine tool in subsequent scheduling is reduced by 75%;

[0141] When situation 4 occurs, the machine tool is in an abnormal operating state and is eliminated from subsequent scheduling.

[0142] The prediction mechanism is summarized as shown in Table 2, where DD, SD, and RD represent the size characteristics, shape and position characteristics, and roughness characteristic element values ​​of the optimal cluster center, respectively, and SR and UR represent the quality characteristic interval of qualified products and the quality characteristic interval of unqualified products, respectively.

[0143] Table 2 Machine tool operation status prediction mechanism

[0144]

[0145]

[0146] Step 5: Based on the machine tool operating status predicted in step 4, build a production process scheduling model and reallocate production processing tasks to normal machine tools.

[0147] Step 5.1 builds the production process code, that is, the matrix code that assigns multiple tasks to the corresponding machine tools, as follows:

[0148] The production process coding is defined in two-dimensional matrix form, and the formula is as follows:

[0149]

[0150] In the formula, each column of the matrix represents the machine tool code vector corresponding to the machining process operation of a task. The number of rows represents the number of tasks. p(n,s) represents the machine tool code used by the nth process operation of the sth task.

[0151] Step 5.2: According to the abnormal machine status predicted in step 4, the machine adjustment rules are formulated as follows:

[0152] Rule 1: If all quality characteristics of parts processed on a particular machine are abnormal, then that machine will be excluded from subsequent production scheduling.

[0153] Rule 2: If one or two quality characteristics of the parts processed on a given machine are abnormal, the probability of that machine being selected in subsequent scheduling is low.

[0154] Rule 3: The scheduling process is implemented in an iterative manner and is triggered when each process operation is completed to save computing resources.

[0155] The improved genetic algorithm is used to reallocate production and processing tasks to available machine tools. The fitness function is as follows:

[0156]

[0157] Fit(solu)=1 / d n ,

[0158] Where, d n Represents the production duration after scheduling. They respectively indicate the last completion time and the earliest start time of the production task. Indicates the last end time and the earliest start time of the process operation of the nth task. Fit(solu) represents the fitness function. n Indicates the construction period of the plan after production scheduling.

[0159] The optimization rules for improving genetic genes are as follows:

[0160] Selection rule: A 2 / 4 selection mechanism is used to select two better encoding matrix chromosomes from two parent chromosomes and two child chromosomes for next-generation crossover and mutation.

[0161] Crossover rule: adopt two-point crossover and select two random columns (or rows) of the encoding matrix, and make possible crossover of columns (rows) between the selected two columns (or rows).

[0162] Mutation rule: a column (or row) is selected, the mutation probability is estimated by an expert, and its elements are randomly valued based on the available machines.

[0163] The scheduling plan with the highest fitness is the optimal production scheduling plan in response to abnormal machine tool operation status.

[0164] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A method for predicting machine tool operation status and workshop scheduling based on product processing quality, characterized in that: The method comprises the following steps: Step 1: According to the operation status of the machine tool, a machine tool, processing procedure and processing quality characteristic mapping model is constructed; Step 2: Construct preliminary clustered quality data for each machine tool in the production process; Step 3: Based on the quality data of the preliminary clustering, a data label sample balancing algorithm is used to generate sufficient processing quality data and balance the quality data of qualified products and unqualified products; Step 4: Analyze and predict the machine tool operation status based on the equalized quality data; Step 5: Based on the predicted machine tool operating status, a production process scheduling model is constructed to reallocate production and processing tasks to normal machine tools; Step 2 includes: Step 2.1: For each machine tool in the production process, collect the processing quality data label samples and define the qualified product quality data and unqualified product quality data labels; Step 2.2: Use the improved K-means algorithm to evaluate the cluster center of the quality data and the corresponding data set to obtain the preliminary clustered quality data; Define the quality data labels as described in step 2.1 as follows: QL i,j =[D i,j ,S i,j ,RR i,j ] (1) In the formula, QL i,j represents the quality data label of the jth process of the part processed by the i-th machine tool; D i,j , S i,j , RR i,j Indicates the ratio of the deviation of the dimensional characteristics, shape and position characteristics, and roughness characteristics from the corresponding tolerance range; When the value of at least one element of a quality data label is within the unqualified product quality characteristic range, it is defined as an unqualified product quality data label; When the values ​​of the quality label elements are all within the qualified product quality characteristic range, it is defined as a qualified product quality data label; In step 2.2, the processing quality data labels are regarded as point coordinates in a Cartesian coordinate system, and the improved K-means algorithm is used to obtain preliminary clustered quality data, specifically including: Step 2.2.1: Each processing quality data label has three elements, each element has two conditions: qualified and unqualified. After combination, eight types of processing quality data labels can be obtained, and random values ​​are taken within the qualified and unqualified product quality characteristic interval as the initial eight cluster centers; Step 2.2.2: Cluster all quality data labels according to step 2.2.1: evaluate the class to which each quality data label belongs, i.e., calculate the Cartesian coordinate distance between the quality data label and the eight cluster centers, and classify the quality data label into the class to which the cluster center has the shortest distance; Step 2.2.3: Update cluster centers: Re-evaluate and update cluster centers based on the quality data set after clustering. The formula is as follows: Where CCX i represents the updated cluster center label element value; i represents the i-th quality data label, j represents the j-th element value; x i,j represents the jth element value of the quality data label belonging to the i-th category; Y i Represents the element value set of the i-th quality data label; n s Indicates the number of quality data labels belonging to the same category; Step 2.2.4: Terminate the preliminary clustering process: iterate and loop through steps 2.2.2 and 2.2.3 until the distance between the Cartesian coordinates of two consecutive updated cluster centers is less than a predetermined threshold, and output the quality data of the preliminary clustering; The step 3 specifically includes: Step 3.1: Eliminate abnormal data: Calculate the average Cartesian coordinate system distance between the quality data labels of the same category and the cluster center, and compare it with the distance between each quality data label and the cluster center. The quality data labels with a distance greater than the average distance are eliminated as abnormal data, and the quality data labels with a distance less than the average distance are retained for generating data samples; Step 3.2: Balance the quality data labels: For classes with a small number of quality data labels, use the following formula to generate quality data label elements: Where GX represents the generated quality data label element value; XR i,j Indicates the difference between the quality data label element value and the cluster center element value; Represents the mean of the difference between the element value of all quality data labels and the element value of the cluster center in the same class; When the number of samples after generating quality data labels reaches the number of classes with the most data labels, the process of generating quality data labels and balancing quality data labels for that class is terminated.

2. The method for predicting machine tool operation status and workshop scheduling based on product processing quality according to claim 1, characterized in that: In step 1, one machine tool can be mapped to multiple parts processing qualities, and the processing quality characteristics include size characteristics, shape and position characteristics, and roughness characteristics.

3. The method for predicting machine tool operation status and workshop scheduling based on product processing quality according to claim 1, characterized in that: The step 4 comprises: Step 4.1: Quality data change trend analysis: Use the improved particle swarm algorithm to evaluate the same cluster center of all quality data. The rules are as follows: The sum of the distances between all quality data labels and the unique cluster center Cartesian coordinate system is defined as the optimization target of the improved particle swarm algorithm. The formula is as follows: Where OB represents the sum of the distances between all quality data labels and the unique cluster center in the Cartesian coordinate system; QL p Indicates quality data label samples; AC k Represents the optimal cluster center; CL represents the quality data label class; According to the optimization goal, the clustering centers of all quality data labels are continuously updated by improving the position and speed rules of the particle swarm algorithm to obtain the optimal clustering center; Step 4.2: Predict the machine tool operating status based on the mapping model constructed in step 1 as follows: Evaluate the processing quality data and obtain the following four situations according to the quality data label element type: Scenario 1: The dimensional characteristic values, shape and position characteristic values, and roughness characteristic values ​​are all within the qualified product quality characteristic range; Scenario 2: Among the dimensional characteristic value, shape and position characteristic value, and roughness characteristic value, one value is within the unqualified product quality characteristic range; Scenario 3: Two of the dimensional characteristic values, shape and position characteristic values, and roughness characteristic values ​​are within the unqualified product quality characteristic range; Scenario 4: Three of the dimensional characteristic values, shape and position characteristic values, and roughness characteristic values ​​are within the unqualified product quality characteristic range; According to the above four situations, the conclusions of predicting the machine tool operation status are as follows: When situation 1 occurs, the machine tool is operating normally; When the second scenario occurs, the machine tool is in a sub-healthy state, and the probability of the machine tool being selected as an available machine tool in subsequent scheduling is reduced by 50%; When situation three occurs, the machine tool is in a sub-healthy state, and the probability of the machine tool being selected as an available machine tool in subsequent scheduling is reduced by 75%; When situation 4 occurs, the machine tool is in an abnormal operating state and is eliminated from subsequent scheduling.

4. The method for predicting machine tool operation status and workshop scheduling based on product processing quality according to claim 1, characterized in that: The step 5 comprises: Step 5.1: Construct the production process code and assign multiple tasks to the matrix code of the corresponding machine tools. Specifically: The production process coding is defined in two-dimensional matrix form, and the formula is as follows: In the formula, each column of the matrix represents the machine tool code vector corresponding to the machining process operation of a task; the number of rows represents the number of tasks; p(n,s) represents the machine tool code used for the nth process operation of the sth task; Step 5.2: Based on the machine tools whose operating status is abnormal as predicted in step 4, the improved genetic algorithm is used to reallocate production and processing tasks to available machine tools. Specifically, the method of reallocating production and processing tasks to available machine tools is as follows: Taking the construction period of the production scheduling plan as the optimization target, the improved genetic algorithm is used for evaluation to calculate the scheduling plan with the shortest construction period; The calculation formula for construction period is as follows: Where, d n Indicates the production duration after scheduling; Respectively indicate the last end time and the earliest start time of the production task; Indicates the last completion time and the earliest start time of the process operation of the nth task; The optimization objective formula is as follows: Fit(only)=1 / d n (7) Where Fit(solu) represents the fitness function; d n It indicates the duration of the plan after production scheduling; According to the above optimization objective formula, the optimal production scheduling plan is obtained by improving the genetic algorithm to optimize the scheduling plan.

Images