A factory production planning design system based on order analysis

Abstract

The present invention belongs to the technical field of production plan design, and discloses a factory production plan design system based on order analysis; the system comprises: obtaining historical production data of a factory and relevant data of pending orders, and pre-processing the historical production data to obtain historical perfect production data; calculating the time urgency of the pending orders according to the relevant data; constructing a production line constraint relationship model, and based on the production line constraint relationship model, constructing a production capacity matrix of the factory using the historical perfect production data; based on the time urgency of the pending orders and the production capacity matrix of the factory, solving using an optimization algorithm to obtain an optimal production plan, applying the optimal production plan to the factory production line, and realizing a globally optimal production plan, thereby significantly improving the delivery guarantee capability of the enterprise.

Description

Technical Field

[0001] The present invention relates to the technical field of production plan design, and more specifically, to a factory production plan design system based on order analysis. Background Art

[0002] The patent application with the publication number CN109426920A discloses a method for optimizing enterprise production plans taking into account predicted orders and actual orders, including the following steps: obtaining historical order task related data and classifying the data as input for order task prediction; constructing an order task prediction model based on an improved time series; combining the predicted order task data and the actual order task data to formulate a corresponding production plan optimization objective function; establishing a hybrid optimization algorithm based on the traditional artificial bee colony algorithm and the Tabu Search algorithm to optimize and solve the optimization objective function, and the optimization result is the production plan details to guide the development and implementation of actual production activities; by dividing the production plan optimization into two stages, the formulation and implementation of the production plan of the textile machinery manufacturing enterprise is more accurately realized, thereby being able to provide effective guidance for the production plan optimization of the corresponding small and medium-sized manufacturing enterprises.

[0003] However, the existing technology still has the problem that when formulating production plans, it often only considers a single dimension such as order delivery time or capacity utilization rate, while ignoring other important influencing factors such as raw material supply and constraints between production lines, resulting in the inability to effectively guarantee the execution efficiency and quality of the plan; secondly, for the collection of multiple data, outlier processing is very important. The existing method relies entirely on manual experience judgment and lacks an automated and intelligent outlier identification mechanism, which is not only inefficient but also prone to subjective bias, affecting the accuracy of subsequent analysis and decision-making; thirdly, the existing technology lacks scientific modeling and quantitative analysis between production lines, and cannot give full play to the synergy of production lines and maximize the utilization of production capacity. At the same time, it cannot provide theoretical support for the formulation of efficient production plans; in addition, due to the lack of reasonable quantitative analysis of the time urgency of orders, it is difficult for existing technologies to scientifically determine the priority of different orders, and the formulated production plans lack pertinence, and the company's delivery guarantee ability is poor, affecting market competitiveness; finally, the production plan formulation process lacks optimization model and algorithm support, making it difficult to efficiently find the optimal solution under multi-objective constraints, and unable to achieve the global optimal production plan, which restricts the modernization and intelligent development of production management.

[0004] In view of this, the present invention proposes a factory production planning design system based on order analysis to solve the above problems. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solutions: A factory production plan design system based on order analysis, comprising: a data acquisition and processing module, for obtaining the factory's historical production data and the relevant data of the orders to be processed, and pre-processing the historical production data to obtain historical perfect production data;

[0006] The time fitting module is used to calculate the time urgency of pending orders based on relevant data;

[0007] Model matrix construction module, used to construct the production line constraint relationship model. Based on the production line constraint relationship model, the factory production capacity matrix is ​​constructed by using historical perfect production data;

[0008] The plan control module uses an optimization algorithm to solve the optimal production plan based on the time urgency of pending orders and the factory's production capacity matrix, and applies the optimal production plan to the factory production line. The various modules are connected by wired and / or wireless means.

[0009] Furthermore, the historical production data includes product data and production data corresponding to orders within a fixed time period in the past, the product data includes the production volume, production cost and inventory cost of various products; the production data includes the production efficiency of each production line and the inventory of raw materials;

[0010] Relevant data include product quantity requirements, order delivery deadlines, types of raw materials required, quantities of raw materials required and product inventory levels.

[0011] Furthermore, the method of preprocessing the historical production data includes:

[0012] Identify outliers from historical production data; analyze the data points corresponding to the identified outliers to obtain analysis results, which include the distribution of outliers, numerical characteristics of outliers and mean differences;

[0013] According to the analysis results, the nature characteristics of the abnormal values ​​are obtained, which are extreme values, outliers, short-term mutations, long-term deviations or abnormal fluctuations;

[0014] Delete or smooth outliers whose characteristics are extreme values ​​or outliers, interpolate outliers whose characteristics are short-term mutations, perform segmentation processing or trend correction on outliers whose characteristics are long-term deviations; denoise outliers whose characteristics are abnormal fluctuations; and obtain historical and complete production data.

[0015] Furthermore, the method of identifying abnormal values ​​from historical production data includes:

[0016] Clean and normalize historical production data;

[0017] The optimization parameters defined include population size, maximum explosion amplitude A, explosion seed constant a, mutation probability pm and maximum number of iterations MA;

[0018] Define the initial N individuals to form the initial population, including normal individuals and abnormal individuals;

[0019] Different types of data points in the historical production data are respectively formed into corresponding data sets; for any data point x in the data set, its average density di(x) is defined;

[0020] ;

[0021] Where μ is the mean vector of the data set, M is the size of the data set, N_k(x) represents the set of data points within the circle with x as the center and k as the radius; y is the data point within N_k(x), k is the radius, t(x) is the timestamp of data point x, t(y) is the timestamp of data point y, and σ is the parameter that controls the attenuation width; is the transpose of xy; is the transpose of x-μ;

[0022] Sort the average density of all data points from small to large, take the first m data points as outlier individuals, and the rest of the data points as normal value individuals;

[0023] Define the fitness function fin; for each individual, calculate the value of its fitness function, record it as the fitness value, arrange the individuals from large to small according to the fitness value, and select the first M3 individuals as elite individuals;

[0024] According to the explosion seed constant a, a random explosion seed r is generated to calculate the explosion radius of the elite individual ;

[0025] A hypersphere region is created with the elite individual as the center and the explosion radius R as the radius. Within the hypersphere region, N2 new candidate individuals are randomly sampled according to Gaussian distribution;

[0026] For each of the N2 candidate individuals, calculate the Euclidean distance DS between it and the elite individual; if DS is greater than R, discard the candidate individual; if DS is less than or equal to R, perform a mutation operation on it according to the Euclidean distance DS to generate a new JP;

[0027] The formula for the mutation operation is: ; Where N(0,1) is a random value that obeys the standard normal distribution, k1 and k2 are mutation intensity parameters, and fin(JY) is the fitness value of the elite individual JY;

[0028] Add new individuals that have not been discarded to the population, and repeat the process of sampling candidate individuals and performing mutation operations on all elite individuals to obtain a new population;

[0029] From the new population, randomly select N4 individuals according to the preset mutation probability pm, and randomly select one or m1 gene positions from these N4 individuals, and perform mutation operation on the selected lth gene position;

[0030] The formula for mutation operation is: , where g_l' is the new gene value of the gene position after mutation, g_l is the original gene value of the gene position, gu_l is the upper boundary value of the gene position, and gw_l is the lower boundary value of the gene position;

[0031] ; Where X is a preset random number in the interval [0, 1], t' is the current iteration number, B is the mutation control parameter, B1 is the fitness control parameter, and fin(JK) is the fitness value of the individual JK to which the mutated gene belongs;

[0032] Repeat until all selected individuals have completed the mutation operation; obtain newly generated individuals; merge the newly generated individuals into the population, and select elite individuals according to the fitness value to form a new generation of population. If the termination condition is met, the outlier detection scheme corresponding to the optimal individual is output; otherwise, it will continue to iterate until the termination condition is met, and the outlier detection scheme is obtained. The outlier detection scheme is applied to the historical production data to identify outliers.

[0033] Furthermore, the formula of the fitness function fin is:

[0034] ; Where TP is the number of correctly detected outliers, FP is the number of normal values ​​that are incorrectly detected as outliers, and FN is the number of undetected outliers; γ and θ are normal parameters; FD is the penalty term; α1, α2, α3, α4, and α5 are the normal weights for adjusting each term;

[0035] Penalty ; Where g(v) is the constraint function and δ is the penalty weight; is the maximum value; v is the decision variable;

[0036] ; Wherein, fi(v) is the original constraint function, β1 and ε1 are positive adjustment parameters, γ1 is an exponential parameter; fi_max is the maximum value of fi(v) in the feasible domain, fi_min is the minimum value of fi(v) in the feasible domain, fi_avg is the average value of fi(v) in the feasible domain, and bi is a constant;

[0037] Original constraint function

[0038] ; Where x_i,q1 represents the number of products assigned to production line q1 for order i, Cap_q1 is the capacity limit of production line q1, δ2 is the allowed excess capacity ratio; ρ_q1(t) is the capacity time function, and t is the time index;

[0039] ; Among them, ω_q1 is the capacity fluctuation coefficient of production line q1, T_q1 is the capacity fluctuation period of production line q1, is the initial phase of the capacity fluctuation of production line q1, ∈_q1(t) is the random disturbance term of production line q1 at time t.

[0040] Furthermore, the calculation method of the time urgency includes:

[0041] Based on the relevant data, the time required for completion and the time required for raw material preparation are calculated; if the product quantity demand Qi is less than or equal to the product inventory Si, the remaining production quantity Ni is defined as 0; if the product quantity demand Qi is greater than the product inventory Si;

[0042] The production quantity still needs to be ; Calculate the time required to complete pending order I ; Where Ni_Ij is the remaining production quantity of product j in pending order I, tu_Ij is the unit production time of product j in pending order I, and tb_j is the unit production time of product j in the bottleneck process;

[0043] Build a supply chain network, use suppliers, factories and raw materials as nodes of the supply chain network, use supply paths as directed edges to connect nodes, and mark the transportation time and transportation cost on each directed edge;

[0044] Obtain the transportation time and cost of the supply route based on the type of raw materials required, the quantity of raw materials required, and the raw materials inventory;

[0045] Build an optimization model and use linear programming methods to solve the optimization model and obtain the optimal solution ;

[0046] For each required raw material f, its preparation time ; Where T_sf is the transportation time for purchasing the required raw material f from supplier s, then the total preparation time t2=max(t_f);

[0047] Based on the time required to complete pending order I and the total preparation time, the capacity-limited time is obtained. ;

[0048] The order delivery deadline and capacity-constrained time TB are comprehensively modeled using a Gaussian mixture model, and the cumulative distribution function FTR of the comprehensive delivery time distribution is fitted. The threshold quantile PN is set, and the numerical optimization algorithm is used to solve FTR=PN to obtain the time-critical threshold Ty.

[0049] Taking the current time t0 as the benchmark, for each pending order I, calculate the remaining delivery time TP_I;

[0050] TP_I=YU_I-t0; where YU_I is the order delivery deadline of pending order I; if TP_I is less than or equal to Ty, the time urgency of pending order I is GL_I;

[0051] ; Wherein, A1 is the maximum value of time urgency, B1 is the offset of time urgency; k3 is the curve steepness parameter; if TP_I is greater than Ty, the time urgency GL_I of pending order I=0.

[0052] Furthermore, the optimization model is constructed in the following manner:

[0053] The basic framework for defining the optimization model includes the objective function LOM and constraints;

[0054] Among them, C sf is the cost of purchasing the raw material f required per unit from supplier s, U sf is the quantity of raw material f purchased from supplier s, t_sf is the delivery time of raw material f purchased from supplier s; Tt is the upper limit of the expected total preparation time, θ1 is the time limit penalty coefficient, e_sf is the environmental impact assessment score of purchasing raw material f from supplier s; τ1 is the environmental impact penalty coefficient;

[0055] The constraints include the safety stock condition function PUY, the supply capacity constraint function PUT and the diversity constraint PUE;

[0056] Safety stock condition function ; Among them, α_f is the safety stock factor of the required raw material f, QL_f is the quantity of the required raw material f, and SL_f is the current inventory of the required raw material f;

[0057] Supply Capacity Constraint Function ; Where G_s is the upper limit of supplier s’s supply capacity; RT_s is the reputation score of supplier s, and RP is the preset maximum weighted purchase quantity;

[0058] Diversity Constraints ; Among them, τ4 is the proportion of the largest single supplier.

[0059] Furthermore, the production line constraint relationship model is constructed in the following manner:

[0060] Construct the production line correlation matrix and initialize a Matrix R, n is the number of production lines;

[0061] For any two production lines q1 and q2, calculate their correlation coefficient as the elements of the matrix R;

[0062] The formula for calculating the correlation coefficient is:

[0063] ;WQ(rg_q1-rq_q2) is the grade difference weight function, rg_q1 is the capacity utilization grade of production line q1, rg_q2 is the capacity utilization grade of production line q2, θ2 is the index parameter, ρ_q1,q2 is the correlation coefficient between production lines q1 and q2; δ3 is the root index parameter;

[0064] Rank difference weight function ; Among them, μ1' is the location parameter of the grade difference, σ1' is the scale parameter of the grade difference;

[0065] Each production line is constructed as a graph node in an undirected graph. For production lines q1 and q2 whose correlation coefficient is not 0, an edge is connected between their corresponding graph nodes. The weight of the edge is 1-ρ_q1,q2, and a weighted undirected graph is obtained.

[0066] Run the minimum spanning tree algorithm in the weighted undirected graph to obtain the minimum spanning tree connecting all graph nodes; count the degree of each graph node in the minimum spanning tree, and take the production line with the largest degree as the key production line;

[0067] According to the production process, all production lines are divided into n5 levels; the key production lines are placed in the middle level of the n5 levels, and a capacity limitation model is established for each level;

[0068] The expression of the capacity constraint model is:

[0069] ;

[0070] Among them, f_K'_q1,q2 is the allocation ratio of the input obtained by the production line q1 of the K'th level from the production line q2 of the K'-1th level, cap_K'_q1 is the production capacity of the production line q1 of the K'th level; cap_(K'-1)_q2 is the production capacity of the production line q2 of the K'-1th level.

[0071] Furthermore, the production capacity matrix is ​​constructed in the following manner:

[0072] Based on the historical production data and production line constraint relationship model, the unit time capacity of each production line is calculated using regression analysis method; an initial matrix is ​​constructed, and the unit time capacity of all production lines is merged into the production capacity matrix. The rows of the matrix correspond to production lines, the columns correspond to different product types, and the element value represents the unit time capacity of the product corresponding to the production line, that is, the production capacity matrix of the factory is obtained.

[0073] Furthermore, the method for obtaining the optimal production plan includes:

[0074] Generate an initial cat group of size N5, where each cat in the initial cat group corresponds to a production plan; define a production matrix E to represent the production plan, where the rows of the production matrix E represent production lines, the columns represent pending orders, and each element value of the production matrix represents the quantity allocated to the corresponding production line to produce the order;

[0075] Define the target optimization function F'(E) of the cat group;

[0076] ;

[0077] Where P'_I(E) is the planned delivery time of pending order I calculated according to the production matrix E, V'_IJ' is the number of pending orders I assigned to the J'th production line, y_J' is the unit time capacity of the J'th production line, and b1 and b2 are the target weight coefficients;

[0078] Calculate the value of each cat for the target optimization function and take the negative number as the survival parameter of the corresponding cat;

[0079] Sort each cat in ascending order according to the survival parameter, select the first M6 cats as member cats, and the rest as prey;

[0080] Define the position of the u-th member cat at the current iteration number s' as YT_u_(s'), and its speed as VT_u_(s'), then the member cat updates its own position according to its own speed and the positions of other member cats;

[0081] Define a tracking random number rd in the interval [0, 1] for each cat, and preset the tracking probability p_track; if rd is less than p_track, update the position of the u-th member cat; the updating formula is:

[0082] ; Where YT_u_(s'+1) is the position of the u-th member cat at iteration number s'+1, r3 is a global random number, and YT_best_(s') is the best historical position at iteration number s';

[0083] For the cats that are prey, random disturbances are performed with a preset disturbance probability; the cats are re-sorted in ascending order according to the size of their survival parameters, and the first N7 cats are selected to form a new generation of cat population; until the target optimization function converges, the production plan corresponding to the cat with the smallest current survival parameter is output, which is the optimal production plan.

[0084] Technical effects and advantages of a factory production planning design system based on order analysis of the present invention:

[0085] The present invention realizes the comprehensive analysis and optimization of multi-dimensional information such as order demand, raw material supply, and production line constraints, greatly improving the execution efficiency and quality of the production plan; secondly, an automatic outlier detection mechanism based on an intelligent optimization algorithm is introduced. Compared with the traditional manual judgment method, the accuracy and efficiency of outlier identification are greatly enhanced; thirdly, a production line constraint relationship model is constructed to quantitatively describe the correlation and constraint relationship between production lines, which is not only conducive to giving full play to the synergy between production lines and maximizing the utilization of production capacity, but also provides an important theoretical basis for formulating efficient production plans; in addition, the quantitative modeling method of order time urgency can reasonably determine the priority of different orders, and model the production plan formulation problem as an optimization problem, effectively find the optimal solution under multi-objective constraints, and realize the global optimal production plan, thereby significantly improving the company's delivery guarantee capability. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 A schematic diagram of a factory production planning design system based on order analysis of the present invention;

[0087] Figure 2 It is a schematic diagram of a factory production plan design method based on order analysis of the present invention. DETAILED DESCRIPTION

[0088] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0089] Example 1

[0090] See also Figure 1 As shown, the factory production planning design system based on order analysis described in this embodiment includes:

[0091] The data acquisition and processing module is used to obtain the factory's historical production data and the relevant data of pending orders, and pre-process the historical production data to obtain historical perfect production data;

[0092] The time fitting module is used to calculate the time urgency of pending orders based on relevant data;

[0093] Model matrix construction module, used to construct the production line constraint relationship model. Based on the production line constraint relationship model, the factory production capacity matrix is ​​constructed by using historical perfect production data;

[0094] The plan control module uses an optimization algorithm to solve the optimal production plan based on the time urgency of pending orders and the factory's production capacity matrix, and applies the optimal production plan to the factory production line. The modules are connected by wired and / or wireless means to achieve efficient production and data transmission between modules.

[0095] Furthermore, historical production data includes product data and production data corresponding to orders within a fixed time period in the past. Product data includes production volume, production cost and inventory cost of various products. Production data includes production efficiency of each production line (comparing output with theoretical full production capacity to calculate production efficiency) and inventory of raw materials.

[0096] Relevant data include product quantity requirements, order delivery deadlines, types of raw materials required, quantities of raw materials required and product inventory levels.

[0097] Furthermore, the methods for preprocessing the historical production data include:

[0098] Identify outliers from historical production data; specifically, clean the historical production data by removing invalid data, including null values ​​and duplicate values; normalize the historical production data and map the data to the same numerical range.

[0099] The optimization parameters defined include population size, maximum explosion amplitude A, explosion seed constant a, mutation probability pm and maximum number of iterations MA;

[0100] Define the initial N individuals to constitute the initial population (individuals are encoded in real numbers), and each individual in the population represents a possible outlier detection scheme; individuals include normal value individuals and outlier individuals; when performing outlier detection, it is necessary to set some parameters, such as anomaly score threshold, number of neighbors, kernel bandwidth, etc. Different parameter combinations represent different outlier detection schemes or strategies.

[0101] Each individual is actually a combination of settings of the outlier detection algorithm parameters, representing a possible detection scheme. By evaluating and evolving the fitness of all individuals in the population, the optimal or approximately optimal combination of outlier detection parameters is automatically searched to obtain the ideal detection effect.

[0102] Different types of data points in the historical production data are respectively formed into corresponding data sets; for any data point x in the data set, its average density di(x) is defined;

[0103] ;

[0104] Where μ is the mean vector of the data set, M is the size of the data set, N_k(x) represents the set of data points within the circle with x as the center and k as the radius; y is the data point within N_k(x), k is the radius, t(x) is the timestamp of data point x, t(y) is the timestamp of data point y, and σ is the parameter that controls the attenuation width; is the transpose of xy; is the transpose of x-μ;

[0105] Sort the average density of all data points from small to large, take the first m data points as outlier individuals; the remaining data points are normal value individuals.

[0106] The formula of fitness function fin is:

[0107] ; Where TP is the number of correctly detected outliers, FP is the number of normal values ​​that are incorrectly detected as outliers, and FN is the number of undetected outliers; γ and θ are normal parameters; FD is the penalty term; α1, α2, α3, α4, and α5 are the normal weights for adjusting each term;

[0108] Penalty ; Where g(v) is the constraint function and δ is the penalty weight; is the maximum value; v is the decision variable;

[0109] ; Wherein, fi(v) is the original constraint function, β1 and ε1 are positive adjustment parameters, γ1 is an exponential parameter; fi_max is the maximum value of fi(v) in the feasible domain, fi_min is the minimum value of fi(v) in the feasible domain, fi_avg is the average value of fi(v) in the feasible domain, and bi is a constant;

[0110] Original constraint function

[0111] ; Where x_i,q1 represents the number of products assigned to production line q1 for order i, Cap_q1 is the capacity limit of production line q1, δ2 is the allowed excess capacity ratio; ρ_q1(t) is the capacity time function, and t is the time index;

[0112] ; Among them, ω_q1 is the capacity fluctuation coefficient of production line q1, T_q1 is the capacity fluctuation cycle of production line q1 (set according to the actual production plan cycle), is the initial phase of the capacity fluctuation of production line q1 (different values ​​are set for different production lines to reflect staggered fluctuations), ∈_q1(t) is the random disturbance term of production line q1 at time t (obeying Gaussian distribution).

[0113] For each individual, calculate the value of its fitness function, record it as the fitness value, arrange the individuals from large to small according to the fitness value, and select the first M3 individuals as elite individuals;

[0114] According to the explosion seed constant a, a random explosion seed r is generated to calculate the explosion radius of the elite individual ;

[0115] A hypersphere region is created with the elite individual as the center and the explosion radius R as the radius. Within the hypersphere region, N2 new candidate individuals are randomly sampled according to Gaussian distribution;

[0116] For each of the N2 candidate individuals, calculate the Euclidean distance DS between it and the elite individual; if DS is greater than R, discard the candidate individual; if DS is less than or equal to R, perform a mutation operation on it according to the Euclidean distance DS to generate a new JP;

[0117] The formula for the mutation operation is: ; Among them, N(0,1) is a random value that obeys the standard normal distribution, k1 and k2 are mutation intensity parameters, and fin(JY) is the fitness value of the elite individual JY.

[0118] Add new individuals that have not been discarded to the population, and repeat the process of sampling candidate individuals and performing mutation operations on all elite individuals to obtain a new population;

[0119] From the new population, randomly select N4 individuals according to the preset mutation probability pm, and randomly select one or m1 gene positions from these N4 individuals, and perform mutation operation on the selected lth gene position;

[0120] The formula for mutation operation is: , where g_l' is the new gene value of the gene position after mutation, g_l is the original gene value of the gene position, gu_l is the upper boundary value of the gene position, and gw_l is the lower boundary value of the gene position;

[0121] ; Where X is a preset random number in the interval [0, 1], t' is the current iteration number, B is the mutation control parameter, B1 is the fitness control parameter, and fin(JK) is the fitness value of the individual JK to which the mutated gene belongs;

[0122] Repeat until all selected individuals have completed the mutation operation; obtain newly generated individuals; merge the newly generated individuals into the population, and select elite individuals according to the fitness value to form a new generation of population. If the termination condition (such as the maximum number of iterations or the objective function value) is met, the outlier detection scheme corresponding to the optimal individual (with the largest fitness value) is output; otherwise, the iteration will continue until the termination condition is met, and the outlier detection scheme is obtained. The outlier detection scheme is applied to the historical production data to identify outliers.

[0123] The data points corresponding to the identified outliers are analyzed to obtain analysis results, which include the distribution of outliers (concentrated or dispersed), the numerical characteristics of outliers (maximum or minimum) and the mean difference (between outliers and normal values).

[0124] According to the analysis results, the nature characteristics of the outliers are obtained, which are extreme values, outliers, short-term mutations, long-term deviations or abnormal fluctuations.

[0125] Delete or smooth outliers whose characteristics are extreme values ​​or outliers, interpolate outliers whose characteristics are short-term mutations, perform segmentation processing or trend correction on outliers whose characteristics are long-term deviations; denoise outliers whose characteristics are abnormal fluctuations; and obtain historical and complete production data.

[0126] The distribution of outliers is obtained using the central tendency measurement method in statistics. Specifically, the interquartile range (IQR) and mean absolute deviation (MAD) are calculated. The smaller the IQR and MAD values ​​are, the higher the degree of data concentration is and the more concentrated the outliers are. A threshold is set. When the IQR or MAD is less than the threshold, the outliers are considered to be concentrated.

[0127] If the maximum / minimum difference between the abnormal value and the normal value exceeds the preset threshold, the property characteristic is an extreme value; if the IQR or MAD is large, and the maximum / minimum difference between the abnormal value and the normal value exceeds the preset threshold, the property characteristic is an outlier; if the difference mean of several consecutive abnormal values ​​is large, but only lasts for a short time, the property characteristic is a short-term mutation; if the difference mean continues to deviate from the normal level for a long time, the property characteristic is a long-term deviation; by calculating indicators such as variance, it is evaluated that the property characteristic is an abnormal fluctuation.

[0128] Furthermore, the calculation method of time urgency includes:

[0129] Relevant data include product quantity requirements, order delivery deadlines, types of raw materials required, quantities of raw materials required and product inventory levels.

[0130] Based on the relevant data, the time required for completion and the time required for raw material preparation are calculated; if the product quantity demand Qi is less than or equal to the product inventory Si, the remaining production quantity Ni is defined as 0; if the product quantity demand Qi is greater than the product inventory Si;

[0131] The production quantity still needs to be ; Calculate the time required to complete pending order I ; Where Ni_Ij is the remaining production quantity of product j in pending order I, tu_Ij is the unit production time of product j in pending order I, and tb_j is the unit production time of product j in the bottleneck process;

[0132] Build a supply chain network, use suppliers, factories and raw materials as nodes of the supply chain network, use supply paths as directed edges to connect nodes, and mark the transportation time and transportation cost on each directed edge;

[0133] Obtain the transportation time and cost of the supply route based on the type of raw materials required, the quantity of raw materials required, and the raw materials inventory;

[0134] Build an optimization model and define the basic framework of the optimization model including the objective function LOM and constraints;

[0135] Among them, C sf is the cost of purchasing the raw material f required per unit from supplier s, U sfis the quantity of raw material f purchased from supplier s, t_sf is the delivery time of raw material f purchased from supplier s; Tt is the expected upper limit of total preparation time (when the delivery time of a supplier exceeds the upper limit of the total time, it will be punished twice to avoid serious time overruns), θ1 is the time overrun penalty coefficient, e_sf is the environmental impact assessment score of purchasing raw material f from supplier s (evaluate the greenhouse gas emissions generated during the production and transportation of purchased raw materials, expressed in carbon dioxide equivalent, and calculate its carbon footprint score to represent the environmental impact assessment score); τ1 is the environmental impact penalty coefficient;

[0136] The constraints include the safety stock condition function PUY, the supply capacity constraint function PUT and the diversity constraint PUE;

[0137] Safety stock condition function ; Among them, α_f is the safety stock factor of the required raw material f, QL_f is the quantity of the required raw material f, and SL_f is the current inventory of the required raw material f;

[0138] Supply Capacity Constraint Function ; Among them, G_s is the upper limit of supplier s's supply capacity; RT_s is supplier s's credibility score (including a weighted score of the supplier's past performance in terms of delivery time, product quality, service attitude, etc.), RP is the preset maximum weighted purchase volume, and suppliers with high credibility can be allocated more purchase volume.

[0139] Diversity Constraints ; Among them, τ4 is the proportion of the largest single supplier, which avoids excessive concentration of the supply chain.

[0140] Use linear programming method to solve the optimization model and get the optimal solution ; For each required raw material f, its preparation time ; Where T_sf is the transportation time for purchasing the required raw material f from supplier s, then the total preparation time t2=max(t_f);

[0141] Based on the time required to complete pending order I and the total preparation time, the capacity-limited time is obtained. ;

[0142] The order delivery deadline and capacity-constrained time TB are comprehensively modeled using a Gaussian mixture model (GMM), and the cumulative distribution function FTR of the comprehensive delivery time distribution is fitted. The threshold quantile PN (such as the 0.9 quantile) is set, and a numerical optimization algorithm (such as the gradient descent method) is used to solve FTR=PN to obtain the time-critical threshold Ty.

[0143] Taking the current time t0 as the benchmark, for each pending order I, calculate the remaining delivery time TP_I;

[0144] TP_I=YU_I-t0; where YU_I is the order delivery deadline of pending order I; if TP_I is less than or equal to Ty, the time urgency of pending order I is GL_I;

[0145] ; Wherein, A1 is the maximum value of time urgency, B1 is the offset of time urgency; k3 is the curve steepness parameter; if TP_I is greater than Ty, the time urgency GL_I of pending order I=0.

[0146] Furthermore, the production line constraint relationship model is constructed in the following ways:

[0147] Construct the production line correlation matrix and initialize a Matrix R, n is the number of production lines;

[0148] For any two production lines q1 and q2, their correlation coefficients are calculated as elements of the matrix R.

[0149] The calculation formula of the correlation coefficient is:

[0150] ;WQ(rg_q1-rq_q2) is the grade difference weight function, rg_q1 is the capacity utilization grade of production line q1, rg_q2 is the capacity utilization grade of production line q2, θ2 is the exponential parameter, ρ_q1,q2 is the correlation coefficient between production lines q1 and q2; δ3 is the root exponential parameter; it is used to control the degree of nonlinear transformation of the correlation coefficient; it should be noted that for the two production lines, their capacity utilization data in the past several time units are X1 and Y1 respectively;

[0151] Where X1=(x1, x2, ..., xn); Y1=(y1, y2, ..., yn); sort the data values ​​in X1 and Y1 from small to large to get their levels, which is the capacity utilization level.

[0152] Rank difference weight function ; Among them, μ1' is the location parameter of the level difference, which controls where the weight function reaches its maximum value, and σ1' is the scale parameter of the level difference, which controls the steepness of the weight function curve.

[0153] Each production line is constructed as a graph node in an undirected graph. For production lines q1 and q2 whose correlation coefficient is not 0, an edge is connected between their corresponding graph nodes. The weight of the edge is 1-ρ_q1,q2, and a weighted undirected graph is obtained.

[0154] Run a minimum spanning tree algorithm, such as Kruskal's algorithm or Prim's algorithm, in a weighted undirected graph to obtain a minimum spanning tree connecting all graph nodes; count the degree of each graph node in the minimum spanning tree, and take the production line with the largest degree as the key production line;

[0155] The degree of each graph node (production line) in the minimum spanning tree (the number of edges directly connected to the graph node) is counted. The larger the degree, the closer the constraint relationship between the production line and other lines. The production line with the largest degree is regarded as the key production line (bottleneck process).

[0156] According to the production process, all production lines are divided into n5 levels. The production lines at the same level are independent of each other; the input of the production line at the upper level comes from the output of the production line at the lower level.

[0157] Place the key production line in the middle level of the n5 levels, and for each level, establish a capacity limitation model (the capacity of the upper production line is limited by the output of the lower production line).

[0158] The expression of the capacity constraint model is:

[0159] ;

[0160] Among them, f_K'_q1,q2 is the allocation ratio of the input obtained by the production line q1 of the K'th level from the production line q2 of the K'-1th level, cap_K'_q1 is the production capacity of the production line q1 of the K'th level; cap_(K'-1)_q2 is the production capacity of the production line q2 of the K'-1th level.

[0161] By cascading the capacity constraint models at each level, we can construct a production line constraint relationship model. The capacity of the top-level production line is limited by the capacity of the bottom-level production line. Through the cascade model, we can fully describe the constraint relationship of the entire production system.

[0162] Based on the historical production data and production line constraint relationship model, the unit time capacity of each production line is calculated using regression analysis method; an initial matrix is ​​constructed, and the unit time capacity of all production lines is merged into the production capacity matrix. The rows of the matrix correspond to production lines, the columns correspond to different product types, and the element value represents the unit time capacity of the product corresponding to the production line, that is, the production capacity matrix of the factory is obtained.

[0163] Furthermore, the optimal production plan is obtained by:

[0164] Generate an initial cat group of size N5 (the number of cats), and each cat in the initial cat group corresponds to a possible production plan.

[0165] Define the production matrix E to represent the production plan. The rows of the production matrix E represent the production lines, the columns represent the pending orders, and each element value of the production matrix represents the quantity allocated to the corresponding production line to produce the order;

[0166] Define the target optimization function F'(E) of the cat group;

[0167] ;

[0168] Where P'_I(E) is the planned delivery time of pending order I calculated according to the production matrix E, V'_IJ' is the number of pending orders I assigned to the J'th production line, y_J' is the unit time capacity of the J'th production line (obtained based on the production capacity matrix), b1 and b2 are target weight coefficients used to adjust the weights of the two operations.

[0169] Calculate the value of each cat for the target optimization function and take the negative number as the survival parameter of the corresponding cat;

[0170] Sort each cat in ascending order according to the survival parameter, select the first M6 cats as member cats, and the rest as prey;

[0171] Define the position of the u-th member cat at the current iteration number s' as YT_u_(s'), and its speed as VT_u_(s'), then the member cat updates its own position according to its own speed and the positions of other member cats;

[0172] Define a tracking random number rd in the interval [0, 1] for each cat, and preset the tracking probability p_track; if rd is less than p_track, update the position of the u-th member cat; the updating formula is:

[0173] ; Among them, YT_u_(s'+1) is the position of the u-th member cat at iteration number s'+1, r3 is a global random number, and YT_best_(s') is the historical best position at iteration number s', that is, the best solution encountered by the u-th cat during its own search process.

[0174] And for the cats used as prey, random perturbations are performed with a preset perturbation probability; all cats (including the original cats and the cats after position update) are re-sorted in ascending order according to the size of the survival parameters, and the first N7 cats are selected to form a new generation of cat population.

[0175] Until the target optimization function converges, the production plan corresponding to the cat with the smallest current survival parameter is output, which is the optimal production plan.

[0176] This embodiment realizes the comprehensive analysis and optimization of multi-dimensional information such as order demand, raw material supply, and production line constraints, which greatly improves the execution efficiency and quality of the production plan; secondly, an automatic detection mechanism for outliers based on an intelligent optimization algorithm is introduced. Compared with the traditional manual judgment method, the accuracy and efficiency of outlier identification are greatly enhanced; thirdly, a production line constraint relationship model is constructed to quantitatively describe the correlation and constraint relationship between production lines, which is not only conducive to giving full play to the synergy between production lines and maximizing the utilization of production capacity, but also provides an important theoretical basis for formulating efficient production plans; in addition, the quantitative modeling method of order time urgency can reasonably determine the priority of different orders, and model the production plan formulation problem as an optimization problem, effectively find the optimal solution under multi-objective constraints, and realize the global optimal production plan, thereby significantly improving the company's delivery guarantee capability.

[0177] Example 2

[0178] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, which provides a factory production plan design method based on order analysis, including:

[0179] S1. Obtain the factory's historical production data and relevant data of pending orders, and pre-process the historical production data to obtain historical perfect production data;

[0180] S2. Calculate the time urgency of the pending orders based on relevant data;

[0181] S3. Construct a production line constraint relationship model. Based on the production line constraint relationship model, the production capacity matrix of the factory is constructed by using historical perfect production data;

[0182] S4. Based on the time urgency of pending orders and the factory's production capacity matrix, the optimization algorithm is used to solve the optimal production plan, and the optimal production plan is applied to the factory production line.

[0183] Example 3

[0184] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the factory production plan design method based on order analysis provided above is implemented.

[0185] Since the electronic device introduced in this embodiment is an electronic device used to implement a factory production plan design method based on order analysis in the embodiment of the present application, based on the factory production plan design method based on order analysis introduced in the embodiment of the present application, the technical personnel of this field can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of the present application is not described in detail here. As long as the technical personnel of this field implement the electronic device used in the factory production plan design method based on order analysis in the embodiment of the present application, it belongs to the scope of protection of this application.

[0186] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0187] The above is only a preferred embodiment 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 technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A factory production planning design system based on order analysis, characterized in that: include: The data acquisition and processing module is used to obtain the factory's historical production data and the relevant data of pending orders, and pre-process the historical production data to obtain historical perfect production data; The time fitting module is used to calculate the time urgency of pending orders based on relevant data; Model matrix construction module, used to construct the production line constraint relationship model. Based on the production line constraint relationship model, the factory production capacity matrix is ​​constructed by using historical perfect production data; The plan control module uses an optimization algorithm to solve the optimal production plan based on the time urgency of pending orders and the factory's production capacity matrix, and applies the optimal production plan to the factory production line. Each module is connected by wired and / or wireless means; The construction method of the production line constraint relationship model includes: Construct the production line correlation matrix and initialize a The matrix R, is the number of production lines; For any two production lines and , calculate their correlation coefficients as elements of the matrix R; The formula for calculating the correlation coefficient is: ;in, is the grade difference weight function, For production line The capacity utilization level, For production line The capacity utilization level, is the exponential parameter, For production line and The correlation coefficient between is the root index parameter; Rank difference weight function ;in, is the location parameter of the grade difference, is the scale parameter of the grade difference; Each production line is constructed as a graph node in an undirected graph. For production lines with a correlation coefficient not equal to 0, and , connect an edge between their corresponding graph nodes, and the weight of the edge is , we get a weighted undirected graph; Run the minimum spanning tree algorithm in the weighted undirected graph to obtain the minimum spanning tree connecting all graph nodes; count the degree of each graph node in the minimum spanning tree, and take the production line with the largest degree as the key production line; According to the production process, all production lines are divided into n5 levels; the key production lines are placed in the middle level of the n5 levels, and a capacity limitation model is established for each level; The expression of the capacity constraint model is: ;in, For the Level production line From Level production line Get the input distribution ratio, For the Level production line production capacity; For the Level production line production capacity; The production capacity matrix is ​​constructed in the following ways: Based on the historical production data and production line constraint relationship model, the unit time capacity of each production line is calculated using regression analysis method; an initial matrix is ​​constructed to merge the unit time capacity of all production lines into the production capacity matrix. The rows of the matrix correspond to production lines, the columns correspond to different product types, and the element value represents the unit time capacity of the corresponding product of the production line, that is, the production capacity matrix of the factory is obtained; The method for obtaining the optimal production plan includes: Generate an initial cat group of size N5, each cat in the initial cat group corresponds to a production plan; define the production matrix Represents the production plan, production matrix The rows represent production lines, the columns represent pending orders, and each element value of the production matrix represents the quantity assigned to the corresponding production line to produce the order; Define the objective optimization function for the cat population ; ;in, According to the production matrix The calculated planned delivery time of pending order I, To assign pending order I to The number of production lines, For the The unit time capacity of the production line, and is the target weight coefficient; Calculate the value of each cat for the target optimization function and take the negative number as the survival parameter of the corresponding cat; Sort each cat in ascending order according to the survival parameter, select the first M6 cats as member cats, and the rest as prey; Definition The number of member cats in the current iteration The location is , whose speed is , then the member cat updates its position according to its own speed and the positions of other member cats; Define a tracking random number rd in the interval [0, 1] for each cat, and preset the tracking probability p_track; if rd is less than p_track, The position of only the member cat is updated; the updating formula is: ;in, For the Only member cats in iteration number location, is a global random number, For the number of iterations The best historical position at that time; For the cats that are prey, random disturbances are performed with a preset disturbance probability; the cats are re-sorted in ascending order according to the size of their survival parameters, and the first N7 cats are selected to form a new generation of cat population; until the target optimization function converges, the production plan corresponding to the cat with the smallest current survival parameter is output, which is the optimal production plan.

2. A factory production planning design system based on order analysis according to claim 1, characterized in that: The historical production data includes product data and production data corresponding to orders within a fixed time period in the past, the product data includes the production volume, production cost and inventory cost of various products; the production data includes the production efficiency of each production line and the inventory of raw materials; Relevant data include product quantity requirements, order delivery deadlines, types of raw materials required, quantities of raw materials required and product inventory levels.

3. A factory production planning design system based on order analysis according to claim 2, characterized in that: The method of preprocessing the historical production data includes: Identify outliers from historical production data; analyze the data points corresponding to the identified outliers to obtain analysis results, which include the distribution of outliers, numerical characteristics of outliers and mean differences; According to the analysis results, the nature characteristics of the abnormal values ​​are obtained, which are extreme values, outliers, short-term mutations, long-term deviations or abnormal fluctuations; Delete or smooth outliers whose characteristics are extreme values ​​or outliers, interpolate outliers whose characteristics are short-term mutations, perform segmentation processing or trend correction on outliers whose characteristics are long-term deviations; denoise outliers whose characteristics are abnormal fluctuations; and obtain historical and complete production data.

4. A factory production planning design system based on order analysis according to claim 3, characterized in that: The method of identifying abnormal values ​​from historical production data includes: Clean and normalize historical production data; Define optimization parameters including population size, maximum explosion amplitude , explosion seed constant , mutation probability and the maximum number of iterations ; Define the initial N individuals to form the initial population, including normal individuals and abnormal individuals; Different types of data points in the historical production data are respectively formed into corresponding data sets; for any data point in the data set , defining its average density ; ;in, is the mean vector of the data set, is the size of the dataset, Indicates Centered on The set of data points within the range of the circle formed by the radius; for The data points within is the radius, For data points timestamp, For data points timestamp, is the parameter that controls the attenuation width; for The transpose of for The transpose of Sort the average density of all data points from small to large, take the first m data points as outlier individuals, and the rest of the data points as normal value individuals; Define the fitness function ; For each individual, calculate the value of its fitness function, record it as the fitness value, arrange the individuals from large to small according to the fitness value, and select the first M3 individuals as elite individuals; According to the explosion seed constant , randomly generate an explosion seed , calculate the explosion radius of elite individuals ; The explosion radius is centered on the elite individual. Create a hypersphere region for the radius, and randomly sample N2 new candidate individuals in the hypersphere region according to Gaussian distribution; For each of these N2 candidate individuals, calculate the Euclidean distance between it and the elite individual ;like Greater than , then discard the candidate individual. Less than or equal to , then according to the Euclidean distance Perform mutation operation on it to generate a new ; The formula for the mutation operation is: ;in, is a random value that follows a standard normal distribution, and is the variation intensity parameter, For elite individuals The fitness value of Add new individuals that have not been discarded to the population, and repeat the process of sampling candidate individuals and performing mutation operations on all elite individuals to obtain a new population; From the new population, according to the preset mutation probability Randomly select N4 individuals, and randomly select one or m1 gene position from these N4 individuals, and Perform mutation operation on each gene position; The formula for mutation operation is: ,in, is the new gene value of the gene position after mutation, is the original genetic value of the gene position, is the upper boundary value of the gene position, is the lower boundary value of the gene position; ;in, is a preset random number in the interval [0, 1], is the current iteration number, is the mutation control parameter, is the fitness control parameter, The individual to whom the mutated gene belongs The fitness value of Repeat until all selected individuals have completed the mutation operation; obtain newly generated individuals; merge the newly generated individuals into the population, and select elite individuals according to the fitness value to form a new generation of population. If the termination condition is met, the outlier detection scheme corresponding to the optimal individual is output; otherwise, it will continue to iterate until the termination condition is met, and the outlier detection scheme is obtained. The outlier detection scheme is applied to the historical production data to identify outliers.

5. A factory production planning design system based on order analysis according to claim 4, characterized in that: The fitness function The formula is: ;in, is the number of correctly detected outliers, is the number of normal values ​​that are incorrectly detected as outliers, is the number of outliers that could not be detected; and is a positive constant parameter; For penalty items; , , , and To adjust the positive constant weights of each item; Penalty ;in, is the constraint function, is the penalty item weight; To obtain the maximum value; is the decision variable; ;in, is the original constraint function, and Positive adjustment parameters, is the index parameter; for The maximum value in the feasible region is for The minimum value in the feasible region, for The mean value in the feasible region is is a constant; Original constraint function ;in, Indicates order Assign to production line The number of products on For production line capacity constraints, is the permitted excess capacity ratio; is the capacity time function, is the time index; ;in, For production line The capacity fluctuation coefficient, For production line The production capacity fluctuation cycle, For production line The initial phase of capacity fluctuation, For production line In time The random disturbance term of .

6. A factory production planning design system based on order analysis according to claim 5, characterized in that: The calculation method of the time urgency includes: Based on the relevant data, calculate the time required for completion and the time required for raw materials to be stocked; if the product quantity required Less than or equal to the product inventory , then define the remaining production quantity Is 0, if the product quantity required Greater than product inventory ; The production quantity still needs to be ; Calculate pending orders Time required for completion ;in, For pending orders Medium Products The remaining production quantity is For pending orders Medium Products Unit product production time, For products Unit production time at the bottleneck process; Build a supply chain network, use suppliers, factories and raw materials as nodes of the supply chain network, use supply paths as directed edges to connect nodes, and mark the transportation time and transportation cost on each directed edge; Obtain the transportation time and cost of the supply route based on the type of raw materials required, the quantity of raw materials required, and the raw materials inventory; Build an optimization model and use linear programming methods to solve the optimization model and obtain the optimal solution ; For each raw material required , its preparation time ;in, For suppliers Purchase the required raw materials The total preparation time is ; Based on pending orders The completion time and total preparation time are used to obtain the capacity-limited time ; Order delivery deadlines and production capacity constraints Perform comprehensive modeling of Gaussian mixture model and fit the cumulative distribution function of comprehensive delivery time distribution , set the threshold quantile , using numerical optimization algorithms to solve , get the time urgency threshold ; Set the current time As a benchmark, for each pending order , calculate the remaining time for delivery ; ;in, For pending orders If Less than or equal to , then pending orders Time urgency ; ;in, is the maximum value of time urgency, is the offset of time urgency; is the curve steepness parameter; if Greater than , then pending orders Time urgency .

7. A factory production planning design system based on order analysis according to claim 6, characterized in that: The optimization model is constructed in the following manner: The basic framework for defining an optimization model includes the objective function and constraints; ;in, For suppliers Raw materials required by the purchasing unit The cost, For suppliers Purchase the required raw materials The number of For suppliers Purchase the required raw materials delivery time; is the upper limit of the expected total lead time, is the time limit penalty coefficient, For suppliers Purchase the required raw materials Environmental impact assessment scores; is the environmental impact penalty coefficient; Constraints include safety stock condition function , Supply Capacity Constraint Function and diversity constraints ; Safety stock condition function ;in, The raw materials needed The safety stock factor, The raw materials needed The number of The raw materials needed Current inventory levels; Supply Capacity Constraint Function ;in, Is a supplier The supply capacity limit; For Suppliers The reputation score, is the preset maximum weighted purchase quantity; Diversity Constraints ;in, It is the largest single supplier.

Images