A heuristic rules approach for orchestrating a cruise ship production design task plan

Abstract

The present application provides a heuristic rule method for scheduling a cruise ship production design task plan, comprising the following steps: 1. quantifying the parallel dependency relationship between tasks based on task logic and task maturity; 2. designing a task man-hour-person relationship function integrating learning curves; 3. constructing a plan scheduling model; and 4. proposing a design plan scheduling rule for scheduling the plan. The present application belongs to the field of ship design technical management. By establishing a task execution logic model between different processes within and between various professions in cruise ship production design, defining the concept of task maturity, and using the execution logic and maturity between tasks to draw a task logic directed acyclic graph (DAG), the logical and quantitative parallel dependency relationship between tasks is clearly expressed. A task man-hour-person relationship function integrating learning curves is constructed, and a heuristic rule method is designed to solve the cruise ship production design task execution plan.

Description

TECHNICAL FIELD

[0001] The present application relates to a heuristic rule method for scheduling a cruise ship production design task plan, belonging to the technical management field of ship design. BACKGROUND

[0002] Cruise ship production design is an important stage of undertaking detailed design and starting on-site construction. From a professional point of view, the cruise ship production design business stage can be divided into structure, outfitting, piping, air pipe, electrical, and hotel engineering. From the perspective of tasks, each professional task is different, and within the same profession, tasks can be divided into pre-preparation, modeling, and drawing tasks. From the production design process, the design process in the production design process is complex, and the shipyard divides the professional tasks into main process-sub-process-design rhythm dimensions, and there is a 1-N task quantity relationship between different dimensions. For example, in the pre-preparation of air pipe tasks, it can be further divided into component library definition, modeling rule definition, and other processes. There are complex execution logics between different professions and within the same profession, for example, component library and material library construction must be completed before starting the modeling task of each profession, and the model must be available before starting the drawing task.

[0003] In general, the process and management within the same profession of cruise ship production design are clear, the management between professions is relatively independent, and the potential connection between professions is complex, so the cross-professional process is complex; the process is complex and variable, the design process of each profession is not unified, and there is overlap and overlap between different design processes; the degree of parallelism of tasks is high, and there is a high degree of dependency between tasks.

[0004] Therefore, the task execution logic of cruise ship production design varies due to different professional business processes, and the parallel and dependent relationship between tasks is difficult to quantify, and the rough division of plan nodes makes the constraints loose. In order to reduce the blank and delay between plans, meet the requirements of plan nodes, and realize the orderly development of tasks, it is necessary to develop a task plan scheduling method considering the parallelism of multi-professional design.

[0005] Task plan scheduling includes task sequencing and plan compilation, which represents the task execution sequence through sequencing and the specific time arrangement of task execution through plan compilation. The commonly used methods in task sequencing problems include priority constraints, graph networks, and design structure matrices, but mainly focus on 0, 1 sequencing, and less involve different parallelism between tasks, and the quantification relationship between complex task parallelism and dependence relationship and plans is not clear.

[0006] The commonly used methods in planning problem include linear programming, nonlinear programming, dynamic programming, heuristic algorithm, meta-heuristic algorithm, etc. Generally, the production of enterprises is taken as the background. Because most enterprises adopt the business characteristics of assembly line production, most planning methods do not consider the influence of data dependence between tasks on planning, so the application effect of the cruise ship production design task planning problem with multiple specialties, multiple processes and complex task parallel dependence relationship is poor. At the same time, the cruise ship design task process needs to match the cruise ship construction and loading process, so the task planning of cruise ship production design for improving the execution speed and minimizing the execution cost is not the best choice. On the contrary, when arranging the plan, the design process needs to be controlled, the relationship between the plan and the personnel needs to be balanced, and the task execution continuity, the personnel change balance and the single person working time stability need to be ensured.

[0007] The heuristic rule algorithm is a kind of heuristic algorithm, which guides the problem solving or decision-making process through a series of rules. The result obtained by this kind of algorithm may not be the optimal solution, but usually a better approximate solution can be found within a reasonable time. Compared with meta-heuristic algorithm, the meta-heuristic algorithm is the product of the combination of random algorithm and local search algorithm, which depends on the selection of population size, parameters and fitness function setting, and needs to be adjusted for many times to avoid the algorithm falling into local optimum, so the efficiency of this kind of algorithm is sometimes low.

[0008] In the cruise ship production design task planning problem, when the meta-heuristic algorithm is used to study the problem, it is difficult to accurately establish a planning model with complex task execution logic and parallel dependence relationship, and it is difficult to solve the optimal value, so the heuristic rule method is more suitable for the cruise ship production design task planning problem. SUMMARY

[0009] The present application is to solve the influence of the execution logic and parallel degree between the quantized parallel tasks on the plan, and at the same time, to pay attention to the planning arrangement requirements of task execution continuity, personnel change balance and single person working time stability, and then to propose a heuristic rule method for arranging cruise ship production design task plan.

[0010] The technical solution adopted by the present application to solve the above problems is that the steps of the present application include:

[0011] Step 1, quantizing the parallel dependence relationship between tasks based on task logic and task maturity;

[0012] Step 2, designing a task working hours-personnel relationship function fused with learning curve;

[0013] Step 3, constructing a planning arrangement model;

[0014] Step 4, proposing a design planning arrangement rule for arranging the plan.

[0015] Further, the parallel dependency between tasks in step 1 is quantified based on the task logic and the maturity between tasks as follows:

[0016] task Xi denotes a design task i in the professional X, task Xi the start time of the completion time of the man-hour of the mth execution of then

[0017] Because the tasks have execution logic and time sequence requirements of business processes, the tasks can be regarded as vertices, and the execution logic between tasks as directed edges, thereby forming a task logic directed acyclic graph (DAG).

[0018] Task maturity denotes the maturity relationship between two adjacent tasks task Xi and task Xj in the task logic DAG connected by a directed edge, that is, the pre-task task Xi must be completed to a certain extent before the post-task tasj Xj can be started. That is, the earliest start time of the post-task tasj Xj is

[0019] Two kinds of task execution logic are defined, denotes that tasj Xj starts after a period of time from the start of tasj Xj , denotes that tasj Xj starts after the completion of tasj Xj ,

[0020] Therefore, through the task execution logic and the maturity relationship between tasks, a task logic DAG graph can be constructed to quantify the parallel dependency between tasks.

[0021] Further, the task man-hour-person relationship function of the designed fusion learning curve in step 2 is:

[0022]

[0023] Learning curve is a kind of mathematical model describing the learning effect of personnel, which was first proposed by Wright and further summarized and named by Yelle as learning curve. At present, learning curve can be divided into logarithmic learning curve, exponential learning curve and hyperbolic learning curve. Considering the application of machine resources such as design modeling software, calculation software and drawing software in the design process, as well as some prior factors such as schemes and rules before starting design, the S learning curve model proposed by Nembhard is selected.

[0024] where, N Xi represents the number of people completing task Xi . NT X represents the number of designers in the professional X, 1≤N Xi ≤NT X , N Xi ∈Z, NT=∑ X NT X . represents the working hours of the mth execution of the design task task Xi , tm represents the average working hours of the first completion of the design task task Xi by a single designer. tr is the task working hour variation rate adjustment parameter, 0<tr<1. M represents the incompressible factor in the design process, that is, the time cannot be reduced to repeat the design task, 0<M<1. 1-M represents the compressible factor in the design process. B is the prior factor existing in the design process, which represents the previous experience. If sufficient preparation is made before the first business, detailed documents can be referred to, and the business completion time will be greatly reduced. Generally, the experience level is divided into 10 parts, 0<B<10. b represents the learning index, that is, the learning ability in the design process, 0<b<1.

[0025] Further, the plan arrangement model is constructed in step 3, which includes:

[0026] Step 301, defining plan node constraints;

[0027] Step 302, constructing a plan arrangement model.

[0028] Further, the plan node constraints are defined in step 301, which are:

[0029] tepn1, tepn2…tepnm represent the plan nodes that need to be met in the task process.

[0030] λ represents a step, that is, the length of a directed edge in the task logic DAG. represents the connection relationship between any two tasks in the task logic DAG, denotes the sequential connection relationship in the task logic DAG with a step length of 1, i.e., from task Xj to task Xj . denotes the reverse sequential connection relationship in the task logic DAG with a step length of 1, i.e., from task Xj to task Xi . Set

[0031] Plan node constraint 1:

[0032] Plan node constraint 2:

[0033] Plan node constraint 3:

[0034]

[0035] wherein, denotes the task earliest start time constraint, i.e., the start time of the task needs to satisfy the inequality denotes the task latest completion time constraint, i.e., the completion time of the task needs to satisfy the inequality

[0036] Two tasks task Xi and task Xj connected by one directed edge on the task logic DAG have direct execution logic, so the completion times of task Xi and task Xj need to satisfy the inequality Therefore, on the task logic DAG, the completion times of two adjacent tasks with a step length of 1 and a sequential connection relationship need to satisfy

[0037] Further, the plan arrangement model constructed in step 302 is:

[0038]

[0039] The goal is as follows:

[0040]

[0041] wherein, Δt X denotes the planned idle time of all tasks in the specialty X, denotes the person number change standard deviation of all tasks in the specialty X, denotes the single-person work duration change standard deviation of all tasks in the specialty X.

[0042] The plan arrangement model constraints are as follows:

[0043]

[0044] 1≤Z Xi ≤NT X ,N Xi ∈Z

[0045]

[0046] Further, the design plan arrangement rule for arranging the plan in step 4 includes:

[0047] Step 401, task plan arrangement sequence rule;

[0048] Step 402, plan arrangement unit rule;

[0049] Step 403, task number and working hour adjustment rule.

[0050] Further, the task plan arrangement sequence rule in step 401 is as follows:

[0051] Indegree Ind and outdegree Outd are important concepts in DAG, the indegree reflects the degree of influence of the vertex by its predecessor vertex, and the outdegree reflects the degree of influence of the vertex on the successor vertex. If the size of the indegree and outdegree is larger, the key degree of the vertex in the DAG is larger. Therefore, the concept of indegree and outdegree is used to design the task plan arrangement sequence rule, including: initial vertex task initial selection rule, anchor point AnchorP X selection rule, anchor point traversal sequence rule, and other vertex traversal sequence rule.

[0052] Rule 1 (initial vertex selection rule):

[0053]

[0054] Rule 2 (anchor point AnchorP X selection rule):

[0055] AnchorP X ={task Xi |Outd(task Xi )>1or Ind(task Xi )>1for the first time

[0056] Rule 3 (anchor point traversal sequence rule): let denote the traversal sequence of the task, denote descending order.

[0057]

[0058] Rule 4 (Other vertex traversal order rules): ... Indicates ascending order.

[0059]

[0060] When scheduling tasks, following the above rules allows us to traverse all vertices of the DAG in a fixed and ordered manner based on the criticality of each vertex and its influence on other vertices, thus providing a sequential basis for scheduling.

[0061] Given this traversal order, there are two scenarios when scheduling tasks.

[0062] Case 1: When calculating the preceding vertex plan using the following vertices, apply the formula to solve for the start time of the preceding vertices in reverse order.

[0063] Case 2: When calculating the vertex plan with in-degree Ind≥2, it is necessary to compare the vertex start times calculated by different preceding tasks and take the optimal one.

[0064] Furthermore, step 402 proposes the following rules for the planning and scheduling units:

[0065] Rule 5 (Task Planning and Scheduling Unit Rules):

[0066] In the task logic DAG, based on the task plan arrangement rules, select the step distance that is smallest and satisfies... Start time constraint or The two vertices that fulfill the time constraint are taken as the first and last endpoints of a Unit. The two endpoints and their intermediate vertex on the task logic DAG are considered as a complete Unit.

[0067] Under the combined effect of rules 1 to 5, the task logic DAG will be divided into groups that exist in a certain order. Unit combination,

[0068] A combination of two Units is obtained by arranging Units according to the task plan and rules.

[0069] Assume that Unit2 contains tasks. Xi ,task Xj ,task Xn ,task Xl Through Unit 1 tasks Xo The plan can calculate the tasks within Unit 2. Xi start time And through The right endpoint task can be calculated step by step. XlThe start time and finish time.

[0070]

[0071] The planned time for the task is the number of people N who will complete the task. Xi The function f(N) Xi ).

[0072] Based on endpoint task Xl The constructor needs to satisfy different constraints. It contains tasks Xi ,task Xj ,task Xn ,task Xl If the right endpoint task Xl The start time constraint needs to be satisfied, let If the right endpoint task Xl The completion time constraint needs to be met, let

[0073] Furthermore, step 403 proposes the following rules for adjusting the number of personnel and working hours:

[0074] The left and right endpoints of a unit must either satisfy a start time constraint or a finish time constraint. In a task logic DAG constructed from task logic, the planned times of tasks affect each other; therefore, there are two scenarios when scheduling tasks.

[0075] Scenario 1: When When the left endpoint does not meet the planned node, adjust the previous one. The task plan, thereby achieving adjustment The purpose of the left endpoint plan.

[0076] Scenario 2: When When the right endpoint does not meet the planned node, adjust... The task plan, thereby achieving adjustment The purpose of the right endpoint plan.

[0077] exist First, calculate the total number of task participants required to satisfy the node constraints for the endpoint tasks; second, calculate the planning function for the endpoint tasks. Calculate the number of people N for each task. Xi N Xj N Xn N Xl The partial derivatives are recorded, and the absolute values ​​Φ of the partial derivative coefficients of each part are recorded. Xi ,Φ Xj ,Φ Xn,Φ Xl Finally, Φ Xi ,Φ Xj ,Φ Xn ,Φ Xl Allocate the total number of personnel for the task proportionally and recalculate. The task plan is as follows. Additionally, for two tasks with a step size of 1, if the completion time of the preceding task is longer than that of the following task, the personnel allocation is adjusted based on the previously adjusted number of personnel to complete the task, until the task plan meets the constraints.

[0078] The specific steps are as follows:

[0079] (1) Through Task time calculation left endpoint task Xi The start time.

[0080] ·like Then continue with step (2).

[0081] ·like make return make Adjustment Task plan.

[0082] (2) Calculation Right endpoint task Xl The plan.

[0083] (3) Let the number of people in the task be N. Xi N Xj ,…=1, calculate

[0084] ·like Then task Xl The node constraints are satisfied. Number of participants N Xi N Xj ,…=1, execute step (8).

[0085] ·like Perform step (4) to adjust the task plan.

[0086] (4) Unify the function variable N Xi N Xj ... is N X N X ∈Z, calculate and set when The minimum total number of people required for the task is N. X The value of .

[0087] (5) Determine N X Does the number of participants meet the limit?

[0088] • If Continue to step (6).

[0089] • If Let Return Let Adjust the task plan.

[0090] (6) Take as a multi-variable function of N Xi ,N Xj …, 1≤N Xi ,N Xj …≤NT X . Take the partial derivative of with respect to N Xi ,N Xj …, respectively, and record the absolute values of the coefficients of each partial derivative Φ Xi ,Φ Xj …

[0091] Take the partial derivative of with respect to N Xi as an example, and the other partial derivatives are calculated in the same way. To facilitate calculation, calculate the absolute value of the partial derivative where Π = M + (1 - M)(m + B) -b is a constant value.

[0092]

[0093] (c) Φ X = Φ Xi + Φ Xj + …

[0094] (7) Distribute the total number of task people N Xi according to the proportions of Φ Xj ,Φ X …. Set the upward rounding symbol to represent the smallest integer greater than or equal to itself. At the same time, according to rule 3, the number of task people needs to satisfy 1≤N Xi ,N Xj …≤NT X .

[0095]

[0096] (8) Within , apply the number of task people obtained by the above steps to calculate the start time and completion time of the task in the unit.

[0097] (9) Compare the completion times of adjacent preceding tasks and subsequent tasks in turn.

[0098] comparing the task Xi , the completion time of task Xj , if , the number of task persons remains unchanged. If , N Xi ++1, 1≤N Xi ≤NT X ; N Xj --1, 1≤N Xj ≤NT X , until the final update of the number of task persons N Xi and N Xj and recalculate the task plan.

[0099] · Step by step comparison and adjustment of the number of task persons and completion time of adjacent preceding and subsequent tasks until the task plan is completed.

[0100] (10) the task calculation is completed. At the same time, the start time of the left end task of can be calculated, and the task plan of is continued to be compiled.

[0101] The beneficial effects of the present application are: the present application solves the influence of the execution logic and parallel degree between quantized parallel tasks on the plan, and at the same time pays attention to the plan arrangement requirements of task execution continuity, number change balance, and single person work time stability. By establishing the task execution logic model between different processes within and between various professions of cruise ship production design, defining the task maturity concept, and using the execution logic and maturity between tasks to draw a task logic directed acyclic graph (DAG), the logical relationship and quantized task parallel dependence between tasks are clearly expressed. The task man-hour-person relationship function integrating the learning curve is constructed, and the heuristic rule method is designed to solve the cruise ship production design task execution plan. The problem that the meta-heuristic algorithm cannot accurately establish a plan model with complex task execution logic and parallel dependence relationship and it is difficult to solve the optimal value is avoided. BRIEF DESCRIPTION OF DRAWINGS

[0102] Figure 1 is a flowchart of the present application;

[0103] Figure 2 is an example of task logic DAG of the present application;

[0104] Figure 3 is an example containing two Unit combinations;

[0105] Figure 4 is an example of the first adjustment method when the task plan is arranged in the present application;

[0106] Figure 5 is the second adjustment method example when the task plan is arranged by the present application;

[0107] Figure 6 is the cruise production design 6 professional task plan scheme arranged by the present application. DETAILED DESCRIPTION

[0108] In combination Figure 1 The steps of the heuristic rule method for arranging the cruise production design task plan in the embodiment include:

[0109] Step 1, quantifying the parallel dependency relationship between tasks based on task logic and task maturity;

[0110] task Xi represents the design task i in the professional X, task Xi the start time of the completion time is the working hours of the mth execution is then

[0111] Because the tasks have execution logic and time sequence requirements of business processes, the tasks can be regarded as vertices, and the execution logic between the tasks can be regarded as directed edges, so as to form a task logic directed acyclic graph (DAG).

[0112] Task maturity represents the maturity relationship between two adjacent tasks task Xi and task Xj in the task logic DAG connected by a directed edge, that is, the antecedent task task Xi must be completed to a certain extent before the subsequent task task Xj can be started. That is, the earliest start time of the subsequent task task Xj is

[0113] Two kinds of task execution logic are defined, task Xj is started after task Xi starts for a period of time, task Xj is started after task Xi is completed,

[0114] Therefore, a task logic DAG graph can be constructed by the task execution logic and the maturity relationship between tasks, and the originally discrete tasks are connected into a DAG network through the DAG directed edges, and any two task vertices in the DAG network can be connected through the DAG directed edges. The construction basis of the task logic DAG is the execution logic between tasks, and only the tasks with data interaction can be connected through the task logic, and the task maturity is the time quantization result of the task logic, and also the time quantization result of the priority constraint of two adjacent tasks connected through the directed edge on the task logic DAG. The task logic DAG obtained by quantizing the parallel dependency relationship between tasks based on the task logic and the maturity between tasks in step 1 is shown in FIG. 2. Figure 2

[0115] Step 2, design a task man-hour-person relationship function fused with a learning curve;

[0116] The learning curve is a mathematical model describing the learning effect of personnel, which was first proposed by Wright and further summarized and named as learning curve by Yelle. At present, the learning curve can be divided into logarithmic learning curve, exponential learning curve and hyperbolic learning curve. Considering the application of machine resources such as design modeling software, calculation software and drawing software in the design process, and some prior factors such as schemes and rules before starting design, the S learning curve model proposed by Nembhard is selected.

[0117] t m =C1[M+(1-M)(m+B) -b ]

[0118] C1 represents the time for producing the first product, M represents the non-compressible factor in the design process, that is, the time for repeatedly performing the design task cannot be reduced, and 0 < M < 1. 1-M represents the compressible factor in the design process. B is the prior factor existing in the design process, which represents the previous experience. If sufficient preparation is made before the first business, and detailed documents can be referred to, the business completion time will be greatly reduced. Generally, the experience degree is divided into 10 parts, and 0 < B < 10. b represents the learning index, that is, the learning ability in the design process, and 0 < b < 1.

[0119] Task man-hours are closely related to the number of people. With the increase of the number of people, the task man-hours gradually decrease. According to the change relationship characteristics between task man-hours and the number of people, the task man-hour formula based on the learning effect of personnel is rewritten to obtain formula (1).

[0120]

[0121] wherein, N Xi represents the number of people completing the task Xi . NT​X N represents the number of designers in the profession X, 1≤N Xi ≤NT X ,N Xi ∈Z, NT=∑ X NT X . tepm represents the man-hour of the mth execution of the design task task Xi . tr represents the average man-hour of the first execution of the design task task Xi by a single designer. tr is a task man-hour variation rate adjustment parameter, 0<tr<1.

[0122] Step 3, constructing a scheduling model; comprising:

[0123] Step 301, defining scheduling node constraints;

[0124] Step 302, constructing a scheduling model.

[0125] Specifically:

[0126] The scheduling node constraints defined in step 301 are:

[0127] tepn1, tepn2…tepnm represent the scheduling nodes that need to be met in the task process. λ represents a step, i.e. the length of a directed edge in the task logic DAG.

[0128] represents the connection relationship between any two tasks in the task logic DAG, represents the sequential connection relationship with a step length of 1 step in the task logic DAG, i.e. from task Xi to task Xj . represents the reverse sequential connection relationship with a step length of 1 step in the task logic DAG, i.e. from task Xj to task Xi . Set

[0129] Scheduling node constraint 1:

[0130] Scheduling node constraint 2:

[0131] Scheduling node constraint 3:

[0132]

[0133] wherein, ​denotes the latest start time constraint of a task, i.e., the start time of the task needs to satisfy the inequality denotes the latest finish time constraint of a task, i.e., the finish time of the task needs to satisfy the inequality

[0134] Two tasks task Xi and task Xj are directly connected by one directed edge on the task logic DAG, thus the finish time of task Xi and task Xj needs to satisfy the inequality Thus, for two adjacent tasks on the task logic DAG with a step of 1 and a sequential connection relationship, their finish times need to satisfy

[0135] The planning scheduling model constructed in step 302 is as follows:

[0136]

[0137] The objectives are as follows:

[0138]

[0139] where Δt X denotes the planned idle time of all tasks in the specialty X, denotes the standard deviation of the number of people of all tasks in the specialty X, denotes the standard deviation of the single-person working time of all tasks in the specialty X.

[0140] The constraints of the planning scheduling model are as follows:

[0141] task Xi ,task Xj ∈D(task Xi ,task Xj )

[0142] 1≤N Xi ≤NT X ,N Xi ∈Z

[0143]

[0144] Step 4, design planning scheduling rules for scheduling plans. Including:

[0145] Step 401, task planning scheduling order rules;

[0146] Step 402, planning scheduling unit rules;

[0147] Step 403, task number and working hour adjustment rule.

[0148] Specifically, the task plan arrangement order rule in step 401 is:

[0149] In-degree and out-degree Outd are important concepts in DAG, in-degree reflects the degree of influence of a vertex on its predecessor vertex, and out-degree reflects the degree of influence of the vertex on its successor vertex. If the in-degree and out-degree are larger, the key degree of the vertex in the DAG is larger. Therefore, the in-degree and out-degree concepts are used to design the task plan arrangement order rule, including: the initial vertex task inital selection rule, anchor point AnchorP X selection rule, anchor point traversal order rule, and other vertex traversal order rule.

[0150] Rule 1 (initial vertex selection rule):

[0151]

[0152] Rule 2 (anchor point AnchorP X selection rule):

[0153] AnchorP X ={task Xi |Outd(task Xi )>1orInd(task Xi )>1for the first time}

[0154] Rule 3 (anchor point traversal order rule): let denote the traversal order of the task, denote descending order.

[0155]

[0156] Rule 4 (other vertex traversal order rule): let denote ascending order.

[0157]

[0158] When arranging the task plan, the above rules can be used to traverse all vertices on the task logic DAG in a fixed order according to the key degree of each vertex and its influence on other vertices, providing a sequence basis for plan arrangement.

[0159] Under this traversal order, there are two cases when arranging the task plan.

[0160] Case 1: When calculating the predecessor vertex plan through the successor vertex, the formula inverse order is used to solve the start time of the predecessor vertex.

[0161] Case 2: When calculating the vertex plan with in-degree Ind≥2, it is necessary to compare the vertex start times calculated by different preceding tasks and take the optimal one.

[0162] The planning and arrangement unit rules in step 402 are as follows:

[0163] Rule 5 (Task Planning and Scheduling Unit Rules):

[0164] In the task logic DAG, based on the task plan arrangement rules, select the step distance that is the smallest and satisfies... Start time constraint or The two vertices that fulfill the time constraint are taken as the first and last endpoints of a Unit. The two endpoints and their intermediate vertex on the task logic DAG are considered as a complete Unit.

[0165] Under the combined effect of rules 1 to 5, the task logic DAG will be divided into groups that exist in a certain order. Unit combination,

[0166] An example of a combination of two Units obtained from the Unit arrangement rules of the task plan is as follows: Figure 3 As shown.

[0167] Assume that Unit2 contains tasks. Xi ,task Xj ,task Xn ,task Xl ,like Figure 3 As shown. This is achieved through the tasks in Unit 1. Xo The plan can calculate the tasks within Unit 2. Xi start time And through The right endpoint task can be calculated step by step. Xl The start time and finish time.

[0168]

[0169] The planned time for the task is the number of people N who will complete the task. Xi The function f(N) Xi ).

[0170] Based on endpoint task Xl The constructor needs to satisfy different constraints. It contains tasks Xi ,task Xj ,task Xn ,taskXl If the right endpoint task Xl The start time constraint needs to be satisfied, let If the right endpoint task Xl The completion time constraint needs to be met, let

[0171] The rules for adjusting the number of workers and working hours in step 403 are as follows:

[0172] The left and right endpoints of a unit must either satisfy a start time constraint or a finish time constraint. In a task logic DAG constructed from task logic, the planned times of tasks are interdependent. Therefore, there are two scenarios when scheduling tasks: Figure 4 and Figure 5 As shown.

[0173] Scenario 1: When When the left endpoint does not meet the planned node, adjust the previous one. The task plan, thereby achieving adjustment The purpose of the left endpoint plan.

[0174] Scenario 2: When When the right endpoint does not meet the planned node, adjust... The task plan, thereby achieving adjustment The purpose of the right endpoint plan.

[0175] Rule 6 (Rules for Adjusting the Number of Personnel and Working Hours):

[0176] exist First, calculate the total number of task participants required to satisfy the node constraints for the endpoint tasks; second, calculate the planning function for the endpoint tasks. Calculate the number of people N for each task. Xi N Xj N Xn N Xl The partial derivatives are recorded, and the absolute values ​​Φ of the partial derivative coefficients of each part are recorded. Xi ,Φ Xj ,Φ Xn ,Φ Xl Finally, Φ Xi ,Φ Xj ,Φ Xn ,Φ Xl Allocate the total number of personnel for the task proportionally and recalculate. The task plan is as follows. Additionally, for two tasks with a step size of 1, if the completion time of the preceding task is longer than that of the following task, the personnel allocation is adjusted based on the previously adjusted number of personnel to complete the task, until the task plan meets the constraints.

[0177] The specific steps are as follows:

[0178] (1) Through Task time calculation left endpoint task Xi The start time.

[0179] ·like Then continue with step (2).

[0180] ·like make return make Adjustment Task plan.

[0181] (2) Calculation Right endpoint task Xl The plan.

[0182] (3) Let the number of people in the task be N. Xi N Xj ,…=1, calculate

[0183] ·like Then task Xl The node constraints are satisfied. Number of participants N Xi N Xj ,…=1, execute step (8).

[0184] ·like Perform step (4) to adjust the task plan.

[0185] (4) Unify the function variable N Xi N Xj ... is N X N X ∈Z, calculate and set when The minimum total number of people required for the task is N. X The value of .

[0186] (5) Determine N X Does the number of participants meet the limit?

[0187] ·like Continue with step (6).

[0188] ·like make return make Adjustment Task plan.

[0189] (6) Consider N XiN Xj … the multi-variable function, 1≤N Xi ,N Xj …≤NT X . Respectively, take the partial derivative of N Xi ,N Xj …, record the absolute value of the coefficient of each partial derivative Φ Xi ,Φ Xj ….

[0190] Take the partial derivative of N Xi as an example, the other partial derivative calculation is the same. For convenience of calculation, calculate the absolute value of the partial derivative Where, Π=M+(1-M)(m+B) -b is a constant value.

[0191]

[0192] (c) Φ X =Φ Xi +Φ Xj +…

[0193] (7) Distribute the total number of task people N X in proportion to Φ Xi ,Φ Xj …, Set the upward rounding symbol to represent the smallest integer greater than or equal to itself. At the same time, according to rule 3, the number of task people needs to meet 1≤N Xi ,N Xj …≤NT X .

[0194]

[0195] (8) Within , apply the number of task people obtained by the above steps to calculate the start time and completion time of the task in the unit.

[0196] (9) Compare the completion times of adjacent preceding and subsequent tasks in turn.

[0197] · Compare the completion times of task Xi ,task Xj , if , the number of task people and the plan remain unchanged. If , N Xi ++1, 1≤N Xi ≤NT X ; N Xj --1, 1≤N Xj ≤NT X , until Final update task number N Xi With N Xj And recalculate Task plan within.

[0198] · Step-by-step comparison and adjustment of adjacent pre-task and post-task task number and completion time, until Task plan within.

[0199] (10) Task calculation is completed. At the same time, the start time of the left end task of Can be calculated, and the task plan of Continue to be compiled.

[0200] In this embodiment, after the design of the heuristic rule method for cruise ship production design task plan compilation is completed, the cruise ship production design 6 professional task plan is solved by using the heuristic rule method, and the specific steps are as follows:

[0201] Step 1: Analyze the standard task execution logic in the 6 professional tasks of cruise ship production design;

[0202] Step 2: According to the AHP hierarchical analysis method, the expert weight and expert score are obtained, and the maturity between tasks is solved;

[0203] Step 3: Draw the task logic DAG according to the task execution logic and the maturity between tasks;

[0204] Step 4: Set the plan node constraint;

[0205] Step 5: Determine the cruise ship production design task plan compilation order according to the task plan compilation order rule;

[0206] Step 6: Determine the cruise ship production design task plan compilation unit according to the task plan compilation Unit rule;

[0207] Step 7: Compile the cruise ship production design task plan according to the task number and working hour adjustment rule;

[0208] After the iteration of the above steps, the plan of the 6 professional tasks of cruise ship production design and the rated number of people completing each task can be obtained. The 6 professional task set of cruise ship production design is shown in Table 1. Among them, task X ={task Xi |X=A,B,C,D,E,F} respectively represent the hull structure, outfitting, piping, air pipe, electrical and hotel engineering six professional task sets.

[0209] Table 1 Cruise ship production design 6 professional task set

[0210]

[0211]

[0212] Cruise ship production design 6 professional task plan scheme as shown in Figure 6 , each task rated number of people as shown in Table 2.

[0213] Table 2 each task rated number of people

[0214]

[0215] The above, only the preferred embodiments of the present application, not any form of the present application, although the present application has been disclosed as above, however, not to limit the present application, any skilled in the art, without departing from the scope of the present application technical solution, when can use the above disclosed technical content made a little more change or modification of equivalent variations of equivalent embodiments.

Claims

1. A heuristic rules method for orchestrating a cruise ship production design task plan, characterized by the steps Comprise: Step 1, based on the task logic and the maturity of the task between the quantitative relationship between the parallel dependence between tasks; Step 2, design task man-hour-personnel relationship function of learning curve fusion; Step 3, construct plan scheduling model; Step 4, propose design plan scheduling rules for scheduling plan; Specifically includes: Step 401, task plan scheduling sequence rule; Ind and Outd are important concepts in DAG, Ind reflects the degree of influence of a vertex by its predecessor, Outd reflects the degree of influence of the vertex on its successor; the greater the size of Ind and Outd, the greater the key degree of the vertex in the DAG; the concept of Ind and Outd is used to design the task planning sequence rules, including: initial vertex task initial selection rule, anchor point AnchorP X selection rule, anchor point, anchor traversal sequence rule, other vertex traversal sequence rule; Rule 1, initial vertex selection rule: Rule 2, Anchor P X Selection rule: AnchorP X = {task Xi | Outd(task Xi ) > 1 or Ind(task Xi ) > 1 for the first time} Rule 3, Anchor Traversal Order Rule: Let denote the traversal order of tasks, denote a descending order; Rule 4, other vertex traversal order rule: to denotes ascending order; When scheduling the task plan, the key degree of each vertex on the task logic DAG and its influence on other vertices can be traversed in a fixed order according to the above rule, providing a sequential basis for plan scheduling; Under this traversal order, there are two cases when scheduling the task plan: Case 1: When calculating the plan of the predecessor vertex through the successor vertex, apply formula reverse order to solve the start time of the predecessor vertex; Case 2: When calculating the plan of a vertex with in-degree Ind≥2, compare the start time of the vertex calculated through different predecessor tasks and take the optimal one; Step 402, plan scheduling unit rule; Rule 5, task plan scheduling Unit rule: On the task logic DAG, according to the task plan arrangement order rule, select two vertices with the minimum step distance and meet the start time constraint or the end time constraint as the front and back two endpoints of the Unit; take the two endpoints and their intermediate vertices on the task logic DAG as a complete Unit; Under the joint action of rules 1 to 5, the task logic DAG will be divided into a certain order Unit combination, According to the task plan scheduling Unit rule, a combination of two Units is obtained; Let Unit2 contain task Xi ,task Xj ,task Xn ,task Xl , the start time of task Xo in Unit2 is calculated by the plan of task Xi in Unit1 The start time and finish time of the right end point task Xl can be calculated step by step through ; The planned time of a task is a function f(N) of the number of people N completing the task Xi . Xi ) of the number of people N completing the task According to the end task task Xl The constructor The task task Xi ,task Xj ,task Xn ,task Xl If the right end task task Xl If the start time constraint needs to be met, let If the right end task task Xl If the completion time constraint needs to be met, let Step 403, task personnel and man-hour adjustment rule; The left and right endpoints of the Unit need to meet the start time constraint or the completion time constraint; On the task logic DAG constructed by the task logic, the plan time of the tasks affects each other, and there are two cases when scheduling the task plan: Case 1: When left endpoint of the plan node does not satisfy the task plan, adjust the previous task plan so as to achieve the purpose of adjusting the left endpoint plan; Case 2: When the right endpoint of the plan node does not satisfy the plan node, adjust the task plan of the right endpoint so that the right endpoint plan is satisfied. the right endpoint plan is satisfied. Rule 6, task personnel and man-hour adjustment rule: In , first, the total number of task persons required to satisfy the node constraint of the end task is calculated; second, the planning function of the end task is solved , and the partial derivatives of the task persons N Xi , N Xj , N Xn , N Xl are solved respectively, and the absolute values of the partial derivative coefficients of each part are recorded as Φ Xi , Φ Xj , Φ Xn , Φ Xl ; finally, the total number of task persons is distributed in proportion to Φ Xi , Φ Xj , Φ Xn , Φ Xl , and the task plan in is recalculated; meanwhile, for two tasks with a step size of 1, if the completion time of the previous task is greater than the completion time of the subsequent task, the personnel allocation is adjusted on the basis of the completed task person number in the above adjustment, and the adjustment is continued until the task plan satisfies the constraint. The specific steps are as follows: (1) By the task time of the left endpoint task task Xi the start time of If then continue with step (2); If Let Return Let Adjust the task plan; (2) calculating right endpoint task task Xl schedule; (3) Let the number of tasks N Xi ,N Xj … = 1, compute If then task Xl satisfies the node constraint; number of task people N Xi ,N Xj ,… = 1, perform step (8); If Step (4) is performed, adjusting the task plan; (4) Unified function variable N Xi ,N Xj ,… is N X ,N X ∈Z, calculate and set the minimum total number of tasks N when X the value of N (5) judging N X whether the number of people is limited; If Continue with step (6); If Let Return Let Adjust the task plan; (6) the viewed as N Xi ,N Xj …-ary function, 1≤N Xi ,N Xj …≤NT X ; respectively, the partial derivative of N Xi ,N Xj …, record the absolute value of the coefficient of each partial derivative Φ Xi ,Φ Xj …; To the derivative of N Xi , other derivative calculation is the same; For easy calculation, the absolute value of the derivative Where, ∏ = M + (1-M)(m+B) -b is a constant value; (a) (b) (c) Φ X = Φ Xi + Φ Xj +... (7) With Φ Xi ,Φ Xj ...the proportion of the total number of people assigned to the task N X Set the round-up symbol. This represents the smallest integer greater than or equal to itself; also, according to rule 3, the number of participants must satisfy 1 ≤ N. Xi N Xj …≤NT X ; (8) In In the above steps, the number of tasks is applied to calculate the start time and completion time of the unit task. (9) Compare the completion times of adjacent predecessor and successor tasks in turn; Comparison task Xi , completion time of task Xj , if the number of task persons remains unchanged; if N Xi ++1, 1≤N Xi ≤NT X ; N Xj --1, 1≤N Xj ≤NT X , until the final update of the number of task persons N Xi and N Xj and recalculation of the task plan in ; Step by step, compare and adjust the number of tasks and completion time of adjacent pre- and post- tasks until The internal task planning is completed; (10) The inner task computation is completed; at the same time, the start time of the left endpoint task of can be calculated, and the task plan of continues to be compiled.

2. A heuristic rules method for orchestrating cruise ship production design task plans according to claim 1, characterized in that: Step 1 specific method is: task Xi denotes a design task i in the specialty X, task Xi the start time of the completion time of the man-hour at the mth execution of then The task has execution logic and time sequence requirements of business process, the task is regarded as a vertex, the execution logic between tasks is regarded as a directed edge, thereby forming a task logic directed acyclic graph DAG; Task maturity In a Directed Acyclic Graph (DAG), two adjacent tasks connected by a directed edge represent the task logic. Xi with task Xj The maturity relationship between them, i.e., the preceding tasks Xi Only after a certain stage has been completed should subsequent tasks begin. Xj That is, the subsequent task. Xj The earliest start time is define two task execution logics, represent task Xj in task Xi start after a period of time, represent task Xj in task Xi start after completion, Through the task execution logic and the maturity relationship between tasks, a task logic DAG graph is constructed to quantify the parallel dependence relationship between tasks.

3. A heuristic rules method for scheduling cruise ship production design task plans according to claim 1, characterized in that: In step 2, the task man-hour-personnel relationship function of learning curve fusion is designed: wherein N Xi represents the number of completed tasks; NT Xi represents the number of designers in the profession X, 1≤N X ≤NT Xi , N X ∈Z, NT=∑ Xi NT X ; X ; represents the man-hours of the mth execution of the design task task Xi , T represents the average man-hours of the first completion of the design task task Xi by a single designer; tr is a task man-hour variation rate adjustment parameter, 0<tr<1; M represents an incompressible factor in the design process, i.e., the time for repeatedly performing the design task cannot be reduced, 0<M<1; 1-M represents a compressible factor in the design process; B is a prior factor existing in the design process, representing previous experience, which is divided into 10 equal parts, 0<B<10; b represents a learning index, i.e., the learning ability in the design process, 0<b<1.

4. A heuristic rules method for orchestrating cruise ship production design task plans according to claim 1, characterized in that: In step 3, the plan scheduling model includes: Step 301, define plan node constraints; Step 302, construct plan scheduling model.

5. A heuristic rules method for scheduling a cruise ship production design task plan according to claim 4, characterized in that: In step 301, the plan node constraints are defined as: representing a plan node that needs to be satisfied during the execution of the task the task; λ denotes a step, i.e. the length of a directed edge in the task logic DAG; denotes the connection relationship between any two tasks in the task logic DAG, denotes the sequential connection relationship with a step length of 1 step in the task logic DAG, i.e. from task Xi to task Xj ; denotes the reverse sequential connection relationship with a step length of 1 step in the task logic DAG, i.e. from task Xj to task Xi ; set Plan node constraint 1: Plan node constraint 2: Plan node constraint 3: wherein, denotes a task earliest start time constraint, i.e. the start time of a task needs to satisfy the inequality denotes a task latest finish time constraint, i.e. the finish time of a task needs to satisfy the inequality Two tasks task Xi and task Xj are directly connected by one directed edge on the task logic DAG Xi and task Xj 's completion times need to satisfy inequality Two adjacent tasks on the task logic DAG with step size 1 and sequential connection relationship need to satisfy 6. A heuristic rules method for scheduling cruise ship production design task plans according to claim 4, characterized in that: In step 302, the plan scheduling model is constructed as: The goal is as follows: where Δt X represents the planned idle time for all tasks in specialty X, represents the standard deviation of the number of people for all tasks in specialty X, represents the standard deviation of the individual work duration for all tasks in specialty X; The constraints of the plan scheduling model are as follows:

Images