Intelligent repairing method after failure of polymer mark production scheduling
By introducing a virtual universal grade mechanism, the problem of polymer grade scheduling failure was solved, automated repair and flexible scheduling solutions were implemented, manual intervention was reduced, and complex scenarios could be adapted.
Patent Information
- Application Number
- CN202510753847.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-06
- Publication Date
- 2025-10-17
AI Technical Summary
Existing technologies cannot provide effective repair suggestions when polymer grade scheduling fails, causing the scheduling process to stagnate and users to repeatedly verify and adjust input parameters.
A virtual universal brand mechanism is introduced. By adding a virtual universal brand and trying to schedule production until it succeeds, the virtual brand is hidden after the scheduling is successful to locate the cause of failure. After the cause of failure is fixed, the scheduling of the original brand is completed.
It enables automated repair of polymer scheduling failures, reduces manual trial and error costs, supports dynamic extension constraints, and provides multiple repair options to adapt to complex scheduling scenarios.
Smart Images
Figure CN120806410A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of chemical intelligent production, in particular to the field of polymer grade production scheduling, and specifically to an intelligent repair method after polymer grade production scheduling failure. BACKGROUND
[0002] Polymer grade production scheduling is a common optimization problem in chemical production, especially in continuous production processes. Such production processes often involve manufacturing multiple different grades of polymer materials, each with specific physical and chemical properties to meet different application requirements.
[0003] In actual production, switching from one grade to another is called "changeover". Polymer grade production scheduling needs to optimize the switching sequence in continuous production to reduce the generation of transition material (non-target product). However, actual production scheduling needs to meet three types of hard constraints:
[0004] Changeover rules: grade switching needs to follow a directed rule (e.g. grade A→grade B is allowed, grade B→grade A may not be allowed);
[0005] First processing grade constraint: the specified initial grade must be produced first;
[0006] Partial sequence constraint: some grades need to be processed in a specified order (e.g. grade C must be processed after grade B, and other grades can be processed between grade B and grade C).
[0007] When the user inputs conflicting constraint parameters, the traditional method cannot provide repair suggestions, causing the production scheduling process to stall.
[0008] Existing solutions mostly use depth-first search (DFS) or neighborhood search algorithms to determine the existence of feasible solutions. Such methods can only return a "no solution" result and cannot guide the user to adjust the input parameters. At this time, the user needs to base on experience and verify repeatedly to successfully troubleshoot the problem. For example, if a key path is missing in the changeover constraint (e.g. grade A→grade C is not defined), and the sequence constraint requires grade C to be processed after grade B, the system will only indicate that there is no solution, but cannot suggest adding the grade A→grade C path or adjusting the sequence constraint after which the algorithm can successfully schedule production. The user needs to troubleshoot whether "grade A→grade C" is not defined reasonably, or whether "grade C is processed after grade B" is reasonable. SUMMARY
[0009] The present application proposes an intelligent repair method after polymer grade production scheduling failure to address the problems in the background art.
[0010] Technical solution:
[0011] A kind of polymer grade production failure intelligent repair method, grade production failure, one by one increase virtual omnipotent grade and try to produce, until production succeeds;The virtual omnipotent grade can be bidirectional production with all other grades, after production succeeds, the two original grades adjacent to virtual omnipotent grade are the original production failure reason, repair failure reason after completing the production of original grade.
[0012] Preferably, the virtual omnipotent grade can be hidden when production result is output, and the hidden virtual omnipotent grade obtains the production result after repairing failure reason.
[0013] Preferably, the repair failure reason is to adjust hard constraint condition to soft constraint condition.
[0014] Specifically, the method specifically includes the following steps:
[0015] 1) grade production failure;
[0016] 2) construct a virtual omnipotent grade, the virtual omnipotent grade can be bidirectional production with all other grades;
[0017] 3) add the new virtual omnipotent grade to the original production plan;
[0018] 4) find out all possible production plans of current grade production;
[0019] 5) if the production plan is empty, i.e. still production failure, step 2) is converted;If the production plan is not empty, i.e. production succeeds, step 6) is converted;
[0020] 6) for each production plan, keep two grades before and after each virtual omnipotent grade as "possible unreasonable condition";
[0021] 7) for each production plan, hide all virtual omnipotent grades, keep the remaining production plan, and output to the user, if the above "possible unreasonable condition" is repaired, the production plan can be obtained;
[0022] 8) end.
[0023] Preferably, in the step 4), all Hamilton paths are found by using backtracking algorithm, and all possible production plans of current grade production are found.
[0024] Advantages of the present application
[0025] The present application can realize automatic positioning of production failure reason and give repair direction by defining virtual omnipotent grade and using inherent production algorithm, and has the following advantages:
[0026] Automatic repair suggestion: through virtual omnipotent grade positioning conflict production grade, reduce the cost of artificial trial and error;
[0027] Flexibility: Supports dynamic expansion of constraints to adapt to complex scheduling scenarios;
[0028] Customer-oriented: Multiple repair options are provided, and customers can choose the best adjustment plan based on their actual needs. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 This is an example diagram of the virtual universal brand of the present invention.
[0030] Figure 2 This is an example diagram of brand production scheduling requirements in a specific implementation method.
[0031] Figure 3 The first virtual universal brand example diagram is added to the specific implementation method.
[0032] Figure 4 This is an example diagram of the virtual universal brand production schedule 1 in a specific implementation method.
[0033] Figure 5 This is an example diagram of the virtual universal brand production schedule 2 in a specific implementation method. DETAILED DESCRIPTION
[0034] The present invention will be further described below with reference to the embodiments, but the scope of protection of the present invention is not limited thereto:
[0035] (1) Information entry into the brand production scheduling system
[0036] Time information: processing start time, expected completion time;
[0037] Basic information of brand: brand name, planned production volume, user demand time, etc.;
[0038] Production transfer relationship: the brand that can be transferred, the duration of the transfer, the transition material, etc.
[0039] First processing grade constraint: specifies that a certain grade must be processed first;
[0040] Partial sequence constraints: There is a sequence relationship when processing some grades.
[0041] (2) Neighborhood search algorithm for brand scheduling (existing solution)
[0042] 1) Problem Modeling
[0043] Objective function: Minimum total transition material;
[0044] Soft constraints (can be violated): The actual completion time should be completed before the expected completion time as much as possible, and the actual completion time of each brand should be completed before the user's required time as much as possible (if a brand is not completed in advance, it is called "brand delay");
[0045] Hard constraints (not violable): first processing grade, processing grades in sequence, transition relationship exists.
[0046] 2) Initial solution generation
[0047] Randomly generate an initial production scheduling plan.
[0048] 3) Neighborhood definition
[0049] Define the "neighborhood" as a small adjustment of the current production scheduling plan. Define the neighborhood by the transition of the transition material, the less the transition of the transition material, the smaller the "neighborhood" of the two grades, which will lead to a better objective function; the more the transition of the transition material, the larger the "neighborhood" of the two grades, which will lead to a worse objective function.
[0050] 4) Neighborhood search
[0051] Find a better production scheduling plan in the neighborhood of the current solution. If the neighborhood solution is better than the current solution, update the current solution.
[0052] 5) Termination condition
[0053] If the objective function value does not improve significantly within a number of iterations, or the solution time limit is reached, the algorithm terminates.
[0054] 6) Result output
[0055] If the production scheduling is successful, the system outputs the production scheduling result, and the user formulates the corresponding production plan according to the result;
[0056] If the production scheduling fails (i.e. the solution time limit is reached and there is no any production scheduling result), go to "virtual grade production scheduling".
[0057] (3) Virtual universal grade definition
[0058] The reason why the grade production scheduling fails is that there are multiple hard constraints that restrict each other, which eventually leads to no feasible solution. When the user analyzes, he usually assumes that, for example, the transition of grade A to grade C is feasible, if the transition of grade C to grade A is also feasible, then the production scheduling can be successful. The essence is to relax some hard constraints (change the one-way transition of hard constraint grade A to grade C to the bidirectional transition of grade A to grade C and grade C to grade A).
[0059] Based on the above analysis, if there is a certain grade that can be transitioned to grade A and grade C, and grade A and grade C can be transitioned to this grade, then it can be realized by taking it as a transit to realize the bidirectional transition of grade A and grade C.
[0060] For example, Figure 1As shown, the hollow circle represents the original grade, the gray circle represents the newly added grade V, and the arrow direction is the conversion direction. Grade C cannot be directly converted to grade A, but it can be achieved through the transition of grade V, i.e., grade C → grade V → grade A.
[0061] Furthermore, when providing the production scheduling scheme, the grade V is hidden, and the user is informed. That is, the grade scheduling constraint conflict exists, but if grade C can be converted to grade A, there is a feasible solution. Therefore, the user checks whether grade C cannot be converted to grade A, which exists in an unreasonable situation.
[0062] Therefore, a virtual universal grade is defined:
[0063] It can be bidirectionally converted with all other grades (i.e., "universal");
[0064] The grade is hidden when the production scheduling result is output (i.e., "virtual");
[0065] The two original grades adjacent to the grade before and after are the "possible unreasonable conditions" set by the user.
[0066] (4) Virtual universal grade scheduling
[0067] The main purpose of virtual universal grade scheduling is to find all possible set unreasonable conditions, so it is not necessary to consider whether the production scheduling plan is optimal, but to find all feasible solutions.
[0068] 1) Grade scheduling fails? If yes, go to 2); if no, go to 8);
[0069] 2) Construct a virtual universal grade;
[0070] 3) Add the newly added virtual universal grade to the original production scheduling plan;
[0071] Grade basic information: planned production quantity is 0, user demand time is infinite;
[0072] Conversion relationship: can be converted from / to any original grade, all conversion time is 0, conversion transition material is 0, etc.;
[0073] First processing grade constraint: no;
[0074] Partial sequence constraint: none.
[0075] 4) Find all Hamiltonian paths using backtracking algorithm to find all possible production scheduling plans of the current grade;
[0076] Among them, the basic idea of backtracking method: through depth-first search (DFS) to gradually build the path, if the current path cannot form a Hamiltonian path, backtrack. Including the following steps:
[0077] 1. Start at an arbitrary vertex, mark it as visited.
[0078] 2. Recursively visit its unvisited neighbors.
[0079] 3. If the path contains all vertices and loops back to the starting vertex, record it as a solution.
[0080] 4. Backtrack, removing the visit mark from the vertices, and continue searching for other possible paths.
[0081] A Hamiltonian path is a path in a graph that visits each vertex exactly once.
[0082] 5) If the scheduling plan is empty, i.e. the scheduling fails, go to 1); if the scheduling plan is not empty, i.e. the scheduling succeeds, go to 6);
[0083] 6) For each scheduling plan, keep the two serial numbers before and after each virtual universal serial number as "possible unreasonable conditions";
[0084] 7) For each scheduling plan, hide all virtual universal serial numbers, keep the remaining scheduling plan, and output it to the user. If the above "possible unreasonable conditions" are repaired, the scheduling plan can be obtained;
[0085] 8) End.
[0086] (5) Example
[0087] 1) A total of six serial numbers A, B, C, D, E, and F are prepared for scheduling, as shown in the figure. Among them, serial number A is the first processing, represented by a dark solid circle; serial numbers C and F are serial processing serial numbers, represented by light solid circles, serial number C is serial number 1, and serial number F is serial number 2, so serial number C must be processed before serial number F; Figure 2
[0088] Other serial numbers are represented by hollow circles; the transfer relationship is shown by the dashed arrow, for example, serial number A and serial number B can be bidirectional transfer, serial number D can be transferred to serial numbers C, E, and F, etc.;
[0089] 2) Serial number A is the first processing serial number and must be processed first; serial number C can only be transferred from other serial numbers, so the first three serial numbers must be A→B→C. The remaining serial numbers D, E, and F cannot be scheduled, and the scheduling fails;
[0090] 3) Add a virtual universal serial number V1, and add serial number V1 to the original scheduling plan; serial number V1 can be bidirectional with serial numbers A, B, C, D, E, and F;
[0091] 4) Reschedule production. One of the production plans is A→B→C→V1→D. Since the remaining brands E and F cannot be transferred to each other, brands E and F cannot be scheduled, and the production scheduling fails. Figure 3 As shown, the solid arrows indicate part of the scheduled production plan;
[0092] 5) Add the virtual universal brand V2 again and add brand V2 to the original production schedule; brand V2 can be transferred to brand A, B, C, D, E, F, and V1 in both directions;
[0093] 6) Rescheduling
[0094] Production schedule 1: There is A→B→C→V1→D. If brand D is switched to brand F, the successful production schedule is A→B→C→V1→D→F→V2→E. Figure 4 As shown in the figure, the two brands before and after virtual universal brand V1 are C and D, and the two brands before and after virtual universal brand V2 are F and E. Therefore, the "possibly unreasonable condition" is [C→D,F→E]. All virtual universal brands are hidden, and the remaining production plan, A→B→C→D→F→E, is retained and output to the user.
[0095] Production schedule 2: The successful production schedule is A→B→C→V1→F→V2→D→E. Figure 5 As shown in the figure, the two brands before and after virtual universal brand V1 are C and F, and the two brands before and after virtual universal brand V2 are F and D. Therefore, the "potentially unreasonable condition" is [C→F, F→D]. All virtual universal brands are hidden, and the remaining production plan, A→B→C→F→D→E, is retained and output to the user.
[0096] 7) The user conducts troubleshooting based on the repair suggestions
[0097] Repair suggestion 1: "Potentially unreasonable conditions" is [C→D, F→E]. If these two groups of brands can be transferred, the production schedule A→B→C→D→F→E can be used;
[0098] Repair suggestion 2: "Possibly unreasonable conditions" are [C→F, F→D]. If these two groups of brands can be converted, the production schedule A→B→C→F→D→E can be used.
[0099] The specific embodiments described herein are merely illustrative of the spirit of the present invention. Persons skilled in the art may make various modifications, additions, or substitutions to the described specific embodiments without departing from the spirit of the present invention or exceeding the scope of the appended claims.
Claims
1. An intelligent repair method after polymer grade production scheduling failure, characterized in that After the production scheduling of a brand fails, virtual universal brands are added one by one and production scheduling is attempted until the production scheduling is successful; the virtual universal brand can be switched to production in both directions with all other brands. After the production scheduling is successful, the two original brands before and after the virtual universal brand are the original reasons for the production scheduling failure. After the reason for the failure is corrected, the production scheduling of the original brand is completed.
2. The method according to claim 1, characterized in that The virtual universal brand can be hidden when the production scheduling result is output. Hiding the virtual universal brand means obtaining the production scheduling result after the cause of the failure is repaired.
3. The method according to claim 1, characterized in that The reason for the repair failure is that the hard constraint condition is adjusted to a soft constraint condition.
4. The method according to claim 1, characterized in that The method specifically comprises the following steps: 1) Brand production scheduling failed; 2) Constructing a virtual universal brand that can be bidirectionally converted to production with all other brands; 3) Add the newly added virtual universal brand to the original production schedule; 4) Find all possible production schedules for the current brand; 5) If the production schedule is empty, that is, the production schedule still fails, go to step 2); if the production schedule is not empty, that is, the production schedule is successful, go to step 6); 6) For each production schedule, the two grades before and after each virtual universal grade are retained as "potentially unreasonable conditions"; 7) For each production schedule, all virtual universal brands are hidden, the remaining production schedule is retained, and the remaining schedule is output to the user. If the above "potentially unreasonable conditions" are corrected, the production schedule can be obtained; 8) End.
5. The method according to claim 4, characterized in that In step 4), a backtracking algorithm is used to find all Hamiltonian paths and all possible production schedules for the current brand are found.