An intelligent production scheduling method for robot cable flexible manufacturing
By constructing a directed acyclic process topology graph and equipment state space vector, and combining discrete particle swarm optimization algorithm and equipment resource locking strategy, the equipment deadlock problem caused by parallel and mutually exclusive processes in robot cable manufacturing was solved, thus improving production efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- FUZHOU YONGTONG WIRE & CABLE
- Filing Date
- 2026-05-13
- Publication Date
- 2026-06-09
AI Technical Summary
Existing production scheduling systems for flexible manufacturing of robotic cables cannot effectively handle parallel and mutually exclusive processes in multi-branch process paths, leading to equipment resource deadlock and production line stagnation.
By generating a directed acyclic process topology graph containing parallel and mutually exclusive processes, and combining it with the equipment state space vector, a discrete particle swarm optimization algorithm is used for scheduling optimization. The particle positions are updated using a neighborhood search strategy of equipment resource locks, thereby generating a production scheduling sequence that satisfies process dependencies and equipment physical constraints.
It effectively avoids equipment resource deadlock, shortens the material flow stagnation time in multi-variety, small-batch mixed production lines, and improves production efficiency.
Smart Images
Figure CN122175327A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing technology, and specifically to an intelligent production scheduling method for flexible manufacturing of robotic cables. Background Technology
[0002] In the field of flexible manufacturing of robotic cables, existing production scheduling systems typically construct fixed process templates based on standard operating procedures and assign tasks using heuristic rules such as first-come, first-served or shortest processing time priority. During operation, the scheduling system reads the process list from the order and, according to a preset sequence, searches the equipment pool for idle equipment with the corresponding processing capabilities to assign processing tasks. For robotic cables with optional multi-core wires and shielding layers, the production process requires movement between different processing lines depending on the material's condition.
[0003] The manufacturing process of robotic cables involves multiple branching process paths, with parallel processing and mutually exclusive equipment usage logic between different processes. Fixed process templates cannot represent the mutual exclusion and parallel dependencies between processes. When different cable orders are processed simultaneously on shared equipment such as extrusion and stranding, heuristic rules based on equipment idle status cannot predict resource usage conflicts in subsequent processes. Due to the lack of a mechanism to jointly constrain process dependencies and equipment physical status, existing technologies, when arranging parallel and mutually exclusive processes, result in multiple processing tasks cyclically waiting for the same equipment, leading to equipment resource deadlock and production line stagnation. Summary of the Invention
[0004] The purpose of this invention is to provide an intelligent production scheduling method for flexible manufacturing of robotic cables, which can effectively solve the problems mentioned in the background art.
[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:
[0006] A smart production scheduling method for flexible manufacturing of robot cables includes: extracting process parameters of the robot cable to be processed and generating a directed acyclic process topology graph containing parallel and mutually exclusive processes.
[0007] Establish a device state space vector for the production line equipment, wherein the device state space vector includes the remaining lifespan of the equipment's processing die head and the current temperature control range of the equipment;
[0008] The directed acyclic process topology graph and the equipment state space vector are input into the discrete particle swarm optimization algorithm. In each iteration of the discrete particle swarm optimization algorithm, each process node is mapped to the equipment node that satisfies the current temperature control range of the equipment according to the hierarchical depth of the directed acyclic process topology graph.
[0009] For nodes in the directed acyclic process topology graph that have mutually exclusive processes, a neighborhood search strategy based on equipment resource locks is used to update the particle positions.
[0010] Output a production scheduling sequence that satisfies the dependency relationship between the physical processing boundary of the device state space vector and the directed acyclic process topology graph.
[0011] Preferably, the step of extracting the process parameters of the robot cable to be processed and generating a directed acyclic process topology diagram containing parallel and mutually exclusive processes includes: parsing the bill of materials and processing documents of the robot cable to be processed, and extracting the core drawing process, insulation extrusion process, shielding braiding process and cable stranding process as basic nodes;
[0012] Detect the material state transition conditions between the basic nodes. When the material state transition conditions satisfy the condition that the same material is processed in parallel between different processes, establish parallel directed edges between the basic nodes.
[0013] When the material state transition condition satisfies that only a single material is allowed to enter the same equipment within the same time period, mutually exclusive directed edges are established between the corresponding basic nodes, and the directed acyclic process topology is generated by combining the basic nodes, the parallel directed edges, and the mutually exclusive directed edges.
[0014] Preferably, the establishment of the equipment state space vector of the production line equipment includes: collecting the cumulative number of screw rotations of the extruder in the production line equipment and the historical processing temperature curve, and calculating the remaining life of the processing die head of the equipment based on the cumulative number of screw rotations;
[0015] The real-time temperature of the barrel heating zone and the real-time temperature of the die head of the extruder are obtained at the current moment, and the current temperature control range of the equipment is constructed based on the real-time temperature of the barrel heating zone and the real-time temperature of the die head.
[0016] The remaining lifespan of the processing die head of the equipment is vector-concatenated with the current temperature control range of the equipment to generate the state space vector of the equipment, wherein the current temperature control range of the equipment includes the highest tolerable temperature threshold and the lowest sustaining temperature threshold.
[0017] Preferably, the step of mapping each process node to a device node that satisfies the current temperature control range of the device according to the hierarchical depth of the directed acyclic process topology graph includes: performing a breadth-first traversal of the directed acyclic process topology graph to determine the hierarchical depth sequence of each process node;
[0018] In the current particle encoding of the discrete particle swarm algorithm, the processing temperature range required for the current process node to be mapped is extracted according to the ascending order of the level depth sequence;
[0019] Perform an intersection operation between the processing temperature range and the current temperature control range of the equipment. If the result of the intersection operation is a non-empty set, then assign the current process node to be mapped to the corresponding equipment node. If the result of the intersection operation is an empty set, then mark the current process node to be mapped as blocked and skip the current iteration mapping.
[0020] Preferably, the step of updating particle positions using a neighborhood search strategy based on device resource locks for nodes with mutually exclusive processes in the directed acyclic process topology graph includes: identifying a first mutually exclusive process node and a second mutually exclusive process node connected by mutually exclusive directed edges in the directed acyclic process topology graph.
[0021] In the particle encoding space of the discrete particle swarm optimization algorithm, a mutex lock flag is assigned to the candidate device node that is jointly pointed to by the first mutex process node and the second mutex process node;
[0022] During the execution of the neighborhood search strategy, when the first mutex process node occupies the candidate device node, the mutex lock flag is set to a locked state, forcing the second mutex process node to shift along a dimension different from the candidate device node in the particle encoding space, thereby generating an updated particle position.
[0023] Preferably, the production scheduling sequence that outputs the physical processing boundary of the device state space vector and the directed acyclic process topology includes: extracting the current global optimal particle code when the discrete particle swarm algorithm reaches the maximum number of iterations or the global optimal fitness value converges.
[0024] The global optimal particle code is expanded along the time axis to parse out the mapping relationship between each process node and the device node, as well as the execution sequence between the process nodes.
[0025] The production scheduling sequence is generated and output after comparing the execution sequence with the directed edge pointing relationship in the directed acyclic process topology graph, and confirming that the remaining lifespan of the processing die head of the equipment is greater than zero and the current temperature control range of the equipment covers the processing temperature range of the process node.
[0026] Preferably, when the material state transition condition satisfies the requirement that the same material is processed in parallel across different processes, establishing parallel directed edges between the basic nodes includes: tracking the physical splitting path of the same material in the bill of materials; when the physical splitting path indicates that a single wire core is diverted to multiple insulation extrusion lines, obtaining the current queue length of the multiple insulation extrusion lines.
[0027] Between the basic node corresponding to the wire core drawing process and the basic node of the insulation extrusion process corresponding to the multiple insulation extrusion production lines, a parallel directed edge carrying the current queue length is established, wherein the weight value of the parallel directed edge is negatively correlated with the current queue length.
[0028] Preferably, the step of obtaining the real-time temperature of the barrel heating zone and the real-time temperature of the die of the extruder at the current moment, and constructing the current temperature control range of the equipment based on the real-time temperature of the barrel heating zone and the real-time temperature of the die, includes: when the extruder is in an idle standby state, reading the environmental cooling curve of the real-time temperature of the barrel heating zone and the heat dissipation gradient of the real-time temperature of the die;
[0029] Based on the ambient cooling curve and the heat dissipation gradient, the minimum sustaining temperature threshold in the current temperature control range of the equipment is constructed. When the real-time temperature of the barrel heating zone is lower than the minimum sustaining temperature threshold, the preheating command of the extruder is triggered, and the lower limit of the current temperature control range of the equipment is updated to the minimum sustaining temperature threshold.
[0030] Preferably, the execution of the neighborhood search strategy further includes: monitoring the completion state parameters of the first mutually exclusive process node in the fitness function of the discrete particle swarm optimization algorithm;
[0031] When the completion status parameter indicates that the processing task corresponding to the first mutually exclusive process node is completed on the candidate device node, calculate the cooling time and material cleaning time of the candidate device node;
[0032] After a waiting period equal to the sum of the cooling time and the material cleaning time, the mutex flag is switched from the locked state to the released state, allowing the second mutex process node to acquire the allocation permission of the candidate device node in subsequent iterations.
[0033] Preferably, when the discrete particle swarm optimization algorithm reaches the maximum number of iterations or the global optimal fitness value converges, the step of extracting the current global optimal particle code includes: setting a stall iteration counter for the global optimal fitness value, and comparing the absolute value of the difference between the current global optimal fitness value and the previous global optimal fitness value after each iteration.
[0034] When the absolute value of the difference is less than the preset perturbation threshold, the stagnation iteration counter is incremented by 1. When the value of the stagnation iteration counter reaches the preset stagnation threshold, it is determined that the global optimal fitness value has converged. At this time, the current global optimal particle code is directly extracted, and the remaining iteration steps of the discrete particle swarm algorithm are skipped.
[0035] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0036] 1. This invention generates a directed acyclic process topology graph containing parallel and mutually exclusive directed edges by extracting process parameters. Combined with the equipment state space vector containing the remaining die life and temperature control range, the process is mapped to equipment nodes satisfying the temperature control range based on the layer-level depth of the topology graph during discrete particle swarm optimization iterations. For nodes with mutually exclusive directed edges, a neighborhood search strategy based on a mutex lock flag is used to update particle positions. This processing method establishes a strong coupling relationship between process dependencies and equipment physical constraints during the scheduling sequence generation stage. The mutex lock flag blocks concurrent occupancy paths of mutually exclusive processes for the same candidate equipment node, eliminating equipment resource deadlock caused by parallel process branches.
[0037] 2. When generating the directed acyclic process topology graph, parallel directed edges are established based on the physical splitting path and the current queue length, and weighted values are assigned. When constructing the current temperature control range of the equipment, the minimum maintenance temperature threshold is determined based on the environmental cooling curve and heat dissipation gradient under no-load conditions. During the execution of the neighborhood search strategy, a waiting period is set based on the sum of the cooling time and the material clearing time to control the timing of the mutex lock flag switching from the locked state to the released state. These additional technical features enable the scheduling sequence to satisfy topological dependencies while conforming to the physical boundary of the extruder temperature control and the actual material flow state of the production line queue, reducing processing blockages caused by substandard equipment temperatures or insufficient material clearing time, and shortening the material flow stagnation time in a multi-variety, small-batch mixed-line production environment. Attached Figure Description
[0038] Figure 1 This is a flowchart of the directed acyclic process topology graph generation process of the present invention;
[0039] Figure 2 Flowchart for constructing the state space vector of the device of this invention;
[0040] Figure 3 This is a flowchart illustrating the process-equipment mapping of the discrete particle swarm optimization algorithm of this invention.
[0041] Figure 4 This is a flowchart of the parallel directed edge weight calculation and flow distribution process of this invention;
[0042] Figure 5 This is a flowchart of the mutex lock flag control and neighborhood search process of the present invention;
[0043] Figure 6 This is a flowchart of the production scheduling sequence generation and verification process of the present invention. Detailed Implementation
[0044] Please refer to the attached document. Figure 1This embodiment provides a production scheduling process for flexible manufacturing of robot cables. First, the process parameters of the robot cable to be processed are extracted, generating a directed acyclic process topology graph containing parallel and mutually exclusive processes. The process parameters of the robot cable to be processed are derived from order process files, bill of materials data, and standardized processing process cards stored in the production line's MES system. These parameters include process name, process sequence dependencies, type of processing equipment required for the process, processing temperature requirements, material specifications for the process, and rated processing time for the process.
[0045] The extracted process parameters are analyzed in a structured manner, and process nodes are constructed with a single processing step as the smallest unit. The processing dependencies between process nodes are identified, including parallel processing dependencies and mutual exclusion dependencies. A directed acyclic process topology graph is generated based on the process nodes and their corresponding dependencies. There are no directed edges with cyclic dependencies in this topology graph, and the processing flow of all process nodes follows the direction of the directed edges to complete the material state transformation.
[0046]
[0047] in, This is a directed acyclic process topology diagram. For a set of process nodes, This represents the total number of processes required to process the robot cable. It is a set of parallel directed edges used to represent the parallel processing dependencies between process nodes; It is a set of mutually exclusive directed edges, used to represent the mutually exclusive occupancy relationship of equipment resources between process nodes.
[0048] Please refer to the attached document. Figure 2 A state space vector for the production line equipment is established. This production line equipment includes specialized robotic cable processing equipment such as extruders, wire drawing machines, braiding machines, and cabling machines. The core processing equipment is the extruder, whose physical state directly determines the feasibility and quality of the processing task. For each piece of equipment in the production line, real-time status data is collected during operation. Two core state parameters are extracted: the remaining lifespan of the processing die and the current temperature control range of the equipment. After standardization of these two core state parameters, vectors are concatenated to generate the corresponding equipment state space vector. The combined state space vectors of all equipment form the overall state matrix of the production line, used to characterize the real-time physical processing boundaries of all equipment on the production line.
[0049]
[0050] in, For the first The state-space vector of a production line device. , This represents the total number of equipment on the production line. The remaining lifespan of the die head for this equipment; This is the lower limit of the current temperature control range of the device, i.e., the minimum maintenance temperature threshold. This is the upper limit of the current temperature control range of the device, i.e., the highest tolerable temperature threshold.
[0051] Please refer to the attached document. Figure 3 The directed acyclic process topology graph and equipment state space vectors are input into the discrete particle swarm optimization (DSO) algorithm to complete the mapping from process nodes to equipment nodes and optimize the scheduling scheme. In the population initialization phase of the DSO algorithm, the dimension of the particle encoding and the population size are set based on the total number of process nodes and the total number of production line equipment. The position encoding of each particle corresponds to a process-equipment allocation scheme, and the velocity encoding of each particle corresponds to the update step size and direction of the particle position in the encoding space.
[0052]
[0053] in, For the first Position encoding vector of each particle , The size of the particle population; For the first The first particle Each process node The mapped device node number, , .
[0054] In each iteration of the Discrete Particle Swarm Optimization (DPSO) algorithm, each process node is first mapped to a device node that meets the current temperature control range of the equipment, based on the hierarchical depth of the directed acyclic process topology graph. The DPSO graph is traversed to determine the hierarchical depth of each process node. The hierarchical depth is used to characterize the order of process nodes in the entire processing flow. The smaller the hierarchical depth of the process node, the earlier its processing sequence is, and the higher the priority should be for completing the equipment mapping.
[0055]
[0056] in, For process nodes The depth of the hierarchy; For process nodes The set of all direct predecessor nodes; for a starting process node without any predecessor nodes, its hierarchy depth. .
[0057] Following the order of increasing hierarchical depth, the processing temperature requirements of the current process node to be mapped are extracted sequentially. The processing temperature requirements are then matched with the current temperature control range of the equipment in the equipment state space vector. Only when the current temperature control range of the equipment meets the processing temperature requirements of the process node is the process node allowed to be mapped to the corresponding equipment node, thus completing the initial process-equipment allocation in the current iteration.
[0058] in, For process nodes Required processing temperature range and the first The result of the intersection calculation of the current temperature control range of the devices; , These are the process nodes. The lower and upper limits of the required processing temperature range; when When it is a non-empty set, determine the first... The equipment meets the process nodes Temperature control matching conditions.
[0059] During the iterative process of the discrete particle swarm optimization algorithm, the velocity and position of the particles are updated according to preset rules. The update process must satisfy temperature control matching constraints and equipment resource occupancy constraints.
[0060] in, For the first The particle in the first The velocity vector in the next iteration; Inertial weight; , These are the individual learning factor and the global learning factor, respectively. , For the range of values within Uniformly distributed random numbers within; For the first The individual historical optimal position vector of each particle; This is the global historical optimal position vector of the population; For the first The particle in the first The position vector in the next iteration.
[0061]
[0062] in, For the first The particle in the first In the nth iteration The positional value of the dimension; This is the rounding function; For the first The particle in the first In the nth iteration The speed value of the dimension; the constraints include the matching conditions of the equipment temperature control range and the equipment resource lock state conditions.
[0063] For nodes with mutually exclusive processes in the directed acyclic process topology graph, a neighborhood search strategy based on equipment resource locks is adopted to update particle positions. First, mutually exclusive occupancy dependencies between process nodes are identified in the directed acyclic process topology graph. Processing tasks corresponding to mutually exclusive process nodes cannot occupy the same equipment for processing within the same time period; otherwise, equipment resource contention and processing conflicts will occur. In the particle encoding space of the discrete particle swarm optimization algorithm, a mutex lock flag is assigned to the candidate equipment node that the mutually exclusive process nodes all point to. The mutex lock flag is used to characterize the resource occupancy status of the candidate equipment node.
[0064] During the execution of the neighborhood search strategy, when one of the mutually exclusive process nodes is assigned to the candidate device node in the current iteration, the corresponding mutex lock flag is set to a locked state. This forces the other process nodes that have a mutual exclusion relationship with this process node to update their positions in the particle encoding space along the dimensions other than the dimension corresponding to the candidate device node, generating updated particle positions that satisfy the mutual exclusion constraints. This avoids multiple mutually exclusive process nodes being assigned to the same device in the same iteration, eliminating the risk of device resource contention.
[0065]
[0066] in, For the first The mutex lock flag of each candidate device node; when When the device is in a locked state, except for the currently occupied mutex process node, other mutex process nodes cannot be allocated to this device node in the current iteration; when When in a released state, process nodes are allowed to be assigned and mapped.
[0067] Output a production scheduling sequence that satisfies the dependency relationship between the physical processing boundary of the equipment state space vector and the directed acyclic process topology graph. When the iterative process of the discrete particle swarm optimization algorithm meets the termination condition, extract the global optimal particle code in the current population, decode the particle code, and parse out the mapping relationship between each process node and the equipment node, as well as the execution sequence between each process node, according to the chronological order of the time axis.
[0068] The execution sequence obtained from the parsing is verified to confirm that it is completely consistent with the directed edge pointing relationship in the directed acyclic process topology graph and that there is no process sequence that violates the predecessor-successor dependency relationship. At the same time, the mapped device nodes are verified to confirm that the remaining lifespan of the processing die head of the corresponding device is greater than zero and that the current temperature control range of the device completely covers the processing temperature range of the process node. After all verification items pass, the corresponding production scheduling sequence is generated and output. This production scheduling sequence can be directly sent to the production line MES system to drive each device to complete the processing and production of robot cables according to the scheduling sequence.
[0069] Table 1. Robot Cable Processing Process Nodes and Topology Parameters
[0070]
[0071] This table clearly defines the process nodes of the entire processing flow of the robot cable to be processed, the temperature control requirements of each process, the predecessor and successor dependencies, and the dependency types of the processes, providing basic data support for the construction of the directed acyclic process topology. Among them, nodes 1 and 2 are parallel process nodes, which can be processed on different equipment within the same time period; nodes 5 and 7 are mutually exclusive process nodes, both of which require the use of extruder equipment to complete processing and cannot use the same extruder within the same time period.
[0072] In this embodiment, by constructing a directed acyclic process topology graph containing parallel and mutually exclusive dependencies, and combining it with the equipment state space vector representing the physical processing boundary of the equipment, a hierarchical depth-first process-equipment mapping rule and a neighborhood search strategy based on equipment resource locks are introduced during the iteration process of the discrete particle swarm algorithm. This ensures that the generated scheduling scheme simultaneously satisfies process dependency constraints and equipment physical state constraints. During the scheduling sequence generation stage, the concurrent occupation path of mutually exclusive processes on the same equipment is blocked, eliminating the triggering conditions for equipment resource deadlock.
[0073] In a preferred embodiment, during the process of generating the directed acyclic process topology graph, the bill of materials and processing technology documents of the robot cable to be processed are first parsed. The bill of materials contains all the material composition, material specifications, material hierarchy and material consumption path of the robot cable to be processed. The processing technology documents contain the processing steps, process parameter requirements, quality inspection standards and equipment adaptation requirements of the corresponding materials.
[0074] The bill of materials and processing documents were structurally analyzed, and the core drawing process, insulation extrusion process, shielding braiding process, and cable stranding process were extracted as basic nodes. These four basic nodes cover the core processes of robot cable processing. Among them, the core drawing process is the initial processing node for materials, which is used to draw the raw metal conductor into cores that meet the specifications. The insulation extrusion process is used to coat the surface of the cores with an insulation layer to achieve electrical isolation between the cores. The shielding braiding process is used to braid a metal shielding layer on the surface of the insulated cores to suppress electromagnetic interference. The cable stranding process is used to strand multiple insulated and shielded cores into a cable core that meets the specifications, completing the main structure processing of the robot cable.
[0075] Please refer to the attached document. Figure 4 The system detects the material state transition conditions between basic nodes. These conditions characterize the physical changes in material form, equipment occupancy requirements, and time matching requirements as the material flows between different processes. When the material state transition conditions satisfy the requirement that the same material is processed in parallel across different processes, parallel directed edges are established between the corresponding basic nodes. Parallel processing of the same material across different processes refers to the process where, after processing by a preceding process, the same batch of material can be divided into multiple independent material units, which simultaneously enter different subsequent process nodes to complete processing. The processing of these multiple material units does not interfere with each other, and the processed material units can then be reassembled and enter the same subsequent process node.
[0076] The physical splitting path of the same material in the bill of materials is tracked. The physical splitting path represents the flow path of a single batch of material being split into multiple independent sub-material units during processing. When the physical splitting path indicates that a single wire core is diverted to multiple insulation extrusion production lines, the current queue length of the multiple insulation extrusion production lines is obtained. The current queue length is the total number of tasks assigned but not yet processed in the corresponding production line. Parallel directed edges carrying the current queue length are established between the basic node corresponding to the wire core drawing process and the basic nodes of the insulation extrusion process corresponding to the multiple insulation extrusion production lines. The weight value of the parallel directed edge is negatively correlated with the current queue length. The shorter the queue length of the production line, the higher the weight value of the corresponding parallel directed edge, and the higher the priority in the process allocation.
[0077]
[0078] in, These are the weights of the parallel directed edges; The current queue length of the target insulation extrusion production line is the number of tasks to be processed on this production line. The weight value is negatively correlated with the queue length. The shorter the queue length, the higher the weight value, and the higher the priority of the corresponding parallel processing path.
[0079] Table 2. Mapping parameters between parallel directed edge weights and production line queue lengths.
[0080]
[0081] This table quantifies the weights and path priorities of parallel directed edges from the wire drawing process node to different insulation extrusion process nodes. The weights are assigned based on the current queue length of the target production line, providing a quantitative basis for the diversion and allocation of parallel processes. This allows materials to be preferentially diverted to production lines with shorter queue lengths and lower loads, reducing the waiting time for materials to be processed.
[0082] When the material state transition condition satisfies that only one material is allowed to enter the same equipment within the same time period, mutually exclusive directed edges are established between the corresponding basic nodes. Situations where only one material is allowed to enter the same equipment within the same time period include situations where the same equipment can only process one type of material at a time and cannot simultaneously handle multiple processing tasks, or where processing different materials requires different die head changes, parameter configurations, and material cleaning, making it impossible to switch between different materials without interrupting processing.
[0083] Mutually exclusive directed edges represent the mutual exclusion relationship of equipment resources between two process nodes. Two process nodes connected by a mutually exclusive directed edge cannot be assigned to the same equipment for processing within the same time period. A directed acyclic process topology graph is generated by combining basic nodes, parallel directed edges, and mutually exclusive directed edges. During the generation process, the cyclic dependencies of the topology graph are checked. If a cyclic directed edge is detected, the corresponding process dependency relationship is corrected to ensure that the final generated topology graph satisfies the acyclic constraint.
[0084] In this embodiment, the core process nodes are extracted by parsing the bill of materials and processing technology documents. Parallel directed edges and mutually exclusive directed edges are established based on the material state transition conditions to complete the construction of the directed acyclic process topology graph. At the same time, the parallel directed edges are assigned corresponding weight values based on the production line queue length, so that the topology graph can not only accurately represent the logical dependencies between processes, but also fit the actual load state of the production line, providing a precise process constraint basis for subsequent process-equipment mapping.
[0085] In a preferred embodiment, during the process of establishing the equipment state space vector of the production line equipment, the cumulative number of screw rotations of the extruder in the production line equipment and the historical processing temperature curve are first collected. The cumulative number of screw rotations of the extruder is collected in real time by the encoder built into the equipment and uploaded to the production line MES system for storage. The cumulative number of screw rotations is the total number of screw rotations of the extruder from the start of its use to the current moment, which directly reflects the wear degree of the processing die.
[0086] Historical processing temperature curves are continuous data showing the temperature changes of the barrel heating zone and die over time during past extrusion processes. These data are used to fit the temperature change and heat dissipation characteristics of the extruder. The remaining lifespan of the processing die is calculated based on the cumulative screw rotations. The processing die is a core, easily worn component of the extruder, and its wear is directly related to the cumulative screw rotations. If the die wear exceeds a certain threshold, it will lead to uneven thickness of the extruded insulation or sheath layer, affecting the electrical and mechanical properties of the cable.
[0087]
[0088] in, For the first The rated cumulative number of rotations of the processing die head on the machine; For the first The cumulative number of rotations of the screw of the machine is acquired in real time through the machine's PLC system; The range of values is ,when When the equipment's die head reaches its service life limit, it is determined that the equipment is no longer capable of undertaking processing tasks.
[0089] The real-time temperature of the barrel heating zone and the real-time temperature of the die head of the extruder are obtained at the current moment. The barrel heating zone is usually divided into multiple independent temperature control sections. Each temperature control section is equipped with a temperature sensor to collect the temperature value of the corresponding area in real time. The real-time temperature of the barrel heating zone is the average of the real-time temperatures of all temperature control sections, and the real-time temperature of the die head is the real-time temperature value collected by the temperature sensor at the die head position.
[0090] The current temperature control range of the equipment is constructed based on the real-time temperatures of the barrel heating zone and the die. This range includes a maximum tolerable temperature threshold and a minimum sustaining temperature threshold. The maximum tolerable temperature threshold is the highest safe temperature that the extruder barrel and die can withstand; exceeding this temperature will lead to damage to equipment components or carbonization of the processed material. The minimum sustaining temperature threshold is the lowest temperature required for the extruder to complete its processing tasks normally; below this temperature, the material will not melt sufficiently, failing to meet the extrusion process requirements. The remaining lifespan of the processing die and the current temperature control range are vector-concatenated to generate an equipment state space vector. During this vector concatenation process, all parameters are standardized to ensure that parameters with different dimensions are within the same numerical range, avoiding numerical deviations in subsequent algorithm calculations.
[0091] When the extruder is in an idle state, the ambient cooling curve of the real-time temperature of the barrel heating zone and the heat dissipation gradient of the real-time temperature of the die are read. The idle state refers to the state where the extruder screw stops rotating, no material enters the barrel, and only the heating system is in standby operation. The ambient cooling curve is a continuous curve showing the temperature change of the barrel heating zone over time after the heating system is turned off in the idle state, and is used to characterize the heat dissipation characteristics of the barrel; the heat dissipation gradient is the rate of change of the die temperature over time, and is used to characterize the heat dissipation capacity of the die.
[0092] Based on the environmental cooling curve and heat dissipation gradient, the minimum sustaining temperature threshold in the current temperature control range of the equipment is constructed. A temperature change function under no-load conditions of the extruder is obtained by fitting historical environmental cooling curves. Based on this function, the decrease in barrel temperature within a preset processing preparation time is calculated. Using the minimum process temperature required for processing as a benchmark, this temperature decrease is added to obtain the minimum sustaining temperature threshold for the current temperature control range of the equipment. When the real-time temperature of the barrel heating zone falls below the minimum sustaining temperature threshold, a preheating command is triggered on the extruder, controlling the extruder's heating system to start, raising the temperature of the barrel and die to above the minimum sustaining temperature threshold, and updating the lower limit of the current temperature control range of the equipment to this minimum sustaining temperature threshold.
[0093]
[0094] in, When the extruder is unloaded, the barrel heating zone is in Temperature value at any given time; The initial temperature of the barrel heating zone at the moment of cooling initiation; The heat dissipation coefficient of the extruder barrel is obtained by fitting historical temperature data; This refers to the duration of the cooling process; This refers to the ambient temperature of the workshop where the production line is located.
[0095] Table 3. State Space Vector Parameters for Extruder Equipment
[0096]
[0097] This table clarifies the core state parameters of each extruder in the production line. The remaining lifespan of the die is calculated based on the cumulative number of screw rotations, and the upper and lower limits of the temperature control range are determined based on real-time temperature and heat dissipation characteristics. All parameters together form the state space vector of the corresponding equipment, providing clear physical boundary constraints for the process-equipment mapping of the discrete particle swarm optimization algorithm. Only when the remaining lifespan of the die of the equipment is greater than zero and the temperature control range meets the processing requirements of the process, is it allowed to undertake the corresponding processing task.
[0098] In this embodiment, the remaining lifespan of the die head is calculated based on the cumulative number of screw rotations, and the current temperature control range of the equipment is constructed based on the real-time temperatures of the barrel and die head. At the same time, the minimum maintenance temperature threshold is determined by combining the ambient cooling curve and heat dissipation gradient under no-load conditions. This completes the construction of the equipment state space vector, so that the representation of the equipment state can fit the actual physical characteristics and operating state of the extruder, avoiding processing interruptions and product quality defects caused by substandard equipment temperature or die head lifespan.
[0099] In a preferred embodiment, in each iteration of the discrete particle swarm optimization algorithm, when mapping each process node to the device node that satisfies the current temperature control range of the device according to the hierarchical depth of the directed acyclic process topology graph, the breadth-first traversal of the directed acyclic process topology graph is first performed to determine the hierarchical depth sequence of each process node.
[0100] Breadth-first traversal starts from the initial process node in the topology graph, visits all direct successor nodes of the initial process node in turn, and then visits the direct successor nodes of each successor node in turn. During the traversal, the level depth of each process node is recorded. Finally, a level depth sequence is generated in ascending order of level depth. Process nodes at the same level in the level depth sequence can complete equipment mapping in parallel within the same iteration cycle. Process nodes at different levels are mapped strictly in ascending order of level depth to ensure that the equipment mapping of the predecessor process node is completed before the mapping of the successor process node, thus avoiding allocation results that violate process dependencies.
[0101] In the current particle encoding of the Discrete Particle Swarm Optimization (DPSO) algorithm, the processing temperature range required by the current process node to be mapped is extracted in ascending order of the hierarchical depth sequence. The processing temperature range is the temperature range that can guarantee processing quality as clearly specified in the processing technology file corresponding to the process node. The processing temperature range is intersected with the current temperature control range of the equipment. If the result of the intersection operation is a non-empty set, it means that the current temperature control range of the equipment can cover the processing temperature requirements of the process node, satisfying the temperature control matching condition, and the current process node to be mapped is assigned to the corresponding equipment node. If the result of the intersection operation is an empty set, it means that the current temperature control range of the equipment cannot meet the processing temperature requirements of the process node, and it does not have the conditions to undertake the processing task. The current process node to be mapped is marked as blocked and the current iteration mapping is skipped. Temperature control matching and equipment allocation are re-performed in the next iteration.
[0102] Please refer to the attached document. Figure 5For nodes with mutually exclusive processes in the directed acyclic process topology graph, when updating particle positions using a neighborhood search strategy based on equipment resource locks, the first mutually exclusive process node and the second mutually exclusive process node connected by mutually exclusive directed edges in the directed acyclic process topology graph are first identified. The first mutually exclusive process node and the second mutually exclusive process node are two process nodes with mutually exclusive equipment resource relationships, and the two cannot occupy the same equipment to complete processing in the same time period.
[0103] In the particle encoding space of the Discrete Particle Swarm Optimization (DPO) algorithm, a mutex lock flag is assigned to the candidate device node jointly pointed to by the first and second mutually exclusive operation nodes. The state of the mutex lock flag is only related to the allocation state of the mutually exclusive operation node and is independent of the allocation state of the non-mutually exclusive operation nodes. During the execution of the neighborhood search strategy, when the first mutually exclusive operation node occupies a candidate device node, the mutex lock flag is set to a locked state, forcing the second mutually exclusive operation node to shift along a dimension different from that of the candidate device node in the particle encoding space. That is, the second mutually exclusive operation node cannot be assigned to this candidate device node in the current iteration and can only be selected from other device nodes that meet the temperature control conditions, generating an updated particle position.
[0104] During the execution of the neighborhood search strategy, it is also necessary to monitor the completion status parameter of the first mutually exclusive process node in the fitness function of the discrete particle swarm algorithm. The completion status parameter is used to characterize the execution progress of the processing task corresponding to the first mutually exclusive process node on the candidate device node. When the completion status parameter indicates that the processing task corresponding to the first mutually exclusive process node is completed on the candidate device node, the cooling time and material cleaning time of the candidate device node are calculated.
[0105] Cooling time is the duration required for the temperature of the equipment to drop to the required temperature range for subsequent processing tasks after the equipment completes the preceding high-temperature processing task. Material cleaning time is the duration required for the equipment to clean residual material in the barrel and die head and replace the processing die head after completing the preceding material processing task. After the waiting period equal to the sum of cooling time and material cleaning time, the mutex flag is switched from locked to released, allowing the second mutex process node to acquire the allocation permission of the candidate equipment node in subsequent iterations.
[0106]
[0107] in, The waiting period required for the mutex lock flag to switch from the locked state to the released state; Cooling time after candidate device nodes complete their preceding processing tasks; The material cleanup time after the candidate device node completes the preceding processing tasks.
[0108] The optimization process of the Discrete Particle Swarm Optimization (DPO) algorithm uses the fitness function as the evaluation criterion. The fitness function is used to quantify the quality of the scheduling scheme corresponding to the particle encoding. The smaller the fitness value, the better the overall performance of the scheduling scheme.
[0109] in, Encoding particle position The corresponding fitness value; This represents the total processing completion time corresponding to the scheduling scheme; This refers to the cumulative lifespan loss of the equipment module corresponding to this scheduling scheme; The device load balancing degree corresponding to this scheduling scheme; , , These are the weight coefficients for each evaluation indicator, and they satisfy... .
[0110] Please refer to the attached document. Figure 6 When outputting a production scheduling sequence that satisfies the dependency relationship between the physical processing boundary of the equipment state space vector and the directed acyclic process topology, the current global optimal particle code is extracted when the discrete particle swarm optimization algorithm reaches its maximum number of iterations or the global optimal fitness value converges. A stagnation iteration counter for the global optimal fitness value is set to record the number of iterations in which the global optimal fitness value remains unchanged. After each iteration, the absolute value of the difference between the current global optimal fitness value and the previous global optimal fitness value is compared. When the absolute value of the difference is less than a preset perturbation threshold, it indicates that the change in the global optimal fitness value is within a negligible range, and the current scheduling scheme has no significant optimization space; the stagnation iteration counter is incremented by 1. When the absolute value of the difference is greater than or equal to the preset perturbation threshold, it indicates that the global optimal fitness value still has significant optimization potential; the stagnation iteration counter is reset to zero. When the value of the stagnation iteration counter reaches the preset stagnation threshold, the global optimal fitness value is determined to have converged. At this point, the current global optimal particle code is directly extracted, the remaining iteration steps of the discrete particle swarm optimization algorithm are skipped, the algorithm iteration process is terminated, unnecessary iteration calculations are reduced, and the efficiency of scheduling scheme generation is improved.
[0111]
[0112] in, For the first The fitness value corresponding to the global optimal position of the population after the next iteration; For the first The fitness value corresponding to the global optimal position of the population after the next iteration; The preset perturbation threshold is used to determine whether the change in fitness value is within a negligible convergence interval.
[0113]
[0114] in, For the first The value of the stall iteration counter after the next iteration; when When the preset stagnation threshold is reached, the global optimal fitness value is determined to have converged, and the iteration process is terminated.
[0115] Table 4. Convergence parameters of the discrete particle swarm optimization algorithm during iterative process.
[0116]
[0117] This table tracks the changes in core parameters during the iteration process of the Discrete Particle Swarm Optimization (DPSO) algorithm. As the number of iterations increases, the global optimal fitness value gradually decreases and tends to stabilize. The fitness difference between adjacent iterations gradually narrows, and the stall iteration counter continues to accumulate. When the stall iteration counter reaches the preset stall threshold of 10, the algorithm is determined to have converged, the iteration process is terminated, and the current global optimal particle code is extracted for subsequent decoding and verification.
[0118] The globally optimal particle code is expanded along the time axis to analyze the mapping relationship between each process node and equipment node, as well as the execution sequence between process nodes. During the analysis, the rated processing time of each process, the equipment's processing preparation time, and the material flow time are combined to determine the processing start time and processing completion time of each process node, forming a complete time axis scheduling scheme. The execution sequence is compared with the directed edge pointing relationship in the directed acyclic process topology graph to confirm that the predecessor node of all process nodes completes processing before the processing start time of the process node, and there is no timing arrangement that violates the process dependency relationship. At the same time, it is confirmed that the remaining life of the processing die head of all mapped equipment nodes is greater than zero, and the current temperature control range of the equipment covers the processing temperature range of the corresponding process node. After all the verification items pass, the production scheduling sequence is generated and output.
[0119] In this embodiment, the hierarchical depth sequence of process nodes is determined by breadth-first traversal, and the temperature control matching mapping between processes and equipment is completed according to the hierarchical depth order. The allocation conflict of mutually exclusive process nodes is handled based on the neighborhood search strategy of equipment resource lock. The release timing of the mutex lock is controlled by combining the processing completion status and waiting period. At the same time, the algorithm is terminated early by using a stall iteration counter. Finally, a production scheduling sequence that meets all constraints is generated and output, so that the scheduling scheme has higher generation efficiency and better scheduling effect while satisfying process dependence and equipment physical constraints.
Claims
1. An intelligent production scheduling method for flexible manufacturing of robotic cables, characterized in that, include: Extract the process parameters of the robot cable to be processed and generate a directed acyclic process topology graph containing parallel and mutually exclusive processes. Establish a device state space vector for the production line equipment, wherein the device state space vector includes the remaining lifespan of the equipment's processing die head and the current temperature control range of the equipment; The directed acyclic process topology graph and the equipment state space vector are input into the discrete particle swarm optimization algorithm. In each iteration of the discrete particle swarm optimization algorithm, each process node is mapped to the equipment node that satisfies the current temperature control range of the equipment according to the hierarchical depth of the directed acyclic process topology graph. For nodes in the directed acyclic process topology graph that have mutually exclusive processes, a neighborhood search strategy based on equipment resource locks is used to update the particle positions. Output a production scheduling sequence that satisfies the dependency relationship between the physical processing boundary of the device state space vector and the directed acyclic process topology graph.
2. The intelligent production scheduling method for flexible manufacturing of robotic cables according to claim 1, characterized in that, The process parameters of the robot cable to be processed are extracted, and a directed acyclic process topology diagram containing parallel and mutually exclusive processes is generated. This includes: parsing the bill of materials and processing documents of the robot cable to be processed, and extracting the core drawing process, insulation extrusion process, shielding braiding process and cable stranding process as basic nodes. Detect the material state transition conditions between the basic nodes. When the material state transition conditions satisfy the condition that the same material is processed in parallel between different processes, establish parallel directed edges between the basic nodes. When the material state transition condition satisfies that only a single material is allowed to enter the same equipment within the same time period, mutually exclusive directed edges are established between the corresponding basic nodes, and the directed acyclic process topology is generated by combining the basic nodes, the parallel directed edges, and the mutually exclusive directed edges.
3. The intelligent production scheduling method for flexible manufacturing of robotic cables according to claim 1, characterized in that, The establishment of the equipment state space vector of the production line equipment includes: collecting the cumulative number of screw rotations of the extruder in the production line equipment and the historical processing temperature curve, and calculating the remaining life of the processing die head of the equipment based on the cumulative number of screw rotations; The real-time temperature of the barrel heating zone and the real-time temperature of the die head of the extruder are obtained at the current moment, and the current temperature control range of the equipment is constructed based on the real-time temperature of the barrel heating zone and the real-time temperature of the die head. The remaining lifespan of the processing die head of the equipment is vector-concatenated with the current temperature control range of the equipment to generate the state space vector of the equipment, wherein the current temperature control range of the equipment includes the highest tolerable temperature threshold and the lowest sustaining temperature threshold.
4. The intelligent production scheduling method for flexible manufacturing of robotic cables according to claim 1, characterized in that, The step of mapping each process node to a device node that satisfies the current temperature control range of the device according to the hierarchical depth of the directed acyclic process topology graph includes: performing a breadth-first traversal of the directed acyclic process topology graph to determine the hierarchical depth sequence of each process node; In the current particle encoding of the discrete particle swarm algorithm, the processing temperature range required for the current process node to be mapped is extracted according to the ascending order of the level depth sequence; Perform an intersection operation between the processing temperature range and the current temperature control range of the equipment. If the result of the intersection operation is a non-empty set, then assign the current process node to be mapped to the corresponding equipment node. If the result of the intersection operation is an empty set, then mark the current process node to be mapped as blocked and skip the current iteration mapping.
5. The intelligent production scheduling method for flexible manufacturing of robotic cables according to claim 1, characterized in that, The step of updating particle positions for nodes with mutually exclusive processes in the directed acyclic process topology graph using a neighborhood search strategy based on device resource locks includes: identifying a first mutually exclusive process node and a second mutually exclusive process node connected by mutually exclusive directed edges in the directed acyclic process topology graph. In the particle encoding space of the discrete particle swarm optimization algorithm, a mutex lock flag is assigned to the candidate device node that is jointly pointed to by the first mutex process node and the second mutex process node; During the execution of the neighborhood search strategy, when the first mutex process node occupies the candidate device node, the mutex lock flag is set to a locked state, forcing the second mutex process node to shift along a dimension different from the candidate device node in the particle encoding space, thereby generating an updated particle position.
6. The intelligent production scheduling method for flexible manufacturing of robotic cables according to claim 1, characterized in that, The production scheduling sequence that outputs the physical processing boundary of the device state space vector and the directed acyclic process topology graph includes: extracting the current global optimal particle code when the discrete particle swarm algorithm reaches the maximum number of iterations or the global optimal fitness value converges. The global optimal particle code is expanded along the time axis to parse out the mapping relationship between each process node and the device node, as well as the execution sequence between the process nodes. The production scheduling sequence is generated and output after comparing the execution sequence with the directed edge pointing relationship in the directed acyclic process topology graph, and confirming that the remaining lifespan of the processing die head of the equipment is greater than zero and the current temperature control range of the equipment covers the processing temperature range of the process node.
7. The intelligent production scheduling method for flexible manufacturing of robotic cables according to claim 2, characterized in that, When the material state transition condition satisfies the requirement that the same material is processed in parallel across different processes, a parallel directed edge is established between the basic nodes, including: tracking the physical splitting path of the same material in the bill of materials; when the physical splitting path indicates that a single wire core is diverted to multiple insulation extrusion lines, obtaining the current queue length of the multiple insulation extrusion lines. Between the basic node corresponding to the wire core drawing process and the basic node of the insulation extrusion process corresponding to the multiple insulation extrusion production lines, a parallel directed edge carrying the current queue length is established, wherein the weight value of the parallel directed edge is negatively correlated with the current queue length.
8. The intelligent production scheduling method for flexible manufacturing of robotic cables according to claim 3, characterized in that, The step of obtaining the real-time temperature of the barrel heating zone and the real-time temperature of the die of the extruder at the current moment, and constructing the current temperature control range of the equipment based on the real-time temperature of the barrel heating zone and the real-time temperature of the die, includes: when the extruder is in an idle standby state, reading the environmental cooling curve of the real-time temperature of the barrel heating zone and the heat dissipation gradient of the real-time temperature of the die; Based on the ambient cooling curve and the heat dissipation gradient, the minimum sustaining temperature threshold in the current temperature control range of the equipment is constructed. When the real-time temperature of the barrel heating zone is lower than the minimum sustaining temperature threshold, the preheating command of the extruder is triggered, and the lower limit of the current temperature control range of the equipment is updated to the minimum sustaining temperature threshold.
9. The intelligent production scheduling method for flexible manufacturing of robotic cables according to claim 5, characterized in that, The process of executing the neighborhood search strategy further includes: monitoring the completion state parameters of the first mutually exclusive process node in the fitness function of the discrete particle swarm algorithm; When the completion status parameter indicates that the processing task corresponding to the first mutually exclusive process node is completed on the candidate device node, calculate the cooling time and material cleaning time of the candidate device node; After a waiting period equal to the sum of the cooling time and the material cleaning time, the mutex flag is switched from the locked state to the released state, allowing the second mutex process node to acquire the allocation permission of the candidate device node in subsequent iterations.
10. The intelligent production scheduling method for flexible manufacturing of robotic cables according to claim 6, characterized in that, When the discrete particle swarm optimization algorithm reaches the maximum number of iterations or the global optimal fitness value converges, the current global optimal particle code is extracted, including: setting a stall iteration counter for the global optimal fitness value, and comparing the absolute value of the difference between the current global optimal fitness value and the previous global optimal fitness value after each iteration. When the absolute value of the difference is less than the preset perturbation threshold, the stagnation iteration counter is incremented by 1. When the value of the stagnation iteration counter reaches the preset stagnation threshold, it is determined that the global optimal fitness value has converged. At this time, the current global optimal particle code is directly extracted, and the remaining iteration steps of the discrete particle swarm algorithm are skipped.