The present invention provides an adaptive scheduling method for a robot manufacturing unit based on an XGBoost algorithm, belongs to the field of automated scheduling of a production line in the manufacturing industry, and in particular relates to a method for performing real-time online scheduling on a robot manufacturing unit with complex constraints by using the XGBoost algorithm. The method comprises the following steps: establishing a sample database, wherein samples stored in the database are production data and an optimal heuristic method for the corresponding production data; applyinga FIPS-XGBoost algorithm heuristic scheduling method classification model based on the sample database; and by monitoring production state information in the robot manufacturing unit, obtaining actual production data, and inputting the actual production data to the heuristic scheduling method classification model to obtain the optimal heuristic method. According to the method provided by the present invention, the optimal feature subset can be quickly and efficiently solved, the performance of the algorithm is generally better than the combined scheduling rule, and the effectiveness of the adaptive scheduling of the algorithm is verified; and by using the classification model provided by the present invention, the optimal heuristic method is obtained according to the actual production data, so that the number of times of no-load movements of the robot can be reduced, the scheduling efficiency can be accelerated, and the throughput within the unit can be improved.

Adaptive scheduling method for robot manufacturing unit based on XGBoost algorithm

Provided in the present invention is a high-throughput material simulation calculation optimisation method based on time prediction, relating to the field of materials science. The method comprises: first constructing a prediction model of task configurations and corresponding time predictions, and using the prediction model to generate the execution time of all of the tasks in a high-throughput material simulation calculation under different conditions; then generating an optimal scheduling plan for each model in the high-throughput material simulation calculation by means of directed graphs; and, according to the optimal scheduling plan for each model, sequentially executing all of the tasks until all of the tasks are completed. The present invention combines the characteristics of high-throughput material calculation and simulation tasks, greatly increasing the efficiency of executing high-throughput tasks and reducing the time needed to design new materials.

A high-throughput material simulation calculation optimization method based on time prediction

The embodiment of the invention relates to a neural network scheduling method and system. The method includes: loading a trained MV-net model, and generating a corresponding precision table and a prediction table; obtaining the performance/quality information data of the system by the scheduling algorithm; obtaining the CPU and GPU cache loss rate; according to the performance model and schedulingalgorithm, under the constraints of QoS and QoR defined by users, selecting the appropriate branch as the actual output of the neural network. The MV-net can be dynamically reconfigured by the scheduling algorithm when parallel processes cause resource occupancy to satisfy the QoS constraints in different applications. The scheduling algorithm of the invention can dynamically configure the most reasonable MV-net branch according to the constraint requirements under the constraint of simultaneously satisfying the QoS and the QoR, avoiding problems that the traditional network mapping in the hardware computation time is too long and the result real-time request of users cannot be guaranteed.

Neural network scheduling method and system

The invention relates to a three-dimensional simulation experiment method based on data driving, and the method comprises the steps: determining the positions of a first experiment body and a second experiment body, and building the attribute list texts of the first experiment body and the second experiment body; when an experiment is carried out, determining whether the relative position relationbetween the first experiment body and the second experiment body is correct or not according to the real-time attribute list text change of the first experiment body and the second experiment body; if the relative position relationship between the first experiment body and the second experiment body is correct, continuing the experiment; if the relative position relationship between the first experimental body and the second experimental body is incorrect, adjusting the attribute list text of the first experimental body or the second experimental body, so that the relative position relationship between the first experimental body and the second experimental body is correct. The second experiment body is dynamically adjusted according to the attribute list sequence of the second experimentbody, so that the relative positions of the first experiment body and the second experiment body are accurate in the experiment process, the experiment error is reduced, and the accuracy of the experiment result is ensured.

Three-dimensional simulation experiment method based on data driving

The invention discloses a job shop dynamic scheduling method. The method module comprises an algorithm-based simulation platform and a sample data acquisition module. The output terminal of the algorithm-based simulation platform is a sample data acquisition module; the output terminal of the sample data acquisition module is provided with an intelligent dynamic scheduling module and an offline data dynamic processing module. The output terminal of the offline data dynamic processing module is a reserve database; a group of input terminals of the reserve database are manual adjusting terminals; the output terminal of the reserve database is provided with a cloud processing module and an intelligent dynamic scheduling module. The cloud processing module and the reserve database are input and output terminal for each other, the output terminal of the cloud processing module is cloud manual adjustment, and the cloud manual adjustment and the cloud processing module are input and output terminals for each other, so that the consistency with real-time dynamic variables of a workshop after processing is ensured, and the accuracy after processing can be improved.

Job shop dynamic scheduling method

The invention provides a production-data-driven dynamic job-shop scheduling rule intelligent selection method and belongs to the manufacturing enterprise job shop production planning and scheduling application field. The method mainly comprises the following steps: introducing a Multi-Pass algorithm simulation mechanism, establishing a job-shop production scheduling simulation platform, and generating production planning and scheduling sample data; screening the obtained sample data and generating a scheduling parameter set; designing BP neural network models for scheduling knowledge learningunder different scheduling targets; optimizing training of the BP neural networks through a new firefly algorithm to obtain NFA-BP models; integrating the NFA-BP models under various scheduling targets into an intelligent scheduling module, which is integrated with a job shop MES system to guide on-line scheduling; manually adjusting online production planning and scheduling deviation and updatingthe scheduling parameter set, and the intelligent scheduling module carrying out online optimization learning; and the intelligent scheduling module adapted to real workshop production status outputting optimal scheduling rules according to current job conflict decision points.

Luo Zhiyong

Papers

Production-data-driven dynamic job-shop scheduling rule intelligent selection method