Intelligent optimization solution for laboratory testing and allocation of lithium battery industry

For more information, please pay attention to the public accounts ("Smart Manufacturing and Smart Scheduling" and "Shangjian Smart")!

1 What is the allocation problem

In the scene of laboratory sample testing, the batch setting of battery cell samples and equipment channels is called arrangement. Samples may need to undergo different types of tests. Each test has certain requirements for temperature and current, and there is a sequence relationship between test items. It is necessary to reasonably allocate each test task to a certain channel of each channel through intelligent optimization calculations. In the time period, the output is accurate to hours, minutes and seconds, and fine-grained experimental plans for each channel. Under the premise of meeting the above constraints, optimize channel utilization, order delivery completion rate, and experimental test cycle.

2 Arrangement features

• Large amount of tasks : more than 150 test tasks per day;
• Many equipment channels : involving two types of equipment, incubator and incubator, with more than 4000 channels;
• Various channel capabilities : high-standard channels can be compatible with low-specification tests;
• Compatible temperature range : the test task meets certain temperature requirements, for variable temperature, a variable temperature box is required;
• Strong failure uncertainty : failure may occur at any time during the experimental stage;
• Long cycle time : some test cycles require thousands of cycles, as long as several years or even decades.

3 Status Quo and Pain Points of Scheduling

• Paper test items : all test items are recorded and distributed by paper documents, and the process automation and transparency are low;
• Appointment time priority : arrange test tasks according to the principle of first-in-first-out (FIFO), lack of optimization;
• Inspection channel status : It is necessary to manually inspect the status of the channel. Once it is found to be idle, the test task will be arranged, and the real-time response is slow;
• Manual channel matching : Manually check the availability of channels, and the matching workload is huge.

4 Intelligent Arrangement Solution

4.1 Overall Architecture

The following figure shows the overall architecture of the intelligent scheduling engine. It takes resources, capabilities, samples, test items, test time, test requirements, and fixtures as inputs, and is based on AI algorithms and operations research models to provide time-based and accurate time-based information based on limited capabilities. Second channel operation plan, output test item plan, sample collection plan, sample test plan, test item delivery plan, etc., supplemented by feedback on test progress and dynamic factors, to achieve a closed loop of test plan release, execution, and feedback. In addition, the system supports a variety of different production scheduling strategies (front row, reverse row, lock, order insertion, order splitting, order combination, etc.), rich visual interfaces (Gantt chart, plan comparison, channel monitoring, etc.), algorithm library (heuristics, metaheuristics, mathematical programming, hyperheuristics, deep learning, reinforcement learning, etc.), which can optimize multiple objectives simultaneously.

4.2 General idea

Based on the three-layer structure system of order form (order)-sample (work order)-test item (process) and area (temperature)-equipment (current)-channel, each test item is regarded as the test process route of the sample, and the arrangement will be The problem is transformed into a scheduling problem, selecting the appropriate channel for each test item.

4.3 Intelligent Arrangement Constraints

• Channel Availability : Through the resource calendar, the allocation management of whether the equipment channel is available can be realized. If there is a failure or maintenance is required, it is necessary to set the time for the channel to return to normal, and this time period will be automatically skipped during scheduling.
• Temperature and range constraints : temperature and range are used as constraints on the available equipment channels in the test item process data.
• Fixture constraints : Fixtures are used as auxiliary resources to avoid being occupied by different test items at the same time.
• The principle of area proximity : assign values ​​to equipment locations, and give priority to the combination of the closest channels when arranging.
• Test duration : Determine the test duration according to the test requirements and test item standards, and occupy the channel according to the test duration.
• Sequence of test items : sequence of use of device channels: short-term > performance > security > storage > long-term.

4.4 Intelligent Algorithm Helps Intelligent Arrangement

In the battery world, testing is a critical step in ensuring battery performance and reliability. However, traditional test scheduling often relies on experience and heuristic methods, and cannot fully consider factors such as variable test tasks, channel constraints, and priorities. The battery test scheduling solution based on deep reinforcement learning realizes intelligent test scheduling optimization by using deep neural network and reinforcement learning algorithm .

Deep reinforcement learning is a machine learning method that learns optimal decision-making strategies through interaction with the environment. In battery test scheduling, deep reinforcement learning can learn information such as the dynamic characteristics of test tasks, resource allocation methods, and test urgency , so as to fully consider the complex relationship of multiple factors and optimize test scheduling in continuous learning and iteration , gradually improve the test efficiency and quality, so that the test task can be completed in the shortest time, and channel resources can be fully utilized. At the same time, the learned strategy is flexible and scalable, and can respond to changes in different test scenarios and requirements.

4.5 Intelligent + Interactive Arrangement

• More flexible test task adjustment, which can adjust the order of scheduled test items and cancel tasks ;

• Support multiple drag-and-drop operations, automatically propagate constraints after saving, and generate feasible arrangement results;

• For drag and drop operations that violate time constraints and equipment availability constraints, provide relevant warning prompts ;

• Support the combination of manual and automatic, only need to complete the drag and drop of some test items, the system can intelligently rearrange the remaining test items to improve operational efficiency.

Battery row with drag

4.6 Real-time scheduling response

When disturbances occur on the experimental site, such as emergency insertion, channel abnormality, test failure, etc., the system adopts efficient decision-making algorithms and scheduling strategies to make corresponding decisions and adjustments in a short time to ensure full utilization of equipment channels.

5 Implementation Value of Intelligent Scheduling

Intelligent scheduling adopts artificial intelligence and optimization algorithm technology, which greatly improves the efficiency and value of battery testing, and can bring value-added to enterprises in the following aspects:

• Maximize the use of channel resources : intelligent scheduling can reasonably schedule and allocate channels according to channel availability and test requirements, maximize the use of equipment channels, and increase channel utilization by more than 5% ;
• Improve test efficiency : intelligent scheduling can realize parallel execution of test tasks and optimization of time allocation by optimizing test task arrangement and channel utilization, which can greatly improve test efficiency and shorten test cycle by more than 15 % ;
• Improve delivery rate : intelligent scheduling can rationally allocate test resources according to the priority and time requirements of test tasks, and adjust scheduling plans in a timely manner based on real-time feedback information, and deal with delayed or priority-changed tasks to ensure delivery Achievement rate of 100% ;
• Improve management transparency : Intelligent scheduling can intelligently generate channel plans with one click, and managers can intuitively understand the future conditions of all channels, including channel utilization, channel idle time, test task delay , etc., thereby improving management transparency.