Search

CN-122016184-A - Workpiece seal detection method, model training method, electronic device, and storage medium

CN122016184ACN 122016184 ACN122016184 ACN 122016184ACN-122016184-A

Abstract

The application discloses a workpiece sealing detection method, a model training method, electronic equipment and a storage medium, wherein the method comprises the steps of obtaining sealing parameters related to a workpiece to be detected in a sealing process; inputting the sealing parameters into a target sealing quality prediction model, predicting the corresponding sealing qualification probability of the workpiece to be detected, and performing sealing detection on the workpiece to be detected based on the sealing qualification probability to obtain a sealing detection result corresponding to the workpiece to be detected.

Inventors

  • WANG ZHEN
  • TANG MIAO

Assignees

  • 惠州亿纬锂能股份有限公司

Dates

Publication Date
20260512
Application Date
20251205

Claims (13)

  1. 1. A method of detecting a seal of a workpiece, comprising: Acquiring sealing parameters related to a workpiece to be tested in a sealing process; inputting the sealing parameters into a target sealing quality prediction model, and predicting the sealing qualification probability corresponding to the workpiece to be tested; And performing seal detection on the workpiece to be detected based on the seal qualification probability to obtain a seal detection result corresponding to the workpiece to be detected.
  2. 2. The method according to claim 1, wherein the performing seal inspection on the workpiece to be inspected based on the seal qualification probability to obtain a seal inspection result corresponding to the workpiece to be inspected, includes: Determining a seal detection type corresponding to the workpiece to be detected based on the seal qualification probability; and performing seal detection on the workpiece to be detected according to the seal detection type to obtain a seal detection result corresponding to the workpiece to be detected.
  3. 3. The method of claim 2, wherein determining the seal inspection type corresponding to the workpiece to be inspected based on the seal eligibility probability comprises: When the seal qualification probability exceeds a preset qualification interval, determining that the seal detection type corresponding to the workpiece to be detected is non-standard seal detection; When the seal qualification probability is positioned in the prediction qualification interval, determining that the seal detection type corresponding to the workpiece to be detected is standard seal detection; And when the seal qualification probability does not reach the preset qualification interval, determining that the seal detection type corresponding to the workpiece to be detected is termination seal detection.
  4. 4. A method according to any one of claims 1 to 3, wherein the sealing process is a welding process, and the seal detection result is obtained by performing an air tightness detection on the workpiece to be detected in an air tightness detection process.
  5. 5. The method of any of claims 1-4, wherein the target seal quality prediction model is obtained by taking a second predicted seal eligibility probability output by a trained first seal quality prediction model as a second training label and model training based on the input first training seal characteristics and the second training label, the first seal quality prediction model being obtained by model training based on the input first training seal characteristics and first training seal detection characteristics.
  6. 6. The method according to any one of claims 1-5, further comprising: Acquiring first training sealing characteristics related to a sample workpiece in a sealing process, first training sealing detection characteristics related to a sealing detection process and a first training label; Performing iterative optimization on a first seal quality prediction model to be trained based on the first training seal characteristic, the first training seal detection characteristic and the first training label to obtain a first seal quality prediction model after training; outputting a second predicted seal qualification probability by the trained first seal quality prediction model aiming at the first training seal characteristic and the first training seal detection characteristic to serve as a second training label; And performing iterative optimization on the second seal quality prediction model to be trained based on the first training seal characteristic, the first training label and the second training label to obtain a target seal quality prediction model after training.
  7. 7. The method of claim 6, wherein the first training label comprises a seal-qualifying classification label and an actual seal leak rate, wherein iteratively optimizing a first seal quality prediction model to be trained based on the first training seal characteristic, the first training seal detection characteristic, and the first training label to obtain a trained first seal quality prediction model comprises: inputting the first training seal characteristic and the first training seal detection characteristic into a first main network and a second auxiliary network in the first seal quality prediction model to be trained; Obtaining a first predicted seal qualification probability through the first main network, and obtaining a predicted seal leakage rate through the second auxiliary network; Determining a first target model loss based on the seal eligibility classification tag, the first predicted seal eligibility probability, the actual seal leak rate, and the predicted seal leak rate; and carrying out iterative optimization on the first seal quality prediction model to be trained based on the first target model loss to obtain a first seal quality prediction model after training.
  8. 8. The method of claim 7, wherein iteratively optimizing the second seal quality prediction model to be trained based on the first training seal characteristic, the first training tag, and the second training tag to obtain a trained target seal quality prediction model comprises: Inputting the first training sealing characteristics into a second sealing quality prediction model to be trained to obtain a third sealing qualification probability; determining a second target model loss based on the seal eligibility classification tag, the second training tag, and a third seal eligibility probability; and performing iterative optimization on the second seal quality prediction model to be trained based on the second target model loss to obtain a target seal quality prediction model after training.
  9. 9. The method of any of claims 5-8, wherein prior to the acquiring the first trained seal feature of the sample workpiece involved in the sealing procedure, the first trained seal detection feature involved in the seal detection procedure, and the first trained label, the method further comprises: Determining an actual process time difference corresponding to each initial sample workpiece based on the seal end time and the seal detection start time respectively corresponding to the plurality of initial sample workpieces; determining an abnormal time sample workpiece from a plurality of initial sample workpieces based on the actual process time difference and the preset standard process time difference of each initial sample workpiece; and screening out the abnormal time sample workpieces from the initial sample workpieces to obtain sample workpieces.
  10. 10. The method of any of claims 5-8, wherein prior to the acquiring the first trained seal feature of the sample workpiece involved in the sealing procedure, the first trained seal detection feature involved in the seal detection procedure, and the first trained label, the method further comprises: Acquiring a reference seal ending time corresponding to a preselected workpiece, wherein the preselected workpiece is a workpiece with successful association of a workpiece identifier and seal data and failed association of the workpiece with seal detection data, the seal data is used for extracting seal characteristics, and the seal detection data is used for extracting seal detection characteristics; determining a target time search range based on the reference seal end time and a preset standard procedure time difference; acquiring a plurality of seal detection data with seal detection starting time within the target time search range, and determining target seal detection data from the plurality of seal detection data; and (3) associating the workpiece identification with the target seal detection data to be a preselected workpiece which is successful as a sample workpiece.
  11. 11. The seal quality prediction model training method is characterized by comprising the following steps of: Acquiring first training sealing characteristics related to a sample workpiece in a sealing process, first training sealing detection characteristics related to a sealing detection process and a first training label; Performing iterative optimization on a first seal quality prediction model to be trained based on the first training seal characteristic, the first training seal detection characteristic and the first training label to obtain a first seal quality prediction model after training; outputting a second predicted seal qualification probability by the trained first seal quality prediction model aiming at the first training seal characteristic and the first training seal detection characteristic to serve as a second training label; And performing iterative optimization on the second seal quality prediction model to be trained based on the first training seal characteristic, the first training label and the second training label to obtain a target seal quality prediction model after training.
  12. 12. An electronic device comprising a processor and a memory, the memory having stored therein a computer program configured to be executed by the processor to implement the steps in the method of any of claims 1 to 11.
  13. 13. A computer storage medium, characterized in that it stores a computer program configured to be executed by a processor to implement the method of any one of claims 1 to 11.

Description

Workpiece seal detection method, model training method, electronic device, and storage medium Technical Field The application relates to the technical field of seal detection, in particular to a workpiece seal detection method, a model training method, electronic equipment and a storage medium. Background In the manufacture of key components such as a power battery cover plate and a housing, the sealing performance of laser sealing welding, for example, is a core quality index for determining the safety and service life of a product. Currently, the industry commonly adopts a quality control mode of combining 'welding process sampling metallographic analysis' with 'helium mass spectrum leak detection equipment total detection'. However, this mode suffers from a systematic deficiency, all workpieces enter the helium mass spectrometer leak detection link with high accuracy but limited throughput indiscriminately, resulting in low seal detection efficiency for the workpieces. Disclosure of Invention Provided are a workpiece seal detection method, a model training method, an electronic device, and a storage medium, which can improve seal detection efficiency. In a first aspect, a method for detecting a seal of a workpiece is provided, including the steps of: Acquiring sealing parameters related to a workpiece to be tested in a sealing process; inputting the sealing parameters into a target sealing quality prediction model, and predicting the sealing qualification probability corresponding to the workpiece to be tested; And performing seal detection on the workpiece to be detected based on the seal qualification probability to obtain a seal detection result corresponding to the workpiece to be detected. In an exemplary embodiment, performing seal detection on a workpiece to be detected based on a seal qualification probability to obtain a seal detection result corresponding to the workpiece to be detected, including: Determining a seal detection type corresponding to the workpiece to be detected based on the seal qualification probability; And performing seal detection on the workpiece to be detected according to the seal detection type to obtain a seal detection result corresponding to the workpiece to be detected. In the embodiment, the seal detection type is determined according to the seal qualification probability of the workpiece to be detected, and corresponding seal detection processing is executed according to the seal detection type, so that the classification processing of the workpiece to be detected can be realized, the accurate seal detection of the workpiece is realized, and the seal detection efficiency and accuracy of the workpiece are improved. In an exemplary embodiment, determining a seal detection type corresponding to a workpiece to be tested based on a seal pass probability includes: when the seal qualification probability exceeds a preset qualification interval, determining that the seal detection type corresponding to the workpiece to be detected is non-standard seal detection; When the seal qualification probability is within the predicted qualification interval, determining that the seal detection type corresponding to the workpiece to be detected is standard seal detection; and when the seal qualification probability does not reach the preset qualification interval, determining that the seal detection type corresponding to the workpiece to be detected is termination seal detection. In the embodiment, the seal detection type is determined according to the seal qualification probability of the workpiece to be detected, and corresponding seal detection processing is executed according to the seal detection type, so that the classification processing of the workpiece to be detected can be realized, the accurate seal detection of the workpiece is realized, and the seal detection efficiency and accuracy of the workpiece are improved. In an exemplary embodiment, the sealing process is a welding process, and the sealing detection result is obtained by performing air tightness detection on the workpiece to be detected in the air tightness detection process. In one exemplary embodiment, the target seal quality prediction model is obtained by taking a second predicted seal eligibility probability output by the trained first seal quality prediction model as a second training label, and model training based on the input first training seal characteristics and the second training label, and the first seal quality prediction model is obtained by model training based on the input first training seal characteristics and the first training seal detection characteristics. In this embodiment, the first seal quality prediction model is trained by using the seal feature and the seal detection feature, so that the first seal quality prediction model can output accurate seal qualification probability according to the seal feature and the seal detection feature. The target seal quality prediction model is trai