Lithium battery Mylar film defect detection method
Technical Field
The invention relates to the field of material detection, in particular to a defect detection method for a Mylar film of a lithium battery.
Background
In recent years, the new energy automobile industry in China is rapidly developed, the innovation and the application of the battery technology in the industrial field are promoted, and the lithium battery is the battery type mainly used by the new energy automobile at present. In the production process, naked electric core needs the cladding one deck Mylar membrane to avoid the aluminum hull to cause the damage to naked electric core before filling in the aluminum hull. Therefore, before the battery is placed into the shell, the defect detection of the Mylar film coated on the battery is necessary.
Patent CN213989009U (a square lithium battery Mylar film and a lithium battery) discloses a structural design method of a square lithium battery Mylar film, which realizes the full wrapping protection of the Mylar film on an electric core, and improves the short circuit risk of the contact between a tab and an aluminum shell caused by the shedding of the existing hot-melt bonded Mylar film.
Patent CN213660510U (a secondary battery Mylar film structure and electric core) discloses a secondary battery Mylar film structure and electric core, including the Mylar film of rectangular structure, which plays the role of reducing the manufacturing process of electric core, reducing cost and ensuring electric core safety.
Patent CN209418561U (Mylar film structure, lithium battery assembly structure and lithium battery for lithium battery) provides a Mylar film structure, lithium battery assembly structure and lithium battery for lithium battery, which plays a good role in protecting battery core.
In summary, only the Mylar film structure design is discussed in the industrial field at the present stage. In the process of coating the battery with the Mylar film, if the Mylar film has defects or the Mylar film is damaged due to equipment of production equipment, the hidden danger of damage of the battery core exists when the battery core is plugged into the aluminum shell. Therefore, the defect detection method has research significance for defect detection after the Mylar film is coated, and the invention provides a systematic solution for defect detection of the Mylar film.
Disclosure of Invention
The invention provides a Mylar film defect detection method, aiming at solving the problem of hidden danger of electric core damage caused by defects of the Mylar film or damage of the Mylar film due to equipment of production equipment in the process of coating a battery with the Mylar film in the prior art.
A method for detecting the defects of a lithium battery M deaf-mute membrane comprises a detection process after a monomer area is set and a Mylar membrane defect detection process.
Preferably, the monomer region detection process includes the following steps:
step SA 1: receiving monomer area detection method process data;
step SA 2: setting a detection area m according to the process data;
step SA 3: setting a detection parameter n according to the process data;
step SA 4: carrying out defect region segmentation by a defect region segmentation method through detecting parameters;
step SA 5: and setting a defect judgment standard according to the divided defect area data.
Through carrying out flow detection to monomer region, detect through a plurality of monitoring regions, every monitoring region sets up different detection parameters and carries out the defect and cuts apart, reduces because of the defect of Mylar membrane leads to the impaired possibility of lithium cell electric core in process of production.
Preferably, the defective region dividing method includes:
the method comprises the following steps: a fixed threshold segmentation method;
the method 2 comprises the following steps: a dynamic threshold segmentation method;
the method 3 comprises the following steps: a deep learning approach is applied.
Dividing gray change difference defects by a fixed threshold dividing method, dividing point-type defects by a dynamic threshold dividing method, and dividing defect types which cannot be solved by the fixed threshold dividing method and the dynamic threshold dividing method by applying a deep learning method; the defect area of the Mylar film is segmented, and the application of deep learning makes up for the defect area segmentation of the traditional image algorithm with unobvious characteristics, so that the detection capability is greatly improved.
Wherein, the deep learning comprises the following contents:
step SC 1: collecting various related information of the Mylar film design, wherein the information comprises the flatness, transparency and mechanical flexibility of the film surface and is used for training a neural network for deep learning;
step SC 2: calculating constraint conditions of various parameter settings of the Mylar film design, and determining the constraint conditions of various parameter settings and loss conditions under unsatisfied conditions by taking the purposes that the Mylar film cannot be damaged due to defects or equipment of production equipment in the use process of the Mylar film and the hidden danger of electric core damage exists when an electric core is plugged into an aluminum shell;
step SC 3: establishing a parameter model of defects generated in the use process of the Mylar film by using the data and the use model provided by the step SC 2;
step SC 4: and verifying the defect parameters of the Mylar film according to the deep learning completion and the learning neural network algorithm thereof to obtain an accurate data threshold value for monitoring the defect parameters, and recording defect region segmentation with unobvious features and storing the features except for data abnormality in a defect state.
Preferably, the Mylar film defect detection process comprises:
step SB 1: setting an image monitoring area n;
step SB 2: setting detection parameters of the region n;
step SB 3: performing defect segmentation on the region n according to set detection parameters to obtain a connected domain;
step SB 4: setting a defect detection standard of the n monomers in the region;
step SB 5: judging whether the connected domain after the region n is divided meets the detection standard or not, if so, judging the quality of the Mylar film is NG and exiting the process; if not, repeating the steps 1-5 for the rest detection areas and setting corresponding parameters;
step SB 6: obtaining a connected domain from each region and merging the connected domains;
step SB 7: setting detection conditions of intensive defects, and performing cluster analysis;
step SB 8: if the detection condition is met, the quality of the Mylar film is judged to be NG; if not, the Mylar film quality defect is determined;
where n is a natural number, n =1,2,3, … n.
Specific detection areas and detection parameters are set according to all parts of the Mylar film, so that the influence of a polishing mode on an image in the actual detection process is reduced, the method is more flexible, and the monitoring efficiency is greatly improved compared with the traditional Mylar film monitoring method.
Preferably, step SB5 includes:
step SB 51: after the detection of each detection area is finished, performing BLOB analysis on the obtained connected domain to obtain various data of the detection area, and comparing the data with the area detection parameters set by the corresponding area;
step SB 52: if one detection area meets the judgment condition, the monomer defect is considered to exist; otherwise, if the detection of all the areas is finished and the judgment condition of a single detection defect is not met, combining the connected domains obtained from all the detection areas, carrying out sequential cluster analysis judgment, and if the detection condition is met, judging that the intensive defect type exists, and if the detection condition is not met, judging that the quality of the Mylar film is qualified.
Preferably, each item of data of the detection area comprises the height, width, area, roundness and aspect ratio of the detected object.
The detected target data are used as parameter comparison parameters of the deep learning method, so that the influence of a polishing mode on the image in the actual detection process is reduced, the method is more flexible, and the monitoring efficiency is greatly improved compared with the traditional Mylar film monitoring method.
Therefore, the invention has the following beneficial effects:
specific detection areas and detection parameters can be set according to all parts of the Mylar film, so that the influence of a polishing mode on an image in the actual detection process is reduced;
through three segmentation methods, the defect region of the Mylar membrane is segmented, and the deep learning is matched with the application of a neural network method, so that the defect region with unobvious characteristics is segmented by the traditional image algorithm, the integral detection capability is greatly improved, and the detection efficiency is improved;
the method is flexible, and if a better detection algorithm is developed subsequently, the method can be introduced as a new segmentation method through a deep learning method, so that algorithm iteration is facilitated.
Drawings
FIG. 1 is a flow chart of the detection process after the monomer area of the Mylar film of the invention is set;
FIG. 2 is an overall flow chart of the Mylar film defect detection of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and the following detailed description.
The invention provides a method for detecting the defects of a lithium battery M deaf-mute membrane, which comprises a detection process after a monomer area is set and a Mylar membrane defect detection process.
As shown in fig. 1, the monomer region detection process preferably includes the steps of:
step SA 1: receiving monomer area detection method process data;
step SA 2: setting a detection area m according to the process data;
step SA 3: setting a detection parameter n according to the process data;
step SA 4: carrying out defect region segmentation by a defect region segmentation method through detecting parameters;
step SA 5: and setting a defect judgment standard according to the divided defect area data.
Through carrying out flow detection to monomer region, detect through a plurality of monitoring regions, every monitoring region sets up different detection parameters and carries out the defect and cuts apart, reduces because of the defect of Mylar membrane leads to the impaired possibility of lithium cell electric core in process of production.
The method for dividing the defect area comprises the following steps:
the method comprises the following steps: a fixed threshold segmentation method;
the method 2 comprises the following steps: a dynamic threshold segmentation method;
the method 3 comprises the following steps: a deep learning approach is applied.
Dividing gray change difference defects by a fixed threshold dividing method, dividing point defects by a dynamic threshold dividing method, and dividing defect types which cannot be solved by the fixed threshold dividing method and the dynamic threshold dividing method by applying a deep learning method; the defect area of the Mylar film is segmented, and the application of deep learning makes up for the defect area segmentation of the traditional image algorithm with unobvious characteristics, so that the detection capability is greatly improved.
Wherein, the deep learning comprises the following contents:
step SC 1: collecting various related information of the Mylar film design, wherein the information comprises the flatness, transparency and mechanical flexibility of the film surface and is used for training a neural network for deep learning;
step SC 2: calculating constraint conditions of various parameter settings of the Mylar film design, and determining the constraint conditions of various parameter settings and loss conditions under the unsatisfied conditions by taking the purposes that the Mylar film is not damaged because of defects or equipment of production equipment in the use process of the Mylar film and the hidden danger of damage of the battery core exists when the battery core is plugged into the aluminum shell as the target;
step SC 3: establishing a parameter model of defects generated in the use process of the Mylar film by using the data and the use model provided by the step SC 2;
step SC 4: and verifying the defect parameters of the Mylar film according to the deep learning completion and the learning neural network algorithm thereof to obtain an accurate data threshold value for monitoring the defect parameters, and recording defect region segmentation with unobvious features and storing the features except for data abnormality in a defect state.
Meanwhile, after a period of deep learning neural network training, prediction model training can be carried out on the Mylar film subjected to preliminary segmentation detection, and the method comprises the following steps:
step SD 1: the initial condition setting unit is used for determining the content and the number of the influence factors, determining the number of variables of the output layer, setting initial conditions such as time scale and the like, inputting the initial parameters into the training model and predicting the defect parameters of various mechanical indexes of the Mylar film;
step SD 2: the training condition confirming unit is used for setting a target function and a constraint condition and determining a training end condition and a target requirement of model training;
step SD 3: the prediction model output unit is used for finishing training when the training result meets the training target requirement and outputting a Mylar film defect detection training design model;
step SD 4: and the retraining unit is used for adjusting the parameter variables to retrain if the implementation result does not meet the training target requirement, and returning to the prediction model building module to perform redesign if the implementation result cannot meet the training target requirement for many times.
As shown in fig. 2, the Mylar film defect detection process includes:
step SB 1: setting an image monitoring area n;
step SB 2: setting detection parameters of the region n;
step SB 3: performing defect segmentation on the region n according to set detection parameters to obtain a connected domain;
step SB 4: setting a defect detection standard of the n monomers in the region;
step SB 5: judging whether the connected domain after the region n is divided meets the detection standard or not, if so, judging the quality of the Mylar film is NG and exiting the process; if not, repeating the steps 1-5 for the rest detection areas and setting corresponding parameters;
step SB 6: obtaining a connected domain from each region and merging the connected domains;
step SB 7: setting detection conditions of intensive defects, and performing cluster analysis;
step SB 8: if the detection condition is met, the quality of the Mylar film is judged to be NG; if not, the Mylar film quality defect is determined;
where n is a natural number, n =1,2,3, … n.
Specific detection areas and detection parameters are set according to all parts of the Mylar film, so that the influence of a polishing mode on an image in the actual detection process is reduced, the method is more flexible, and the monitoring efficiency is greatly improved compared with the traditional Mylar film monitoring method.
Wherein step SB5 includes:
step SB 51: after the detection of each detection area is finished, performing BLOB analysis on the obtained connected domain to obtain various data of the detection area, and comparing the data with area detection parameters set in the corresponding area;
step SB 52: if one detection area meets the judgment condition, the monomer defect is considered to exist; otherwise, if the detection of all the areas is finished and the judgment condition of a single detection defect is not met, combining the connected domains obtained from all the detection areas, carrying out sequential cluster analysis judgment, and if the detection condition is met, judging that the intensive defect type exists, and if the detection condition is not met, judging that the quality of the Mylar film is qualified.
Each item of data of the detection area comprises the height, width, area, roundness and aspect ratio of the detected target.
The detected target data are used as parameter comparison parameters of the deep learning method, so that the influence of a polishing mode on the image in the actual detection process is reduced, the method is more flexible, and the monitoring efficiency is greatly improved compared with the traditional Mylar film monitoring method.
The structure, features and effects of the present invention have been described in detail with reference to the embodiments shown in the drawings, but the above embodiments are merely preferred embodiments of the present invention, and it should be understood that technical features related to the above embodiments and preferred modes thereof can be reasonably combined and configured into various equivalent schemes by those skilled in the art without departing from and changing the design idea and technical effects of the present invention; therefore, the invention is not limited to the embodiments shown in the drawings, and all the modifications and equivalent embodiments that can be made according to the idea of the invention are within the scope of the invention as long as they are not beyond the spirit of the description and the drawings.