CN-122017005-A - Online detection method and system for defects of cigarette lining paper based on eddy current

CN122017005ACN 122017005 ACN122017005 ACN 122017005ACN-122017005-A

Abstract

The invention discloses an on-line detection method and system for defects of cigarette lining paper based on eddy current, and relates to the technical field of cigarette detection; extracting the characteristics of the voltage signals to obtain characteristic vectors, inputting the characteristic vectors into a pre-trained defect recognition model, outputting defect types and confidence of the defect types, and marking cigarettes according to defect recognition results. The scheme of the invention can accurately identify, locate and accurately remove the defects of the lining paper of the tobacco package made of aluminum foil.

Inventors

Mu Pingli
WANG LEI
WANG QIAOJUAN
ZHAO QI
XIE JUNJUN
SHI XIANGYANG
XU YUJIANG
LI CHUNXIA
WANG XIAOTING
WU ZHENHUA
QI LIPING

Assignees

河南中烟工业有限责任公司

Dates

Publication Date: 20260512
Application Date: 20260306

Claims (10)

1. An on-line detection method for defects of cigarette lining paper based on electric vortex is characterized by comprising the following steps: Collecting voltage signals of tobacco bale lining paper through a double-probe transmission type electric vortex sensor; Extracting the characteristics of the voltage signals to obtain characteristic vectors; Inputting the feature vector into a pre-trained defect recognition model, and outputting the defect type and the defect type confidence; and marking the cigarettes according to the defect identification result.
2. The method for online detection of defects of cigarette lining paper based on eddy current according to claim 1, wherein the step of extracting features of the voltage signal to obtain feature vectors comprises the steps of: Preprocessing the voltage signal to obtain a standard voltage signal; Uniformly sampling the standard voltage signal in a preset frequency band to obtain N frequency points; and extracting the amplitude values of the N frequency points to obtain an N-dimensional feature vector with time domain statistical features.
3. The method for online detection of defects of cigarette lining paper based on eddy current according to claim 1, wherein inputting the feature vector into a pre-trained defect recognition model, outputting the defect type and the confidence comprises: extracting a local frequency domain characteristic value of the characteristic vector by the defect identification model; mapping the local frequency domain characteristic values to obtain defect types; and calculating the normalized probability of the defect type to obtain the confidence coefficient of the defect type.
4. The method for online detection of defects in cigarette lining paper based on eddy currents according to claim 1, wherein the presence of defects in the cigarette lining paper of any defect type is determined when the confidence level of the defect type is greater than a preset threshold.
5. The method for online detection of defects of cigarette lining paper based on eddy current according to claim 1, wherein the defect types comprise missing, broken and folded.
6. The method for online detection of defects in cigarette lining paper based on eddy current according to claim 1, further comprising: when the defective cigarette packet with the mark reaches the rejecting station, the rejecting mechanism rejects the defective cigarette packet with the mark from the conveying channel.
7. The method for online detection of defects of cigarette lining paper based on eddy current according to claim 1, wherein the training process of the defect recognition model comprises the following steps: Collecting detection signals of historical normal cigarette packets and detection signals of historical defective cigarette packets; Carrying out frequency domain feature extraction on the detection signals of the history normal tobacco packages and the detection signals of the history defective tobacco packages after fast Fourier transformation to obtain a training data set with labels; inputting the training data set into a one-dimensional convolutional neural network for model training, and performing parameter optimization by using a cross entropy loss function to obtain a defect identification model.
8. An eddy current based cigarette liner defect online detection system, wherein a method as claimed in any one of claims 1 to 7 is performed, the system comprising: a tobacco packet conveying channel (1); The eddy current detection module (2) is arranged on the tobacco bale conveying channel (1); a rejecting mechanism (3) arranged on the tobacco bale conveying channel (1); the eddy current detection module (2) and the rejection mechanism (3) are electrically connected with the programmable logic controller.
9. The eddy current-based cigarette lining paper defect online detection system according to claim 8, wherein the eddy current detection module (2) adopts a double-probe transmission type eddy current sensor, is symmetrically arranged at two sides of a cigarette packet conveying channel (1) and is used for emitting an alternating magnetic field penetrating through an aluminum foil at the outer layer of the cigarette packet and receiving a magnetic field signal modulated by lining paper.
10. The eddy current based cigarette liner paper defect online detection system of claim 8, further comprising: and the removing and collecting box (4) is fixed at one side of the tobacco bale conveying channel (1).

Description

Online detection method and system for defects of cigarette lining paper based on eddy current Technical Field The invention relates to the technical field of cigarette detection, in particular to an on-line detection method and system for defects of cigarette lining paper based on eddy current. Background The tobacco lining paper is used as the initial packing layer of the cigarettes, and the integrality of the tobacco lining paper directly influences the fragrance retention, moisture resistance and appearance quality of the cigarette products. The lining paper is wrapped by more than ten packing stations of raw materials of coiled bobbin dress through processes such as indentation, cutting, transportation, folding, in the process of wrapping the lining paper, because the compact space of packagine machine die box design is narrow, detection device can't cover whole production flow by a full process, this just leads to lining paper to tear the tongue and miss, skew and the damaged phenomenon of lining paper that cause when a plurality of stations cut, fold can't effectively be controlled to the dampproofing, the mould proof of lining paper has been lost and the effect of preventing cigarette fragrance loss has been lost. The method has the advantages that the method has a certain effect, but has obvious defects that the shielding problem is that along with product upgrading, aluminum foil is coated on the surface of the outer hard box packaging surface of many cigarette brands, high-frequency electromagnetic fields are difficult to penetrate through the aluminum foil layer, so that an embedded inductance sensor is invalid and cannot detect the lining paper, the detection blind area is limited by the inner space of a packaging machine, the sensor cannot cover the whole area of the lining paper of a cigarette packet, so that defects such as folding, scratching and local breakage cannot be detected, and meanwhile, the traditional switching value sensor only can detect metal, can not quantitatively identify the form and the size of the defects and is insensitive to the tiny defects. Disclosure of Invention In view of the above, the present invention aims to provide an online detection method and system for defects of a cigarette lining paper based on eddy current, so as to solve the above-mentioned technical problems. The technical scheme adopted by the invention is as follows: the invention provides an on-line detection method for defects of cigarette lining paper based on eddy current, which comprises the following steps: Collecting voltage signals of tobacco bale lining paper through a double-probe transmission type electric vortex sensor; Extracting the characteristics of the voltage signals to obtain characteristic vectors; Inputting the feature vector into a pre-trained defect recognition model, and outputting the defect type and the defect type confidence; and marking the cigarettes according to the defect identification result. Optionally, extracting features of the voltage signal to obtain a feature vector, including: Preprocessing the voltage signal to obtain a standard voltage signal; Uniformly sampling the standard voltage signal in a preset frequency band to obtain N frequency points; and extracting the amplitude values of the N frequency points to obtain an N-dimensional feature vector with time domain statistical features. Optionally, inputting the feature vector into a pre-trained defect recognition model, outputting the defect type and the confidence, including: extracting a local frequency domain characteristic value of the characteristic vector by the defect identification model; mapping the local frequency domain characteristic values to obtain defect types; and calculating the normalized probability of the defect type to obtain the confidence coefficient of the defect type. Optionally, when the confidence of any defect type is greater than a preset threshold, determining that the cigarette lining paper has the defect of the type. Optionally, the defect type includes a defect, a breakage, a fold. Optionally, the method for online detecting defects of the cigarette lining paper based on the eddy current further comprises the following steps: when the defective cigarette packet with the mark reaches the rejecting station, the rejecting mechanism rejects the defective cigarette packet with the mark from the conveying channel. Optionally, the defect recognition model training process includes: Collecting detection signals of historical normal cigarette packets and detection signals of historical defective cigarette packets; Carrying out frequency domain feature extraction on the detection signals of the history normal tobacco packages and the detection signals of the history defective tobacco packages after fast Fourier transformation to obtain a training data set with labels; inputting the training data set into a one-dimensional convolutional neural network for model training, and performing