A winding state monitoring device for PET sheet production based on visual detection

Abstract

The present application relates to PET sheet production equipment technical field, especially in a kind of based on visual inspection's PET sheet production and winding state monitoring equipment, comprising: winding mechanism, visual inspection device and processing unit;Visual inspection device includes structured light projector, image collector and laser vibration meter;Processing unit is configured as: according to the phase shift information of structured light in reflection image, the first feature information that characterizes the abnormal distribution of internal thickness of roll is generated;According to the characteristics of high-frequency component in vibration signal, the second feature information that characterizes the interface slip state in roll interior is generated;And based on the first feature information and second feature information, determine whether winding defect exists in roll interior;The present application can detect the abnormal thickness of initial layer of roll core and interface slip state in real time by structured light phase analysis and laser doppler vibration measurement, realize early warning and intervention of winding defect.

Description

Technical Field

[0001] This invention relates to the field of PET sheet production equipment technology, and in particular to a vision-based detection-based device for monitoring the winding status of PET sheets. Background Technology

[0002] PET (polyethylene terephthalate) sheets are widely used in packaging, electronics, medical and other fields due to their excellent mechanical properties, transparency and processing performance. In the production process of PET sheets, the winding process is one of the key links that determines the quality of the final product. The quality of the roll formed during the winding process directly affects the effect of subsequent slitting, storage and use.

[0003] Currently, existing PET sheet winding status monitoring equipment mainly adopts visual inspection technology. By installing cameras above or to the side of the winding machine, images of the outer surface or end face of the roll are collected, and image processing algorithms are used to identify surface defects, uneven end faces, exposed wrinkles, and other problems. This type of equipment is effective in detecting visible defects on the outer surface.

[0004] However, in practical applications of existing technology, during the initial winding stage, the ends of the PET sheet are fixed to the core surface using tape or suction. If the adhesion is weak or the initial tension is delayed, microscopic slippage (latency) or initial layer wrinkles will occur inside the core. Because these defects occur in the first few dozen layers of the roll and are tightly wrapped by the outer sheet, the optical cameras of existing vision devices cannot penetrate multiple layers of PET material for observation, resulting in defects being completely in a blind spot during the initial winding stage. During the winding process, the microscopic slippage or wrinkles in the initial layer will gradually evolve into chrysanthemum-like wrinkles, internal lumps, or end-capsule wrinkles as the roll diameter increases and radial pressure is applied. Severe defects such as surface misalignment occur; when these defects develop to the point where they are visible on the outer circumference, the entire roll of PET sheet, which is thousands of meters long, is scrapped due to internal damage, resulting in huge waste of raw materials and increased production costs; furthermore, existing equipment relies solely on optical images for judgment, and factors such as changes in ambient light, flying lint, and mechanical vibration at the winding site can easily lead to false detections or missed detections, making it difficult to meet the requirements of high-reliability production; in addition, existing technology cannot obtain information on the physical state at the interface between the core and the sheet (such as friction and micro-slippage), it can only "see" the surface and cannot "perceive" the internal dynamic process, thus it cannot intervene in the early stages of defect formation; In summary, how to achieve real-time and reliable monitoring of internal initial layer defects and interface slippage during the winding process of PET sheets is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] In order to overcome the problems of existing PET sheet winding equipment in application, such as inability to detect defects in the initial layer of the core, delayed defect detection, susceptibility to interference in single-modal detection, and insufficient reliability.

[0006] The technical solution of this invention is: a vision-based inspection device for monitoring the winding status of PET sheet production, comprising: A winding mechanism is used to wind PET sheets onto a core to form a roll; A visual inspection device is disposed on one side of the winding mechanism, and its inspection field of view covers the outer surface of the material roll; The processing unit is communicatively connected to the visual inspection device and is used to receive inspection data and determine the winding status. The visual inspection device includes: A structured light projector is used to project structured light with a preset pattern onto the outer surface of the roll. An image acquisition device is used to acquire the reflected image of the structured light after it has been reflected by the outer surface of the roll. A laser vibrometer is used for non-contact measurement of vibration signals on the outer surface of the material roll; The processing unit is configured as follows: Based on the phase shift information of the structured light in the reflected image, first feature information characterizing the abnormal thickness distribution inside the roll is generated; Based on the high-frequency component characteristics in the vibration signal, a second feature information characterizing the slip state of the internal interface of the material roll is generated; Based on the first feature information and the second feature information, it is determined whether there is a winding defect inside the material roll.

[0007] Preferably, when the processing unit generates the first feature information based on the phase shift information of the structured light in the reflected image, it is configured as follows: Extract the phase offset of multiple pixels in the reflected image; Based on a preset optical thickness-mechanical thickness mapping relationship, the phase offset is converted into a thickness anomaly value; Based on the distribution of the thickness anomalies on the surface of the roll, a thickness anomaly heatmap is generated, which is used to indicate the degree of thickness anomaly at different locations inside the roll.

[0008] Preferably, when the processing unit generates the second feature information based on the high-frequency component characteristics in the vibration signal, it is configured as follows: The vibration signal is decomposed into time and frequency components to separate high-frequency vibration components within a preset frequency band. The preset frequency band corresponds to the elastic wave frequency band generated by the stick-slip effect at the interface between the PET sheet and the core. Extract the energy characteristics and temporal characteristics of the high-frequency vibration components; Based on the energy and timing characteristics, a slip characteristic index is generated, which is used to characterize the severity of interface slippage inside the coil.

[0009] Preferably, when the processing unit determines whether there is a winding defect inside the roll based on the first feature information and the second feature information, it is configured to: The first feature information is compared with a first preset threshold to generate a first confidence level; The second feature information is compared with a second preset threshold to generate a second confidence level; When the first confidence level is greater than the first threshold and the second confidence level is greater than the second threshold, a serious winding defect is determined to exist; When the first confidence level is greater than the first threshold or the second confidence level is greater than the second threshold, it is determined that there is a potential winding defect.

[0010] Preferably, the vision inspection device is mounted on the follower arm, which is linked with the winding mechanism to maintain the working distance between the vision inspection device and the outer surface of the roll within a preset range during the winding process.

[0011] As a preferred option, it also includes: The linkage control interface is communicatively connected to the control system of the winding mechanism; The processing unit is further configured to: when a winding defect is determined to exist, send a control command to the winding mechanism through the linkage control interface, the control command including at least one of a speed reduction command, a pressure increase command, or a stop command.

[0012] As a preferred option, it also includes: A marking device is disposed on the discharge side of the winding mechanism and is communicatively connected to the processing unit; The processing unit is further configured to: when a winding defect is determined, determine the spatial coordinates of the defect location on the roll based on the first feature information, and control the marking device to apply a removable mark to the outer surface of the roll at the position corresponding to the spatial coordinates.

[0013] Preferably, the structured light projected by the structured light projector is coded structured light or sinusoidal striped structured light; the image acquisition device is a high frame rate infrared camera with a frame rate of not less than 1000 frames / second.

[0014] Preferably, the laser vibrometer is a laser Doppler vibrometer with a sampling frequency of not less than 100kHz.

[0015] Preferably, the processing unit is further configured as follows: After the winding process is completed, an internal quality report of the material roll is generated based on the first feature information and the second feature information generated during the winding process. The internal quality report includes three-dimensional distribution information of defect locations and timeline information of slippage events.

[0016] The beneficial effects of this invention are: 1. This invention projects structured light onto the surface of the PET roll using a structured light projector. Utilizing the semi-transparent properties of PET material and the phase modulation effect of its multi-layer structure on the light, the processing unit generates a thickness anomaly thermal map based on the phase shift information of the structured light in the reflected image. This enables non-destructive, real-time detection of internal thickness anomalies such as wrinkles and bubbles in the initial layer of the core. This detection principle solves the problem that existing technologies cannot penetrate multiple layers of PET sheet to observe internal defects, advancing the defect detection time to the moment the inner layer of the PET sheet is formed, thus realizing the detection of the internal state of the PET sheet during winding. 2. This invention introduces a laser vibrometer on the basis of visual inspection for non-contact measurement of vibration signals on the outer surface of the roll. By extracting high-frequency component features to generate a slip characteristic index, it realizes real-time monitoring of micro-slip (latent change) at the core interface. On this basis, the processing unit uses dual confidence and / or logic for fusion judgment: a serious defect is only determined when both structured light detection and vibration detection indicate an abnormality. This multi-modal redundancy mechanism avoids false alarms or missed alarms caused by a single sensor due to factors such as changes in ambient light, flying fluff, and mechanical impact, enabling the monitoring system to operate stably in complex production environments and improving detection reliability and anti-interference capability. 3. This invention communicates with the winding mechanism control system through a linkage control interface, and can automatically send speed reduction and pressure increase commands when potential defects are detected to suppress defect expansion; it can automatically send a stop command when serious defects are detected to avoid the continuous formation of waste rolls; at the same time, a removable mark is applied to the defect location by a marking device, providing a precise basis for rejection in the subsequent slitting process, realizing non-destructive screening, reducing raw material waste and production costs. 4. After winding is completed, the present invention generates an internal quality report containing three-dimensional distribution information of defect locations and timeline information of slippage events. This report can not only be used for quality judgment of a single roll, but also, through statistical analysis of multiple batches of data, identify the correlation between winding process parameters (such as initial tension, pressure roller pressure, acceleration curve) and the occurrence of internal defects, providing data support for continuous optimization of the production process. Attached Figure Description

[0017] Figure 1 The diagram shown is a three-dimensional structural schematic of the vision-based inspection-based PET sheet production winding status monitoring device of the present invention. Figure 2The diagram shown is a cross-sectional view of the vision-based inspection device for monitoring the winding status of PET sheet production according to the present invention. Figure 3 The diagram shown is a partial structural schematic of the vision-based inspection-based PET sheet production winding status monitoring device of the present invention. Explanation of reference numerals in the attached drawings: 1. Winding mechanism; 2. Visual inspection device; 3. Follower arm; 4. Marking device; 21. Structured light projector; 22. Image acquisition device; 23. Laser vibration meter; 31. Fixed base; 32. First swing arm; 33. Second swing arm; 34. End mounting base. Detailed Implementation

[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0019] Example 1: Please see Figure 1 , Figure 2 and Figure 3 This invention provides an embodiment: a vision-based inspection device for monitoring the winding status of PET sheet production, comprising: The winding mechanism 1 is used to wind PET sheets onto a core to form a roll; The visual inspection device 2 is located on one side of the winding mechanism 1, and its inspection field covers the outer surface of the material roll. The processing unit is communicatively connected to the vision inspection device 2 and is used to receive inspection data and determine the winding status. The visual inspection device 2 includes: Structured light projector 21 is used to project structured light with a preset pattern onto the outer surface of the roll; Image acquisition unit 22 is used to acquire the reflected image of structured light after it is reflected from the outer surface of the roll; Laser vibration meter 23 is used for non-contact measurement of vibration signals on the outer surface of material rolls; The processing unit is configured as follows: Based on the phase shift information of the structured light in the reflected image, the first feature information characterizing the abnormal thickness distribution inside the roll is generated; Based on the high-frequency component characteristics in the vibration signal, a second feature information characterizing the slip state of the internal interface of the material roll is generated. Based on the first and second feature information, it is determined whether there are winding defects inside the material roll.

[0020] Furthermore, when the processing unit generates the first feature information based on the phase shift information of the structured light in the reflected image, it is configured as follows: Extract the phase offset of multiple pixels in the reflected image; Based on the preset optical thickness-mechanical thickness mapping relationship, the phase offset is converted into a thickness anomaly value; Based on the distribution of thickness anomalies on the surface of the coil, a thickness anomaly heatmap is generated. The thickness anomaly heatmap is used to indicate the degree of thickness anomaly at different locations inside the coil.

[0021] The structured light projector 21 projects sinusoidal stripes or coded structured light onto the surface of the PET sheet. When the structured light shines on the PET sheet surface, due to the semi-transparent nature of the PET material, some light penetrates the surface layer and undergoes multiple reflections and refractions within the multilayer structure. When initial layer wrinkles, bubbles, or foreign objects are present inside the PET sheet, the thickness of the multilayer PET film in that area changes at the micrometer level, resulting in an optical thickness anomaly. This optical thickness anomaly causes a change in the optical path of the reflected light, which in turn causes a phase shift in the structured light in the reflected image. The image acquisition unit 22 acquires the reflected image at a high frame rate (≥1000 frames / second), and the processing unit extracts the phase shift of each pixel using a multi-frame phase shift algorithm. Since there is a definite physical mapping relationship between the optical thickness change and the mechanical thickness change (e.g., Δφ = ...), ... (4π / λ)·Δd·n, where Δφ is the phase shift, λ is the wavelength, Δd is the mechanical thickness change, and n is the refractive index of the PET material), the processing unit can convert the phase shift into a thickness anomaly value based on a preset calibration model; finally, the thickness anomaly value is mapped to the surface coordinate system of the roll according to the axial coordinate and circumferential angle to generate a thickness anomaly heat map; this heat map can intuitively display the location distribution and severity of internal defects, and realize the detection of initial layer wrinkles.

[0022] Furthermore, when the processing unit generates the second feature information based on the high-frequency component characteristics in the vibration signal, it is configured as follows: The vibration signal is decomposed into time and frequency components to separate the high-frequency vibration components within a preset frequency band. The preset frequency band corresponds to the elastic wave frequency band generated by the stick-slip effect at the interface between the PET sheet and the core. Extracting the energy and temporal characteristics of high-frequency vibration components; Based on energy and temporal characteristics, a slip characteristic index is generated, which is used to characterize the severity of interface slippage within the roll.

[0023] The laser vibration meter 23 (preferably a laser Doppler vibration meter with a sampling frequency ≥100kHz) non-contactly measures the instantaneous vibration velocity of the outer surface of the PET roll. When the interface between the PET sheet and the core undergoes microscopic relative sliding (i.e., "stick-slip effect"), high-frequency elastic waves (usually in the 20kHz-50kHz frequency band) are excited at the interface. These elastic waves propagate along the sheet to the outer surface, causing changes in surface micro-vibrations. After receiving the vibration signal, the processing unit uses time-frequency analysis methods such as variational mode decomposition (VMD) or wavelet packet decomposition to decompose the vibration signal into multiple intrinsic mode components, from which high-frequency vibration components within a preset frequency band (corresponding to the elastic wave frequency band of the stick-slip effect) are separated. The energy characteristics (such as root mean square value and energy integral) and temporal characteristics (such as pulse density and duration) of these high-frequency components are further extracted, and a slip characteristic index is generated through weighted fusion. The higher the index value, the more severe the interface slip. This scheme realizes non-contact real-time monitoring of "latent slip" inside the core.

[0024] Furthermore, when the processing unit determines whether there is a winding defect inside the roll based on the first feature information and the second feature information, it is configured to: The first feature information is compared with a first preset threshold to generate a first confidence level; The second feature information is compared with a second preset threshold to generate a second confidence level; When the first confidence level is greater than the first threshold and the second confidence level is greater than the second threshold, a serious winding defect is determined to exist; When the first confidence level is greater than the first threshold or the second confidence level is greater than the second threshold, a potential winding defect is determined to exist.

[0025] The processing unit employs a multimodal fusion judgment strategy. The first feature information (thickness anomaly heatmap) is quantified to obtain a confidence level α reflecting the degree of thickness anomaly, and the second feature information (slippage feature index) is normalized to obtain a confidence level β reflecting the severity of slippage. A graded judgment is performed by setting dual thresholds: when both α and β exceed their respective thresholds, it indicates that there are both structural anomalies (folds) and dynamic anomalies (slippage) inside, which is judged as a serious defect and requires immediate emergency intervention. When only α or only β exceeds the threshold, it is judged as a potential defect and requires preventive intervention. This fusion judgment mechanism avoids false alarms caused by a single sensor due to ambient light interference (affecting structured light) or mechanical impact (affecting vibration), thus improving the reliability of detection.

[0026] Furthermore, the vision inspection device 2 is mounted on the follower arm 3, which is linked with the winding mechanism 1 to maintain the working distance between the vision inspection device 2 and the outer surface of the roll within a preset range during the winding process.

[0027] One end of the follower arm 3 is fixed to the winding machine frame, and the other end is equipped with a vision inspection device 2. The follower arm 3 is equipped with a displacement sensor and a servo drive mechanism, which are linked with the roll diameter detection device of the winding mechanism 1. As the diameter of the roll gradually increases during the winding process, the follower arm 3 automatically swings outward, so that the working distance between the vision inspection device 2 and the outer surface of the roll is always kept within a preset range of 200±50mm, ensuring the stability and consistency of structured light projection and vibration signal acquisition.

[0028] Furthermore, it also includes: The linkage control interface is connected to the control system of the winding mechanism 1. The processing unit is also configured to send a control command to the winding mechanism 1 through the linkage control interface when a winding defect is determined. The control command includes at least one of the following: a speed reduction command, a pressure increase command, or a shutdown command.

[0029] The linkage control interface uses an industrial fieldbus to connect with the PLC control system of the winding machine. When the processing unit determines that there is a potential defect, it sends a speed reduction command (such as reducing the speed to 30% of the set value) and a pressure increase command (such as increasing the pressure of the pressure roller by 10%) through this interface to suppress the further expansion of the defect. When a serious defect is determined, a stop command is sent to urgently stop the winding action and prevent the continued formation of waste rolls.

[0030] Furthermore, it also includes: The marking device 4 is located on the discharge side of the winding mechanism 1 and is communicatively connected to the processing unit; The processing unit is also configured to: when a winding defect is determined, determine the spatial coordinates of the defect location on the roll based on the first feature information, and control the marking device 4 to apply a removable mark at the position corresponding to the spatial coordinates on the outer surface of the roll.

[0031] The marking device 4 is an inkjet printer or labeler, installed behind the winding pressure roller on the discharge side after the material roll is formed. The processing unit calculates the precise spatial coordinates (axial position and circumferential angle) of the defect on the material roll based on the roll diameter encoder reading and axial position information at the moment the defect occurs. When the defect position rotates to the marking station with the winding, the processing unit controls the marking device 4 to spray a removable QR code or color mark on the outer surface of the material roll. This mark can be identified by the subsequent slitting process, guiding the operator to accurately remove the defective section and achieve non-destructive screening.

[0032] Furthermore, the structured light projector 21 projects coded structured light or sinusoidal striped structured light; the image acquisition unit 22 is a high frame rate infrared camera with a frame rate of not less than 1000 frames / second.

[0033] Among them, coded structured light or sinusoidal fringe structured light has good phase resolution characteristics, which makes it easy to extract sub-pixel level phase shift information; high frame rate infrared cameras (frame rate ≥ 1000 frames / second) can capture dynamic images during high-speed winding, while the infrared band can effectively reduce ambient light interference and improve imaging quality.

[0034] Furthermore, the laser vibrometer 23 is a laser Doppler vibrometer with a sampling frequency of not less than 100kHz.

[0035] Among them, the laser Doppler vibrometer uses the Doppler effect to perform non-contact measurement of the vibration velocity of the object surface. The sampling frequency is no less than 100kHz, which can cover the high-frequency elastic wave band (20kHz-50kHz) generated by the stick-slip effect, ensuring the complete capture of slip characteristics.

[0036] Furthermore, the processing unit is also configured as follows: After the winding process is completed, an internal quality report of the coil is generated based on the first and second feature information generated during the winding process. The internal quality report includes three-dimensional distribution information of defect locations and timeline information of slippage events.

[0037] The processing unit archives and processes information such as thickness anomaly heatmap sequences, slip characteristic index time series data, and alarm event records recorded throughout the winding process, generating a complete internal quality report. This report displays the distribution of defects in the axial, circumferential, and depth directions of the roll in a three-dimensional graphical manner, and shows the occurrence sequence of slip events in a timeline manner, providing data support for quality traceability and process optimization.

[0038] Through the above steps, this invention projects structured light onto the surface of the roll using a structured light projector 21. Utilizing the semi-transparent properties of PET material and the phase modulation effect of its multi-layer structure on the light, the processing unit generates a thickness anomaly thermal map based on the phase shift information of the structured light in the reflected image. This achieves non-destructive, real-time detection of internal thickness anomalies such as wrinkles and bubbles in the initial layer of the roll core. This detection principle solves the problem that existing technologies cannot penetrate multiple layers of PET sheet to observe internal defects, advancing the defect detection time to the instant the inner layer of the PET sheet roll is formed, thus achieving detection of the internal state of the PET sheet during winding. Based on visual inspection, this invention introduces a laser vibrometer 23 for non-contact measurement of vibration signals on the outer surface of the roll. By extracting high-frequency component features to generate a slip characteristic index, real-time monitoring of microscopic slip (latency) at the roll core interface is achieved. Furthermore, the processing unit uses dual-confidence and / or logic for fusion judgment: a serious defect is only determined when both structured light detection and vibration detection simultaneously indicate an anomaly. This multi-modal redundancy mechanism avoids… This invention addresses the issue of false alarms or missed alarms caused by single sensors due to factors such as changes in ambient light, obstruction by flying fluff, and mechanical impact. It enables the monitoring system to operate stably in complex production environments, improving detection reliability and anti-interference capabilities. Through a linkage control interface, the invention communicates with the control system of the winding mechanism 1, automatically sending speed-reducing and pressure-increasing commands when potential defects are detected to suppress defect expansion. It also automatically sends a stop command when serious defects are detected, preventing the continuous formation of defective rolls. Simultaneously, a removable marker is applied to the defect location using a marking device 4, providing a precise basis for subsequent slitting processes, achieving non-destructive screening, and reducing raw material waste and production costs. After winding, the invention generates an internal quality report containing three-dimensional distribution information of defect locations and a timeline of slippage events. This report can not only be used for quality judgment of individual rolls but also, through statistical analysis of multiple batches of data, identify the correlation between winding process parameters (such as initial tension, pressure roller pressure, and acceleration curve) and the occurrence of internal defects, providing data support for continuous optimization of the production process.

[0039] Example 2: Optionally, this embodiment provides a vision-based PET sheet production winding status monitoring device, which includes a winding mechanism 1, a vision inspection device 2, and a processing unit.

[0040] The winding mechanism 1 is used to wind PET sheets onto a core to form a roll; specifically, the winding mechanism 1 includes a winding shaft, a pressure roller, a tension control device, and a roll diameter detection device; the winding shaft is driven by a servo motor, the pressure roller presses onto the surface of the roll with controllable pressure, and the roll diameter detection device detects the roll diameter in real time and feeds it back to the control system.

[0041] The visual inspection device 2 is located on one side of the winding mechanism 1, and its inspection field covers the outer surface of the material roll. The visual inspection device 2 is installed at the end of the follower arm 3, which is linked with the winding mechanism 1. During the winding process, the working distance between the visual inspection device 2 and the outer surface of the material roll is kept within a preset range.

[0042] The visual inspection device 2 includes a structured light projector 21, an image acquisition device 22, and a laser vibration meter 23.

[0043] The structured light projector 21 is used to project a pre-defined pattern of structured light onto the outer surface of the roll. In this embodiment, the structured light projector 21 uses a digital light processing (DLP) projection module to project sinusoidal stripe structured light with a stripe period of 2mm and a projection angle of 15° with the normal direction of the roll surface.

[0044] Image acquisition device 22 is used to acquire the reflected image of structured light after reflection from the outer surface of the roll. In this embodiment, image acquisition device 22 is a high-speed industrial camera equipped with a narrow-band filter to match the wavelength of structured light projector 21 (center wavelength 850nm). The camera frame rate is set to 2000 frames / second and the resolution is 1280×1024 pixels.

[0045] The laser vibration meter 23 is used for non-contact measurement of the vibration signal on the outer surface of the material roll. In this embodiment, the laser vibration meter 23 is a laser Doppler vibration meter, and its working principle is based on the Doppler effect: when a laser beam irradiates the surface of a vibrating object, the frequency of the reflected light will shift, and the amount of shift is proportional to the vibration velocity of the object surface. The sampling frequency of the laser Doppler vibration meter is set to 200kHz, and the measurement resolution is better than 0.01μm / s.

[0046] The processing unit is connected to the vision inspection device 2 to receive inspection data and determine the winding status; the processing unit adopts a high-performance industrial computer, equipped with a GPU acceleration card, and runs real-time data processing algorithms.

[0047] The processing unit is configured to: generate first feature information characterizing the abnormal thickness distribution inside the roll based on the phase offset information of the structured light in the reflected image; generate second feature information characterizing the interface slippage state inside the roll based on the high-frequency component features in the vibration signal; and determine whether there is a winding defect inside the roll based on the first feature information and the second feature information.

[0048] In this embodiment, the processing unit processes the reflected image as follows: First, the acquired reflected image sequence is input into the processing unit, and the phase distribution is extracted using a four-step phase-shifting method. The basic principle of the four-step phase-shifting method is to project four sinusoidal fringes with a phase difference of π / 2, and the corresponding reflected light intensity can be expressed as: I1(x,y) = I0(x,y) + I1(x,y)cos[φ(x,y)] I2(x,y) = I0(x,y) + I1(x,y)cos[φ(x,y) + π / 2] I3(x,y) = I0(x,y) + I1(x,y)cos[φ(x,y) + π] I4(x,y) = I0(x,y) + I1(x,y)cos[φ(x,y) + 3π / 2] Where I0(x,y) is the background light intensity, I1(x,y) is the modulation amplitude, and φ(x,y) is the phase distribution to be solved; the above four equations can be used to obtain: φ(x,y) = arctan[(I4 - I2) / (I1 - I3)] When initial layer wrinkles or foreign objects are present inside the coil, the local optical thickness changes, causing a phase shift Δφ(x,y) in the reflected light; the relationship between this phase shift and the optical thickness change Δh is as follows: Δφ(x,y) = (4π / λ)·Δh(x,y)·n Wherein, λ is the structured light wavelength (850nm in this embodiment), and n is the refractive index of PET material in the 850nm band (measured value is 1.57); there is a linear mapping relationship between the optical thickness change Δh(x,y) and the mechanical thickness change Δd(x,y). A calibration curve is established through a pre-calibration experiment to convert the phase offset into a thickness anomaly value.

[0049] The processing unit maps the thickness anomaly values ​​according to the axial coordinates and circumferential angles of the material roll surface to generate a thickness anomaly heat map, which is the first feature information.

[0050] Simultaneously, the processing unit processes the vibration signal collected by the laser vibrometer 23; assuming the vibration signal is v(t) and the sampling frequency is f_s = 200kHz; the processing unit uses the variational mode decomposition (VMD) method to decompose v(t) into K eigenmode components u_k(t), where K is set to 8; VMD is achieved by solving the following constrained variational problem: min_{ {u_k},{ω_k}} { Σ_k || ∂_t [ (δ(t) + j / (πt)) u_k(t) ] e^{-jω_k t} ||2²} st Σ_k u_k(t) = v(t) Where u_k(t) is the k-th modal component, and ω_k is its center frequency; after decomposition, modal components with center frequencies in the range of 20kHz-50kHz are extracted. This frequency band corresponds to the elastic wave frequency band generated by the stick-slip effect at the interface between the PET sheet and the core; the extracted high-frequency component is denoted as v_hf(t), and its energy characteristic E is calculated: E = ∫_{t0}^{t1} v_hf²(t) dt Simultaneously, temporal features are calculated, including pulse density ρ (the number of pulses exceeding a threshold per unit time) and duration τ (the average duration of a pulse); these features are then weighted and fused to generate the slip characteristic index S. S = w1·(E / E0) + w2·(ρ / ρ0) + w3·(τ / τ0) Where E0, ρ0, and τ0 are baseline values ​​under normal conditions, obtained through pre-calibration; w1, w2, and w3 are weighting coefficients, satisfying w1+w2+w3=1, and in this embodiment, they are taken as 0.4, 0.3, and 0.3 respectively. The slip characteristic index S is the second feature information.

[0051] The processing unit determines defects based on the first feature information and the second feature information. Specifically, it extracts the maximum thickness anomaly value Δd_max from the thickness anomaly heatmap, compares it with the first preset threshold T1=0.05mm, and generates a first confidence level α=min(Δd_max / T1, 1). It compares the slip feature index S with the second preset threshold T2=0.7 and generates a second confidence level β=min(S / T2, 1). When α>0.8 and β>0.8, it determines that there is a serious winding defect. When α>0.8 or β>0.8, it determines that there is a potential winding defect.

[0052] Example 3: Optionally, this embodiment further explains the method by which the processing unit generates the first feature information based on embodiment 2.

[0053] In this embodiment, when the processing unit generates the first feature information based on the phase offset information of the structured light in the reflected image, it is configured to: extract the phase offset of multiple pixels in the reflected image; convert the phase offset into thickness anomaly values ​​based on a preset optical thickness-mechanical thickness mapping relationship; and generate a thickness anomaly heat map based on the distribution of thickness anomaly values ​​on the surface of the roll.

[0054] Specifically, the processing unit first performs phase unwrapping processing on the acquired reflection image sequence. Since the phase φ(x,y) calculated by the four-step phase shift method is truncated within the interval (-π, π], a phase unwrapping algorithm is needed to restore the continuous phase distribution. This embodiment employs a minimum norm-based phase unwrapping algorithm to solve the following optimization problem: min_{φ_unwrap} Σ_{(i,j)} |φ_unwrap(i,j) - φ_unwrap(i-1,j) - Δφ_x(i,j)|² + |φ_unwrap(i,j) - φ_unwrap(i,j-1) - Δφ_y(i,j)|² Here, Δφ_x and Δφ_y are the phase differences in the x and y directions, respectively.

[0055] After obtaining the continuous phase distribution, the phase offset Δφ(x,y) = φ_unwrap(x,y) - φ_ref(x,y) is calculated, where φ_ref(x,y) is the reference phase distribution of the defect-free region, which is obtained by scanning the surface of the initial roll before winding begins.

[0056] According to optical principles, the relationship between phase shift and optical thickness change is as follows: Δφ(x,y) = (4π / λ)·Δh(x,y) Where λ is the structured light wavelength (850nm), and Δh(x,y) is the optical thickness change; the relationship between the optical thickness change Δh and the mechanical thickness change Δd is obtained through pre-calibration; the calibration method is as follows: a set of stacked PET sheet samples with known thickness differences of 0.01mm, 0.02mm, 0.03mm, 0.05mm, 0.07mm, and 0.10mm are prepared, the corresponding phase shifts are measured, and the mapping curve is obtained by fitting: Δd = k·Δh Where k is the calibration coefficient, and the measured value in this embodiment is 0.82.

[0057] Therefore, the processing unit converts the phase offset of each pixel into a mechanical thickness anomaly value Δd(x,y); and transforms the thickness anomaly value according to the axial coordinate x (along the width direction of the roll, in mm) and the circumferential angle θ (along the circumference direction of the roll, in degrees) of the roll surface; assuming the current diameter of the roll is D, the correspondence between the circumferential position and the pixel coordinate is as follows: θ = (y / (πD)) × 360° Where y is the vertical coordinate of a pixel in the image; the final generated thickness anomaly heatmap M(x,θ)=Δd(x,θ) is presented in color image form, with red areas indicating large thickness anomalies and blue areas indicating normal values.

[0058] Example 4: Optionally, this embodiment further explains the method by which the processing unit generates the second feature information based on embodiment 2.

[0059] In this embodiment, when the processing unit generates the second feature information based on the high-frequency component characteristics in the vibration signal, it is configured to: perform time-frequency decomposition on the vibration signal to separate the high-frequency vibration components within a preset frequency band; extract the energy characteristics and temporal characteristics of the high-frequency vibration components; and generate a slip characteristic index based on the energy characteristics and temporal characteristics.

[0060] Specifically, the vibration signal v(t) acquired by the laser Doppler vibration meter is pre-amplified and filtered before being input into the processing unit. The processing unit performs time-frequency decomposition using wavelet packet transform. The Daubechies 4 wavelet is selected as the mother wavelet, and the decomposition level is set to 6 levels, resulting in 64 frequency band subspaces. The frequency band range corresponding to each subspace is: f_band(k) = [ (k-1)·f_s / 2^{J+1}, k·f_s / 2^{J+1} ] Where J=6 is the number of decomposition layers, f_s=200kHz is the sampling frequency, and k=1,2,...,64 is the subspace index.

[0061] The subspaces with center frequencies in the range of 20kHz-50kHz are extracted, corresponding to k=7 to 16 (20kHz-25kHz), k=17 to 26 (25kHz-31.25kHz), k=27 to 36 (31.25kHz-37.5kHz), k=37 to 46 (37.5kHz-43.75kHz), and k=47 to 56 (43.75kHz-50kHz). The reconstructed signals from these subspaces are superimposed to obtain the high-frequency vibration component v_hf(t).

[0062] Extracting energy feature E: E = ∫_{tT}^{t} v_hf²(τ) dτ Where T is the sliding time window, and in this embodiment, T = 0.1 seconds.

[0063] Extracting pulse density feature ρ: Set pulse threshold v_th=0.5μm / s, count the number of pulses N_pulse in which v_hf(t) exceeds v_th within the time window T, then pulse density ρ = N_pulse / T.

[0064] Extract duration feature τ: Calculate the duration τ_i of each pulse and take the average value τ = (1 / N_pulse)·Στ_i.

[0065] The above features are normalized and then fused into a slip feature index S: S = tanh( (w_E·(E / E0) + w_ρ·(ρ / ρ0) + w_τ·(τ / τ0)) ) Wherein, tanh is the hyperbolic tangent function, used to map the output to the interval [0,1); E0, ρ0, and τ0 are the baseline values ​​for the normal state, calculated from the data of the first 5 seconds after the start of winding; w_E, w_ρ, and w_τ are weighting coefficients, which are 0.5, 0.3, and 0.2 respectively in this embodiment.

[0066] Example 5: Optionally, this embodiment further explains the defect determination method of the processing unit based on embodiment 1.

[0067] In this embodiment, when the processing unit determines whether there is a winding defect inside the roll based on the first feature information and the second feature information, it is configured to: compare the first feature information with a first preset threshold to generate a first confidence level; compare the second feature information with a second preset threshold to generate a second confidence level; when the first confidence level is greater than the first threshold and the second confidence level is greater than the second threshold, it is determined that there is a serious winding defect; when the first confidence level is greater than the first threshold or the second confidence level is greater than the second threshold, it is determined that there is a potential winding defect.

[0068] Specifically, the processing unit compares the maximum value Δd_max in the thickness anomaly heatmap M(x,θ) with the first preset threshold T1=0.05mm, and calculates the first confidence level α: α = min(Δd_max / T1, 1.0) When α > 0.8, it indicates that the thickness is abnormally significant.

[0069] The slip characteristic index S is compared with the second preset threshold T2 = 0.7 to calculate the second confidence level β: β = min(S / T2, 1.0) When β > 0.8, the slip phenomenon is significant.

[0070] The judgment rules are as follows: If α > 0.8 and β > 0.8, it is judged as a serious winding defect, triggering a level three alarm; If α > 0.8 or β > 0.8, it is determined to be a potential winding defect, triggering a level 2 alarm; If α ≤ 0.8 and β ≤ 0.8, it is considered a normal state.

[0071] Example 6: Optionally, this embodiment adds a follower arm 3, a linkage control interface, and a marking device 4 to embodiment 2.

[0072] In this embodiment, the vision inspection device 2 is mounted on the follower arm 3, which is linked with the winding mechanism 1 to maintain the working distance between the vision inspection device 2 and the outer surface of the roll within a preset range during the winding process.

[0073] Specifically, the follower arm 3 is a two-section articulated arm structure, including a fixed base 31, a first swing arm 32, a second swing arm 33, and an end mounting base 34. The fixed base 31 is mounted on the winding machine frame. The first swing arm 32 is connected to the fixed base 31 via a rotary joint, the second swing arm 33 is connected to the first swing arm 32 via a rotary joint, and the end mounting base 34 is connected to the second swing arm 33 via a universal joint. The vision inspection device 2 is fixed on the end mounting base 34. Each rotary joint has a built-in servo motor and angle encoder, which are linked with the roll diameter detection device of the winding mechanism 1. The roll diameter detection device detects the roll diameter D_roll in real time. The processing unit calculates the required angle θ1 of the first swing arm 32 and the angle θ2 of the second swing arm 33 based on D_roll, and controls the servo motor to drive the swing arm movement so that the angle between the vision inspection device 2 and the normal of the roll surface is maintained at 15°±2°, and the working distance is maintained within the range of 200mm±10mm.

[0074] This embodiment also includes a linkage control interface, which is communicatively connected to the control system of the winding mechanism 1. The linkage control interface adopts the EtherCAT industrial Ethernet protocol to establish real-time communication with the PLC controller of the winding machine. The processing unit is also configured to send control commands to the winding mechanism 1 through the linkage control interface when a winding defect is detected.

[0075] Specifically, when a potential winding defect is detected, the processing unit sends a speed reduction command and a pressure increase command through the linkage control interface: the speed reduction command reduces the winding speed from the current value V_cur to V_cur×0.3; the pressure increase command increases the pressure of the pressure roller from the current value P_cur to P_cur×1.1; when a serious winding defect is detected, the processing unit sends a stop command to urgently stop the operation of the winding shaft and the traction roller.

[0076] This embodiment also includes a marking device 4, which is located on the discharge side of the winding mechanism 1 and is communicatively connected to the processing unit. The marking device 4 is a piezoelectric inkjet printer, with the printhead installed about 500mm behind the winding pressure roller and about 10mm away from the outer surface of the roll.

[0077] The processing unit is also configured to: when a winding defect is determined, determine the spatial coordinates of the defect location on the roll based on the first feature information, and control the marking device 4 to apply a removable mark at the position corresponding to the spatial coordinates on the outer surface of the roll.

[0078] Specifically, let the time of defect occurrence be t0, at which time the roll diameter detection device measures the roll diameter as D0, and the roll diameter encoder records the wound length as L0; the angular position θ_def of the defect in the circumferential direction of the roll is determined by the pixel position in the structured light image; the processing unit calculates the spatial coordinates of the defect position on the roll: the axial coordinate x_def is directly obtained by mapping the lateral pixel position in the image; the circumferential coordinate is based on the time of defect occurrence, considering the delay between the detection point and the marking device 4, and calculates the angle through which the defect position rotates at the marking time t_mark; let the winding speed be V(t), then the angle through which the roll rotates from t0 to t_mark is: Δθ = ∫_{t0}^{t_mark} [V(t) / (π·D(t))] × 360° dt Where D(t) is the roll diameter that varies with time. The processing unit uses the axial coordinate x_def and circumferential angle θ_def+Δθ of the defect as the target marking position; when this position rotates to below the printhead of the marking device 4, the processing unit sends a trigger signal, and the inkjet printer sprays a removable water-based ink QR code on the surface of the roll. The QR code encoding contains the defect type, coordinates, and occurrence time information.

[0079] Example 7: Optionally, based on Embodiment 2, this embodiment further explains the specific parameters of the structured light projector 21, the image acquisition unit 22, and the laser vibrometer 23, as well as the post-processing function of the processing unit.

[0080] In this embodiment, the structured light projected by the structured light projector 21 is either coded structured light or sinusoidal striped structured light. Specifically, this embodiment uses sinusoidal striped structured light with a stripe period of 1.6 mm, a phase shift step of 4 steps, a projection resolution of 1024×768 pixels, and a projection frame rate synchronized with the image acquisition device 22, set to 1000 frames / second.

[0081] Image acquisition unit 22 is a high frame rate infrared camera with a frame rate of not less than 1000 frames / second. In this embodiment, a CMOS sensor infrared camera with a resolution of 1280×1024 pixels and a frame rate of 1000 frames / second is used. It is equipped with an 850nm narrowband filter (10nm half-width) to effectively suppress ambient light interference.

[0082] The laser vibrometer 23 is a laser Doppler vibrometer with a sampling frequency of not less than 100kHz. In this embodiment, a helium-neon laser (wavelength 632.8nm) is used as the light source, with a maximum output power of 1mW, a sampling frequency of 200kHz, a velocity range of ±1m / s, and a frequency response range of 0-100kHz.

[0083] In this embodiment, the processing unit is further configured to generate an internal quality report of the roll based on the first feature information and the second feature information generated during the winding process after the winding process is completed.

[0084] Specifically, the processing unit archives the data recorded throughout the entire winding process, including: 3D distribution information of defect locations: All pixels with thickness anomalies exceeding 0.05mm are reconstructed in 3D according to axial coordinate x, circumferential angle θ, and time (or roll length) t to generate a 3D scatter plot; Defect depth information is estimated through the mapping relationship between thickness anomalies and the number of layers: Assuming the thickness of a single layer of PET sheet is h_single, then the depth of the defect in the number of layers n_layer ≈ Δd / h_single; In this embodiment, h_single = 0.3mm, then the number of layers corresponding to a thickness anomaly of 0.05mm is approximately 0.17 layers (i.e., the initial intralayer defect); Slippage event timeline information: The curve of slippage characteristic index S changing with time is presented in the form of a line graph, marking the time period exceeding the threshold of T2=0.7, as well as the corresponding process parameters such as winding speed and pressure roller pressure; Alarm event log: Records information such as the time, defect type, location coordinates, and intervention measures taken for each level 2 or level 3 alarm triggered.

[0085] The internal quality report is stored in PDF format on the local hard drive of the processing unit and can be uploaded to the factory's MES system for quality management personnel to use for traceability analysis and process optimization.

[0086] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A vision-based inspection device for monitoring the winding status of PET sheet production, characterized in that: include: The winding mechanism (1) is used to wind PET sheets onto a core to form a roll; A visual inspection device (2) is disposed on one side of the winding mechanism (1), and its inspection field covers the outer surface of the roll. The processing unit is communicatively connected to the visual inspection device (2) and is used to receive inspection data and determine the winding status; The visual inspection device (2) includes: Structured light projector (21) is used to project structured light with a preset pattern onto the outer surface of the roll; Image acquisition device (22) is used to acquire the reflected image of the structured light after it is reflected by the outer surface of the roll; A laser vibration meter (23) is used for non-contact measurement of the vibration signal on the outer surface of the material roll; The processing unit is configured as follows: Based on the phase shift information of the structured light in the reflected image, first feature information characterizing the abnormal thickness distribution inside the roll is generated; Based on the high-frequency component characteristics in the vibration signal, a second feature information characterizing the slip state of the internal interface of the material roll is generated; Based on the first feature information and the second feature information, it is determined whether there is a winding defect inside the material roll.

2. The vision-based inspection-based PET sheet production winding status monitoring device according to claim 1, characterized in that: When the processing unit generates the first feature information based on the phase shift information of the structured light in the reflected image, it is configured to: Extract the phase offset of multiple pixels in the reflected image; Based on a preset optical thickness-mechanical thickness mapping relationship, the phase offset is converted into a thickness anomaly value; Based on the distribution of the thickness anomalies on the surface of the roll, a thickness anomaly heatmap is generated, which is used to indicate the degree of thickness anomaly at different locations inside the roll.

3. The vision-based inspection-based PET sheet production winding status monitoring device according to claim 1, characterized in that: When the processing unit generates the second feature information based on the high-frequency component characteristics in the vibration signal, it is configured as follows: The vibration signal is decomposed into time and frequency components to separate high-frequency vibration components within a preset frequency band. The preset frequency band corresponds to the elastic wave frequency band generated by the stick-slip effect at the interface between the PET sheet and the core. Extract the energy characteristics and temporal characteristics of the high-frequency vibration components; Based on the energy and timing characteristics, a slip characteristic index is generated, which is used to characterize the severity of interface slippage inside the coil.

4. The vision-based inspection-based PET sheet production winding status monitoring device according to claim 1, characterized in that: When the processing unit determines whether there is a winding defect inside the roll based on the first feature information and the second feature information, it is configured to: The first feature information is compared with a first preset threshold to generate a first confidence level; The second feature information is compared with a second preset threshold to generate a second confidence level; When the first confidence level is greater than the first threshold and the second confidence level is greater than the second threshold, a serious winding defect is determined to exist; When the first confidence level is greater than the first threshold or the second confidence level is greater than the second threshold, it is determined that there is a potential winding defect.

5. A vision-based inspection-based PET sheet production winding status monitoring device according to claim 1, characterized in that: The visual inspection device (2) is mounted on the follower arm (3), which is linked with the winding mechanism (1) to maintain the working distance between the visual inspection device (2) and the outer surface of the roll within a preset range during the winding process.

6. The vision-based inspection-based PET sheet production winding status monitoring device according to claim 1, characterized in that: Also includes: The linkage control interface is connected in communication with the control system of the winding mechanism (1); The processing unit is further configured to: when a winding defect is determined to exist, send a control command to the winding mechanism (1) through the linkage control interface, wherein the control command includes at least one of a speed reduction command, a pressure increase command, or a stop command.

7. The vision-based inspection-based PET sheet production winding status monitoring device according to claim 1, characterized in that: Also includes: The marking device (4) is located on the discharge side of the winding mechanism (1) and is communicatively connected to the processing unit; The processing unit is further configured to: when a winding defect is determined, determine the spatial coordinates of the defect location on the roll based on the first feature information, and control the marking device (4) to apply a removable mark to the outer surface of the roll at the position corresponding to the spatial coordinates.

8. A vision-based inspection-based PET sheet production winding status monitoring device according to claim 1, characterized in that: The structured light projector (21) projects coded structured light or sinusoidal stripe structured light; the image acquisition device (22) is a high frame rate infrared camera with a frame rate of not less than 1000 frames / second.

9. A vision-based inspection-based PET sheet production winding status monitoring device according to claim 1, characterized in that: The laser vibration meter (23) is a laser Doppler vibration meter with a sampling frequency of not less than 100kHz.

10. A vision-based inspection-based PET sheet production winding status monitoring device according to claim 1, characterized in that: The processing unit is further configured to: After the winding process is completed, an internal quality report of the material roll is generated based on the first feature information and the second feature information generated during the winding process. The internal quality report includes three-dimensional distribution information of defect locations and timeline information of slippage events.

Images