Wood floor surface embossing equipment and embossing method

Abstract

The invention relates to the technical field of wood floor processing, in particular to wood floor surface embossing equipment and an embossing method, and the wood floor surface embossing equipment comprises an embossing roller module, an optical detection module, a neural network processor, a closed-loop control unit and a data storage module. A wood surface vibration frequency spectrum is collected through a high-frequency vibration sensor array, embossing edge features are scanned in combination with an ultraviolet-visible light dual-band imaging unit, and embossing parameters are dynamically predicted and calibrated by using a lightweight convolutional neural network model. The invention aims to overcome the defects in the prior art, solve or at least alleviate the problem that the sensitivity is insufficient when a traditional plywood defect detection method is used for distinguishing a tiny degumming area and normal bonding textures, and improve the problem that frequency response overlapping of non-uniform adhesive layer curing and wood natural textures cannot be effectively distinguished by traditional acoustic detection. The invention provides a plywood defect detection system and a detection method.

Description

Technical Field

[0001] The invention belongs to the technical field of wood floor processing, and in particular relates to a wood floor surface embossing device and an embossing method. Background Art

[0002] Currently, the natural grain of wood flooring can cause the embossing to be too shallow in low-grain density areas and too deep in high-grain density areas, affecting the aesthetics and overall quality of the embossing. Furthermore, fiber rebound can blur the edges of the pattern, reducing the sharpness of the embossed pattern and failing to meet decorative requirements. Therefore, multi-parameter detection and dynamic adjustment are needed to improve embossing quality.

[0003] Some existing technologies improve embossing quality by adding a pressure control module or an optical detection module. However, current pressure control is mostly static, resulting in slow response to wood floors with varying grain densities and adjustment lag. Optical detection uses only a single wavelength band to distinguish between fiber breakage and natural reflections, leading to misjudgment of grain density and affecting accuracy. For example, excessively high or low grain density can interfere with the optical signal, affecting the determination of sharpness. The natural patterns of wood flooring grain make existing equipment less suitable for wood floors.

[0004] For example, patent publication number CN102852311A proposes a method for pressing wood flooring using a patterned steel roller. While emphasizing the importance of controlling embossing depth, it fails to address intelligent monitoring and dynamic adjustment. Another example is patent publication number CN104290517A, which proposes a device for automatically adjusting embossing pressure. However, its core focus remains on mechanical structure optimization, lacking a data-driven closed-loop control system. These technologies have limitations in addressing the issues of texture density variations and fiber rebound in embossed wood flooring surfaces. Summary of the Invention

[0005] The purpose of the present invention is to overcome the shortcomings of the existing technology, solve or at least alleviate the problem of insufficient sensitivity of traditional plywood defect detection methods in distinguishing between tiny debonding areas and normal bonding textures, and at the same time improve the problem of frequency response overlap between traditional acoustic detection and the inability to effectively distinguish between uneven curing of the glue layer and the natural texture of the wood, and provide a plywood defect detection system and detection method.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a wood floor surface embossing device, comprising:

[0007] The embossing roller module has a high-frequency vibration sensor array embedded in its outer surface to obtain the vibration spectrum of the wood surface in real time;

[0008] The optical detection module is installed at the embossing outlet and includes a UV-visible dual-band imaging unit for scanning the edge features of the embossing area;

[0009] A neural network processor connects the high-frequency vibration sensor array and the ultraviolet-visible dual-band imaging unit via a data transmission line, and has a built-in lightweight convolutional neural network model;

[0010] The closed-loop control unit receives the prediction instructions output by the neural network processor and dynamically adjusts the pressure parameters of the embossing roller module;

[0011] Data storage module, recording vibration spectrum, optical characteristics and predicted deviation data;

[0012] Among them, the embossing roller module is fixed on the frame through a bearing seat, and the optical detection module is installed at the embossing outlet end through a slide rail, and its position can be adjusted along the slide rail to adapt to embossing paths of different lengths.

[0013] Preferably, the closed-loop control unit includes a forward control channel and a reverse calibration channel. The forward control channel transmits the predicted sharpness value to the embossing roller module in real time through a signal cable, and the reverse calibration channel feeds back the measured data of the optical detection module to the neural network processor through optical fiber communication.

[0014] A method for embossing wooden floors based on the above-mentioned device comprises the following steps:

[0015] Step S1: obtaining the wood texture vibration spectrum in real time through the embossing roller vibration sensor, and dynamically adjusting the embossing pressure according to the spectrum characteristics;

[0016] Step S2: Scan the embossed edge using a multispectral imaging system, calculate the edge sharpness coefficient, and mark unqualified areas;

[0017] Step S3: Correlating vibration data with edge sharpness using a neural network model to perform forward prediction and reverse calibration;

[0018] Step S4: According to the deviation between the predicted results and the measured data, the embossing parameters are dynamically optimized and the neural network model is updated.

[0019] In order to further realize the present invention, the following technical solutions may be preferably used:

[0020] Preferably, the step S1 includes:

[0021] Step S101: collecting the main frequency energy distribution of the wood surface vibration during the embossing process;

[0022] Step S102: when a low texture density area is detected and a predicted sharpness warning is received, the pressure of the current area is increased;

[0023] Step S103: When a high texture density area is detected and the predicted sharpness is too high, suppressing the pressure fluctuation amplitude.

[0024] Preferably, step S2 includes:

[0025] Step S201: synchronously obtaining the fiber break reflectivity in the ultraviolet band and the contour geometric characteristics in the visible light band;

[0026] Step S202: Calculate the dual-band reflectivity difference ratio to generate a sharpness coefficient;

[0027] Step S203: When the sharpness coefficient is continuously lower than the threshold, it is marked as an area requiring compensation and triggers historical data analysis.

[0028] Preferably, step S3 includes:

[0029] Step S301: Input the vibration spectrum features into the neural network and output the predicted sharpness value and confidence level;

[0030] Step S302: comparing the deviation between the predicted sharpness value and the measured sharpness value;

[0031] Step S303: When the deviation exceeds the tolerance, the texture density determination threshold is adjusted in the reverse direction.

[0032] Preferably, the step S301 specifically includes:

[0033] When the predicted sharpness value is lower than the warning threshold, the embossing roller pressure compensation instruction is triggered in advance;

[0034] When the confidence level is lower than a reliable threshold, the neural network parameter update is paused.

[0035] Preferably, the step S4 includes:

[0036] Step S401: transmitting the prediction instruction to the embossing actuator in real time to dynamically allocate the pressure ratio of the main / auxiliary embossing rollers;

[0037] Step S402: When the measured data and the predicted data continuously exceed the tolerance, the emergency calibration procedure is started.

[0038] Preferably, it also includes:

[0039] Periodically store forecast deviation data and environmental parameters;

[0040] When the amount of stored abnormal data reaches a threshold, adversarial training samples are automatically generated;

[0041] Update neural network weights through incremental learning.

[0042] Preferably, in the forward control loop, the response time from vibration data to pressure adjustment is less than the embossing stroke time;

[0043] In the reverse control loop, the delay from optical detection data to texture determination calibration is less than the embossing start time of the next block.

[0044] The beneficial effects of the present invention are:

[0045] The present invention is based on the wood surface vibration spectrum obtained by a high-frequency vibration sensor array and the embossing edge feature information collected by an ultraviolet-visible dual-band imaging unit, including but not limited to the influence of fiber rebound and texture density changes of the embossed edge on the embossing quality. By jointly analyzing the vibration spectrum information and optical feature information, as well as a lightweight convolutional neural network model, the dynamic prediction of the embossing depth and edge sharpness is achieved, and the calibration unit is driven to calibrate the embossing equipment; through the forward control channel and reverse calibration channel of the closed-loop control unit, the embossing parameters are adjusted in real time and the model weights are updated online, thereby improving the stability and uniformity of the embossing process; by analyzing the vibration spectrum of different texture density areas, the embossing pressure is dynamically adjusted, and then the embossing depth is adjusted, thereby solving the problem of excessively deep or shallow embossing caused by high texture density and high fiber content; by calculating the sharpness coefficient through the dual-band reflectivity difference ratio, the blurriness of the pattern edge is improved, thereby solving the problem of unclear pattern edge caused by fiber rebound, which affects the beauty and decorativeness of the embossed pattern. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a block diagram of the device of the present invention.

[0047] Figure 2 It is a block diagram of the embossing roller module of the present invention.

[0048] Figure 3 This is an input-output relationship diagram of the neural network processor of the present invention.

[0049] Figure 4 This is the internal structure and signal flow of the closed-loop control unit of the present invention.

[0050] Figure 5 Flow chart of the method of the present invention.

[0051] Figure 6 This is a flowchart of the neural network processing of the present invention. DETAILED DESCRIPTION

[0052] In the description of the present invention, it should also be noted that, unless otherwise expressly specified or limited, the terms "disposed," "installed," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; they may refer to mechanical connections or electrical connections; they may refer to direct connections or indirect connections through an intermediate medium; and they may refer to internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0053] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0054] Example 1

[0055] Currently, the natural grain of wood flooring can cause the embossing to be too shallow in low-grain density areas and too deep in high-grain density areas, affecting the aesthetics and overall quality of the embossing. Furthermore, fiber rebound can blur the edges of the pattern, reducing the sharpness of the embossed pattern and failing to meet decorative requirements. Therefore, multi-parameter detection and dynamic adjustment are needed to improve embossing quality.

[0056] Some existing technologies improve embossing quality by adding a pressure control module or an optical detection module. However, current pressure control is mostly static, resulting in slow response to wood floors with varying grain densities and adjustment lag. Optical detection uses only a single wavelength band to distinguish between fiber breakage and natural reflections, leading to misjudgment of grain density and affecting accuracy. For example, excessively high or low grain density can interfere with the optical signal, affecting the determination of sharpness. The natural patterns of wood flooring grain make existing equipment less suitable for wood floors.

[0057] This embodiment discloses a wood floor surface embossing device, such as Figure 1 As shown, the embossing roller module is placed on the frame, and its outer surface is covered with a high-frequency vibration sensor array for collecting the vibration spectrum of the wood surface. The optical detection module installed at the embossing outlet includes an ultraviolet-visible light dual-band imaging unit for obtaining the edge features of the embossing area. The neural network processor is connected to the high-frequency vibration sensor array and the ultraviolet-visible light dual-band imaging unit through a data transmission line and has a built-in lightweight convolutional neural network. The closed-loop control unit is connected to the neural network processor to receive its prediction instructions and then adjust the pressure of the embossing roller module. The data storage module records the vibration spectrum, optical characteristics and prediction deviation.

[0058] like Figure 2 As shown, the high-frequency vibration sensor array is installed on the outer surface of the embossing roller module, and each of the high-frequency vibration sensors is connected to the neural network processor through a signal line. The two ends of the embossing roller module are fixed to the frame through bearing seats, so that the embossing roller module keeps rotating during operation. The high-frequency vibration sensor array is installed on the outer surface of the embossing roller module and is arranged at equal intervals along the circumference of the embossing roller module. By optimizing its position and spacing, the signal collected by the high-frequency vibration sensor array can cover the entire embossing area without signal blind spots. The frequency of collecting the vibration signal of the wood surface is 1000 times per second. When the wood surface vibrates, the vibration of the wood surface will trigger the high-frequency vibration sensor installed on the surface of the embossing roller module to vibrate. The high-frequency vibration sensor collects the vibration signal of the wood surface and transmits the signal to the neural network processor. The neural network processor transmits the received signal to the data transmission line.

[0059] like Figure 3 As shown, the UV-visible dual-band imaging unit is mounted on a slide rail, which is fixed to the frame at the embossing outlet. The UV-visible dual-band imaging unit includes two imaging lenses, which are used to collect the UV-band fiber break reflectivity and the visible light band contour geometric characteristics respectively. The slide rail enables the UV-visible dual-band imaging unit to adjust its position along the embossing path to adapt to embossing areas of different lengths. The UV-visible dual-band imaging unit is connected to the neural network processor via an optical fiber to transmit the collected image data to the neural network processor for processing. The adjustment range of the slide rail is set to 0-500mm, with an adjustment accuracy of 0.1mm to meet different process requirements.

[0060] like Figure 4 As shown in the figure, the forward control channel and reverse calibration channel of the closed-loop control unit together constitute a closed-loop control system. The forward control channel transmits the predicted sharpness value output by the neural network processor to the embossing roller module in real time through a signal cable for adjusting the embossing pressure. The reverse calibration channel feeds back the measured data of the optical detection module to the neural network processor through optical fiber communication for updating the neural network weight parameters. The response time of the forward control channel is less than the embossing stroke time, thereby ensuring that the adjustment instruction can participate in the embossing process while taking effect. The delay time of the reverse calibration channel is less than the start time of the embossing of the next block, thereby realizing rapid calibration of the texture density judgment threshold.

[0061] Example 2

[0062] like Figure 5 and Figure 6 As shown, in order to better enable the system of embodiment 1 to work better, this embodiment provides a method for embossing a wooden floor, comprising the following steps:

[0063] Step S1: obtaining the wood texture vibration spectrum in real time through the embossing roller vibration sensor, and dynamically adjusting the embossing pressure according to the spectrum characteristics;

[0064] Step S2: Scan the embossed edge using a multispectral imaging system, calculate the edge sharpness coefficient, and mark unqualified areas;

[0065] Step S3: Correlating vibration data with edge sharpness using a neural network model to perform forward prediction and reverse calibration;

[0066] Step S4: According to the deviation between the predicted results and the measured data, the embossing parameters are dynamically optimized and the neural network model is updated.

[0067] The step S1 specifically includes:

[0068] Step S101: collecting the vibration signal of the embossing roller at a sampling interval of 0.1 ms and extracting the main frequency energy ratio E (E=main frequency amplitude / total amplitude);

[0069] Step S102: when E>70%, it is determined to be a high-density texture area; when E<40%, it is determined to be a low-density texture area;

[0070] Step S103: Receive the predicted sharpness output by the neural network like And the current low-density texture area, increase the pressure to 120%-140% of the baseline value;

[0071] Step S104: If And the current area is a high-density texture area, which limits the pressure fluctuation range to within ±5%.

[0072] The step S2 specifically includes:

[0073] Step S201: Scan the fiber fracture surface in the ultraviolet band (365±10nm) to obtain the reflection intensity I_uv;

[0074] Step S202: Scan the pattern profile in the visible light band (550±20nm) to obtain the edge gradient G_vis;

[0075] Step S203: Calculate the sharpness coefficient S = K·(I_uv / G_vis), where K is the calibration coefficient (0.8-1.2);

[0076] Step S204: When the detection value of S is less than 0.8 for three consecutive times, the area is marked and the compensation historical data backtracking is triggered.

[0077] The neural network execution in step S3 includes:

[0078] Step S301: input three characteristics of the vibration spectrum: main frequency f (kHz), energy entropy H, and waveform kurtosis Q;

[0079] Step S302: CNN network outputs predicted sharpness and confidence C (0-1 scale);

[0080] Step S303: When and when C>0.85, a pressure compensation instruction is generated to the embossing roller;

[0081] Step S304: When the measured S and , adjust the texture threshold inversely by the Δ value (E±5%).

[0082] The waveform kurtosis Q is calculated as follows: Q = frac{μ_4}{σ^4}-3 (μ_4 is the fourth-order central moment, σ is the standard deviation);

[0083] When the confidence level C is less than 0.7, the neural network weight update is frozen;

[0084] When the ambient humidity is greater than 75%, the sharpness will be automatically predicted. The judgment threshold is lowered by 0.05.

[0085] The step S4 comprises:

[0086] Step S401: The pressure distribution ratio between the main embossing roller and the auxiliary micro roller is adjusted to 7:3;

[0087] Step S402: When a pressure compensation instruction is received, the pressure of the main roller is increased to 85%, and the pressure of the auxiliary roller is reduced to 15%;

[0088] Step S403: 3 times in a row When the emergency calibration is started:

[0089] Pause the embossing process for 10 seconds;

[0090] Recalibrate the white balance of the optical detection module;

[0091] Reset the initial weights of the neural network.

[0092] The model evolution includes:

[0093] Three types of abnormal data are stored: Type A: predicted deviation points with |Δ|>0.2; Type B: points where S is still less than 0.7 after pressure compensation; Type C: points where the ambient humidity suddenly changes by more than 15% / min;

[0094] After every 50 embossing operations, incremental training is performed, and 20% abnormal data is extracted from the storage module to generate adversarial samples; the loss function weight distribution is: vibration feature weight α = 0.6, environmental parameter weight β = 0.4; the learning rate decays to 30% of the initial value.

[0095] The forward control response time is ≤300ms (covering the time required for the embossing roller to rotate 15° arc length), the reverse calibration delay is ≤100ms (to ensure that the threshold adjustment is completed before the embossing of the next block is started), the scanning resolution of the optical detection module is ≥1200dpi, and the positioning error is <0.1mm.

[0096] Example 3

[0097] The above embodiment only detects the texture density and does not address the risk of equipment damage caused by the resonance of the embossing roller itself. Therefore, a safety monitoring clause for the resonance frequency of the embossing roller is added.

[0098] The steps for safely monitoring the resonance frequency of the embossing roller are as follows:

[0099] SA01: Resonance risk assessment;

[0100] Real-time collection of embossing roller bearing seat vibration signal, calculation of 5-8kHz frequency band energy integral Er, when Er

[0101] >60% is marked as high-risk resonance;

[0102] SA02: Security Response Execution;

[0103] For the high-risk resonance area of ​​the mark, reduce the embossing speed by 20%. When Er>80%, perform the emergency calibration procedure of step S402)

[0104] Step SA03: life prediction maintenance;

[0105] ΣEr·t is accumulated and calculated. When Σ is greater than the threshold, an early warning is generated and stored in the data storage module.

[0106] Improve equipment safety, avoid roller breakage caused by resonance, ensure production continuity, reduce unnecessary downtime with graded response, and accurately guide spare parts replacement with life prediction, making maintenance intelligent.

[0107] The deployment scheme of the lightweight convolutional neural network model in Example 2 is:

[0108] Step SB01: depthwise separable convolution reconstruction;

[0109] The standard 3×3 convolution is split into depthwise convolution (channel separation) and 1× point-by-point convolution, the fully connected layer is removed, and the global average pooling output layer is used instead, which compresses the model parameters to 0.48M±0.02M;

[0110] Step SB02: dynamic INT8 quantization;

[0111] Using the sharpness coefficient dataset generated in step S202 as a calibration benchmark, determine the FP32→INT8 scaling factor using the KL divergence algorithm, with a maximum quantization error of <0.5%;

[0112] Step SB03: embedded hardware acceleration;

[0113] A fixed 200MB memory pool is allocated to the Jetson Nano processor, the TensorRT inference engine is configured, and the BN layer and ReLU activation layer are integrated;

[0114] Step SB04: Runtime accuracy maintenance;

[0115] When the confidence level C in step S301 is less than 0.7, the system automatically switches to FP16 calculation mode and performs offline calibration every 24 hours using stored abnormal data to maintain the Top-1 error less than 0.8%.

[0116] It meets the control needs of high-speed production lines, improves real-time assurance, and maintains prediction reliability under temperature and humidity fluctuations, thereby maintaining accuracy stability.

[0117] Example 4

[0118] In order to enable relevant personnel in this technical field to better understand and implement the present invention, the implementation principle of the present invention is further described in detail below in conjunction with a specific application scenario of the present invention.

[0119] In the actual processing process, the wooden floor to be embossed is placed on the conveyor belt, and the conveying mechanism sends the wooden floor into the working area of ​​the embossing mechanism, such as Figure 1 As shown in the figure, the embossing roller module is fixed to the frame via a bearing block. Its outer surface is embedded with a high-frequency vibration sensor array. When the wooden floor enters the embossing zone, the high-frequency vibration sensor array collects vibration spectrum data from the wood surface at a frequency of 1000 times per second and transmits this data to the neural network processor via a signal line, which serves as the basis for subsequent processing. The high-frequency vibration sensor array can sense the vibration changes caused by different texture densities on the wood surface, thereby determining the distribution of low- and high-texture density areas. Next is the optical detection module. The UV-visible dual-band imaging unit moves along the slide rail to the appropriate position, collecting the UV fiber break reflectance and visible light contour geometric features of the embossing zone. The data is then transmitted via optical fiber to the neural network processor, which calculates the difference ratio of the dual-band reflectance to generate a sharpness coefficient. This method can distinguish between fiber breakage and changes in natural texture reflectance, avoiding the errors caused by using a single-band detection method. The UV-visible dual-band imaging unit's slide rail has an adjustment range of 0-500mm, with an adjustment accuracy of 0.1mm, to accommodate embossing paths of varying lengths.

[0120] The collected vibration signals are sent to the neural network processor by the data acquisition module for analysis, as shown in the figure Figure 6 The vibration spectrum features are sent to the compressed and optimized convolutional neural network model as input signals, and the model feature parameters are gradually extracted through multi-layer convolution operation, and finally the predicted sharpness value and confidence are output, and the neural network processor judges whether the current embossing has the problem of texture density abnormality or insufficient edge sharpness according to the prediction result, when the predicted sharpness value is lower than the early warning value, the pressure compensation instruction is sent to the embossing roller module in advance to reduce the embossing failure degree; when the confidence is lower than the reliable value, stop weight updating to avoid model deviation and increase embossing failure degree. The guiding ideology of the closed-loop control unit is as above, the predicted sharpness value is sent to the embossing roller module through the signal cable in the forward control channel to guide it to adjust the embossing pressure parameter and change the embossing behavior, for example, increase the pressure in the low texture density area to ensure the embossing depth, and suppress the pressure fluctuation in the high texture density area to avoid over-embossing. The measured data measured by the optical detection module are sent to the neural network processor in the reverse calibration channel to realize online updating of the model weight. The response time of the forward control channel should be less than the embossing stroke time to ensure that the adjusted instruction can take effect in time during the embossing process; the delay time of the reverse calibration channel should be less than the start time of the next block embossing to realize rapid calibration of the texture density determination threshold.

[0121] After one embossing is completed, the data storage module fixes the vibration spectrum, the scanning result of the optical module (i.e. the two-dimensional code) and the predicted deviation parameter, and periodically stores the environmental parameters. When the stored abnormal data reaches a specified number, the adversarial training sample is generated and the neural network weight is updated by using the incremental learning method to improve the prediction accuracy and adaptability of the model.

[0122] The present application provides dynamic optimization of wood floor surface embossing by the above-mentioned manner. High-frequency vibration sensor array is used to monitor the vibration of the wood surface in real time, and the scanning result of the ultraviolet-visible light dual-band imaging unit is used to judge the influence of texture density difference and fiber rebound on the embossing quality, the neural network processor is used to jointly analyze the vibration data and optical feature parameters, the embossing parameters are dynamically adjusted, and the model weight is updated to ensure the consistency of the embossing depth and sharpness. Through partition pressure adjustment, the influence of texture density difference on over-embossing or under-embossing is reduced, the fiber rebound is judged by using the dual-band reflectivity difference ratio, the blurriness of the pattern edge is reduced, and the embossing pattern beauty and decoration are improved.

[0123] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can make equivalent replacement or change according to the technical solution and inventive concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A wood floor surface embossing device, characterized in that: include: The embossing roller module has a high-frequency vibration sensor array embedded in its outer surface to obtain the vibration spectrum of the wood surface in real time; The optical detection module is installed at the embossing outlet and includes a UV-visible dual-band imaging unit for scanning the edge features of the embossing area; A neural network processor connects the high-frequency vibration sensor array and the ultraviolet-visible dual-band imaging unit via a data transmission line, and has a built-in lightweight convolutional neural network model; The closed-loop control unit receives the prediction instructions output by the neural network processor and dynamically adjusts the pressure parameters of the embossing roller module; Data storage module, recording vibration spectrum, optical characteristics and predicted deviation data; Among them, the embossing roller module is fixed on the frame through a bearing seat, and the optical detection module is installed at the embossing outlet end through a slide rail, and its position can be adjusted along the slide rail to adapt to embossing paths of different lengths.

2. The device according to claim 1, characterized in that , The closed-loop control unit includes a forward control channel and a reverse calibration channel. The forward control channel transmits the predicted sharpness value to the embossing roller module in real time through a signal cable, and the reverse calibration channel feeds back the measured data of the optical detection module to the neural network processor through optical fiber communication.

3. A method for embossing wooden floors based on the apparatus of claim 1, characterized in that: The following steps are involved: Step S1: obtaining the wood texture vibration spectrum in real time through the embossing roller vibration sensor, and dynamically adjusting the embossing pressure according to the spectrum characteristics; Step S2: Scan the embossed edge using a multispectral imaging system, calculate the edge sharpness coefficient, and mark unqualified areas; Step S3: Correlating vibration data with edge sharpness using a neural network model to perform forward prediction and reverse calibration; Step S4: According to the deviation between the predicted results and the measured data, the embossing parameters are dynamically optimized and the neural network model is updated.

4. The method according to claim 3, characterized in that The step S1 comprises: Step S101: collecting the main frequency energy distribution of the wood surface vibration during the embossing process; Step S102: when a low texture density area is detected and a predicted sharpness warning is received, the pressure of the current area is increased; Step S103: When a high texture density area is detected and the predicted sharpness is too high, suppressing the pressure fluctuation amplitude.

5. The method according to claim 3, characterized in that The step S2 comprises: Step S201: synchronously obtaining the fiber break reflectivity in the ultraviolet band and the contour geometric characteristics in the visible light band; Step S202: Calculate the dual-band reflectivity difference ratio to generate a sharpness coefficient; Step S203: When the sharpness coefficient is continuously lower than the threshold, it is marked as an area requiring compensation and triggers historical data analysis.

6. The method according to claim 3, characterized in that The step S3 comprises: Step S301: Input the vibration spectrum features into the neural network and output the predicted sharpness value and confidence level; Step S302: comparing the deviation between the predicted sharpness value and the measured sharpness value; Step S303: When the deviation exceeds the tolerance, the texture density determination threshold is adjusted in the reverse direction.

7. The method according to claim 6, characterized in that The step S301 specifically includes: When the predicted sharpness value is lower than the warning threshold, the embossing roller pressure compensation instruction is triggered in advance; When the confidence level is lower than a reliable threshold, the neural network parameter update is paused.

8. The method according to claim 3, characterized in that The step S4 comprises: Step S401: transmitting the prediction instruction to the embossing actuator in real time to dynamically allocate the pressure ratio of the main / auxiliary embossing rollers; Step S402: When the measured data and the predicted data continuously exceed the tolerance, the emergency calibration procedure is started.

9. The method according to claim 3, characterized in that Also includes: Periodically store forecast deviation data and environmental parameters; When the amount of stored abnormal data reaches a threshold, adversarial training samples are automatically generated; Update neural network weights through incremental learning.

10. The method according to claim 3, characterized in that ,In the forward control loop, the response time from vibration data to pressure adjustment is ,less than the embossing stroke time; In the reverse control loop, the delay from optical detection data to texture determination calibration is less than the embossing start time of the next block.

Images