Big data bamboo tableware manufacturing system based on production process monitoring

Abstract

The invention provides a big data bamboo tableware manufacturing system based on production process monitoring, and relates to the field of intelligent detection systems.The big data bamboo tableware manufacturing system comprises an acquisition module, a prediction module, a preliminary distribution module, a disturbance response module, an optimization decision module and an abnormity self-inspection module. A dynamic stability index is calculated in combination with a dynamic image sequence under disturbance response, a rejection target is judged through an optimization decision-making module after scores are fused, a Bayesian method is introduced in the optimization process to dynamically adjust multiple decision-making parameters, a rejection action instruction is generated in real time, and an abnormity self-checking module monitors the rejection rate and score stability. Software and hardware signals are combined to realize multi-level fault diagnosis and system protection control; according to the invention, high-precision identification and dynamic elimination of bamboo tableware defects are realized, and the automation level and the production reliability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent detection, and particularly to a big data bamboo tableware manufacturing system based on production process monitoring. Background Art

[0002] In the manufacturing and quality inspection links of bamboo tableware, manual inspection or a primary automated detection system based on static image processing is mostly used, mainly for identifying defects on the surface of tableware, such as cracks, burrs, deformations, etc.

[0003] However, as a natural inhomogeneous material, bamboo is prone to structural defects such as hidden cracks, internal cracks, delamination, mildew, and insect channels inside. Although these defects are not easily detected directly by surface images, they will seriously affect the mechanical strength and use safety of the products. Existing detection means based on static images are difficult to accurately identify the above internal or dynamic response type defects, resulting in ineffective rejection of some unqualified products, restricting the further improvement of product quality and the intelligent development of the manufacturing process. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a big data bamboo tableware manufacturing system based on production process monitoring, which improves the recognition accuracy of bamboo tableware defects through a two-stage scoring mechanism that combines image recognition and disturbance response.

[0005] To solve the above technical problem, the technical solution of the present invention is as follows: A big data bamboo tableware manufacturing system based on production process monitoring, including a transmission line and a turning mechanism, further including: An acquisition module, configured to collect image data of tableware through high-speed cameras deployed on both sides of the turning mechanism; A prediction module, configured to output an initial confidence score for each tableware based on the image data through a pre-trained prediction model; A preliminary diversion module, configured to set a first decision threshold based on the initial confidence score. If the initial confidence score is lower than the first decision threshold, it is determined as a high-probability unqualified product, and the unqualified tableware directly enters the rejection area. If the initial confidence score is equal to or higher than the threshold, the tableware is conveyed to the disturbance section of the transmission line; A disturbance response module, configured to apply a preset air flow disturbance to the tableware entering the disturbance section, obtain at least one perspective dynamic image sequence of the tableware under the air flow disturbance, calculate the dynamic stability index of the tableware through optical flow analysis, fuse the initial image features of the tableware and the dynamic stability index, and perform processing through an ensemble learning algorithm to output a fused confidence score; An optimization decision-making module is used to set a second decision threshold and fuzzy decision interval parameters, determine the final rejection target based on the comparison result of the fusion confidence score and the fuzzy decision interval, obtain historical rejection data and real-time production line status data, and take minimizing the mis-rejection rate and ensuring that the defect detection rate meets the requirements as the optimization goal. Based on the historical rejection data and real-time production line status data, the first decision threshold, the second decision threshold, and the fuzzy decision interval parameters are dynamically determined through the Bayesian optimization method. The updated first decision threshold parameter is sent to the preliminary shunt module in real time, and the second decision threshold and the fuzzy decision interval parameters are reserved for self-use. Based on the fusion confidence score and the dynamically adjusted second decision threshold, an action instruction for the rejection mechanism is generated.

[0006] Furthermore, an abnormal self-checking module is used to continuously monitor the tableware rejection rate within a set time window and set a dynamic safety threshold. The dynamic safety threshold is dynamically updated based on the historical rejection rate statistical model. When the real-time rejection rate continuously exceeds the dynamic safety threshold for a predetermined time length, a multi-level self-checking program is automatically triggered.

[0007] Furthermore, the prediction model architecture adopts a deep residual network, and its main structure includes 4 residual block groups, each group consisting of 2 residual blocks; two convolutional layers are set in each residual block and cross-layer identity mapping operations are performed to solve the problem of deep network degradation through skip connection methods, and are used to extract surface defects and geometric deformation features of bamboo tableware; a multi-scale feature fusion structure is embedded at the front end of the residual network: horizontal connection channels are respectively established at the output ends of the second and third residual block groups, and the low-level high-resolution feature map and the high-level strong semantic feature map are input into the feature pyramid network; the resolution of the high-level feature map is aligned through bilinear interpolation upsampling operation and element-wise addition fusion is performed with the corresponding low-level feature map at the same level; the fused multi-scale feature map is output to simultaneously capture the local microscopic defects and overall morphological features of the tableware; atrous convolutional layers are embedded in the third residual block group, and the dilation rate is set to be not less than 2 to expand the receptive field and enhance the sensitivity to subtle fracture defects in bamboo textures without reducing the resolution; a self-attention mechanism is integrated at the output end of the last residual block group: by calculating the correlation weight matrix between the spatial positions of the feature map, the feature weights of each region are dynamically adjusted to make the model focus on the high-risk defect regions; The output layer of the prediction model adopts a multi-label classification framework: the fully connected layer contains 512 neurons and receives the feature vectors weighted by self-attention; the output layer sets up independent sub-output layers with the same number as the types of defects, and each sub-output layer uses the Sigmoid activation function to output the probability value of the corresponding defect; according to the preset defect risk weights, the probabilities of each defect are weighted and summed to generate an initial confidence score in the range of [0,1], and the lower the score, the higher the defect risk; the training process of the prediction model adopts a supervised learning strategy: using a data set containing labeled samples, and minimizing the multi-task cross-entropy loss function through the Adam optimizer.

[0008] Furthermore, the dynamic stability index of the tableware is calculated through optical flow analysis, including: Based on the continuous image sequence of the tableware under airflow disturbance obtained by the perturbation response module, the image acquisition frame rate is not less than 200 frames per second; the optical flow tracking algorithm is used to track the displacements of at least five preset key geometric feature points on the surface of the tableware in the image sequence, and the key geometric feature points include the center point of the tableware, the point with the maximum edge curvature, and the joint part points, to generate the motion trajectory data of each feature point; three feature parameters are extracted from the motion trajectory data: the first is the maximum offset amplitude, by calculating the straight-line distance from the initial position of all feature points to the maximum displacement point during the airflow disturbance, and taking the maximum value among all the distance values of the feature points as the maximum offset amplitude; the second is the attitude recovery time, starting from the end moment of the airflow disturbance, calculating the time interval when the displacement of all feature points continuously stays within the range of plus or minus 0.5 mm from the initial position for 100 milliseconds, with the unit of milliseconds; the third is the resonance intensity coefficient, performing spectral analysis on the motion trajectory of the feature points, and extracting the proportion of the energy of the main frequency component in the amplitude spectrum to the total energy to obtain the resonance intensity coefficient; finally, the maximum offset amplitude, the attitude recovery time, and the resonance intensity coefficient are combined into the dynamic stability index.

[0009] Furthermore, the decision parameters are dynamically determined through the Bayesian optimization method, including: Define the objective function as minimizing the misjudgment rejection rate, and at the same time set the defect detection rate not less than the predetermined standard value as the constraint condition; based on the fusion confidence score distribution, the occurrence frequencies of multiple types of defects, and the real-time transmission speed characteristics of the production line in the historical rejection data, use the Gaussian process regression algorithm to construct the mapping relationship between the decision parameters and the optimization objective, and the decision parameters include the first decision threshold, the second decision threshold, and the width of the fuzzy decision interval; Dynamic adjustment is achieved by iteratively executing the following process: periodically generating multiple sets of candidate parameter combinations within the range of decision parameter values, using the Gaussian process regression algorithm to predict the error rate and defect detection rate corresponding to each combination, screening candidate combinations that meet the defect detection rate constraint and have an error rate lower than the currently used parameters, and selecting combinations with lower error rates to update parameter settings; outputting the updated first decision threshold to the preliminary diversion module, and outputting the second decision threshold and the fuzzy decision interval width to the optimization decision module.

[0010] Furthermore, the rejection mechanism action instructions generated by the optimization decision module include multiple control parameters for driving the rejection execution structure, and the control parameters include: the rejection nozzle airflow injection angle dynamically determined according to the real-time posture data of the target tableware on the transmission line; the pulse air pressure intensity level set according to the degree of deviation between the fusion confidence score and the currently used second decision threshold, and at least three air pressure intensities are set, and the greater the degree of deviation, the higher the corresponding air pressure intensity level; the nozzle rejection trigger timing determined based on the production line transmission speed and the predicted position of the target tableware.

[0011] Furthermore, the multi-level self-checking procedure includes: Conduct signal integrity and operational stability diagnosis on the high-speed camera, disturbance airflow nozzle array, and transmission line encoder that constitute the disturbance section, completing equipment layer self-test; During the device-level self-test or when no abnormalities are found during the device-level self-test, drift detection analysis is performed on the initial confidence score sequence output by the prediction module, and stability analysis is performed on the fused confidence score sequence output by the disturbance response module to complete the algorithm-level self-test; When both the equipment-level self-inspection and the algorithm-level self-inspection confirm the existence of an abnormality and the cause of the abnormality has not been located, a graded alarm signal is generated and pushed to the central monitoring platform to trigger manual intervention, and a pause operation instruction is sent to the production line control system.

[0012] Furthermore, if no abnormality is found in the device layer self-test, the algorithm layer self-test is performed, including: Comparing and analyzing the real-time initial confidence score sequence currently output by the prediction module with a pre-established reference sequence of initial confidence scores during a stable operation benchmark period, and determining that the performance of the prediction model has drifted when the difference in statistical characteristics of the two sequences exceeds a preset drift threshold; The real-time fusion confidence score sequence currently output by the disturbance response module is compared with the pre-established fusion confidence score reference sequence during the stable operation benchmark period for fluctuation analysis. When the fluctuation of the real-time sequence exceeds the preset fluctuation threshold or the corresponding indicator of the reference sequence, it is determined that the stability of the fusion confidence score output has decreased; Perform pattern recognition on at least one of the real-time initial confidence score sequence output by the prediction module and the real-time fusion confidence score sequence output by the perturbation response module. When it is detected that there are multiple consecutive identical score values in the sequence, there is an abnormal amplitude mutation in the score value, or extremely high or low score values frequently appear within a short time window, it is determined that there is an abnormal pattern in the score output; Regularly input the pre-calibrated tableware reference samples into the prediction module and the perturbation response module; compare the initial confidence score output by the prediction module and the fusion confidence score output by the perturbation response module with the preset reference scores corresponding to the reference samples respectively; when the score deviation exceeds the preset allowable range, record the abnormal information and send an algorithm performance abnormal signal to the optimization decision module.

[0013] Furthermore, the device layer self-checking program includes: The device layer self-checking program specifically performs the following detection operations: sending a test signal to the high-speed camera to analyze the timing continuity and pixel integrity of the image frame; real-time collecting the air pressure sensor readings of the perturbation air nozzle array and comparing them with the target air pressure intensity set by the optimization decision module; reading the displacement pulses of the transmission line encoder, synchronously correlating the tableware pose data captured by the perturbation response module to calculate the position offset; recording the actual response time of the rejection instruction trigger moment and the rejection nozzle feedback signal.

[0014] Furthermore, the abnormal self-checking module jointly analyzes the device layer detection abnormal signal, the real-time rejection rate fluctuation data, and the score stability result output by the algorithm layer self-checking; when at least two detections in the device layer are continuously abnormal, the real-time rejection rate exceeds the dynamic safety threshold, and the algorithm layer self-checking confirms that the initial confidence score of the prediction module drifts or the stability of the fusion confidence score decreases, a downgrade instruction is sent to the optimization decision module and a protection mechanism is triggered: dynamically increasing at least one of the first decision threshold and the second decision threshold, or pausing the current rejection action, or switching to a simplified decision strategy based on the historical average confidence score.

[0015] The above solution of the present invention has at least the following beneficial effects: The above solution of the present invention, by introducing a two-stage scoring mechanism that combines image recognition and perturbation response, is beneficial to improving the recognition accuracy of bamboo tableware defects and increasing the detection rate of internal or structural defects such as hidden cracks and delamination; Adopting the Bayesian optimization algorithm to dynamically adjust the decision threshold and the fuzzy interval is beneficial to reducing the mis-rejection rate and adapting to the recognition requirements under different batches and working conditions; By integrating the dynamic stability index of optical flow analysis, the recognition of structural abnormalities is improved, and the stability of the production line is increased; The abnormal self - inspection module combines the algorithm layer and the device - layer monitoring mechanism to achieve real - time diagnosis and response to system performance drift and operation anomalies in the present invention, ensuring the continuous and stable operation of the production line. Brief Description of the Drawings

[0016] Figure 1 It is a schematic diagram of the module relationship provided by the present invention.

[0017] Figure 2 It is a schematic diagram of the production line structure in the present invention.

[0018] In the figure: 101, transmission line; 102, turning mechanism; 103, high - speed camera; 104, rejection nozzle; 105, rectangular box; 106, disturbing air - flow nozzle; 107, encoder. Detailed Embodiments

[0019] Hereinafter, exemplary embodiments of the present disclosure will be described in more detail with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0020] As Figures 1 to 2 shown, an embodiment of the present invention provides a big - data bamboo tableware manufacturing system based on production process monitoring, including a transmission line 101 and a turning mechanism 102, and further including: An acquisition module, configured to collect image data of tableware through high - speed cameras 103 deployed on both sides of the turning mechanism 102; A prediction module, configured to output an initial confidence score for each tableware based on the image data through a pre - trained prediction model; A preliminary shunting module, configured to set a first decision threshold based on the initial confidence score. If the initial confidence score is lower than the first decision threshold, it is determined as a high - probability non - conforming product, and the non - conforming tableware directly enters the rejection area. If the initial confidence score is equal to or higher than the threshold, the tableware is conveyed to the disturbance section of the transmission line 101; A disturbance response module, configured to apply a preset air - flow disturbance to the tableware entering the disturbance section, obtain a dynamic image sequence of at least one view of the tableware under the air - flow disturbance, calculate the dynamic stability index of the tableware through optical flow analysis, fuse the initial image features of the tableware and the dynamic stability index, and process them through an ensemble learning algorithm to output a fused confidence score; The optimization decision-making module is used to set the second decision threshold and the fuzzy decision interval parameters, determine the final rejection target based on the comparison result of the fusion confidence score and the fuzzy decision interval, obtain the historical rejection data and the real-time production line status data, take minimizing the mis-rejection rate and ensuring that the defect detection rate meets the requirements as the optimization goal, and dynamically determine the first decision threshold, the second decision threshold and the fuzzy decision interval parameters based on the historical rejection data and the real-time production line status data through the Bayesian optimization method. The updated first decision threshold parameter is sent to the preliminary sorting module in real time, and the second decision threshold and the fuzzy decision interval parameters are reserved for self-use. Based on the fusion confidence score and the dynamically adjusted second decision threshold, a rejection mechanism action instruction is generated. The abnormal self-checking module is used to continuously monitor the tableware rejection rate within a set time window, set a dynamic safety threshold, and the dynamic safety threshold is dynamically updated based on the historical rejection rate statistical model. When the real-time rejection rate continuously exceeds the dynamic safety threshold for a predetermined time length, a multi-level self-checking program is automatically triggered.

[0021] In the embodiment of the present invention, the tableware to be tested is placed at the left end of the transmission line 101 by the loading robotic arm. The transmission line 101 conveys the tableware to be tested from left to right. High-speed cameras 103 are symmetrically deployed on both sides of the turning mechanism 102, and high-speed cameras 103 are arranged above the disturbance section. The high-speed cameras 103 are used to collect the image data of the tableware.

[0022] In the prediction module, the pre-trained prediction model is trained based on 200,000 pieces of bamboo tableware image data with defect types and position information marked collected historically. The prediction model adopts a deep residual network architecture and extracts the complex surface features of the tableware through a cross-layer identity mapping structure.

[0023] In the preliminary sorting module, the first decision threshold is set to 0.6. When the initial confidence score output by the prediction module is lower than 0.6, it is determined as a high-probability non-conforming product, and the high-probability non-conforming tableware is rejected through the rejection areas arranged on both sides of the turning mechanism 102; if the initial confidence score is equal to or higher than 0.6, the tableware is conveyed to a 2-meter-long disturbance section through the transmission line 101.

[0024] In the disturbance response module, through an array of multi-disturbance air nozzles 106 deployed above the disturbance section, a pulsed air flow disturbance with a duration of 0.5 seconds and a flow rate of 15 m / s is applied to the tableware. The air flow direction is inclined downward at an angle of 30 degrees perpendicular to the surface of the transmission line 101. The high-speed camera 103 obtains a dynamic image sequence of the tableware under the air flow disturbance at a frame rate of 200 frames per second, calculates the dynamic stability index of the tableware through optical flow analysis, fuses the initial image features of the tableware and the dynamic stability index, and processes them through an ensemble learning algorithm to output a fusion confidence score.

[0025] In the optimization decision-making module, the second decision threshold is set to 0.75, the fuzzy decision interval parameter is [0.7, 0.8]. Based on the comparison result of the fusion confidence score and the fuzzy decision interval, the final rejection target is determined. The historical rejection data for the past 7 days and the real-time status data of the production line are obtained. With the goal of minimizing the mis-rejection rate and ensuring that the defect detection rate is not lower than the preset value (95% in this embodiment), based on the historical rejection data and the real-time status data of the production line, the first decision threshold, the second decision threshold, and the fuzzy decision interval parameter are dynamically determined through the Bayesian optimization method. The updated first decision threshold parameter is sent to the preliminary shunt module in real time, and the second decision threshold and the fuzzy decision interval parameter are reserved for self-use. Based on the fusion confidence score and the dynamically adjusted second decision threshold, the optimal rejection mechanism action instruction is generated.

[0026] In the abnormal self-check module, the tableware rejection rate within 1 hour is continuously monitored. When the real-time rejection rate continuously exceeds the dynamic safety threshold for 10 minutes, a multi-level self-check program is automatically triggered. The dynamic safety threshold is determined based on the moving average of the rejection rate in the past 24 hours, and the dynamic safety threshold is the benchmark threshold plus or minus 20%, so as to construct a statistical model to dynamically update the safety threshold.

[0027] It should be noted that: in this embodiment, the first decision threshold is set to 0.6, the second decision threshold is set to 0.75, and the fuzzy decision interval is set to [0.7, 0.8], which are only example values and can be appropriately adjusted according to the production line conditions in actual applications; In this embodiment, the surface of the transmission line 101 carrying bamboo tableware is meshed, which is convenient for the rejection mechanism to blow the bamboo tableware off the transmission line 101 through the ejected air flow passing through the meshed surface, so as to complete the rejection action.

[0028] A rectangular box 105 is arranged on the transmission line 101. The left and right vertical surfaces of the rectangular box 105 are through-shaped. The transmission line 101 is arranged inside the rectangular box 105. The high-speed camera 103 and the disturbance air nozzle 106 are installed on the inner top surface of the rectangular box 105. The section of the transmission line 101 located inside the rectangular box 105 is the disturbance section; In this embodiment, the rejection mechanism is the rejection nozzles 104 arranged in a rectangular array below the transmission line 101. The area on the transmission line 101 directly projected by the rejection nozzles 104 is the rejection area.

[0029] The pre-trained prediction model outputs the initial confidence score of each tableware, including: The prediction model architecture adopts a deep residual network, whose main structure contains 4 residual block groups, with each group consisting of 2 residual blocks; two convolutional layers are set within each residual block and perform cross-layer identity mapping operations, and the deep network degradation problem is solved through skip connection methods, which is used to extract the surface defects and geometric deformation features of bamboo tableware; a multi-scale feature fusion structure is embedded at the front end of the residual network: horizontal connection channels are respectively established at the output ends of the second and third residual block groups, and the low-level high-resolution feature map and the high-level strong semantic feature map are input into the feature pyramid network; the resolution of the high-level feature map is aligned through bilinear interpolation upsampling operations and element-wise addition fusion is performed with the corresponding-level low-level feature map; the fused multi-scale feature map is output to synchronously capture the local microscopic defects and overall morphological features of the tableware; atrous convolutional layers are embedded within the third residual block group, and the dilation rate is set to be not less than 2 to expand the receptive field, and the sensitivity to subtle fracture defects in bamboo textures is enhanced without reducing the resolution; a self-attention mechanism is integrated at the output end of the last residual block group: by calculating the correlation weight matrix between the spatial positions of the feature map, the feature weights of each region are dynamically adjusted to enable the model to focus on the high-risk defect regions; The output layer of the prediction model adopts a multi-label classification framework: the fully connected layer contains 512 neurons and receives the feature vector weighted by self-attention; the output layer sets independent sub-output layers with the same number as the number of defect types, and each sub-output layer uses the Sigmoid activation function to output the probability value of the corresponding defect; the probabilities of each defect are weighted and summed according to the preset defect risk weights to generate an initial confidence score in the range of [0,1], and the lower the score, the higher the defect risk; the training process of the prediction model adopts a supervised learning strategy: using a data set containing labeled samples, the multi-task cross-entropy loss function is minimized through the Adam optimizer.

[0030] In the embodiment of the present invention, the prediction model adopts the ResNet-50 architecture of the deep residual network, and different layers are connected through a cross-layer identity mapping structure to extract the complex surface features of the tableware. A multi-scale feature pyramid is constructed at the front end of the deep residual network. In the embodiment of the present invention, the multi-scale feature fusion structure exemplarily adopts 16×16, 32×32, and 64×64 convolutional kernels to capture local microscopic defects and overall morphology.

[0031] Embed a dilated convolutional layer in the residual block of the deep residual network. In this embodiment, the dilation rate is set to 2 to enhance the sensitivity to bamboo defects. Integrate a self-attention mechanism at the end of the deep residual network to focus on the defect-related regions by calculating the attention weights of different regions. The output layer of the prediction model adopts a multi-label classification framework, presetting defect types such as cracks, burrs, deformations, and thickness anomalies, and setting risk weights according to the impact degree of each defect on the quality of tableware. In this embodiment, the crack risk weight is set to 0.4, the burr risk weight is 0.3, the deformation risk weight is 0.2, and the thickness anomaly is 0.1. In actual applications, it can be adjusted according to requirements, and the probabilities of each defect are aggregated to generate an initial confidence score; The specific training process of the prediction model is as follows: (1) Uniformly adjust the real bamboo tableware images containing no less than 200,000 artificially labeled defect types and location information to a size of 224×224 pixels, and randomly divide them into a training set (160,000 images) and a validation set (40,000 images) with a ratio of 8:2; (2) Perform data augmentation on the training set images, including random rotation (0-15 degrees), random horizontal flipping, random brightness adjustment (±10%), and random cropping (90% of the original image size and then restored to 224×224 size), and perform pixel normalization to the range of [0,1]; (3) Use the Adam optimizer for model training, set the initial learning rate to 0.001, and the standard parameters of the Adam optimizer are β1 = 0.9, β2 = 0.999, epsilon = 1e-8. Adopt the Warm-up learning rate preheating strategy, and the learning rate gradually increases from 0.0001 to 0.001 in the first 5 epochs. At the end of the 30th, 60th, and 90th epochs, the learning rate is reduced to 0.1 times the original; (4) The training process is carried out for a maximum of 100 epochs, with a batch size of 64 for each epoch. The prediction model calculates Precision, Recall, and F1-score on the validation set, and stops training if the validation set loss value has not improved for 10 consecutive epochs; (5) Take the model parameters corresponding to the highest F1-score on the validation set as the final prediction model parameters. The training and validation processes use the well-known TensorFlow framework in the field of deep learning and are completed on a server configured with an NVIDIA Tesla V100 GPU.

[0032] Calculate the dynamic stability index of the tableware through optical flow analysis, including: Based on the continuous image sequence of tableware under airflow disturbance obtained by the disturbance response module, the image acquisition frame rate is not less than 200 frames per second; the optical flow tracking algorithm is used to track the displacement of at least five preset key geometric feature points on the surface of the tableware in the image sequence. The key geometric feature points include the center point of the tableware, the point with the maximum edge curvature, and the joint part points, and the motion trajectory data of each feature point is generated; three feature parameters are extracted from the motion trajectory data: the first is the maximum offset amplitude, which is obtained by calculating the straight-line distance from the initial position of all feature points to the maximum displacement point during the airflow disturbance, and taking the maximum value among all the distance values of the feature points as the maximum offset amplitude; the second is the attitude recovery time, which is calculated from the end moment of the airflow disturbance, and the time interval when the displacement of all feature points continuously remains within ±0.5 mm of the initial position for 100 milliseconds, with the unit of millisecond; the third is the resonance intensity coefficient, which performs spectral analysis on the motion trajectory of the feature points, extracts the proportion of the energy of the main frequency component in the amplitude spectrum, and multiplies the proportion by the main frequency value and divides by 10 to obtain the resonance intensity coefficient; finally, the maximum offset amplitude, the attitude recovery time, and the resonance intensity coefficient are combined into a dynamic stability index.

[0033] In the embodiment of the present invention, the Lucas-Kanade optical flow algorithm is used to calculate the optical flow field of the obtained dynamic image sequence of the tableware. The image acquisition frame rate is 200 frames per second, and the displacement vector and vibration frequency information of not less than 5 key geometric feature points preset on the surface of the tableware are extracted. In this embodiment, the displacement vector and vibration frequency information of 50 key geometric feature points on the surface of the tableware are extracted; Based on the above calculation results, it is further clarified that: The maximum offset amplitude is obtained by comparing the lengths of the displacement vectors of all 50 key feature points and selecting the maximum value of the displacement vector lengths; The attitude recovery time is measured starting from the end of the airflow disturbance, and the time taken when the displacements of all 50 key feature points continuously return to and remain within ±0.5 mm of the initial position for 100 milliseconds is determined; The resonance intensity coefficient is obtained by performing spectral analysis on the motion trajectories of all 50 feature points to determine the main frequency component in the amplitude spectrum, multiplying the energy proportion of the main frequency by the main frequency value and dividing by 10. The step of dividing by 10 is an exemplary setting to facilitate the normalization of the resonance intensity coefficient value, and it can be adjusted according to requirements in actual applications.

[0034] The decision parameters are dynamically determined by the Bayesian optimization method, including: Define the objective function to minimize the misjudgment rate, and set the defect detection rate not less than the predetermined standard value as the constraint condition; based on the fusion confidence score distribution, the occurrence frequencies of multiple types of defects, and the real-time transmission speed characteristics of the production line in the historical rejection data, use the Gaussian process regression algorithm to construct the mapping relationship between the decision parameters and the optimization objective. The decision parameters include the first decision threshold, the second decision threshold, and the width of the fuzzy decision interval; realize dynamic adjustment by iteratively executing the following process: periodically generate multiple groups of candidate parameter combinations within the value range of the decision parameters, use the Gaussian process regression algorithm to predict the misjudgment rate and defect detection rate corresponding to each combination, screen the candidate combinations that meet the defect detection rate constraint and have a misjudgment rate lower than the currently used parameters, and select the combination with a lower misjudgment rate from them to update the parameter settings; output the updated first decision threshold to the preliminary shunt module, and output the second decision threshold and the width of the fuzzy decision interval to the optimization decision module.

[0035] In the embodiment of the present invention, the process of dynamically determining the decision parameters by the Bayesian optimization method is as follows: Define the optimization objective as minimizing the misrejection rate, and the constraint condition is that the defect detection rate is not less than 95%. The input data for the optimization process includes the statistical distribution of the fusion confidence scores in the historical rejection data of the past 7 days, the frequencies of various defect types in the historical data, and the real-time throughput speed of the current production line (in units of the number of tableware processed per minute).

[0036] During the optimization process, first use the above historical data and real-time production line data, and adopt the Gaussian process regression algorithm to construct a surrogate model between the decision parameters (the first decision threshold, the second decision threshold, and the width of the fuzzy decision interval) and the misrejection rate and defect detection rate. Then, perform a parameter optimization iteration every 30 minutes: periodically generate multiple candidate sets of parameter combinations within the allowable value range of the decision parameters (in this embodiment, 10 candidate combinations are generated each time); use the mapping relationship constructed by the Gaussian process regression algorithm to predict the misrejection rate and defect detection rate of each candidate parameter combination; screen out the candidate parameters that simultaneously meet the defect detection rate constraint (not less than 95%) and have a predicted misrejection rate lower than the currently used parameter combination from the candidate parameter combinations; select the parameter combination with the lowest misrejection rate as the new decision parameter and update it to the system in real time, where the updated first decision threshold is sent to the preliminary shunt module, and the second decision threshold and the width of the fuzzy decision interval are reserved for the optimization decision module itself to use.

[0037] The rejection mechanism action instructions generated by the optimization decision module include multiple control parameters for driving the rejection execution structure, and the control parameters include: the airflow injection angle of the rejection nozzle 104 dynamically determined according to the real-time posture data of the target tableware on the transmission line 101; the pulse air pressure intensity level set according to the degree of deviation between the fusion confidence score and the currently used second decision threshold, and at least three air pressure intensities are set, and the greater the degree of deviation, the higher the corresponding air pressure intensity level; the nozzle rejection trigger timing determined based on the production line transmission speed and the predicted position of the target tableware.

[0038] In an embodiment of the present invention, the optimal rejection mechanism action instructions are generated as follows: the real-time posture data of the target tableware on the transmission line 101 is obtained by a laser sensor installed above the transmission line 101, and the airflow injection angle is dynamically determined based on the posture data. For example, when the center of the tableware deviates from the center line of the conveyor belt by 5 mm, the injection angle is adjusted to 30 degrees with the running direction of the transmission line 101.

[0039] The intensity level of the pulsed air pressure is determined based on the degree of deviation between the fusion confidence score and the currently used second decision threshold. The deviation is set as the absolute value of the difference between the fusion confidence score and the second decision threshold. When the deviation is less than or equal to 0.1, the intensity level is level 1, and the air pressure is 0.2 MPa; when the deviation is less than or equal to 0.1, the intensity level is level 2, and the air pressure is 0.4 MPa; when the deviation is greater than 0.2, the intensity level is level 3, and the air pressure is 0.6 MPa. The timing of the rejection nozzle 104 is based on the production line speed (collected in real time by the encoder 107 of the transmission line 101) and the target tableware position. The rejection trigger time point is determined by the formula t = (Lx) / v, where L is the distance from the rejection nozzle 104 to the rejection zone, x is the distance from the current position of the target tableware to the rejection nozzle 104, v is the production line speed, and t is the trigger time.

[0040] The multi-level self-test program includes: Perform signal integrity and operational stability diagnosis on the high-speed camera 103, the array of disturbance airflow nozzles 106, and the encoder 107 of the transmission line 101 that constitute the disturbance section, completing the equipment layer self-test; During the equipment-level self-inspection or when no abnormality is found in the equipment-level self-inspection, drift detection analysis is performed on the initial confidence score sequence output by the prediction module, and stability analysis is performed on the fused confidence score sequence output by the disturbance response module to complete the algorithm-level self-inspection; when both the equipment-level self-inspection and the algorithm-level self-inspection confirm the existence of an abnormality and the cause of the abnormality is not located, a graded alarm signal is generated and pushed to the central monitoring platform to trigger manual intervention, and a pause operation instruction is sent to the production line control system.

[0041] The device layer self-test procedure includes: The self-check program of the device layer specifically performs the following detection operations: sending a test signal to the high-speed camera 103 to analyze the timing continuity and pixel integrity of the image frames; collecting the air pressure sensor readings of the disturbance air nozzle 106 array in real time and comparing them with the target air pressure intensity set by the optimization decision module; reading the displacement pulses of the encoder 107 of the transmission line 101 and calculating the position offset by synchronously correlating the tableware pose data captured by the disturbance response module; recording the actual response time between the trigger moment of the rejection instruction and the feedback signal of the rejection nozzle 104.

[0042] When no abnormality is found in the self-check of the device layer, the self-check of the algorithm layer is executed, including: Comparing and analyzing the current real-time initial confidence score sequence output by the prediction module with the initial confidence score reference sequence of the pre-established stable operation reference period. When the statistical characteristics difference between the two sequences exceeds the preset drift threshold, it is determined that the performance of the prediction model has drifted; Comparing and analyzing the fluctuation degree of the current real-time fusion confidence score sequence output by the disturbance response module with the fusion confidence score reference sequence of the pre-established stable operation reference period. When the fluctuation degree of the real-time sequence exceeds the preset fluctuation threshold or the corresponding index of the reference sequence, it is determined that the output stability of the fusion confidence score decreases; Performing pattern recognition on at least one of the current real-time initial confidence score sequence output by the prediction module and the current real-time fusion confidence score sequence output by the disturbance response module. When it is detected that there are continuously multiple identical score values in the sequence, the score value has an abnormal amplitude mutation, or extremely high or extremely low score values frequently appear within a short time window, it is determined that there is an abnormal pattern in the score output; Regularly inputting the pre-calibrated tableware reference samples into the prediction module and the disturbance response module; comparing the initial confidence score output by the prediction module and the fusion confidence score output by the disturbance response module with the preset reference scores corresponding to the reference samples respectively; when the score deviation exceeds the preset allowable range, recording the abnormal information and sending an algorithm performance abnormal signal to the optimization decision module.

[0043] The abnormal self-check module jointly analyzes the device layer detection abnormal signal, the real-time rejection rate fluctuation data, and the score stability result output by the algorithm layer self-check; when at the same time at least two detections in the device layer are continuously abnormal, the real-time rejection rate exceeds the dynamic safety threshold, and the algorithm layer self-check confirms that the initial confidence score of the prediction module drifts or the fusion confidence score stability decreases, a downgrade instruction is sent to the optimization decision module and a protection mechanism is triggered: dynamically increasing at least one of the first decision threshold and the second decision threshold, or pausing the current rejection action, or switching to a simplified decision strategy based on the historical average confidence score.

[0044] In the embodiment of the present invention, the device layer self-check program specifically performs the following detection operations: sending a test signal to the high-speed camera 103, continuously sending 100 frames of test images, analyzing the timing continuity of the image frames, requiring that the standard deviation of the time interval between adjacent frames does not exceed 0.5 ms, and the pixel integrity is calculated by the proportion of incorrect pixels in each frame of the image, requiring that it does not exceed 0.05%. Real-time collect the air pressure sensor readings of the disturbance air nozzle 106 array, collect 10 times per minute, compare with the target air pressure intensity set by the optimization decision module, calculate the average deviation, and allow the average deviation not to exceed ±3%. Read the displacement pulses of the encoder 107 of the transmission line 101, record the number of pulses once per second, synchronously associate the tableware pose data captured by the disturbance response module, and calculate the average value of the position offset per minute. In this embodiment, the allowable position offset is not more than ±0.5 mm, and record that the average value of the rejection instruction response time is less than or equal to 15 ms to ensure the unity and operability of the device layer self-check standard.

[0045] Record the actual response time between the triggering moment of the rejection instruction and the feedback signal of the rejection nozzle 104, and count the average value of the response time once per hour, requiring that it does not exceed 15 ms to ensure the stable operation of the device.

[0046] For the real-time initial confidence score sequence currently output by the prediction module, compare and analyze it with the initial confidence score reference sequence established in advance during the stable operation benchmark period (selecting the data of the normal production period in the past 7 days), calculate the mean difference and variance difference between the two. When the mean difference exceeds the preset drift threshold (set to 0.1 in this embodiment) or the variance difference exceeds the preset drift threshold (set to 0.05 in this embodiment), it is determined that the performance of the prediction model has drifted.

[0047] For the real-time fusion confidence score sequence currently output by the disturbance response module, compare and analyze it with the fusion confidence score reference sequence established in advance during the stable operation benchmark period, calculate the standard deviation of the real-time sequence. When the standard deviation of the real-time sequence exceeds the preset fluctuation threshold of the reference sequence standard deviation (set to 15% in this embodiment), it is determined that the stability of the fusion confidence score output has decreased.

[0048] Perform pattern recognition on at least one of the real-time initial confidence score sequence output by the prediction module and the real-time fused confidence score sequence output by the perturbation response module. If five identical score values appear consecutively in the score sequence, or the numerical difference between adjacent scores exceeds a preset amplitude threshold (set to 0.2 in this embodiment), it is determined that there is an abnormal pattern in the score output. Regularly (once a week), input a set of 100 pre-calibrated tableware reference samples into the prediction module and the perturbation response module, and compare the initial confidence score output by the prediction module and the fused confidence score output by the perturbation response module with the preset reference scores corresponding to the reference samples respectively. When the score deviation exceeds ±0.1, record the abnormal information and send an algorithm performance abnormal signal to the optimization decision module.

[0049] The abnormal self-check module jointly analyzes the device layer detection abnormal signal, the real-time rejection rate fluctuation data, and the score stability result output by the algorithm layer self-check.

[0050] When at least two of the following three detections in the device layer are continuously abnormal (the continuous time exceeds 5 minutes) and the following conditions are also met simultaneously: the high-speed camera 103 in the device layer has abnormal image frame timing continuity, the air pressure sensor readings of the perturbation air nozzle 106 array deviate beyond the allowable value, and the position offset of the encoder 107 of the transmission line 101 exceeds the standard, and at the same time, the following conditions are met: the real-time rejection rate exceeds the dynamic safety threshold and the algorithm layer self-check confirms that the initial confidence score of the prediction module drifts (the mean difference exceeds 0.15) or the stability of the fused confidence score decreases (the standard deviation exceeds 20% of the reference sequence), send a downgrade instruction to the optimization decision module and trigger the protection mechanism. In this embodiment, the dynamic safety threshold is calculated based on the average rejection rate within the past 24 hours and is determined as the reference threshold for real-time monitoring by floating ±20% of this average value: Dynamically increase the first decision threshold and the second decision threshold, increase the first decision threshold and the second decision threshold dynamically according to a preset increment (0.1 in this embodiment), pause the current rejection action for 5 minutes, and at the same time switch to a simplified decision strategy based on the historical average confidence score (the average confidence score in the past 1 hour). If the abnormal situation has not been alleviated after the simplified decision strategy runs for 10 minutes, send a suspension operation instruction to the production line control system.

[0051] It should be noted that: the above-mentioned modules can be deployed in an industrial control platform with computing and control functions. The platform includes components such as a processor, a memory, a communication interface, and an execution control interface. Among them: the processor is used to execute algorithm programs such as image analysis, prediction of prediction models, and optimization of dynamic parameters; the memory is used to store image data, model parameters, historical rejection information, and control instructions; the communication interface establishes data communication with an image acquisition module (such as a high-speed camera 103), an execution module (such as a nozzle controller, controlling a rejection nozzle 104 and a disturbing air flow nozzle 106), a production line PLC system, etc.; the execution control interface is used to issue rejection instructions and collect rejection feedback signals. This is prior art and will not be elaborated in detail; In this embodiment, the turning mechanism 102 is a manipulator. The manipulator turns the bamboo tableware through a vacuum suction cup group arranged on its surface. The working principles and usage processes of the transmission line 101, the turning mechanism 102, the high-speed camera 103, the disturbing air flow nozzle 106, the rejection nozzle 104, and the encoder 107 are all prior art and will not be elaborated in detail; The optimization period for Bayesian optimization parameter adjustment is exemplarily set to be executed once every 30 minutes, and can be flexibly set according to requirements during the actual production process; As shown in the appendix Figure 2 As shown in the figure, the turning mechanism 102 is arranged on the transmission line 101. High-speed cameras 103 are symmetrically arranged on both sides of the turning mechanism 102. A rectangular box 105 is arranged on the right side of the turning mechanism 102. A disturbing air flow nozzle 106 and a high-speed camera 103 are installed inside the rectangular box 105. There are three groups of rejection nozzles 104, which are respectively arranged on the left side of the turning mechanism 102, between the turning mechanism 102 and the rectangular box 105, and on the right side of the rectangular box 105, so as to facilitate the transmission, detection, and rejection of bamboo tableware.

[0052] The above is the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A big data bamboo tableware manufacturing system based on production process monitoring, including a transmission line and a turning mechanism, characterized in that: It further includes: An acquisition module for collecting image data of tableware through high-speed cameras deployed on both sides of the turning mechanism; A prediction module for outputting an initial confidence score for each piece of tableware based on the image data through a pre-trained prediction model; A preliminary shunt module for setting a first decision threshold based on the initial confidence score. If the initial confidence score is lower than the first decision threshold, it is determined as a high-probability non-conforming product, and the non-conforming tableware directly enters the rejection area. If the initial confidence score is equal to or higher than the threshold, the tableware is conveyed to the perturbation section of the transmission line; A perturbation response module for applying a preset airflow perturbation to the tableware entering the perturbation section, obtaining a dynamic image sequence of the tableware at at least one viewing angle under the airflow perturbation, calculating the dynamic stability index of the tableware through optical flow analysis, fusing the initial image features of the tableware with the dynamic stability index, and processing through an ensemble learning algorithm to output a fused confidence score; An optimization decision module for setting a second decision threshold and fuzzy decision interval parameters, determining the final rejection target based on the comparison result of the fused confidence score and the fuzzy decision interval, obtaining historical rejection data and real-time production line status data, taking minimizing the mis-rejection rate and ensuring that the defect detection rate meets the requirements as the optimization goal, dynamically determining the first decision threshold, the second decision threshold, and the fuzzy decision interval parameters based on the historical rejection data and the real-time production line status data through the Bayesian optimization method, sending the updated first decision threshold parameter to the preliminary shunt module in real time, keeping the second decision threshold and the fuzzy decision interval parameters for its own use, and generating a rejection mechanism action instruction based on the fused confidence score and the dynamically adjusted second decision threshold; 2. The big data bamboo tableware manufacturing system based on production process monitoring according to claim 1 is characterized in that: It further includes: An abnormal self-check module for continuously monitoring the tableware rejection rate within a set time window, setting a dynamic safety threshold, which is dynamically updated based on a historical rejection rate statistical model. When the real-time rejection rate continuously exceeds the dynamic safety threshold for a predetermined time length, a multi-level self-check program is automatically triggered.

3. The big data bamboo tableware manufacturing system based on production process monitoring according to claim 1, characterized in that, The pre-trained prediction model outputs an initial confidence score for each piece of tableware, including: The prediction model architecture adopts a deep residual network, and its main structure includes 4 residual block groups, each group consisting of 2 residual blocks; Two convolutional layers are set in each residual block and cross-layer identity mapping operations are performed to extract surface defects and geometric deformation features of bamboo tableware; Lateral connection channels are respectively established at the output ends of the second residual block group and the third residual block group, and the low-level high-resolution feature map and the high-level strong semantic feature map are input into the feature pyramid network; The resolution of the high-level feature map is aligned through bilinear interpolation upsampling operation and element-wise addition fusion is performed with the corresponding low-level feature map at the same level to output a fused multi-scale feature map; Atrous convolutional layers are embedded in the third residual block group, and the correlation weight matrix between the spatial positions of the feature map is calculated to dynamically adjust the feature weights of each region; The output layer of the prediction model uses a multi-label classification framework: the fully connected layer contains 512 neurons and receives feature vectors weighted by self-attention. The output layer has the same number of independent sub-output layers as the number of defect types. Each sub-output layer uses a sigmoid activation function to output the probability value of the corresponding defect. The probability of each defect is weighted and summed according to the preset defect risk weight to generate an initial confidence score in the range of [0,1].

4. The big data bamboo tableware manufacturing system based on production process monitoring according to claim 1, characterized in that, The dynamic stability index of tableware is calculated by optical flow analysis, including: Based on a continuous image sequence of tableware under airflow disturbance obtained by the disturbance response module, the image acquisition frame rate is not less than 200 frames per second; an optical flow tracking algorithm is used to track the displacement of at least five key geometric feature points preset on the surface of the tableware in the image sequence, wherein the key geometric feature points include the center point of the tableware, the point of maximum edge curvature, and the joint point, and motion trajectory data of each feature point is generated; three characteristic parameters are extracted based on the motion trajectory data: the first is the maximum offset amplitude, which is obtained by calculating the straight-line distance of all feature points from the initial position to the maximum displacement point during the airflow disturbance, and taking the maximum value of the distance values of all feature points as the maximum offset amplitude; the second is the posture recovery time, which is calculated from the moment the airflow disturbance ends, and the time interval in milliseconds during which the displacement of all feature points remains within the range of ±0.5 mm from the initial position for 100 milliseconds; the third is the resonance intensity coefficient, which is obtained by performing spectral analysis on the motion trajectory of the feature points and extracting the ratio of the energy of the main frequency component in the amplitude spectrum to the total energy to obtain the resonance intensity coefficient; and finally, the maximum offset amplitude, posture recovery time, and resonance intensity coefficient are combined to form a dynamic stability index.

5. The big data bamboo tableware manufacturing system based on production process monitoring according to claim 1, characterized in that Dynamically determining decision parameters through Bayesian optimization methods includes: The objective function is defined as minimizing the rejection misjudgment rate, while setting a constraint that the defect detection rate is no less than a predetermined standard value. Based on the fusion confidence score distribution in historical rejection data, the frequency of multiple defect types, and the real-time transmission speed characteristics of the production line, a Gaussian process regression algorithm is used to construct a mapping relationship between decision parameters and optimization objectives. The decision parameters include a first decision threshold, a second decision threshold, and the width of the fuzzy decision interval. Dynamic adjustment is achieved by iteratively executing the following process: periodically generating multiple sets of candidate parameter combinations within the range of decision parameter values, using the Gaussian process regression algorithm to predict the error rate and defect detection rate corresponding to each combination, screening candidate combinations that meet the defect detection rate constraint and have an error rate lower than the currently used parameters, and selecting combinations with lower error rates to update parameter settings; outputting the updated first decision threshold to the preliminary diversion module, and outputting the second decision threshold and the fuzzy decision interval width to the optimization decision module.

6. The big data bamboo tableware manufacturing system based on production process monitoring according to claim 1 is characterized in that: The rejection mechanism action instructions generated by the optimization decision module include multiple control parameters for driving the rejection execution structure. The control parameters include: the air jet angle of the rejection nozzle dynamically determined according to the real-time pose data of the target tableware on the conveyor line; the pulse air pressure intensity level set by grading according to the deviation degree between the fusion confidence score and the currently used second decision threshold, with at least three air pressure intensities set, and the greater the deviation degree, the higher the corresponding air pressure intensity level; the nozzle rejection trigger timing determined based on the production line transmission speed and the predicted position of the target tableware.

7. The big data bamboo tableware manufacturing system based on production process monitoring according to claim 2, wherein, The multi-level self-checking program includes: Diagnosing the signal integrity and operation stability of the high-speed camera, the disturbance air nozzle array, and the conveyor line encoder that constitute the disturbance section to complete the equipment-level self-check; While or when no abnormality is found in the equipment-level self-check, performing drift detection analysis on the initial confidence score sequence output by the prediction module and performing stability analysis on the fusion confidence score sequence output by the disturbance response module to complete the algorithm-level self-check; When abnormalities are confirmed in both the equipment-level self-check and the algorithm-level self-check and the cause of the abnormality is not located, generating a hierarchical alarm signal and pushing it to the central monitoring platform to trigger manual intervention, and at the same time sending a suspension operation instruction to the production line control system.

8. The big data bamboo tableware manufacturing system based on production process monitoring according to claim 7, wherein When no abnormality is found in the equipment-level self-check, performing the algorithm-level self-check, including: Comparing and analyzing the current real-time initial confidence score sequence output by the prediction module with the initial confidence score reference sequence of the pre-established stable operation reference period. When the statistical characteristics difference between the two sequences exceeds the preset drift threshold, it is determined that the performance of the prediction model has drifted; Comparing the fluctuation degree of the current real-time fusion confidence score sequence output by the disturbance response module with the fusion confidence score reference sequence of the pre-established stable operation reference period. When the fluctuation degree of the real-time sequence exceeds the preset fluctuation threshold or the corresponding index of the reference sequence, it is determined that the stability of the fusion confidence score output has decreased; Performing pattern recognition on at least one of the real-time initial confidence score sequence output by the prediction module and the real-time fusion confidence score sequence output by the disturbance response module. When it is detected that there are multiple consecutive identical score values, abnormal amplitude mutations occur in the score values, or extremely high or extremely low score values frequently appear within a short time window in the sequence, it is determined that there is an abnormal pattern in the score output; Regularly inputting the pre-calibrated tableware reference sample into the prediction module and the disturbance response module; comparing the initial confidence score output by the prediction module and the fusion confidence score output by the disturbance response module with the preset reference score corresponding to the reference sample respectively; when the score deviation exceeds the preset allowable range, recording the abnormal information and sending an algorithm performance abnormality signal to the optimization decision module.

9. The big data bamboo tableware manufacturing system based on production process monitoring according to claim 8 is characterized in that: The equipment-level self-checking program includes: The self-check program of the device layer specifically performs the following detection operations: sending a test signal to the high-speed camera to analyze the timing continuity and pixel integrity of the image frames; collecting the air pressure sensor readings of the turbulent air nozzle array in real time and comparing them with the target air pressure intensity set by the optimization decision module; reading the displacement pulses of the transmission line encoder and calculating the position offset by synchronously correlating the tableware pose data captured by the disturbance response module; recording the actual response time between the trigger moment of the rejection instruction and the feedback signal of the rejection nozzle.

10. The big data bamboo tableware manufacturing system based on production process monitoring according to claim 7, characterized in that, The abnormal self-check module jointly analyzes the device layer detection abnormal signal, the real-time rejection rate fluctuation data, and the scoring stability result output by the algorithm layer self-check; when at least two detections in the device layer are continuously abnormal, the real-time rejection rate exceeds the dynamic safety threshold, and the algorithm layer self-check confirms that the initial confidence score of the prediction module drifts or the stability of the fusion confidence score decreases, a downgrading instruction is sent to the optimization decision module and a protection mechanism is triggered: dynamically increasing at least one of the first decision threshold and the second decision threshold, or pausing the current rejection action, or switching to a simplified decision strategy based on the historical average confidence score.

Images