Carbon fixation fiberboard production quality regulation system and method
By using multi-frequency ultrasonic feature sensing and a pre-trained carbonization quality diagnostic model, combined with adaptive adjustment of process parameters, the problem of low efficiency and consistency in quality control during the production of solid carbon fiber sheets has been solved, achieving efficient and non-destructive quality inspection and process optimization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- WUHAN UNIV OF TECH
- Filing Date
- 2026-03-26
- Publication Date
- 2026-06-12
AI Technical Summary
In existing technologies, the quality control of solid carbon fiber boards relies on destructive sampling inspection, which is inefficient and cannot reflect the true performance of individual products. This results in problems such as high defect rates, waste of carbon source, and difficulty in ensuring product performance consistency.
A multi-frequency ultrasonic feature sensing module is used to collect ultrasonic pulses of different frequencies. Combined with a pre-trained carbonization quality diagnostic model, the predicted bending strength and carbonization state of the solid carbon fiber board are diagnosed through acoustic feature vectors. The process parameter adaptive adjustment module adjusts the carbonization process parameters on the production line according to the diagnostic results.
This has enabled a shift from open-loop to closed-loop production quality improvement in solid carbon fiber boards, increasing product qualification rates, reducing carbon source waste, and ensuring consistent product performance.
Smart Images
Figure CN122196691A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent manufacturing technology, and in particular to a quality control system and method for the production of solid carbon fiber sheets. Background Technology
[0002] Solid carbon fiber boards are high-performance green building materials produced through carbonization curing technology, which converts industrial solid waste into carbon dioxide. Their core mechanical properties and durability directly depend on the completeness and uniformity of the internal carbonization reaction. Therefore, accurate determination and stable control of the degree of carbonization are crucial to ensuring the reliability of the product structure and realizing its environmental value.
[0003] In existing technologies, quality control of solid carbon fiber sheets relies on destructive sampling inspection, which involves taking a small number of samples from a batch for mechanical destructive testing. This method is inefficient, destructive, and cannot reflect the true performance of individual solid carbon fiber sheets, posing a serious risk of missed quality inspections. It has become a bottleneck restricting the automation of production lines and the improvement of product quality consistency. To achieve non-destructive testing, existing technologies such as infrared thermography and conventional ultrasonic thickness measurement have emerged. Infrared thermography is sensitive to surface conditions and environmental interference and cannot detect internal chemical changes. Conventional ultrasonic thickness measurement or velocity measurement mainly reflects the overall density and elastic modulus of the material, and cannot distinguish the performance improvement brought about by the carbonization reaction from the impact of other process fluctuations. Moreover, the production quality control of solid carbon fiber sheets is generally in an open-loop state, making it impossible to trace the process root cause of quality fluctuations, let alone achieve proactive optimization of the production process.
[0004] It is evident that existing technologies suffer from high defect rates, wasteful carbon source consumption, and difficulty in ensuring product performance consistency. Summary of the Invention
[0005] In view of this, it is necessary to provide a quality control system and method for the production of solid carbon fiber boards to solve the technical problems of high defect rates, waste of carbon source consumption, and difficulty in ensuring product performance consistency in the existing technology.
[0006] To address the aforementioned problems, in a first aspect, the present invention provides a quality control system for the production of solid carbon fiber sheets, comprising: The multi-frequency ultrasonic feature sensing module is used to transmit ultrasonic pulses of different preset frequencies to the solid carbon fiber board on the solid carbon fiber board production line, receive the transmitted ultrasonic pulses after the ultrasonic pulses are transmitted through the solid carbon fiber board, and determine the acoustic feature vector of the solid carbon fiber board based on the transmitted ultrasonic pulses. The carbonization state perception module is used to diagnose the acoustic feature vector using a pre-trained carbonization quality diagnostic model to determine the predicted bending strength value and carbonization state of the solid carbon fiber plate. The process parameter adaptive adjustment module is used to determine the carbonization trend of the carbon fiber sheets on the carbon fiber sheet production line based on the predicted bending strength value and carbonization state of a continuous preset number of carbon fiber sheets, and to adjust the carbonization process parameters on the carbon fiber sheet production line with the aim of keeping the carbonization trend within a preset target threshold range.
[0007] In one possible implementation of the present invention, the multi-frequency ultrasound feature sensing module includes: A signal generator is used to generate multiple ultrasonic pulses of different frequencies; An ultrasonic transmitting probe, which consists of multiple probes, is used to transmit multiple ultrasonic pulses of different frequencies to a solid carbon fiber plate. The signal receiving unit corresponds one-to-one with the ultrasonic transmitting probe and is used to receive the transmitted ultrasonic pulses after they have passed through the solid carbon fiber plate. An ultrasonic signal processor is used to calculate the acoustic feature vector of a solid carbon fiber plate based on transmitted ultrasonic pulses and the structure of the solid carbon fiber plate.
[0008] In one possible implementation of the present invention, the ultrasonic signal processor, when calculating the acoustic feature vector of the solid carbon fiber plate based on the transmitted ultrasonic pulse and the structure of the solid carbon fiber plate, is used for: The propagation speed of ultrasonic pulses of different frequencies in solid carbon fiber plates was calculated based on the propagation time of ultrasonic pulses of different frequencies in solid carbon fiber plates and the thickness of solid carbon fiber plates. The energy attenuation coefficient of the ultrasonic pulse in the solid carbon fiber plate is calculated based on the signal amplitude of the transmitted ultrasonic pulse at different frequencies and the reference signal amplitude under the preset state of the unconsolidated carbon fiber plate. By integrating the propagation velocity and energy attenuation coefficient of ultrasonic pulses of different frequencies in the solid carbon fiber plate, the acoustic characteristic vector of the solid carbon fiber plate is obtained.
[0009] In one possible implementation of the present invention, ultrasonic transmitting probes and signal receiving units are installed one-to-one on both sides of the solid carbon fiber board production line, and the distance between the ultrasonic transmitting probes and signal receiving units is adjustable to accommodate the thickness of the solid carbon fiber board.
[0010] In one possible implementation of the present invention, the training process of the pre-trained carbonization quality diagnostic model includes: Construct a training data sample set. A sample data in the training data sample set includes the acoustic feature vector of the sample solid carbon fiber plate and the measured bending strength value and carbonization degree of the sample solid carbon fiber plate. Using the acoustic feature vector of the sample solid carbon fiber board as input, and the measured bending strength and carbonization degree of the sample solid carbon fiber board as output, the preset machine learning model is trained to obtain a pre-trained carbonization quality diagnostic model.
[0011] In one possible implementation of the present invention, the process parameter adaptive adjustment module, when determining the carbonization trend of the solid carbon fiber sheets on the solid carbon fiber sheet production line based on the predicted flexural strength values and carbonization state of a continuous preset number of solid carbon fiber sheets, is used to: The average predicted bending strength and trend of the carbon fiber sheets on the carbon fiber sheet production line are calculated based on the predicted bending strength values of a continuous preset number of carbon fiber sheets. The carbonization trend of carbon fiber sheets on the carbon fiber sheet production line is calculated based on the carbonization status level of a continuous, preset number of carbon fiber sheets.
[0012] In one possible implementation of the present invention, when the process parameter adaptive adjustment module adjusts the process parameters of the carbonization process on the solid carbon fiber board production line with the aim of keeping the carbonization trend within a preset target threshold range, it is used to: When the predicted bending strength trend continues to decline and the average predicted bending strength is lower than the preset average bending strength, or when the carbonization trend is insufficient or insufficient carbonization uniformity, adjust the process parameters in the carbonization process on the solid carbon fiber board production line. The process parameters include carbon dioxide concentration, carbonization duration, and carbonization humidity.
[0013] In one possible implementation of the present invention, the multi-frequency ultrasound feature sensing module is further configured to: The acoustic characteristic vector of the reference solid carbon fiber plate is measured according to the preset production cycle measurement standard. The process parameter adaptive adjustment module is also used for: The acoustic characteristic vector of the solid carbon fiber board on the solid carbon fiber board production line is compensated and corrected based on the measurement deviation between the measured acoustic characteristic vector and the standard reference acoustic characteristic vector of the standard reference solid carbon fiber board.
[0014] In one possible implementation of the present invention, the carbonization state sensing module is further configured to: Acoustic feature vectors, predicted bending strength values, and carbonization states of solid carbon fiber sheets on the solid carbon fiber sheet production line are collected according to the preset model iteration optimization cycle to construct a model iteration optimization dataset. The pre-trained carbonization quality diagnostic model was retrained using model iteration optimization dataset.
[0015] Secondly, the present invention also provides a method for controlling the production quality of solid carbon fiber boards, applicable to any of the aforementioned implementations of the solid carbon fiber board production quality control method system, comprising: The acoustic characteristic vector of the carbon fiber board is determined based on the transmitted ultrasonic pulses after passing through the carbon fiber board on the carbon fiber board production line, and ultrasonic pulses of different preset frequencies are collected. A pre-trained carbonization quality diagnostic model is used to diagnose the acoustic feature vectors and determine the predicted bending strength and carbonization state of the solid carbon fiber plate. The carbonization trend of the carbon fiber sheets on the carbon fiber sheet production line is determined based on the predicted bending strength and carbonization state of a continuous preset number of carbon fiber sheets, and the process parameters of the carbonization process on the carbon fiber sheet production line are adjusted with the aim of keeping the carbonization trend within the preset target threshold range.
[0016] The beneficial effects of this invention are as follows: The solid carbon fiber board production quality control system provided by this invention acquires transmitted ultrasonic pulses of different frequencies after they are transmitted through the solid carbon fiber board via a multi-frequency ultrasonic feature sensing module, and determines the acoustic feature vector of the solid carbon fiber board based on these transmitted ultrasonic pulses. This solves the problem that single ultrasonic measurement is easily interfered with by various confounding factors, ensuring the accuracy of the acoustic feature vector acquisition of the solid carbon fiber board. The carbonization state sensing module diagnoses the acoustic feature vector through a pre-trained carbonization quality diagnostic model, determining the predicted flexural strength value and carbonization state of the solid carbon fiber board. Unlike traditional simple pass / fail detection, this invention diagnoses the acoustic feature vector of the solid carbon fiber board through a model, determining the predicted flexural strength value and carbonization state, which facilitates the subsequent adjustment of process parameters. The process parameter adaptive adjustment module determines the carbonization trend of the solid carbon fiber board on the solid carbon fiber board production line based on the predicted flexural strength value and carbonization state of a continuous preset number of solid carbon fiber boards, and adjusts the carbonization process parameters on the solid carbon fiber board production line with the aim of keeping the carbonization trend within a preset target threshold range. This technology enables the improvement of the quality inspection of solid carbon fiber boards from open-loop to closed-loop, as well as the adaptive control of the production process parameters of solid carbon fiber boards, thereby increasing the pass rate of solid carbon fiber boards on the production line, reducing carbon source waste, and ensuring consistent product performance. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is a schematic diagram of a quality control system for the production of solid carbon fiber boards provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a multi-frequency ultrasonic feature sensing module provided in an embodiment of the present invention; Figure 3 A flowchart illustrating an acoustic feature vector determination method provided in an embodiment of the present invention; Figure 4A schematic flowchart of a model training method provided in an embodiment of the present invention; Figure 5 A schematic flowchart of a carbonization quality prediction method provided in an embodiment of the present invention; Figure 6 A flowchart illustrating a model optimization method provided in an embodiment of the present invention; Figure 7 This is a flowchart illustrating a method for controlling the production quality of solid carbon fiber boards, as provided in an embodiment of the present invention. Detailed Implementation
[0019] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.
[0020] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.
[0021] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.
[0022] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.
[0023] A specific embodiment of the present invention, such as Figure 1 As shown, a quality control system for the production of solid carbon fiber boards is disclosed, comprising: The multi-frequency ultrasonic feature sensing module 101 is used to transmit ultrasonic pulses of different preset frequencies to the solid carbon fiber board on the solid carbon fiber board production line, receive the transmitted ultrasonic pulses after the ultrasonic pulses are transmitted through the solid carbon fiber board, and determine the acoustic feature vector of the solid carbon fiber board based on the transmitted ultrasonic pulses. The carbonization state perception module 102 is used to diagnose the acoustic feature vector using a pre-trained carbonization quality diagnosis model to determine the predicted bending strength value and carbonization state of the solid carbon fiber plate. The process parameter adaptive adjustment module 103 is used to determine the carbonization trend of the solid carbon fiber board on the solid carbon fiber board production line based on the predicted bending strength value and carbonization state of a continuous preset number of solid carbon fiber boards, and to adjust the carbonization process parameters on the solid carbon fiber board production line with the aim of keeping the carbonization trend within a preset target threshold range.
[0024] In this embodiment of the invention, the provided solid carbon fiber board production quality control system is used to adaptively control the process parameters of the solid carbon fiber board production line to ensure the product quality of the solid carbon fiber board.
[0025] In this embodiment of the invention, the multi-frequency ultrasonic feature sensing module is the hardware foundation for achieving specific sensing of the carbonization state. Its core function is to acquire, online and non-destructively, multi-dimensional acoustic feature signals that can sensitively reflect the evolution of the internal carbonization microstructure of the solidified carbon fiber plate. In this embodiment, ultrasonic pulses of different frequencies are selected from low and high frequencies. The low frequency is selected from 100-500 kHz, as the ultrasonic wavelength in this band is relatively long and sensitive to changes in the overall elastic modulus and macroscopic density of the material. Its propagation speed mainly reflects the overall rigidity and basic density of the skeleton formed after carbonization curing of the solidified carbon fiber plate, and the size of the micron-sized calcium carbonate crystals generated in the reaction is on the same order of magnitude. The high frequency is selected from 1 MHz-3 MHz. High-frequency sound waves encounter a large number of newly formed crystal-matrix interfaces during propagation, producing a strong scattering effect. Therefore, the attenuation coefficient at high frequencies is extremely sensitive to the number, size, and uniformity of distribution of calcium carbonate crystals, and is a specific indicator characterizing the carbonization reaction process. The ultrasonic parameters of a single frequency are easily affected by various confounding factors such as fluctuations in the initial density and moisture content of the solidified carbon fiber plate. This invention constructs a multi-dimensional feature vector by jointly measuring the acoustic characteristics of low and high frequencies. It can utilize the differences in sensitivity of different frequency parameters to different influencing factors, and effectively separate and extract the signal features contributed by the "degree of carbonization" through subsequent intelligent diagnostic models, thereby overcoming the shortcomings of traditional single-parameter methods in terms of poor anti-interference ability and weak specificity.
[0026] In this embodiment of the invention, the carbonization state sensing module is the intelligent hub of the system. Its core function is to transform the raw acoustic signals collected by the multi-frequency ultrasonic feature sensing module into a quantitative and executable engineering evaluation of the carbonization degree and mechanical properties of the solid carbon fiber plate. It completes the mapping and diagnosis from multi-dimensional physical characteristics to the chemical-mechanical state of the material through an embedded dedicated machine learning model. The pre-trained carbonization quality diagnostic model is preferably a random forest or gradient boosting tree model, as it has good handling capabilities for mixed-type features and nonlinear relationships, and is less prone to overfitting. The pre-trained carbonization quality diagnostic model is used to diagnose the acoustic feature vectors, determining the predicted flexural strength value and carbonization state of the solid carbon fiber plate. The carbonization state can be directly used for sorting decisions, such as rejecting under-carbonized products. Simultaneously, the predicted flexural strength value provides a quantitative basis for process control, realizing the decoupling and synergy of sorting and control functions.
[0027] In this embodiment of the invention, the adaptive process parameter adjustment module is the execution unit for achieving closed-loop quality control. Its core function is to intelligently generate and execute adjustment instructions for upstream carbonization curing process parameters based on the historical trends of diagnostic results, transforming quality control from passive rejection to proactive optimization. Specific adjustment strategies for process parameters will be described in detail later in this invention.
[0028] The carbon fiber sheet production quality control system provided by this invention acquires transmitted ultrasonic pulses of different frequencies after they are transmitted through the carbon fiber sheet using a multi-frequency ultrasonic feature sensing module. Based on these transmitted ultrasonic pulses, the acoustic feature vector of the carbon fiber sheet is determined, solving the problem of single ultrasonic measurement being easily interfered with by various confounding factors and ensuring the accuracy of the acoustic feature vector acquisition. The carbonization state sensing module diagnoses the acoustic feature vector using a pre-trained carbonization quality diagnostic model to determine the predicted flexural strength value and carbonization state of the carbon fiber sheet. Unlike traditional simple pass / fail detection, this invention diagnoses the acoustic feature vector of the carbon fiber sheet using a model to determine the predicted flexural strength value and carbonization state, which is then used to adjust subsequent process parameters. The adaptive process parameter adjustment module determines the carbonization trend of the carbon fiber sheets on the production line based on the predicted flexural strength values and carbonization states of a continuous preset number of carbon fiber sheets, and adjusts the carbonization process parameters on the production line with the aim of keeping the carbonization trend within a preset target threshold range. This technology enables the improvement of the quality inspection of solid carbon fiber boards from open-loop to closed-loop, as well as the adaptive control of the production process parameters of solid carbon fiber boards, thereby increasing the pass rate of solid carbon fiber boards on the production line, reducing carbon source waste, and ensuring consistent product performance.
[0029] In some possible embodiments of the present invention, such as Figure 2 As shown, the multi-frequency ultrasound feature sensing module includes: Signal generator 201 is used to generate multiple ultrasonic pulses of different frequencies; Ultrasonic transmitting probe 202, there are multiple ultrasonic transmitting probes, used to transmit multiple ultrasonic pulses of different frequencies to the solid carbon fiber plate; The signal receiving unit 203 corresponds one-to-one with the ultrasonic transmitting probe and is used to receive the transmitted ultrasonic pulses after they have passed through the solid carbon fiber plate. An ultrasonic signal processor 204 is used to calculate the acoustic feature vector of a solid carbon fiber plate based on transmitted ultrasonic pulses and the structure of the solid carbon fiber plate.
[0030] In this embodiment of the invention, the signal generator of the multi-frequency ultrasonic feature sensing module is used to generate ultrasonic pulse excitation signals of specific frequencies and pulse widths. The signal generator is connected to an ultrasonic transmitting probe, which includes at least two independent ultrasonic transmitting probes. The center operating frequency of the first transmitting probe is set in the low-frequency band, and the center operating frequency of the second transmitting probe is set in the high-frequency band. The probe type is an air-coupled ultrasonic transducer, which does not require a coupling agent. The signal receiving unit corresponds one-to-one with the transmitting probe and is precisely aligned to receive the ultrasonic signals after penetrating the solid carbon fiber plate under test. When the solid carbon fiber plate under test enters the detection station with the conveyor belt, the detection program is triggered by the photoelectric sensor. The signal generator sequentially or synchronously drives the low-frequency transmitting probe and the high-frequency transmitting probe to emit ultrasonic pulses to the solid carbon fiber plate, and the corresponding receiving probes receive the low-frequency and high-frequency ultrasonic signals transmitted through the solid carbon fiber plate, respectively.
[0031] This invention, through the setting of a multi-frequency ultrasonic feature sensing module, enables the measurement of solid carbon fiber plates by multi-frequency ultrasonic pulse signals, thus eliminating the problem of inaccurate measurement caused by interference from multiple factors with a single pulse wave.
[0032] In some possible embodiments of the present invention, such as Figure 3 As shown, the ultrasonic signal processor is used to calculate the acoustic eigenvectors of the solid carbon fiber plate based on the transmitted ultrasonic pulses and the structure of the solid carbon fiber plate, for: S301, Calculate the propagation speed of ultrasonic pulses of different frequencies in a solid carbon fiber plate based on the propagation time of ultrasonic pulses of different frequencies in the solid carbon fiber plate and the thickness of the solid carbon fiber plate. S302, calculate the energy attenuation coefficient of the ultrasonic pulse in the solid carbon fiber plate based on the signal amplitude of the transmitted ultrasonic pulse at different frequencies and the preset reference signal amplitude in the state of the solid carbon fiber plate. S303, by integrating the propagation speed and energy attenuation coefficient of ultrasonic pulses of different frequencies in the solid carbon fiber plate, the acoustic characteristic vector of the solid carbon fiber plate is obtained.
[0033] In this embodiment of the invention, for transmitted ultrasonic pulses, the acoustic characteristic vector of the solidified carbon fiber plate is accurately calculated using a time-domain analysis algorithm. Specifically, based on the known thickness of the solidified carbon fiber plate and the measured sound wave propagation time, the propagation speed of ultrasonic pulse signals of different frequencies in the solidified carbon fiber plate is calculated. By comparing the received signal amplitude with the reference signal amplitude measured without the solidified carbon fiber plate, the energy attenuation coefficient of ultrasonic pulse signals of different frequencies in the solidified carbon fiber plate is obtained. The formula for the energy attenuation coefficient is:
[0034] in, denoted as the energy attenuation coefficient of the ultrasonic pulse signal in the solid carbon fiber plate, and d is the thickness of the solid carbon fiber plate. A is the amplitude of the reference signal measured under the condition of a solid carbon fiber plate, and A is the amplitude of the transmitted pulse signal.
[0035] Furthermore, for each tested carbon fiber plate, a multidimensional acoustic feature vector F containing four core elements is generated, expressed as follows: ,in, The propagation speed of low-frequency ultrasonic pulse signals in solid carbon fiber plates. This represents the propagation speed of a high-frequency ultrasonic pulse signal in a solid carbon fiber plate. The energy attenuation coefficient of low-frequency ultrasonic pulse signals in solid carbon fiber plates. The energy attenuation coefficient of high-frequency ultrasonic pulse signals in solid carbon fiber plates.
[0036] This invention measures the acoustic characteristics of solid carbon fiber boards using ultrasonic pulse signals of different frequencies and constructs an acoustic feature vector for the solid carbon fiber boards, which facilitates subsequent quality inspection of the solid carbon fiber boards.
[0037] In some possible embodiments of the present invention, ultrasonic transmitting probes and signal receiving units are installed one-to-one on both sides of the solid carbon fiber board production line, and the distance between the ultrasonic transmitting probes and signal receiving units is adjustable to accommodate the thickness of the solid carbon fiber board.
[0038] In this embodiment of the invention, the ultrasonic transmitting probe and signal receiving unit in the foregoing embodiments correspond one-to-one and are securely installed on both sides of the production line conveyor belt, ensuring that the sound beam axis is perpendicular to the surface of the solid carbon fiber plate, and the probe spacing is adjustable to adapt to different plate thicknesses.
[0039] In some possible embodiments of the present invention, such as Figure 4 As shown, the training process of the pre-trained carbonization quality diagnostic model includes: S401, Construct a training data sample set. A sample data in the training data sample set includes the acoustic feature vector of the sample solid carbon fiber plate and the measured bending strength value and carbonization degree of the sample solid carbon fiber plate. S402 uses the acoustic feature vector of the sample solid carbon fiber plate as input and the measured bending strength and carbonization degree of the sample solid carbon fiber plate as output to train the preset machine learning model and obtain the pre-trained carbonization quality diagnosis model.
[0040] In this embodiment of the invention, the pre-trained carbonization quality diagnostic model is a machine learning model. This model is trained on a large number of laboratory samples, with the training data consisting of the acoustic feature vectors of the samples and their corresponding measured flexural strength values and carbonization degrees. The model is trained and validated using hundreds of sets of sample data covering the full range from under-carbonization to full carbonization. Random forest regression and classification ensemble models or gradient boosting decision tree models are preferred because they excel at handling nonlinear relationships, are insensitive to feature scale, and effectively prevent overfitting.
[0041] In this embodiment of the invention, a sample data in the training data sample set includes the acoustic feature vector of a sample solid carbon fiber board, as well as the measured flexural strength value and carbonization degree of the sample solid carbon fiber board. The measured flexural strength value and carbonization degree serve as the label values of the sample. Using the acoustic feature vector of the sample solid carbon fiber board as input and the measured flexural strength value and carbonization degree of the sample solid carbon fiber board as output, a pre-set machine learning model is trained to obtain a pre-trained carbonization quality diagnostic model. Specifically, a batch of solid carbon fiber board samples with different carbonization degrees are prepared. By controlling process parameters such as carbonization reaction time, carbon dioxide pressure, and temperature, samples ranging from under-carbonized, moderately carbonized to fully carbonized are obtained. Acoustic signals are acquired for each sample, an acoustic feature vector is constructed, and the measured flexural strength value and carbonization degree are collected to construct a sample dataset. The constructed sample data is randomly divided into a training set and a validation set. This embodiment of the invention uses a random forest regression and classification ensemble model. Specifically, a random forest model with two output heads is constructed: a regression head to predict the bending strength value and a classification head to predict the degree of carbonization. The degree of carbonization can be considered a continuous value or predicted through regression; if it is a discrete level, classification is used. The model is trained using training set data. The model takes acoustic feature vectors as input and outputs the final predicted values for bending strength and carbonization by integrating the prediction results of multiple decision trees. During training, the model learns the complex nonlinear mapping relationship between acoustic features and the two target values. After training, the model's performance is evaluated using validation set data, with mean absolute error and coefficient of determination used as evaluation metrics. After obtaining the pre-trained model, it can be deployed on the production line. When a new solid carbon fiber board needs quality diagnosis, simply follow the method described in the previous embodiment to collect its acoustic signal and extract the same acoustic feature vector. Input this vector into the trained carbonization quality diagnosis model, and the model can simultaneously output the predicted values for bending strength and carbonization of the fiber board within milliseconds, thereby achieving rapid, non-destructive, and intelligent diagnosis of product quality.
[0042] The embodiments of the present invention facilitate subsequent quality diagnosis of solid carbon fiber plates by training a carbonization quality diagnostic model.
[0043] In some possible embodiments of the present invention, such as Figure 5 As shown, the adaptive adjustment module for process parameters, when determining the carbonization trend of the carbon fiber sheets on the carbon fiber sheet production line based on the predicted flexural strength and carbonization state of a continuous preset number of carbon fiber sheets, is used for: S501, Calculate the average predicted bending strength and trend of the carbon fiber sheets on the carbon fiber sheet production line based on the predicted bending strength values of a continuous preset number of carbon fiber sheets. S502, calculate the carbonization trend of solid carbon fiber sheets on the solid carbon fiber sheet production line based on the carbonization state level of a continuous preset number of solid carbon fiber sheets.
[0044] In this embodiment of the invention, based on the predicted flexural strength values of a continuous preset number of solid carbon fiber sheets, the average predicted flexural strength and the predicted flexural strength trend of the solid carbon fiber sheets on the solid carbon fiber sheet production line are calculated. The average predicted flexural strength value can be calculated by averaging the predicted flexural strength values of the most recent 50 solid carbon fiber sheets on the production line. The predicted flexural strength trend is used to characterize the direction and rate of change of flexural strength over time. Methods for calculating the trend may include, but are not limited to, linear regression slope method, difference method, and sign determination method. Furthermore, based on the carbonization state levels of a continuous preset number of solid carbon fiber sheets, the carbonization state trend of the solid carbon fiber sheets on the solid carbon fiber sheet production line is calculated. The carbonization state level can be a discrete level divided according to the degree of carbonization, such as level 1 (under-carbonized), level 2 (qualified carbonization), level 3 (fully carbonized), etc. Through the above two steps, this module integrates the flexural strength trend and the carbonization state trend to comprehensively grasp the dynamic changes in product quality on the production line. For example, if the average bending strength is lower than the set threshold and shows a downward trend, and the carbonization level also shows a downward trend, it indicates that the product is drifting towards "under-carbonization and insufficient strength", and the process parameters need to be adjusted in time to correct the deviation.
[0045] In some possible embodiments of the present invention, when the process parameter adaptive adjustment module adjusts the process parameters of the carbonization process on the solid carbon fiber board production line with the aim of keeping the carbonization trend within a preset target threshold range, it is used to: When the predicted bending strength trend continues to decline and the average predicted bending strength is lower than the preset average bending strength, or when the carbonization trend is insufficient or insufficient carbonization uniformity, adjust the process parameters in the carbonization process on the solid carbon fiber board production line. The process parameters include carbon dioxide concentration, carbonization duration, and carbonization humidity.
[0046] In this embodiment of the invention, when a trend indicator exceeds a preset control threshold, for example, if the undercarbonization rate exceeds 5% for three consecutive windows, or the average intensity continues to decrease beyond a set value, the control decision-maker is triggered. The control decision-maker generates a fine-tuning instruction based on a predetermined regulation-response relationship model. If an undercarbonization trend is diagnosed, the instruction is to increase the CO2 concentration setpoint of the carbonization curing vessel by ΔC%, for a duration of T minutes. If a deterioration in carbonization uniformity is diagnosed, the instruction may be to simultaneously fine-tune the humidity setpoint by ΔH%. The instruction is sent to the carbonization curing vessel control system via an industrial communication interface, which then performs the final control of the actuators (such as CO2 flow valves and humidifiers).
[0047] In some possible embodiments of the present invention, the multi-frequency ultrasound feature sensing module is further used for: The acoustic characteristic vector of the reference solid carbon fiber plate is measured according to the preset production cycle measurement standard. The process parameter adaptive adjustment module is also used for: The acoustic characteristic vector of the solid carbon fiber board on the solid carbon fiber board production line is compensated and corrected based on the measurement deviation between the measured acoustic characteristic vector and the standard reference acoustic characteristic vector of the standard reference solid carbon fiber board.
[0048] In this embodiment of the invention, to ensure the long-term stability of the solid carbon fiber board production quality control system, the system automatically initiates a calibration process after every N solid carbon fiber boards produced (e.g., N=200) or at the start of each shift. The control program drives the mechanical actuator to move the standard reference solid carbon fiber board to a fixed position at the testing station. The system performs a complete multi-frequency ultrasonic measurement and diagnosis on the standard solid carbon fiber board. The measured acoustic characteristics are compared with the reference characteristic values of the standard solid carbon fiber board to calculate the current system measurement deviation. In subsequent online testing, all measured values are automatically compensated and corrected to eliminate system errors caused by sensor drift and changes in ambient temperature and humidity.
[0049] Furthermore, such as Figure 6 As shown, the carbonization state sensing module is also used for: S601: Collect acoustic feature vectors, predicted bending strength values and carbonization states of solid carbon fiber boards on the solid carbon fiber board production line according to the preset model iteration optimization cycle, and construct the model iteration optimization dataset. S602 uses model iteration to optimize the dataset and retrain the pre-trained carbonization quality diagnostic model.
[0050] In this embodiment of the invention, in order to improve the long-term accuracy of the carbonization quality diagnostic model, new data pairs containing feature vectors and actual final sampling intensity are periodically exported from the online rolling database. The machine learning model in the diagnostic module is incrementally learned or retrained so that the model can adapt to raw material batch fluctuations or seasonal environmental changes, and achieve continuous evolution of diagnostic capabilities.
[0051] In this embodiment of the invention, the overall workflow of the carbon fiber plate production quality control system is as follows: In the solid carbon fiber board production line, the mechanical frame of the multi-frequency ultrasonic feature sensing module is fixedly installed on both sides of the conveyor belt behind the outlet of the carbonization curing kettle. The positions of the low-frequency ultrasonic probe pair and the high-frequency probe pair are precisely adjusted to ensure that the sound beam axis is perpendicular to the conveyor belt plane and centered. The probe spacing is set and locked according to the product thickness. The industrial communication interface of the process parameter adaptive control module is physically connected and configured with the distributed control system of the carbonization curing kettle. In a laboratory environment, no less than 300 sets of solid carbon fiber board samples covering different carbonization process conditions are collected. The sensing module of this system is used to measure the multi-dimensional acoustic feature vector of each sample, and its true bending strength and carbonization degree data are obtained simultaneously through destructive testing. A random forest ensemble learning model is trained using this dataset, and the optimized model software is deployed to the industrial computing unit of the carbonization state intelligent diagnosis module. The entire collaborative control system is started. Under no-load conditions, the transmission and reception references of each ultrasonic probe are calibrated. Subsequently, a batch of known qualified solid carbon fiber boards are passed through the detection station at a uniform speed, and the distribution range of their acoustic feature vectors is recorded and analyzed. The center value of this distribution is set as the initial reference reference for online diagnosis. Simultaneously, a standard reference carbon fiber plate with stable acoustic performance is installed and fixed in the dynamic calibration unit. Once the system enters full online operation, the sensing module automatically collects the acoustic signal of each carbon fiber plate passing through the inspection station and calculates and generates a feature vector in real time. The diagnostic module receives the feature vector, uses an embedded machine learning model for real-time inference, and outputs the predicted flexural strength and carbonization state level of the carbon fiber plate. The results are displayed in real time and stored in a rolling database. The trend analysis engine within the process parameter adaptive control module continuously monitors the temporal changes of the diagnostic results. The engine calculates the distribution of carbonization state levels (e.g., the proportion of under-carbonized levels) and the moving average and standard deviation of the predicted strength within a moving time window (e.g., 50 consecutive carbon fiber plates). When a preset control threshold is triggered, the control decision-maker generates a draft process adjustment instruction based on a fuzzy rule base. The control decision-maker adjusts the draft instruction, for example, by increasing the CO2 concentration setpoint in the curing area by 8%, and sends it to the human-machine interface for safety confirmation, or automatically approves it after a set delay. Approved instructions are transmitted in real time to the actuators of the carbonization curing autoclave, such as CO2 flow control valves and humidifiers, via an industrial communication interface, thereby altering the process environment within the autoclave. The system automatically initiates a dynamic calibration program after each predetermined online cycle (e.g., after testing 200 solid carbon fiber plates, or every 8 hours). The mechanical actuator delivers a standard reference solid carbon fiber plate to the testing station, where the system performs a complete measurement and diagnostic. The measured eigenvector is compared with the reference vector of the standard solid carbon fiber plate to calculate the system deviation. In subsequent online measurements, the system automatically compensates and corrects the original data to eliminate drift errors.Regularly (e.g., weekly or monthly), new data pairs containing feature vectors and actual final sampling intensity are exported from the online rolling database. This data is then used for incremental learning or retraining of the machine learning model within the diagnostic module, enabling the model to adapt to fluctuations in raw material batches or seasonal environmental changes, thus achieving continuous evolution of diagnostic capabilities. These steps constitute a complete monitoring-diagnosis-control-calibration closed loop, allowing the system to operate continuously and automatically 24 hours a day. Simultaneously, the system automatically compiles quality data for production shifts, days, and months (such as pass rates, intensity distribution, and process parameter adjustment records), generating visualized quality reports. When the system self-diagnoses and detects abnormal missing ultrasonic signals, communication interruptions, or excessively low diagnostic confidence, it automatically triggers audible and visual alarms and records fault information. If necessary, it can interlock and suspend sorting or control actions. The probe surface is regularly cleaned, and mechanical components are lubricated to ensure long-term reliable system operation.
[0052] This invention, through a collaborative control mechanism of multi-frequency ultrasonic feature sensing, intelligent condition diagnosis, and process parameter feedback, fundamentally solves the core problems of traditional non-destructive testing methods, which can only achieve post-process sorting and cannot intervene in the production process, as well as the reliance on experience-based setting of conventional carbonization process parameters and the lack of online quality feedback. While achieving online, non-destructive, and accurate determination of carbonization degree and mechanical properties, it completes a paradigm shift in quality control from passively rejecting defective products to proactively optimizing the production process. Due to the adoption of a multi-frequency ultrasonic sensing strategy that specifically responds to the microstructure of carbonization products, combined with intelligent diagnosis using machine learning models, the system significantly improves the prediction accuracy of the flexural strength of solid carbon fiber sheets and can effectively distinguish between insufficient carbonization and other process defects. The built-in dynamic calibration mechanism ensures the measurement stability of the system during long-term operation, overcomes the problem of online sensor drift, and makes the judgment results reliable and highly reproducible.
[0053] This invention also provides a method for controlling the production quality of solid carbon fiber boards, such as... Figure 7 As shown, the carbon fiber plate production quality control method system applicable to any of the foregoing embodiments includes: S701: Collects the transmitted ultrasonic pulses of different preset ultrasonic pulses after they are transmitted through the solid carbon fiber board on the solid carbon fiber board production line, and determines the acoustic characteristic vector of the solid carbon fiber board based on the transmitted ultrasonic pulses. S702 uses a pre-trained carbonization quality diagnostic model to diagnose acoustic feature vectors and determine the predicted bending strength and carbonization state of the solid carbon fiber plate. S703 determines the carbonization trend of the solid carbon fiber plates on the solid carbon fiber plate production line based on the predicted bending strength value and carbonization state of a continuous preset number of solid carbon fiber plates, and adjusts the carbonization process parameters on the solid carbon fiber plate production line with the aim of keeping the carbonization trend within a preset target threshold range.
[0054] The working principle of the solid carbon fiber board production quality control method provided in this embodiment is the same as that of the solid carbon fiber board production quality control system in any of the foregoing embodiments. The specific implementation process of the solid carbon fiber board production quality control method will not be described here.
[0055] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.
Claims
1. A quality control system for the production of solid carbon fiber boards, characterized in that, include: The multi-frequency ultrasonic feature sensing module is used to transmit ultrasonic pulses of different preset frequencies to the solid carbon fiber board on the solid carbon fiber board production line, receive the transmitted ultrasonic pulses after the ultrasonic pulses are transmitted through the solid carbon fiber board, and determine the acoustic feature vector of the solid carbon fiber board based on the transmitted ultrasonic pulses. The carbonization state perception module is used to diagnose the acoustic feature vector using a pre-trained carbonization quality diagnostic model to determine the predicted bending strength value and carbonization state of the solid carbon fiber plate. The process parameter adaptive adjustment module is used to determine the carbonization trend of the solid carbon fiber board on the solid carbon fiber board production line based on the predicted bending strength value and carbonization state of a continuous preset number of solid carbon fiber boards, and to adjust the carbonization process parameters on the solid carbon fiber board production line with the aim of keeping the carbonization trend within a preset target threshold range.
2. The carbon fiber plate production quality control system according to claim 1, characterized in that, The multi-frequency ultrasound feature sensing module includes: A signal generator is used to generate multiple ultrasonic pulses of different frequencies; An ultrasonic transmitting probe, wherein there are multiple ultrasonic transmitting probes, are used to transmit multiple ultrasonic pulses of different frequencies to the solid carbon fiber plate; A signal receiving unit, which corresponds one-to-one with the ultrasonic transmitting probe, is used to receive transmitted ultrasonic pulses after being transmitted through the solid carbon fiber plate. An ultrasonic signal processor is used to calculate the acoustic feature vector of the solid carbon fiber plate based on the transmitted ultrasonic pulse and the structure of the solid carbon fiber plate.
3. The carbon fiber plate production quality control system according to claim 2, characterized in that, When the ultrasonic signal processor calculates the acoustic feature vector of the solid carbon fiber plate based on the transmitted ultrasonic pulse and the structure of the solid carbon fiber plate, it is used for: The propagation speed of ultrasonic pulses of different frequencies in the solid carbon fiber plate is calculated based on the propagation time of ultrasonic pulses of different frequencies in the solid carbon fiber plate and the thickness of the solid carbon fiber plate. The energy attenuation coefficient of the ultrasonic pulse in the solid carbon fiber plate is calculated based on the signal amplitude of the transmitted ultrasonic pulse at different frequencies and the reference signal amplitude in the preset state of the unconsolidated carbon fiber plate. By integrating the propagation speed and energy attenuation coefficient of the ultrasonic pulses of different frequencies in the solid carbon fiber plate, the acoustic feature vector of the solid carbon fiber plate is obtained.
4. The carbon fiber plate production quality control system according to claim 3, characterized in that, The ultrasonic transmitting probe and the signal receiving unit are installed on both sides of the solid carbon fiber board production line in a one-to-one correspondence, and the distance between the ultrasonic transmitting probe and the signal receiving unit is adjustable to adapt to the thickness of the solid carbon fiber board.
5. The carbon fiber plate production quality control system according to claim 1, characterized in that, The training process of the pre-trained carbonization quality diagnostic model includes: Construct a training data sample set, wherein a sample data in the training data sample set includes the acoustic feature vector of the sample solid carbon fiber plate and the measured bending strength value and carbonization degree of the sample solid carbon fiber plate; Using the acoustic feature vector of the sample solid carbon fiber plate as input, and the measured bending strength and carbonization degree of the sample solid carbon fiber plate as output, a preset machine learning model is trained to obtain a pre-trained carbonization quality diagnostic model.
6. The carbon fiber plate production quality control system according to claim 1, characterized in that, When determining the carbonization trend of the carbon fiber sheets on the carbon fiber sheet production line based on the predicted flexural strength and carbonization state of a continuous preset number of carbon fiber sheets, the process parameter adaptive adjustment module is used for: The average predicted bending strength and trend of the carbon fiber sheets on the carbon fiber sheet production line are calculated based on the predicted bending strength values of a continuous preset number of carbon fiber sheets. The carbonization state trend of the solid carbon fiber boards on the solid carbon fiber board production line is calculated based on the carbonization state level of a continuous preset number of solid carbon fiber boards.
7. The carbon fiber plate production quality control system according to claim 6, characterized in that, When the adaptive adjustment module adjusts the process parameters of the carbonization process on the solid carbon fiber board production line with the aim of keeping the carbonization trend within a preset target threshold range, it is used to: When the predicted bending strength trend continues to decline and the average predicted bending strength is lower than the preset average bending strength, or when the carbonization state trend is insufficient carbonization or insufficient carbonization uniformity, the process parameters in the carbonization process on the solid carbon fiber board production line are adjusted. The process parameters include carbon dioxide concentration, carbonization duration, and carbonization humidity.
8. The carbon fiber plate production quality control system according to claim 1, characterized in that, The multi-frequency ultrasonic feature sensing module is also used for: The acoustic characteristic vector of the reference solid carbon fiber plate is measured according to the preset production cycle measurement standard. The adaptive adjustment module for process parameters is also used for: The acoustic characteristic vector of the solid carbon fiber board on the solid carbon fiber board production line is compensated and corrected based on the measurement deviation between the measured acoustic characteristic vector and the standard reference acoustic characteristic vector of the standard reference solid carbon fiber board.
9. The carbon fiber plate production quality control system according to claim 1, characterized in that, The carbonization state sensing module is also used for: According to the preset model iteration optimization cycle, the acoustic feature vector, predicted bending strength value and carbonization state of the solid carbon fiber board on the solid carbon fiber board production line are collected to construct the model iteration optimization dataset. The pre-trained carbonization quality diagnostic model is retrained using the model to iteratively optimize the dataset.
10. A method for quality control in the production of solid carbon fiber boards, characterized in that, The method system for controlling the production quality of solid carbon fiber sheets according to any one of claims 1 to 9 includes: The acoustic characteristic vector of the solid carbon fiber board is determined based on the transmitted ultrasonic pulses after passing through the solid carbon fiber board on the solid carbon fiber board production line, and ultrasonic pulses of different preset frequencies are collected. The acoustic feature vector is diagnosed using a pre-trained carbonization quality diagnostic model to determine the predicted bending strength and carbonization state of the solid carbon fiber plate. The carbonization trend of the carbon fiber sheets on the carbon fiber sheet production line is determined based on the predicted flexural strength and carbonization state of a continuous preset number of carbon fiber sheets, and the carbonization process parameters on the carbon fiber sheet production line are adjusted with the aim of keeping the carbonization trend within a preset target threshold range.