Multi-label online monitoring method for composite gas cylinder in curing forming and service states

Abstract

The invention relates to a multi-label online monitoring method for a composite gas cylinder in curing forming and service states. In the forming process of the composite material gas cylinder, the electronic tag is embedded into the gas cylinder, an RFID reader-writer is used for reading the RSSI value of the tag, corresponding experimental samples are created according to different composite material gas cylinder material systems to obtain the material curing degree, a relation model of the RSSI value and the curing degree is obtained through polynomial fitting, and the relation model of the RSSI value and the curing degree is obtained. In the actual gas cylinder manufacturing process, the curing state of the gas cylinder is monitored through the model. For health monitoring of the gas cylinder, data classification is carried out on RSSI values under different structural damages of the label monitoring gas cylinder, feature learning is carried out on the data through a variational Bayesian inference algorithm, a logic regression model is established, and the structural damage type of the gas cylinder is judged through the model in the service process of the gas cylinder. For multi-label monitoring, a time division multiple access anti-collision protocol based on a unit average constant false alarm rate scheme is provided, the number of labels is estimated, a reader-writer adjusts transmission parameters in time, and collision is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of composite material manufacturing and health monitoring, and in particular to a multi-label online monitoring method for composite material gas cylinders in a curing and service state. Background Art

[0002] Composite gas cylinders have been widely used in the automotive, aerospace, industrial gas storage and other fields due to their light weight and high strength. With the increase in demand, the requirements for the quality and performance of gas cylinders are also increasing. The quality of gas cylinders is affected by many factors, among which the degree of curing of gas cylinders directly affects its quality and performance. Insufficient curing may cause defects such as pores and cracks in gas cylinders, thereby reducing their strength and durability, and even affecting their safety of use. Previous studies on curing degree monitoring have found that traditional monitoring methods have many limitations. Dielectric analysis (DEA) technology is widely used to monitor the curing process because it can detect all transition phases and critical points in the curing process. However, a major disadvantage of this method is that the sensors embedded in the material may have a negative impact on the mechanical properties of the composite material. These sensors are usually removed after the curing monitoring is completed, so they cannot provide additional value during the life cycle of the product. In addition, the wired nature of the sensor increases the difficulty of installation and may become a potential weakness for vacuum accumulation, allowing air to penetrate into the component, which in turn has a negative impact on the material performance. To circumvent these problems, wireless methods have been proposed. The use of ultrasound for curing treatment monitoring has been recognized. Nevertheless, the implementation of this method of curing monitoring is extremely demanding, as interference factors such as temperature, pressure and sensor variability need to be taken into account and the measuring equipment needs to be calibrated accordingly. As a result, this measurement method is sensitive and requires precise set-up by experts, which also leads to additional costs. The cost issue is even more prominent if the equipment cannot be reused for the further product life of the composite material. Fiber optic Bragg sensors are a promising alternative that are used to monitor most stages of the curing process and can remain in the component for structural health monitoring. However, these sensors are considered to be very fragile and susceptible to mechanical stress, resulting in a higher than average risk of damage and failure, leading to increased costs and reduced efficiency. Even if future developments can compensate for these shortcomings, the method is still limited because these sensors, as well as the sensors used in DEA, must be closely connected, which can also lead to potential weaknesses in vacuum build-up.

[0003] The use of RFID technology to monitor the curing degree and health of gas cylinders solves many problems in traditional methods and brings significant innovations and advantages. First, RFID technology realizes non-contact real-time monitoring of the curing degree of gas cylinders during curing, avoids the interference of traditional detection methods on the production process, and improves the monitoring efficiency and accuracy. At the same time, after the RFID tag is embedded in the gas cylinder, it can continuously monitor the health status of the gas cylinder in service, and timely identify structural damage such as cracks and delamination through data changes such as RSSI values. The innovation lies in that RFID technology not only provides remote monitoring capabilities, but also combines multi-tag distribution to locate the damage location and improve the intelligence level of the monitoring system. Compared with traditional mechanical and chemical detection methods, the RFID monitoring system has the advantages of rapid response, low cost, and remote monitoring, thereby realizing intelligent health monitoring of the entire life cycle of gas cylinder production and use, greatly improving the safety and service life of gas cylinders. Summary of the invention

[0004] The purpose of the present invention is to provide an online monitoring method for composite gas cylinders during molding and service. The method is a low-cost real-time non-destructive testing method that allows sensors to remain in the composite material without adversely affecting the mechanical properties of the material. It can effectively monitor the degree of curing and health monitoring of the gas cylinder molding process.

[0005] To achieve the above purpose, the technical solution adopted by the present invention is: a multi-tag online monitoring method for composite gas cylinders in curing and service state, specifically including embedding multiple tags in different positions of the gas cylinder during the curing and forming process of the composite gas cylinder, using an RFID reader to read the RSSI value of the tag in real time, agreeing that each tag ID number is unique, and for different composite gas cylinder material systems, the resin ion viscosity change curve is obtained through dielectric analysis (DEA) experiments, the curing degree is obtained through calculation, and finally the equation of RSSI value and curing degree is obtained through polynomial fitting, and the curing state of each part of the gas cylinder including the head position and the barrel section of the gas cylinder is monitored according to the model. For the health monitoring of the gas cylinder, based on the variational Bayesian logistic regression algorithm, the RSSI value of the gas cylinder measured by the RFID tag under different structural damage is classified, and the RSSI value data is feature learned through the variational Bayesian inference algorithm. Finally, a logistic regression model is established, and the structural damage of the gas cylinder is judged according to the model. This method realizes the whole process health monitoring of the gas cylinder during curing and service state by monitoring the change of the RSSI value parameter. For multi-tag monitoring, a time-division multiple access anti-collision protocol based on the unit average constant false alarm rate scheme is proposed. The unit average constant false alarm rate technology is introduced to estimate the number of tags, so that the reader can adjust the transmission parameters in time to avoid collision.

[0006] The method for online monitoring of labels used in the curing and forming process of composite gas cylinders specifically includes the following steps:

[0007] Step 1: During the wrapping process of the composite gas cylinder, the RFID tags are embedded in different positions of the gas cylinder, and the tags are evenly placed at the gas cylinder head position and the transition between the head and the cylinder body, so that the tags are placed in a triangular array to ensure the accuracy of the collected signals;

[0008] Step 2: Before the resin is cured, measure the initial RSSI value of each tag as the baseline data. During the entire process of the cylinder curing and molding, use the RFID reader to read the RSSI value of the embedded tag in real time;

[0009] Step 3: During the actual production process of the gas cylinder, the real-time curing state of the gas cylinder is judged according to the mathematical model of RSSI value and curing degree obtained in the experiment, so as to realize the online monitoring of the curing molding of the composite material gas cylinder.

[0010] The equation of the curing degree and RSSI value obtained by polynomial fitting in step 3 can be expressed by the following mathematical expression:

[0011] y=Intercept+A1*x+A2*x 2

[0012] The variable y represents the RSSI value, the variable x represents the degree of solidification, Intercept is the intercept term, and A1 and A2 are fitting coefficients.

[0013] After the mathematical model is obtained by polynomial fitting, the goodness of fit (R square value) is introduced to evaluate the fitting effect of the model. The closer the R square value is to 1, the better the model fitting effect is.

[0014] When the gas cylinder is in service, the online monitoring method using the tag includes the following steps:

[0015] Step 1: By embedding RFID tags in gas cylinders, a baseline measurement of RSSI values ​​is performed when the gas cylinders are intact, and the data distribution and characteristics under normal conditions are recorded;

[0016] Step 2: When the gas cylinder is in service, read the RSSI value of the RFID tag regularly or in real time;

[0017] Step 3: When the gas cylinder is damaged, if the RSSI value of the tag in a specific area changes significantly, while the tag signal in other areas is stable, the specific location of the damage can be determined;

[0018] Step 4: Based on the established logistic regression model that can accurately classify damage states, determine the type of structural damage that occurs in the gas cylinder, thereby achieving health monitoring of composite gas cylinders in service.

[0019] Among them, in step 4, a regression model is established based on the variational Bayesian logistic regression algorithm to determine the type of structural damage to the gas cylinder. The specific steps are as follows:

[0020] Step 41, continuously collect RSSI data under different time and environmental conditions through the RFID reader / writer to ensure that the data covers possible damage conditions, such as matrix cracking, delamination damage, fiber breakage, and gas leakage;

[0021] Step 42: annotate the collected RSSI data, record the damage type corresponding to each set of data, remove noise and outliers in the RSSI data, and normalize the data;

[0022] Step 43: Select the variational Bayesian logistic regression model for data modeling, assume that there is a logical relationship between the structural damage data and the RSSI value extracted by the RFID system, and use the Sigmoid function for classification;

[0023] Step 44, using the variational Bayesian inference method, approximating the posterior distribution, and optimizing the model parameters by maximizing the variational bound (ELBO);

[0024] Step 45: Use the iterative optimization algorithm in variational Bayesian inference to continuously update the model parameters until convergence;

[0025] Step 46: Use the trained model to classify the real-time monitoring data and identify the type of structural damage.

[0026] A time division multiple access anti-collision protocol based on unit average constant false alarm rate technology is used for multiple tags. The characteristic is that the unit average constant false alarm rate technology is introduced to estimate the number of physical layer tags, so that the reader can adjust its transmission parameters accordingly to avoid collision.

[0027] Given an accurate estimate of the number of tags, DFSA anti-collision can be programmed with an optimal frame, reducing the number of time slots required to identify a tag in the system.

[0028] The signal power received by the reader is analyzed by the unit average constant false alarm rate detector, and the MAP detector is used to estimate the number of tags at this power level. The specific steps are as follows:

[0029] Step 1: The reader receives power signals from different tags and divides the received signal power into units;

[0030] Step 2: setting a threshold according to the normalized average power and the constant false alarm rate;

[0031] Step 3: Compare the signal strength with the threshold. If it is greater than the threshold, a level change is detected. Otherwise, it will return to compare again.

[0032] When detecting a tag, in the AWGN channel, the signal frame received by the reader can be described as follows:

[0033]

[0034] Where N s is the number of slots in the frame, and n(t) is additive Gaussian white noise. s (t) is the received signal in a given time slot, with a duration of T s , g s (t) can be expressed as:

[0035]

[0036] where b i is the bit information, i is the label index, τ i To follow the uniform distribution The label delay, N t is the number of tags responded in this slot, is the duration T c Pulse information waveform.

[0037] The received signal power Z(t) can be written as:

[0038] Z(t)=|r(t) 2

[0039] Step 4: After detecting the power change, use the MAP detector to estimate the number of tags at the power level, and obtain the detection threshold of the tag according to the likelihood ratio.

[0040] The MAP detector is used to estimate the number of tags, and its signal model is expressed as:

[0041]

[0042] According to the likelihood ratio Λ(y), we can get a label detection threshold:

[0043] BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 Schematic diagram of the gas cylinder monitoring system based on RFID;

[0045] Figure 2 This is a schematic diagram of the layout of RFID tags inside gas cylinders;

[0046] Figure 3 This is the flow chart of RFID-based gas cylinder monitoring; Figure 4 It is a flow chart of RFID tag detection and quantity estimation; DETAILED DESCRIPTION

[0047] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0048] The method for online monitoring of labels used in the curing and forming process of composite gas cylinders specifically includes the following steps:

[0049] Step 1: During the wrapping process of the composite gas cylinder, the RFID tags are embedded in different positions of the gas cylinder, and the tags are evenly placed at the gas cylinder head position and the transition between the head and the cylinder body, so that the tags are placed in a triangular array to ensure the accuracy of the collected signals;

[0050] Step 2: Before the resin is cured, measure the initial RSSI value of each tag as the baseline data. During the entire process of the cylinder curing and molding, use the RFID reader to read the RSSI value of the embedded tag in real time;

[0051] Step 3: During the actual production process of the gas cylinder, the real-time curing state of the gas cylinder is determined based on the RSSI value and the mathematical model of the curing degree obtained in the experiment, thereby realizing online monitoring of the curing molding of the composite material gas cylinder.

[0052] Among them, the equation of curing degree and RSSI value obtained by polynomial fitting is expressed as follows:

[0053] y=Intercept+A1*x+A2*x 2

[0054] The variable y represents the RSSI value, the variable x represents the degree of solidification, Intercept is the intercept term, and A1 and A2 are fitting coefficients.

[0055] After the mathematical model is obtained by polynomial fitting, the goodness of fit (R square value) is introduced to evaluate the fitting effect of the model. The closer the R square value is to 1, the better the model fitting effect is.

[0056] When the gas cylinder is in service, the online monitoring method using the tag includes the following steps:

[0057] Step 1: By embedding RFID tags in gas cylinders, a baseline measurement of RSSI values ​​is performed when the gas cylinders are intact, and the data distribution and characteristics under normal conditions are recorded;

[0058] Step 2: When the gas cylinder is in service, read the RSSI value of the RFID tag regularly or in real time;

[0059] Step 3: When the gas cylinder is damaged, if the RSSI value of the tag in a specific area changes significantly, while the tag signal in other areas is stable, the specific location of the damage can be determined;

[0060] Step 4: Based on the established logistic regression model that can accurately classify damage states, determine the type of structural damage that occurs in the gas cylinder, thereby achieving health monitoring of composite gas cylinders in service.

[0061] Among them, in step 4, a regression model is established based on the variational Bayesian logistic regression algorithm to determine the type of structural damage to the gas cylinder. The specific steps are as follows:

[0062] Step 41, continuously collect RSSI data under different time and environmental conditions through the RFID reader / writer to ensure that the data covers possible damage conditions, such as matrix cracking, delamination damage, fiber breakage, and gas leakage;

[0063] Step 42: annotate the collected RSSI data, record the damage type corresponding to each set of data, remove noise and outliers in the RSSI data, and normalize the data;

[0064] Step 43: Select the variational Bayesian logistic regression model for data modeling, assume that there is a logical relationship between the structural damage data and the RSSI value extracted by the RFID system, and use the Sigmoid function for classification;

[0065] Step 44, using the variational Bayesian inference method, approximating the posterior distribution, and optimizing the model parameters by maximizing the variational bound (ELBO);

[0066] Step 45: Use the iterative optimization algorithm in variational Bayesian inference to continuously update the model parameters until convergence;

[0067] Step 46: Use the trained model to classify the real-time monitoring data and identify the type of structural damage.

[0068] A time division multiple access anti-collision protocol based on unit average constant false alarm rate technology is used for multiple tags, which is characterized by:

[0069] The signal power received by the reader is analyzed using the unit average constant false alarm rate technique, and the MAP detector is used to estimate the number of tags at that power level.

[0070] At the beginning, the reader sends a broadcast message, and all tags send a time slot signal to the reader. The received signal power represents the sum of the responses of the tags within the reader's reading range. By detecting the change in power level, the number of tags can be estimated. The specific steps are as follows:

[0071] Step 1: The reader receives power signals from different tags and divides the received signal power into units;

[0072] Step 2: setting a threshold according to the normalized average power and a certain false alarm rate;

[0073] Step 3: Compare the calculated power with the threshold. If the power is greater than the threshold, a level change is detected. Otherwise, the comparison is returned again.

[0074] In the AWGN channel, the signal frame received by the reader can be described as follows:

[0075]

[0076] Where N s is the number of slots in the frame, and n(t) is additive Gaussian white noise. s (t) is the received signal in a given time slot, with a duration of T s , g s (t) can be expressed as:

[0077]

[0078] where b i is the bit information, i is the label index, τ i To follow the uniform distribution The label delay, N t is the number of tags responded in this slot, is the duration T c Pulse information waveform.

[0079] The received signal power Z(t) can be written as:

[0080] Z(t)=|r(t) 2

[0081] Step 4: After detecting the power change, use the MAP detector to estimate the number of tags at the power level, and obtain the detection threshold of the tag according to the likelihood ratio.

[0082] The MAP detector is used to estimate the number of tags, and its signal model is expressed as:

[0083]

[0084] According to the likelihood ratio Λ(y), we can get a label detection threshold:

[0085] .

Claims

1. A multi-tag online monitoring method for structural damage of composite gas cylinders during curing and service, specifically including embedding multiple tags into the head and barrel of the gas cylinder during the curing and forming process of the composite gas cylinder, using an RFID reader to read the RSSI value of the tag in real time, agreeing that each tag ID number is unique, creating corresponding experimental samples for different composite gas cylinder material systems to calculate the material curing degree over time curve, and obtaining the relationship model between RSSI value and curing degree through polynomial fitting. In the actual manufacturing process of the gas cylinder, the model is used to monitor the curing state of various parts of the gas cylinder including the head position and the barrel of the gas cylinder. For the health monitoring of the gas cylinder, based on the variational Bayesian logistic regression algorithm, the RSSI values ​​of the gas cylinder under different structural damages monitored by the RFID tag are classified, and the RSSI value data is feature learned through the variational Bayesian inference algorithm. Finally, a logistic regression model is established, and the type of structural damage of the gas cylinder is judged according to the model when the gas cylinder is in service. This method realizes the whole process health monitoring of the gas cylinder during curing and service by monitoring the change of the RSSI value. For multi-tag monitoring, a time-division multiple access anti-collision protocol based on the unit average constant false alarm rate scheme is proposed. The unit average constant false alarm rate technology is introduced to estimate the number of tags, so that the reader can adjust the transmission parameters in time to avoid collision.

2. The method according to claim 1, wherein the method first determines the specific material system contained in the composite material gas cylinder used, and the material system of the composite material gas cylinder is mainly composed of a matrix material and a reinforcing fiber material. Provide at least one fiber material layer, embed at least one RFID electronic tag in the fiber material layer, impregnate the fiber material layer with the material constituting the matrix, establish a data communication link between the RFID electronic tag and the reader, determine the curing degree of the material system by analyzing the signal exchanged between the RFID electronic tag and the reader, and obtain a relationship model between the RSSI value and the curing degree through data analysis. The tag online monitoring method used in the actual composite gas cylinder curing molding process specifically includes the following steps: Step 21, during the winding process of the composite gas cylinder, labels are evenly placed at the end cap position of the gas cylinder and at the transition between the end cap and the barrel section, so that the labels are placed in a triangular array to ensure the accuracy of the collected signals; Step 22, before curing, measure the initial RSSI value of each tag as baseline data, and use the RFID reader to read the RSSI value of the embedded tag in real time during the entire process of curing the gas cylinder; Step 23: During the actual production process of the gas cylinder, the real-time curing state of the gas cylinder is determined according to the mathematical model of the RSSI value and curing degree obtained in the experiment, thereby realizing online monitoring of the curing molding of the composite material gas cylinder.

3. The equation of curing degree and RSSI value obtained by polynomial fitting according to claim 2 is expressed as follows: y=Intercept+A1*x+A2*x 2 The variable y represents the RSSI value, the variable x represents the degree of solidification, Intercept is the intercept term, and A1 and A2 are fitting coefficients. After the mathematical model is obtained by polynomial fitting, the goodness of fit (R square value) is introduced to evaluate the fitting effect of the model. The closer the R square value is to 1, the better the model fitting effect is.

4. According to the method described in claim 1, in order to realize the health monitoring of gas cylinders, the RSSI values ​​of the gas cylinders in health and different damage states are first measured by using the RFID system to obtain RSSI data sets in healthy state and RSSI data sets in structural damage state. The data samples of the two states are input into the variational Bayesian logistic regression model to perform data training and complete data classification. The RSSI value data of the gas cylinder under different structural damages are trained by the variational Bayesian inference method, the data features are extracted, and a logistic regression model that can accurately classify the damage state is established. The online monitoring method of the tags used in the actual service state of the gas cylinder specifically includes the following steps: Step 41, using the RFID tag embedded in the gas cylinder, a baseline measurement of the RSSI value is performed when the gas cylinder is intact, and the distribution and characteristics of the RSSI value under normal conditions are recorded; Step 42: When the gas cylinder is in service, the RSSI value of the RFID tag is read periodically or in real time; Step 43: When the gas cylinder is damaged, if the RSSI value of the tag in a certain area changes significantly, while the tag signals in other areas are stable, the specific location of the damage can be determined; Step 44: According to the established logistic regression model that can accurately classify the damage state, the type of structural damage in the gas cylinder is determined, so as to achieve health monitoring of the composite gas cylinder in service.

5. The time division multiple access anti-collision protocol based on unit average constant false alarm rate technology according to claim 1 is characterized in that The introduction of unit average constant false alarm rate technology estimates the number of physical layer tags, allowing the reader to adjust its transmission parameters accordingly to avoid collisions. Given an accurate estimate of the number of tags, DFSA anti-collision can be programmed with optimal frames to reduce the number of time slots required to identify system tags. The signal power received by the reader is analyzed by the unit average constant false alarm rate detector, and the MAP detector is used to estimate the number of tags at this power level. The specific steps are as follows: Step 51, the reader receives power signals from different tags and divides the received signal power into units; Step 52: setting a threshold according to the normalized average power and the constant false alarm rate; Step 53: compare the signal strength with the threshold. If the signal strength is greater than the threshold, a level change is detected. Otherwise, the signal strength is returned and compared again. Step 54: After detecting the power change, use the MAP detector to estimate the number of tags at the power level, and obtain the detection threshold of the tag according to the likelihood ratio.

Images