Heat seal quality detection method and system based on dynamic temperature distribution

Abstract

The invention discloses a heat seal quality detection method based on dynamic temperature distribution, which comprises the following steps: step 1, temperature data acquisition: triggering an infrared temperature measurement module according to a production line beat signal, and acquiring temperature data of a seal area at a frequency greater than or equal to 50-80-120Hz; step 2, obtaining a temperature array: reading the coverage rate calculation of the temperature on the whole sealing area; under the condition that the linear speed is fixed, the area of a measurement area is increased by adjusting the linear distance between the temperature measurement module and the packaging bag sealing area, the theoretical coverage rate can be further increased, and two data unstable points before and after are abandoned. According to the method, edge temperature interference is eliminated through a dynamic array filling method, a proper algorithm is automatically selected based on a mathematical model, model characteristics are calculated, temperature changes are predicted through the model, and then double logics of threshold value judgment and distribution judgment are combined, so that high-precision judgment of the sealing quality is achieved, and temperature collection deviation caused by vibration of a production line is effectively eliminated.

Description

Technical Field

[0001] The present invention relates to the field of industrial detection technology, and in particular to a heat seal quality detection method and system based on dynamic temperature distribution. Background Art

[0002] Heat sealing is widely used in food and pharmaceutical packaging. Heat is applied to melt the contacting surfaces of materials (such as plastic films) and then pressurize them together to create a seal. This process is often used in food and pharmaceutical packaging to ensure the integrity and tightness of the package.

[0003] Heat sealing includes heating system, transmission system, pressing mechanism and control system.

[0004] The current heat sealing device has the following defects:

[0005] 1. Large interference in temperature collection: Mechanical errors in the production line lead to distortion of the temperature data at the beginning and end (such as Figure 2 Error data area);

[0006] 2. Single decision logic: Relying solely on fixed thresholds, it is unable to identify hidden defects caused by uneven temperature distribution.

[0007] 3. Slow algorithm response speed: Image-based analysis requires high computing power and is difficult to meet the needs of high-speed production lines. Summary of the Invention

[0008] The purpose of the present invention is to solve the shortcomings in the prior art.

[0009] In order to achieve the above object, the present invention adopts the following technical solutions:

[0010] The heat seal quality detection method based on dynamic temperature distribution includes the following steps:

[0011] Step 1: Temperature data collection:

[0012] The infrared temperature measurement module is triggered according to the production line beat signal to collect the sealing area temperature data at a frequency of ≥50-80-120Hz;

[0013] Step 2: Get the temperature array:

[0014] Read the temperature to calculate the coverage of the entire sealing area;

[0015] Taking the speed of 42 packets per minute as an example,

[0016] Measure width L 测 =10mm, reading frequency f 读 =100Hz, packaging bag linear speed v l =1.25m / s,

[0017] The width of the packaging bag is 200mm, and the measured passing time is 160ms.

[0018] Measuring coverage,

[0019]

[0020] Under the condition of constant line speed, by adjusting the straight-line distance from the temperature measurement module to the sealing area of the packaging bag, increasing the area of the measurement area, the theoretical coverage rate can be further improved, and the two unstable data points before and after are discarded. 测 *f 读 ≤v l , theoretically, discarding one reading point before and after can ensure data stability;

[0021] Step 3: Eigenvalue calculation:

[0022] Calculate the arithmetic mean μ and standard deviation σ of a dynamic array;

[0023] Select the appropriate physical model based on the data results;

[0024] Based on the model, calculate the model weight;

[0025] Generate a temperature prediction function based on the model weights and obtain the predicted temperature through the current time series;

[0026] The specific weights are related to the automatically selected model;

[0027] Newton's law of cooling model,

[0028] Weight: cooling attenuation coefficient, ambient temperature (medium temperature);

[0029] Weight function: T t =T a +(T a -T0)e -kt ;

[0030] Polynomial model,

[0031] Weight: coefficients of polynomials of various orders;

[0032] Weight function: T t =a n t n +a n-1 t n-1 +a n-2 t n-2 +…+a2t 2 +a1t 1 +a0.

[0033] Step 4: Double quality determination:

[0034] Threshold determination: If the difference between the current temperature and the predicted temperature is too large, a level 1 alarm will be triggered;

[0035] Distribution judgment: The difference between the current product average and the historical average is too large, triggering a secondary alarm.

[0036] The system involved in the heat seal quality detection method based on dynamic temperature distribution, the length of the dynamic array in step 2 is adaptive to the production line speed; the two-level alarm signal in step 4 is output to the PLC control system through the optocoupler isolation circuit.

[0037] The system involved in the heat seal quality detection method based on dynamic temperature distribution includes:

[0038] Infrared temperature measurement module: using thermopile, temperature measurement range 0-500℃, accuracy ±0.5℃;

[0039] Edge computing unit: integrated processor, dynamic temperature window interception mechanism, multiple judgment procedures;

[0040] Alarm control module: Contains RS485 communication interface and relay output circuit, supports graded response of alarm signals.

[0041] The system involved in the heat seal quality detection method based on dynamic temperature distribution includes:

[0042] The edge computing unit directly reads the infrared sensor data through the DMA channel with a delay of ≤10μs;

[0043] The system supports local storage of historical temperature data;

[0044] The selection of a suitable physical model lies in analyzing the relationship between the temperature data collected by the sensor and the time series, and choosing to use Newton's cooling law fitting or polynomial fitting.

[0045] The physical model selection is based on the temperature-time series relationship, and the model is adaptively switched according to the variance of the temperature data. When the variance is greater than the threshold, the polynomial model is enabled. The basis for model selection is based on the set cooling rate threshold.

[0046] The beneficial effects of the present invention are:

[0047] The present invention eliminates edge temperature interference through a dynamic array filling method, automatically selects an appropriate algorithm based on a mathematical model, calculates model characteristics, and predicts temperature changes through the model. Combining the dual logic of threshold determination and distribution determination, it achieves high-precision judgment of sealing quality and effectively eliminates temperature acquisition deviations caused by production line vibration.

[0048] The present invention: establishes a standard deviation-mean joint judgment model to solve the problem of high false alarm rate of single threshold judgment;

[0049] This invention achieves deep collaboration between algorithms and hardware, supports real-time response in microseconds, and reduces the detection error rate from 30% of traditional methods to below 0.2%, making it suitable for high-speed production lines for food and pharmaceutical packaging. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a diagram showing the relationship between the measurement area and distance of the present invention (taking the temperature measurement module H20 as an example);

[0051] Figure 2 This is a schematic diagram of the distortion of temperature acquisition data in the present invention, where indexes 6 and 34 are the distorted temperatures;

[0052] Figure 3 This is a timing diagram of the present invention;

[0053] Figure 4 It is the state transition diagram of the algorithm of the present invention;

[0054] Figure 5 This is a flow chart of the log file ping-pong cache management of the present invention. DETAILED DESCRIPTION

[0055] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, those skilled in the art can choose other machine learning methods according to their needs, including but not limited to Transformer, random forest, support vector machine (SVM), and their core ideas still fall within the scope of protection of this patent. All other embodiments obtained by ordinary technicians without making creative work are within the scope of protection of the present invention.

[0056] The following describes the specific embodiments of the present disclosure in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present disclosure and are not intended to limit the present disclosure.

[0057] Example 1

[0058] To achieve the above objectives, the present invention discloses a heat seal quality detection method based on dynamic temperature distribution, comprising the following steps:

[0059] Step 1: Temperature data collection:

[0060] The infrared temperature measurement module is triggered according to the production line beat signal to collect the sealing area temperature data at a frequency of ≥50-80-120Hz;

[0061] Step 2: Get the temperature array:

[0062] Read the temperature to calculate the coverage of the entire sealing area;

[0063] Taking the speed of 42 packets per minute as an example,

[0064] Measure width L 测 =10mm, reading frequency f 读 =100Hz, packaging bag linear speed v l =1.25m / s,

[0065] The width of the packaging bag is 200mm, and the measured passing time is 160ms.

[0066] Measuring coverage,

[0067]

[0068] Under the condition of constant line speed, by adjusting the straight-line distance from the temperature measurement module to the sealing area of the packaging bag, the area of the measurement area can be increased, which can further improve the theoretical coverage rate; discarding the two unstable data points before and after, due to the acquisition coverage speed L 测 *f 读 ≤v l , theoretically, discarding one reading point before and after can ensure data stability;

[0069] Step 3: Eigenvalue calculation:

[0070] Calculate the arithmetic mean μ and standard deviation σ of a dynamic array;

[0071] Select the appropriate physical model based on the data results;

[0072] Based on the model, calculate the model weight;

[0073] Generate a temperature prediction function based on the model weights and obtain the predicted temperature through the current time series;

[0074] The specific weights are related to the automatically selected model;

[0075] Newton's law of cooling model,

[0076] Weight: cooling attenuation coefficient, ambient temperature (medium temperature);

[0077] Weight function: T t =T a +(T a -T0)e -kt ;

[0078] Polynomial model,

[0079] Weight: coefficients of polynomials of various orders;

[0080] Weight function: T t =a n t n +a n-1 t n-1 +a n-2 t n-2 +…+a2t 2 +a1t 1 +a0.

[0081] Step 4: Double quality determination:

[0082] Threshold determination: If the difference between the current temperature and the predicted temperature is too large, a level 1 alarm will be triggered;

[0083] Distribution judgment: The difference between the current product average and the historical average is too large, triggering a secondary alarm.

[0084] The system involved in the heat seal quality detection method based on dynamic temperature distribution, the length of the dynamic array in step 2 is adaptive to the production line speed; the two-level alarm signal in step 4 is output to the PLC control system through the optocoupler isolation circuit.

[0085] The system involved in the heat seal quality detection method based on dynamic temperature distribution includes:

[0086] Infrared temperature measurement module: using thermopile, temperature measurement range 0-500℃, accuracy ±0.5℃;

[0087] Edge computing unit: integrated processor, dynamic temperature window interception mechanism, multiple judgment procedures;

[0088] Alarm control module: Contains RS485 communication interface and relay output circuit, supports graded response of alarm signals.

[0089] The system involved in the heat seal quality detection method based on dynamic temperature distribution includes:

[0090] The edge computing unit directly reads the infrared sensor data through the DMA channel with a delay of ≤10μs;

[0091] The system supports local storage of historical temperature data;

[0092] The selection of a suitable physical model lies in analyzing the relationship between the temperature data collected by the sensor and the time series, and choosing to use Newton's cooling law fitting or polynomial fitting.

[0093] The physical model selection is based on the temperature-time series relationship, and the model is adaptively switched according to the variance of the temperature data. When the variance is greater than the threshold, the polynomial model is enabled. The basis for model selection is based on the set cooling rate threshold.

[0094] Model selection can be achieved through automated algorithms (such as least squares goodness of fit assessment) or manual intervention (such as switching models based on statistical characteristics of historical data).

[0095] like Figure 3 As shown in (program sequence diagram), the technical breakthroughs of the present invention include:

[0096] 1. Dynamic temperature window interception mechanism:

[0097] When the temperature in the measurement area changes, it starts collecting temperature data at 80-120Hz, automatically skipping the unstable points at the beginning and end of the temperature collection, and automatically judging whether the temperature reading is completed. When the temperature of multiple consecutive points does not exceed the threshold, the array collection is completed and the temperature array is saved. According to the speed parameters of the packaging machine, the theoretical temperature array length at the speed is calculated. When the difference between the actual array length and the theoretical array length is too large, the array is automatically discarded and a level 1 alarm is issued.

[0098] 2. Multiple judgment logic optimization:

[0099] Threshold determination: The predicted temperature is obtained based on the model prediction, and a comprehensive judgment is made based on the actual measured value and the threshold limit;

[0100] Distribution determination: Determine the overall temperature slip of the device based on the model weight to reduce misjudgment.

[0101] 3. Software and hardware collaboration:

[0102] The edge computing module regularly wakes up the temperature reading logic to ensure the stability of temperature acquisition;

[0103] The entire reading and recognition logic requires the edge computing module to read DMA data at a fixed frequency (80-120Hz), encapsulate the data reading logic into a function, and use a timer to wake up the reading function at an interval of 10ms to ensure the stability of data acquisition;

[0104] Automatically identify whether temperature collection is completed and perform data analysis after collection is completed to prevent blockage of temperature reading;

[0105] The data collection process duration is determined by the bag width (for example, if the packaging bag linear speed is 1.25m / s, the data collection time for a 200mm wide bag is about 200ms), and the data analysis and processing time is about 30ms.

[0106] The temperature data log is recorded through the ping-pong cache structure. Figure 5Two log files a and b with fixed lengths are placed locally. When log a is not full, records are recorded line by line in log a. When log a is full, records are switched to log b line by line. When log b is full, all records in log a are cleared and records are switched to log a line by line again. The two log files are written alternately, and continuous temperature log files can be seamlessly written to with fixed memory space. In the event of uncontrollable abnormalities during the production process, the original temperature data can be provided through the local log files.

[0107] 4. Learning + Prediction Mode:

[0108] Learning mode: When the data is completed, analyze whether the current data is normal. If it is normal, the algorithm manager will calculate the temperature data distribution characteristics;

[0109] A: When the cooling rate is too fast, use Newton's law of cooling, which is as follows:

[0110] T t =T a +(T a -T0)e -kt

[0111] T t The temperature of the object at time t;

[0112] T a is the ambient temperature (temperature of the surrounding medium);

[0113] T0 is the initial temperature;

[0114] k is the cooling coefficient, which indicates the cooling rate of an object and is related to factors such as the object's material, surface area, and environmental conditions;

[0115] After preprocessing, the attenuation coefficient k is calculated using the least squares method.

[0116] B: When the material cooling rate is not fast, the time interval is obvious, and the temperature data distribution is more affected by the heat seal block itself, it will lead to multiple maximum or minimum values in the temperature data distribution, too large variance, and uneven temperature changes. The polynomial fitting formula is as follows:

[0117] T t =a n t n +a n-1 t n-1 +a n-2 t n-2 +…+a2t 2 +a1t 1 +a0

[0118] Through Taylor's high-order expansion, any continuous differentiable function can be theoretically approximated. Where is the coefficient of the contrast expansion term;

[0119] Calculate and generate the Vandermonde matrix, solve the coefficients through the normal equations, use the Gaussian elimination method to solve the linear equations, and obtain the coefficients of each expanded term.

[0120] Prediction Mode: Data is collected and a prediction curve is generated based on the parameters returned by the learning mode algorithm. The curve is shifted up or down based on the difference between the current average value and the learning mode average value to address temperature drift issues (such as unstable ambient temperature and temperature drift of the heater block before and after shutdown). The current value is compared with the predicted value, and a level 1 alarm is issued when the difference exceeds a threshold. The variance of the current temperature data is analyzed compared with the variance of the historical data. If the difference is too large, a level 2 alarm is issued.

[0121] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings. However, the present disclosure is not limited to the specific details in the above embodiments. Within the technical concept of the present disclosure, various changes, modifications, replacements and variations can be made to these embodiments, and these changes, modifications, replacements and variations all fall within the scope of protection of the present disclosure.

[0122] Example 2

[0123] Take the hot pot packaging line (speed 42 bags / minute) as an example:

[0124] 1. Parameter initialization:

[0125] Configure the plan for selecting detection products (based on the upper and lower limits collected historically, stored as configuration files) and automatically set the data filtering conditions for the learning mode.

[0126] The lower limit of the length of the temperature array is automatically set according to the speed of the packaging machine to prevent skewed packaging bags from passing through the prediction mode.

[0127] 2. Real-time detection process:

[0128] The infrared stack is sampled at 50-80-120Hz;

[0129] The edge computing module reads the temperature at 80-120Hz;

[0130] When the data collection completion conditions are met, determine whether to execute the learning mode or the prediction mode;

[0131] The learning mode generates a prediction curve and automatically jumps to the prediction mode;

[0132] The prediction mode automatically compensates for the current temperature and substitutes the time t of each temperature data into the prediction curve;

[0133] Compare the difference between the input curve and the predicted curve. If the difference of any temperature data is greater than the threshold, a level 1 alarm is triggered.

[0134] According to the product detection plan, a secondary alarm is issued when the difference between the historical average temperature of the current product and the average temperature of the current product is continuously too large.

[0135] 3. Exception handling:

[0136] After the secondary alarm is triggered, the system automatically saves the original data of the current cycle;

[0137] Upload abnormal data to the MES system via the MQTT protocol to generate a quality analysis report.

[0138] It should also be noted that the various specific technical features described in the above specific embodiments may be combined in any suitable manner, unless there is any contradiction, and these combinations shall also be considered as the contents disclosed in this disclosure. To avoid unnecessary repetition, this disclosure will not further describe various possible combinations. The technical scope of this application is not limited to the contents of the specification and must be determined based on the scope of the claims.

Claims

1. A heat seal quality detection method based on dynamic temperature distribution, characterized in that: The following steps are involved: Step 1: Temperature data collection: The infrared temperature measurement module is triggered according to the production line beat signal to collect the sealing area temperature data at a frequency of ≥50-80-120Hz; Step 2: Get the temperature array: Read the temperature to calculate the coverage of the entire sealing area; Taking the speed of 42 packets per minute as an example, Measure width L 测 =10mm, reading frequency f 读 =100Hz, packaging bag linear speed v l =1.25m / s, The width of the packaging bag is 200mm, and the measured passing time is 160ms. Measuring coverage, Under the condition of constant line speed, by adjusting the straight-line distance from the temperature measurement module to the sealing area of the packaging bag, increasing the area of the measurement area, the theoretical coverage rate can be further improved, and the two unstable data points before and after are discarded. 测 *f 读 ≤v l , theoretically, discarding one reading point before and after can ensure data stability; Step 3: Eigenvalue calculation: Calculate the arithmetic mean μ and standard deviation σ of a dynamic array; Select the appropriate physical model based on the data results; Based on the model, calculate the model weight; Generate a temperature prediction function based on the model weights and obtain the predicted temperature through the current time series; The specific weights are related to the automatically selected model; Newton's law of cooling model, Weight: cooling attenuation coefficient, ambient temperature; Weight function: T t =T a +(T a -T0)e -kt ; Polynomial model, Weight: coefficients of polynomials of various orders; Weight function: T t =a n t n +a n-1 t n-1 +a n-2 t n-2 +…+a2t 2 +a1t 1 +a0; Step 4: Double quality determination: Threshold determination: If the difference between the current temperature and the predicted temperature is too large, a level 1 alarm will be triggered; Distribution judgment: The difference between the current product average and the historical average is too large, triggering a secondary alarm.

2. The system according to claim 1, wherein the method for detecting heat seal quality based on dynamic temperature distribution is characterized in that: In the step 2, the length of the dynamic array is adaptive to the production line speed; in the step 4, the two-level alarm signal is output to the PLC control system through the optocoupler isolation circuit.

3. The system according to claim 2, wherein the method for detecting heat seal quality based on dynamic temperature distribution is characterized in that: include: Infrared temperature measurement module: using thermopile, temperature measurement range 0-500℃, accuracy ±0.5℃; Edge computing unit: integrated processor, dynamic temperature window interception mechanism, multiple judgment procedures; Alarm control module: Contains RS485 communication interface and relay output circuit, supports graded response of alarm signals.

4. The system according to claim 3, wherein: include: The edge computing unit directly reads the infrared sensor data through the DMA channel with a delay of ≤10μs; The system supports local storage of historical temperature data; The selection of a suitable physical model lies in analyzing the relationship between the temperature data collected by the sensor and the time series, and choosing to use Newton's cooling law fitting prediction, polynomial fitting prediction, Kalman recursive filtering prediction, AI model prediction, etc.

5. The system according to claim 4, wherein: The physical model selection is based on the temperature-time series relationship, and the model is adaptively switched according to the variance of the temperature data. When the variance is greater than the threshold, the polynomial model is enabled. The basis for model selection is based on the set cooling rate threshold.

Images