A woven bag production process safety monitoring system

Abstract

The present application belongs to the technical field of industrial control system, and particularly relates to a woven bag production process safety monitoring system, comprising: a data acquisition module, which acquires multi-dimensional data after pretreatment in the workshop; a long-scale prediction module, which predicts the data trend in a preset long-time cycle to obtain a predicted value; a microscopic baseline segmentation module, which obtains the microscopic baseline and the microscopic standard deviation of the data in a sliding observation window to calculate the absolute deviation degree, and updates the microscopic baseline when the adaptive boundary constructed based on the microscopic standard deviation is exceeded; a cross-scale clamping threshold acquisition module, which calculates the difference between the microscopic baseline and the predicted value to obtain the background drift rate, calculates the safety threshold based on the microscopic baseline when the background drift rate does not exceed the allowable limit, and calculates the threshold value based on the cross-scale clamping of the predicted value when the background drift rate exceeds the allowable limit; and a linkage control module, which issues an instruction when any dimensional data continuously crosses the safety threshold. The balance between monitoring sensitivity and anti-interference ability is achieved.

Description

Technical Field

[0001] This invention relates to the field of industrial control system technology. More specifically, this invention relates to a safety monitoring system for the woven bag production process. Background Technology

[0002] The safe and stable operation of the plastic woven bag production line is crucial for the continuous production of the entire factory. The workshop contains multiple process sections with vastly different physical and chemical properties, such as raw material cutting, high-temperature melt extrusion, and high-frequency interlacing circular weaving. In actual continuous operation scenarios, due to the thermal degradation of polymers, complex multi-source dust and harmful volatile chemical gases are generated in the workshop. Once the dust continues to accumulate and exceeds the lower limit of flammability and explosion, or if toxic gases leak slowly, it can easily cause serious fires, explosions, and personnel poisoning accidents. Therefore, real-time monitoring of workshop environmental parameters is the core of preventing major accidents.

[0003] Currently, the monitoring of workshop environmental parameters mostly adopts static monitoring methods based on fixed threshold comparison. For example, Chinese patent application CN117446286A discloses a packaging method for carrying flammable volatile chemical chips. During system operation, it compares data such as combustible gas concentration and electrostatic charge collected by sensors in real time with fixed alarm thresholds. Once the parameters exceed the threshold, an alarm is triggered and operation is stopped. This ignores the complex changes in actual workshop operating conditions at multiple time scales and is a static empirical model that is very prone to false alarms or missed alarms.

[0004] However, the difference in production capacity load between different shifts causes the background dust concentration to fluctuate several times between day and night shifts; the differences in characteristics of different processes mean that the release rate of harmful gases in the melt extrusion stage is much higher than that in the weaving stage, while the dust generation in the raw material cutting stage is higher than that in the printing stage; under the above-mentioned fluctuations in operating conditions at multiple time scales, a single static threshold cannot simultaneously adapt to all time periods and processes: in high-capacity processes, the parameter background itself is high, and the fixed threshold is easily touched frequently, resulting in a large number of false alarms; in low-activity processes, the parameter background is low, and if a real anomaly occurs but the magnitude does not reach the fixed threshold, it will be missed. Summary of the Invention

[0005] To address the technical problem of false alarms and missed alarms caused by fixed thresholds under time-varying operating conditions in the prior art, the present invention provides solutions in the following aspects.

[0006] This invention provides a safety monitoring system for woven bag production processes. The system includes the following modules: a data acquisition module for synchronously acquiring multi-dimensional data after preprocessing in the workshop; a long-scale prediction module for predicting the trend of each dimension of data within a preset long-term period at each moment, and obtaining the predicted value of each dimension of data at each moment; a micro-baseline segmentation module for calculating the micro-baseline and micro-standard deviation of each dimension of data within a sliding observation window at each moment, calculating the absolute deviation of each dimension of data relative to the micro-baseline at each moment, and executing de-jittering logic to update the micro-baseline when the absolute deviation exceeds the adaptive boundary constructed based on the micro-standard deviation; a cross-scale clamping threshold acquisition module for calculating the difference between the updated micro-baseline and the predicted value to obtain the background drift rate; when the background drift rate does not exceed the preset allowable limit, calculating the safety threshold of each dimension of data based on the micro-baseline; when it exceeds the preset allowable limit, executing cross-scale clamping based on the predicted value, and calculating the safety threshold of each dimension of data using the predicted value as a benchmark; and a linkage control module for issuing control commands to associated equipment when it is determined that any dimension of data continuously exceeds its corresponding safety threshold.

[0007] This invention acquires data through a data acquisition module and uses a long-scale prediction module to lock in the overall macroscopic health baseline of the workshop. It combines a micro-baseline segmentation module to dynamically adapt to transient operating conditions and suppress background noise fluctuations. Furthermore, by constructing a cross-scale clamping mechanism, the system has adaptive inclusiveness during normal physical drift and rigid clamping with the help of predicted values ​​when chronic faults worsen. This dynamic and static multi-dimensional evaluation method solves the technical problem of difficulty in setting thresholds in complex and variable physical environments and improves the reliability of safety monitoring in the woven bag production process.

[0008] Preferably, the predicted values ​​of each dimension of the data at each time point satisfy the expression: In the formula, For the first The moment of the first Predicted values ​​for dimensional data; For the first Within the long-term period of the first moment The average value of the dimensional data; For the first The moment of the first Predicted values ​​for dimensional data; The preset smoothing factor is obtained by testing and optimizing historical stable operating data.

[0009] This invention introduces an exponentially weighted moving average recursive model and assigns high weights to historical forecast values. Without increasing the computational overhead of the system hardware, it smooths out the lag effect of long-term physical transition periods from the underlying mathematical logic, providing a macro-prediction baseline for subsequent forecasts.

[0010] Preferably, the step of calculating the micro baseline and micro standard deviation of each dimension data within the sliding observation window at each time step includes: acquiring the data of each dimension at all times within the sliding observation window at each time step; calculating the average value of each dimension data within the sliding observation window and using it as the micro baseline; and calculating the standard deviation of each dimension data within the sliding observation window and using it as the micro standard deviation.

[0011] Preferably, the step of calculating the absolute deviation of each dimension of data at each time point relative to the micro baseline includes: calculating the absolute value of the difference between each dimension of data at each time point and the micro baseline, and using the absolute value as the absolute deviation.

[0012] Preferably, the execution of debouncing logic to update the micro baseline includes: based on the Wright criterion, to As an adaptive boundary, where For the first Within the sliding observation window at time n, the first The micro standard deviation of the dimension; determine whether the absolute deviation is greater than the adaptive boundary: if yes, accumulate the over-limit de-jitter count; if no, clear the over-limit de-jitter count to zero; when the over-limit de-jitter count reaches the set de-jitter cycle threshold, clear the first... The predicted value is used as the updated micro baseline by taking all historical data within a sliding observation window at each time point.

[0013] This invention uses the Wright criterion in statistical control to dynamically define adaptive boundaries. The system can automatically expand and contract according to the inherent noise level of the current workstation, avoiding missegmentation caused by heavy load noise. Combined with the time-domain jitter reduction and counting mechanism, the system can distinguish between transient artifact noise and real process steps, avoiding baseline contamination caused by the mixing of new and drastically different working condition data in the sliding window, and ensuring the authenticity of the micro baseline under time-varying working conditions.

[0014] Preferably, the calculation of the difference between the updated micro baseline and the predicted value to obtain the background drift rate includes: calculating the difference between the updated micro baseline and the predicted value, dividing the difference by the predicted value, and the resulting quotient is the background drift rate.

[0015] This invention employs a dimensionless ratio difference method to obtain the deviation ratio of the micro baseline relative to the predicted value, characterizing the degree of chronic deterioration deviation of the local microenvironment in the workshop from the macro-historical health basis, and providing data basis for identifying hidden faults such as chronic wear of machine parts.

[0016] Preferably, the security thresholds for each dimension of data are calculated based on the micro baseline, including the following expressions: In the formula, For the first The moment of the first Dimensional security threshold; For the first The moment of the first The micro baseline of the dimension; For the first The moment of the first The micro standard deviation of the dimension.

[0017] Preferably, the execution of cross-scale clamping based on predicted values, using the predicted values ​​as a benchmark to calculate the safety threshold for each dimension of data, includes the following expression: In the formula, For the first The moment of the first Dimensional security threshold; This is the preset allowable drift limit; For the first The moment of the first The micro standard deviation of the dimension; For the first The moment of the first Predicted values ​​for dimensional data.

[0018] When the baseline is confirmed to be contaminated, this invention fixes the safety threshold at the predicted value and its allowable drift limit, cutting off the assimilation path of the alarm baseline floating infinitely with abnormal data. This ensures that the system retains a rigid safety baseline when facing long-term chronic faults, and improves the accuracy of capturing minor hidden dangers.

[0019] Preferably, when it is determined that any dimension data continuously exceeds its corresponding safety threshold, the control command issued to the associated device includes: the dimension data refers to the comprehensive dust concentration and the concentration of harmful gases; if the comprehensive dust concentration continuously exceeds its corresponding safety threshold, the issued control command includes increasing the operating frequency of the main dust collection fan, shortening the backflushing interval of the filter cartridge, and reducing the operating frequency of the fiber supply traction motor; if the concentration of harmful gases continuously exceeds its corresponding safety threshold, the issued control command includes turning on the fresh air unit and the exhaust tower, and triggering the safety warning device.

[0020] Preferably, the method for obtaining the preprocessed multi-dimensional data is as follows: the method for obtaining the comprehensive dust concentration is to collect the number of dust particles within a preset discrete particle size range, and substitute the number of dust particles in each particle size range into the corresponding Mie optical scattering cross section coefficient for weighted fusion to obtain the preprocessed comprehensive dust concentration; the method for obtaining the harmful gas concentration is to collect the initial gas concentration, temperature and humidity, substitute the temperature and humidity into a compensation polynomial containing a quadratic term of temperature, a quadratic term of humidity and a temperature and humidity cross-coupling term to calculate the interference offset, and take the difference between the initial gas concentration and the interference offset as the preprocessed harmful gas concentration.

[0021] The beneficial effects of this invention are as follows: This invention constructs a dual-scale benchmark assessment and cross-scale clamping mechanism. On the one hand, through micro-baseline adaptive segmentation and time-domain de-jittering logic, the system can suppress high-frequency transient disturbances such as single-machine jamming and gusts, reducing frequent false alarms under high-capacity load processes. On the other hand, for progressive hidden faults such as chronic blockage of exhaust filters, the system introduces long-scale macro-predictive values ​​as an absolute health reference. Once the baseline drift rate exceeds the limit, cross-scale forced clamping is immediately triggered to avoid the assimilation loophole of the safety alarm line slowly rising with abnormal data, reduce the risk of missed detection of chronic progressive faults, and improve the reliability of production safety monitoring. Attached Figure Description

[0022] Figure 1 This is a schematic diagram illustrating a safety monitoring system for the woven bag production process according to the present invention.

[0023] Figure 2 This diagram schematically illustrates a comparison of the anomaly identification performance between the traditional fixed threshold and the cross-scale clamping threshold monitoring method of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0026] This invention discloses a safety monitoring system for the woven bag production process, referring to... Figure 1 A safety monitoring system for woven bag production process includes modules 101-105: The data acquisition module 101 is used to collect multi-dimensional data after preprocessing in the workshop.

[0027] It should be noted that the physical environment of a plastic woven bag workshop is extremely complex. If the original electrical signals are not corrected to fit the physical characteristics of the workshop, subsequent monitoring results will be distorted. For example, regarding dust concentration, the dust in the woven bag workshop is not from a single source, but rather a mixture highly dependent on specific processes: in the raw material mixing and cutting process, large plastic fragments are mainly generated, with aerodynamic equivalent particle sizes mostly around 10μm; in the high-speed circular weaving process, medium-sized fibrous abrasive dust is mainly generated, mostly distributed around 2.5μm; and in the high-temperature melt extrusion process, polymer volatiles condense in the air to form extremely fine aerosol plumes. These particles are mostly distributed around 0.5 μm. Since conventional laser dust sensors are calibrated for a single standard particle size, and the Mie scattering efficiency of light on the surface of particles of different sizes has a serious nonlinear difference, if they are not calculated separately, the readings will deviate significantly from the true mass concentration. Similarly, for harmful gases, there are not only sudden changes in temperature and humidity in the workshop, but high temperature and high humidity often occur simultaneously. The reaction rate of the electrolyte inside the electrochemical gas sensor is affected by the nonlinear coupling of temperature and humidity, which makes its zero-point drift not a simple linear superposition. If cross-compensation is not performed to restore the true gas release, the monitoring data will be directly used for subsequent analysis, resulting in unreliable monitoring results.

[0028] Specifically, the integrated controller synchronously scans and acquires the raw data stream output by the distributed sensor array within the workshop at a preset sampling rate via an industrial bus. The data stream signals specifically include: Laser dust sensors are installed at key locations in the raw material mixing and cutting process, the high-speed circular weaving process, and the high-temperature melt extrusion process. The system uses a built-in laser particle counter in the sensor to obtain the number of dust particles in real time, which are divided into three core discrete intervals: 0~0.5μm, 0.5~2.5μm, and 2.5~10μm. Electrochemical gas sensors and temperature and humidity sensors are installed at key locations in the production workshop to collect the initial concentration of harmful gases, temperature, and humidity in real time. In this embodiment of the invention, the sampling rate is 1Hz, which can be adjusted by the implementers based on the actual situation.

[0029] After receiving the above data, the controller calls the pre-configured cross-compensation model to perform parallel preprocessing: Based on the well-known Mie scattering theory, the controller extracts the number of dust particles in each process area and particle size range acquired by the sensor in real time, and substitutes them into the corresponding Mie optical scattering cross section coefficient for weighted fusion calculation to obtain the comprehensive dust concentration at each moment. This operation converts the number of particles in the sensor into the actual dust concentration, offsetting the optical measurement deviation caused by the mixing of multi-source dust.

[0030] The system pre-determines the drift curves of the electrochemical gas sensor under different temperature and humidity gradients in a constant temperature and humidity calibration chamber, and constructs a compensation polynomial containing quadratic terms of temperature, humidity, and temperature-humidity cross-coupling terms by fitting the data using the least squares method. During actual monitoring, the controller uses the initial concentration of harmful gas at each moment as the base, and substitutes the real-time collected ambient temperature and humidity into the above polynomial model to calculate the false zero-point offset of the initial concentration of harmful gas caused by temperature and humidity interference. Subsequently, the difference between the initial concentration of harmful gas at each moment and the offset is calculated to obtain the concentration of harmful gas at each moment.

[0031] After the above preprocessing, the data sequence that suppresses physical interference in the underlying environment is extracted, thus avoiding false alarms in monitoring caused by physical interference.

[0032] At this point, the overall dust concentration and harmful gas concentration at each moment are obtained.

[0033] The long-scale prediction module 102 is used to obtain the predicted values ​​of each dimension of data at each time point.

[0034] It should be noted that the woven bag production workshop has significant multi-dimensional time-varying scene characteristics: the intermittent dust from the cutting process will generate intermittent physical burrs and noise on a second to minute scale; while the difference in production capacity between day shift and night shift and the overall ventilation changes in the factory will cause a generally smooth macroscopic background drift on an hourly scale; since there is a difference in magnitude between the second-level pulse and the hourly shift jump, this step sets a fixed long-term sampling period as a spatiotemporal scale isolator to achieve the initial extraction of the benchmark level of each dimension of the workshop while ensuring extremely low computational overhead of the microcontroller.

[0035] Specifically, a fixed long-term period is preset, and the controller continuously acquires data sequences of various dimensions within each long-term period at each moment, including: comprehensive dust concentration sequence and harmful gas concentration sequence.

[0036] It should be emphasized that in a woven bag workshop, the complete physical weaving cycle of a standard circular loom producing a standard specification woven bag is about 45 to 60 seconds. Therefore, in this embodiment, the fixed long cycle is set to 60 seconds, which precisely includes a complete physical closed loop of high-frequency interlacing, complete periodic fluctuations of yarn tension, and frictional heating and heat dissipation of machine parts, and can encompass a single weaving physical cycle.

[0037] Considering that the device has just been powered on and is in the first long-term period, and since the complete long-term period data cannot be obtained at these times, a linear interpolation algorithm is used to complete the missing data of each dimension in the long-term period at these times.

[0038] For each time point, the controller first calculates the average value of each dimension of the data over the long-term period at that time, and then uses an exponentially weighted moving average model to calculate the predicted value of each dimension of the data at that time; the specific calculation formula is as follows:

[0039] In the formula, For the first The moment of the first Predicted values ​​for dimensional data; For the first Within the long-term period of the first moment The average value of the dimensional data; For the first The moment of the first Predicted values ​​for dimensional data; This is the preset smoothing factor.

[0040] The smoothing factor was obtained as follows: During the system deployment phase, 30 consecutive days of stable operation test data from the woven bag workshop were collected and preprocessed to construct a test dataset. Considering that an excessively large smoothing factor would make the predicted values ​​highly susceptible to adverse effects from localized spikes such as single-machine yarn breakage, while an excessively small smoothing factor would lose its ability to track normal global operating condition changes such as shift changes and speed adjustments, and considering that after the workshop experiences day-night shift changes or significant adjustments in overall production capacity, due to the delay in thermodynamic conduction and the physical inertia of dust gravity settling, the physical transition time for environmental parameters to reach a new steady state is typically between 20 and 200 minutes, the smoothing factor was determined based on the formula for the equivalent memory window length of the exponentially weighted moving average algorithm. ,in, The equivalent window length is in minutes. To ensure the system's memory covers the aforementioned physical transition period of 20 to 200 minutes, the formula is derived in reverse by substituting the values. The constraint range is within [0.01, 0.1]. Based on this constraint, the system will... Within the restricted interval, rolling simulation tests were conducted with multiple sets of parameters at a fixed step size of 0.005. The residual values ​​between the predicted and actual curves were calculated for each set of parameters. The value corresponding to the minimum residual value was obtained as the smoothing factor. Historical data were given higher weights based on the obtained smoothing factor, making the calculated... The moment of the first The predicted values ​​of the dimensional data will not change drastically due to local short-term disturbances, ensuring that the extracted values... Represents the first The stable state of the workshop under normal operation over a long period of time.

[0041] At this point, the predicted values ​​of each dimension of the data at each moment have been obtained.

[0042] The micro baseline segmentation module 103 is used to calculate the absolute deviation of each dimension of data relative to the micro baseline at each time step, and to execute de-jittering logic to update the micro baseline when the absolute deviation exceeds the adaptive boundary constructed based on the micro standard deviation.

[0043] It should be noted that the woven bag production process involves several technological stages with drastically different physical and chemical properties, such as raw material cold cutting, high-temperature melt extrusion, and high-speed circular weaving. When the workshop switches processes according to the production schedule or the unit stops to change rolls, the background of environmental parameters will undergo a step change. If the monitoring system uses a rigid fixed-length sliding window to span the switching point, the window will be mixed with data from the two completely different working conditions, resulting in serious distortion of the calculated working condition benchmark. Therefore, the system must also use the statistical characteristics of the data itself to construct an adaptive discrimination boundary mechanism, so as to filter out high-frequency transient noise and confirm the real process step within seconds.

[0044] Specifically, the data processing unit maintains a real-time sliding observation window in memory with a total length of a preset microscopic observation period. In this embodiment, the continuous consumption life of a single roll of flat yarn in a circular loom is limited to 20 minutes. The implementer can adjust this based on the life performance of the circular loom.

[0045] The controller calculates the mean and standard deviation of each dimension's data within the sliding observation window at each time in real time. The mean is used as the micro baseline for each dimension at each time, and the standard deviation is used as the micro standard deviation for each dimension at each time. Based on the difference between each dimension's data at each time and the micro baseline, the absolute deviation of each dimension's data at each time is calculated. The specific calculation formula is as follows:

[0046] In the formula, For the first The moment of the first The absolute deviation of the dimensional data; For the first The moment of the first Dimensional data; For the first The moment of the first The micro baseline of the dimension; To take the absolute value.

[0047] in, Reflects the first The degree of deviation of the transient physical environment of the workshop from the microscopic baseline within a given moment; the larger the value, the more severe the deviation. The more intense the mechanical vibration or thermodynamic abrupt change that the local microenvironment of the workshop is experiencing at any given moment, the more severe the change.

[0048] Subsequently, the Wright criterion from statistical control was invoked to... As an adaptive boundary, where For the first Within the sliding observation window at time n, the first The micro-standard deviation of the dimension is used to execute time-domain debouncing logic: like This increments the internal debouncing counter step by one. like Then, the debounce count that exceeds the limit will be forcibly reset to zero instantly.

[0049] in, Reflects the first The limiting physical noise envelope space of the production process at a given moment in steady state; the larger this value, the more significant the difference between the first and second moments. The more intense the fluctuations in the normal physical operation of the workstation at any given moment, the more this adaptive mechanism allows the system to automatically provide greater leeway when facing heavy-load processes with high background noise, thereby preventing normal mechanical noise from being misjudged as changes in operating conditions.

[0050] When the count of excessive de-jitter reaches the preset de-jitter cycle threshold, the system confirms that the process has been substantially split and immediately clears the first... The sliding observation window at time point 1 will collect all historical data and reset the over-limit debounce counter to zero, and then... The predicted values ​​of each dimension of the data at time step n are used as the first time step n. Micro baselines in each dimension at each moment.

[0051] The preset de-jittering period threshold reflects the system's tolerance to the temporal physical lifecycle of spurious abnormal fluctuations. The larger the threshold, the longer the natural dissipation time of occasional transient disturbances such as factory gusts and single-wire jams that the system needs to contain and filter. Since the extreme value of the natural dissipation time of a single physical disturbance in the workshop is within 8 seconds, the de-jittering period threshold must be strictly greater than 8 seconds. In this embodiment, 10 seconds is used, which can encapsulate the concurrent artifact noise of a single physical disturbance on the time axis. The implementer can adjust it based on the actual workshop environment.

[0052] At this point, the micro-baselines of each dimension at each moment have been obtained.

[0053] The cross-scale clamping threshold acquisition module 104 is used to calculate the difference between the updated micro baseline and the predicted value to obtain the background drift rate; when the background drift rate does not exceed the preset allowable limit, the safety threshold of each dimension of data is calculated based on the micro baseline; when it exceeds the preset allowable limit, cross-scale clamping based on the predicted value is performed, and the safety threshold of each dimension of data is calculated using the predicted value as the benchmark.

[0054] It should be noted that, for example, chronic wear and heating of the main bearing of a circular loom and gradual clogging of the exhaust filter can cause local parameters to rise slowly over several hours at an extremely small slope. Because this deterioration is extremely slow, the short-scale microscopic observation window will continuously follow it, causing the microscopic alarm line to rise along with the abnormal data, which is very easy to miss. Therefore, the predicted values ​​of data of each dimension extracted at each moment within a long time period are introduced as a mandatory constraint to perform cross-scale clamping interception on the baseline of the microscopic observation window, thereby ensuring that the calculated safety threshold is more reasonable.

[0055] Specifically, based on the difference between the micro baselines of each dimension and the predicted values ​​at each time point, the background drift rate of each dimension at each time point is calculated, and the specific expression is as follows:

[0056] In the formula, For the first The moment of the first The baseline drift rate of dimensional data; For the first The moment of the first The micro baseline of the dimension; For the first The moment of the first Predicted values ​​for dimensional data.

[0057] in, Reflects the first The degree of deviation of the workshop's local short-term operating conditions from the macro-historical healthy baseline at a given moment; the larger this value, the more pronounced the deviation. The more severe the assimilation effect of chronic wear and heating or gradual dust accumulation in the bearings of a circular loom at any given moment, the more it means that the first... The microenvironment of the workshop at that moment may have been continuously contaminated and deviated from its historical healthy level.

[0058] Then, Compared with the preset allowable drift limit The comparison is performed, and the following threshold calculation model is applied for traffic splitting: It should be added that, firstly, based on the Arrhenius equation characterizing chemical and physical volatilization rates, the typical diurnal temperature variation in a woven bag workshop, usually between 10°C and 12°C, will cause an 8% to 10% natural physical increase in the thermodynamic activity of polypropylene surface molecules and the release background of normal mechanical friction particles; therefore, the drift limit must be set above 10% to absolutely encapsulate and avoid false alarms caused by macroscopic fluctuations due to the natural temperature difference between day and night shifts; secondly, based on the mechanical and thermal decay characteristics of polymer materials, standard circular looms are designed for PP flat yarns... The safety margin for the tensile strength limit is usually around 15% of the rated tensile strength. When local chronic heat accumulation or abnormal wear of machine parts causes the background rise of environmental dust and volatile gases to reach 15% over a long period, it indicates that the surface hardness of the polypropylene material has entered a non-linear and rapidly decreasing region. Based on the above analysis, in this embodiment, the allowable drift limit is set at 15%, which not only provides sufficient safety margin for the natural changes in the workshop day and night, but also suppresses it before the physical phase transition critical point of softening and deformation of the polypropylene material, which leads to large-scale fiber breakage and shutdown.

[0059] when When, explain the first The moment of the first The micro-baseline is uncontaminated, and the normal envelope is executed:

[0060] when When, explain the first The moment of the first The dimensional micro baseline has been contaminated by a progressive fault, triggering a forced clamping envelope:

[0061] In the formula, For the first The moment of the first Dimensional security threshold; This is the preset allowable drift limit; For the first The moment of the first The micro standard deviation of the dimension; For the first The moment of the first Predicted values ​​for dimensional data; For the first The moment of the first The micro baseline of the dimension.

[0062] Among them, when the system discovers the first The moment of the first When the background drift rate of a dimension exceeds the limit, the decision-making power of the contaminated micro baseline is forcibly suppressed, and it is mapped to the vicinity of the predicted value of each dimension data extracted at each time point within a long time period, thereby ensuring that the obtained safety threshold achieves a balance between identifying hidden faults and suppressing high-frequency false alarms.

[0063] At this point, the security thresholds for each dimension at each moment have been obtained.

[0064] The linkage control module 105 is used to issue control commands to associated devices when any dimension of data continuously exceeds its corresponding safety threshold.

[0065] It should be noted that dust exceeding the limit indicates uncontrolled diffusion of particles from physical cutting or weaving; excessive harmful gases indicate the accumulation of volatile chemicals in the extrusion process. If a single strategy of shutting down the entire line is adopted, it will lead to the scrapping of work-in-process and the inability to expel toxic gases. Therefore, a differentiated hardware linkage strategy should be issued to match different dimensions.

[0066] Specifically, if data in any dimension exceeds its safety threshold for multiple consecutive time periods, this embodiment preferably sets it to 5 consecutive time periods, which can be adjusted by the implementer based on actual production monitoring requirements; if the system determines that any dimension of data is abnormal, the system issues differentiated instructions to the corresponding control equipment through the industrial communication bus: For example, if the overall dust concentration exceeds the limit abnormally, the system determines that the dust is in an uncontrolled diffusion state. The PLC issues a high-frequency exhaust command, which fully activates the variable frequency dust collection main fan and shortens the filter cartridge backflushing interval. Simultaneously, the frequency of the wire feeding traction motor is reduced to suppress dust generation by decreasing the physical shear rate. If the concentration of harmful gases exceeds the limit abnormally, the system determines that there is a dangerous accumulation of polymer thermal degradation volatiles. The PLC issues a command to start all fresh air units and exhaust towers to dilute the plume and triggers a safety warning device to remind relevant personnel to pay attention to safety.

[0067] For example, Figure 2 This image compares the anomaly identification performance of the traditional fixed threshold monitoring method and the cross-scale clamping threshold method of this invention. Parallel tests were conducted on continuous operating data from a woven bag production workshop under complex environmental conditions, including alternating day and night shifts, high-frequency wire breakage and jamming, and chronic clogging of exhaust filters. The results are as follows: In terms of overall accuracy, the traditional method achieves only 68.5%, while the method of this invention jumps to 98.2%. For high-frequency transient disturbances, this invention utilizes an adaptive micro-jitter removal mechanism, reducing the false alarm rate from 24.3% of the traditional method to 1.1%, thus avoiding interference with normal production scheduling. Regarding the prevention of chronic progressive faults, the traditional method has a false alarm rate as high as 18.7%, while this invention, relying on cross-scale forced clamping envelope, reduces the false alarm rate to 0.4%. Therefore, this invention achieves a balance between monitoring sensitivity and anti-interference capability, improving the accuracy of safety monitoring in the woven bag production process.

Claims

1. A safety monitoring system for the woven bag production process, characterized in that, The system includes the following modules: The data acquisition module is used to synchronously collect multi-dimensional data after preprocessing in the workshop; The long-scale prediction module is used to predict the trend of data in each dimension within a preset long-term period at each moment, and obtain the predicted value of data in each dimension at each moment. The micro baseline segmentation module is used to calculate the micro baseline and micro standard deviation of each dimension of data within the sliding observation window at each time step, calculate the absolute deviation of each dimension of data relative to the micro baseline at each time step, and execute de-jittering logic to update the micro baseline when the absolute deviation exceeds the adaptive boundary constructed based on the micro standard deviation. The cross-scale clamping threshold acquisition module is used to calculate the difference between the updated micro baseline and the predicted value to obtain the background drift rate; when the background drift rate does not exceed the preset allowable limit, the safety threshold of each dimension of data is calculated based on the micro baseline. When the preset allowable limit is exceeded, cross-scale clamping based on the predicted value is executed, and the safety threshold of each dimension of data is calculated based on the predicted value. The linkage control module is used to issue control commands to associated devices when it is determined that any dimension of data continuously exceeds its corresponding safety threshold.

2. The safety monitoring system for woven bag production process according to claim 1, characterized in that, The predicted values ​​of each dimension of the data at each time point satisfy the expression: ； In the formula, For the first The moment of the first Predicted values ​​for dimensional data; For the first Within the long-term period of the first moment The average value of the dimensional data; For the first The moment of the first Predicted values ​​for dimensional data; The preset smoothing factor is obtained by testing and optimizing historical stable operating data.

3. The safety monitoring system for woven bag production process according to claim 1, characterized in that, The calculation of the micro baseline and micro standard deviation of each dimension of data within the sliding observation window at each time point includes: Acquire the data for each dimension at all times within the sliding observation window at each time point; calculate the average value of the data for each dimension within the sliding observation window and use it as the micro baseline; calculate the standard deviation of the data for each dimension within the sliding observation window and use it as the micro standard deviation.

4. The safety monitoring system for woven bag production process according to claim 1, characterized in that, The calculation of the absolute deviation of each dimension of data relative to the micro baseline at each time point includes: Calculate the absolute value of the difference between each dimension of data at each time point and the micro baseline, and use the absolute value as the absolute deviation.

5. The safety monitoring system for woven bag production process according to claim 1, characterized in that, The execution of the debouncing logic to update the micro baseline includes: Based on the Wright criterion, As an adaptive boundary, where For the first Within the sliding observation window at time n, the first The micro standard deviation of the dimension; determine whether the absolute deviation is greater than the adaptive boundary: if yes, accumulate the over-limit de-jitter count; if no, clear the over-limit de-jitter count to zero; when the over-limit de-jitter count reaches the set de-jitter cycle threshold, clear the first... The predicted value is used as the updated micro baseline by taking all historical data within a sliding observation window at each time point.

6. The safety monitoring system for woven bag production process according to claim 1, characterized in that, The difference between the calculated updated micro baseline and the predicted value is used to obtain the background drift rate, including: The difference between the updated micro baseline and the predicted value is calculated, and the difference is divided by the predicted value. The resulting quotient is the background drift rate.

7. The safety monitoring system for woven bag production process according to claim 1, characterized in that, The security thresholds for each dimension of data are calculated based on the micro baseline, including the following expressions: ； In the formula, For the first The moment of the first Dimensional security threshold; For the first The moment of the first The micro baseline of the dimension; For the first The moment of the first The micro standard deviation of the dimension.

8. A safety monitoring system for woven bag production process according to claim 1, characterized in that, The execution of cross-scale clamping based on predicted values, using predicted values ​​as a benchmark to calculate the safety thresholds for data in each dimension, includes the following expressions: ； In the formula, For the first The moment of the first Dimensional security threshold; This is the preset allowable drift limit; For the first The moment of the first The micro standard deviation of the dimension; For the first The moment of the first Predicted values ​​for dimensional data.

9. A safety monitoring system for woven bag production process according to claim 1, characterized in that, When it is determined that data in any dimension continuously exceeds its corresponding safety threshold, a control command is issued to the associated device, including: The dimensional data refers to the combined dust concentration and the concentration of harmful gases. If the combined dust concentration continuously exceeds its corresponding safety threshold, the control instructions issued include increasing the operating frequency of the main dust collection fan, shortening the backflushing interval of the filter cartridge, and reducing the operating frequency of the wire feeding traction motor. If the concentration of harmful gases continuously exceeds its corresponding safety threshold, the control instructions issued include starting the fresh air unit and the exhaust tower, and triggering the safety warning equipment.

10. A safety monitoring system for woven bag production process according to claim 1, characterized in that, The method for obtaining the preprocessed multi-dimensional data is as follows: The method for obtaining the comprehensive dust concentration is as follows: collect the number of dust particles within a preset discrete particle size range, substitute the number of dust particles in each particle size range into the corresponding Mie optical scattering cross section coefficient for weighted fusion, and obtain the pre-processed comprehensive dust concentration; the method for obtaining the harmful gas concentration is as follows: collect the initial gas concentration, temperature and humidity, substitute the temperature and humidity into a compensation polynomial containing a quadratic term of temperature, a quadratic term of humidity and a temperature and humidity cross-coupling term to calculate the interference offset, and take the difference between the initial gas concentration and the interference offset as the pre-processed harmful gas concentration.

Images