A polyester staple fiber oil content detection method and system

Abstract

The application provides a polyester staple fiber oil content detection method and system, and relates to the technical field of chemical fiber production detection, wherein the method comprises the following steps: inserting a detection head of a detection pen into a polyester staple fiber sample; obtaining an electrical signal changed due to the change of fiber oil content through the detection head; comparing the electrical signal with preset oil content reference data; and outputting oil content information obtained based on the comparison result. The application realizes on-site rapid detection of polyester staple fiber oil content, greatly improves production efficiency, and eliminates subjective errors of human experience judgment through objective electrical signal measurement, thereby providing instant and reliable quantitative basis for process adjustment.

Description

Technical Field

[0001] This invention relates to the field of chemical fiber production testing technology, and in particular to a method and system for detecting the oil content of polyester staple fiber. Background Technology

[0002] In the production process of polyester staple fiber, the oil content is a key indicator that directly affects the spinnability of the fiber, the quality of subsequent processing, and the hand feel of the final product.

[0003] Currently, the industry mainly relies on two testing methods: one is laboratory testing using the Soxhlet extraction method. Although this method is accurate, the process is cumbersome and time-consuming, taking about 2 hours to complete a single test, resulting in serious data lag and failing to meet the needs of real-time adjustments to process parameters on the production site; the other is relying on the judgment of experienced personnel, that is, estimating the oil content by touch and sight. Although this method is quick to respond, it is completely limited by personal subjective experience and has inherent defects such as inconsistent judgment standards, inability to quantify results, and large differences in judgment among different personnel, making it difficult to serve as a reliable basis for process adjustments.

[0004] Therefore, existing technologies cannot simultaneously meet the requirements of speed, objectivity and accuracy for on-site testing, and there is an urgent need for a solution for detecting the oil content of polyester staple fibers that can be fast, objective and easy to use on-site. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for detecting the oil content of polyester staple fibers, so as to solve the problems pointed out in the background art.

[0006] In a first aspect, an embodiment of the present invention provides a method for detecting the oil content of polyester staple fiber, comprising: Insert the testing head of the testing pen into the polyester staple fiber sample; The detection head acquires electrical signals that change due to variations in fiber oil content. The electrical signal is compared with preset oil content benchmark data; Output the oil content information obtained based on the comparison results.

[0007] Optionally, the detection head has a surface-treated contact portion, the surface treatment making the contact portion oleophobic or oleophilic; the electrical signal is the capacitance or resistance value of the detection head.

[0008] Optionally, the step of comparing the electrical signal with preset oil content benchmark data is executed by the main control chip inside the detection pen, and the main control chip has pre-stored the mapping relationship data between the electrical signal and the oil content.

[0009] Optionally, the steps for outputting oil content information include: displaying the oil content level in different colors using multiple LED indicators, or displaying the percentage value of oil content on an LCD digital screen.

[0010] Secondly, an embodiment of the present invention provides a polyester staple fiber oil content detection system, comprising: A detection pen, the detection pen being configured to perform the method described in any of the first aspects.

[0011] Optionally, the detection pen includes: Pen body; The detection head is located at the front end of the pen body; A circuit module housed inside the pen body, the circuit module being electrically connected to the detection head; The display unit is located on the outside of the pen body and is electrically connected to the circuit module.

[0012] Optionally, the circuit module includes: A signal conversion circuit is used to receive and process electrical signals from the detection head; The main control chip is connected to the signal conversion circuit and is used to query the mapping relationship data and determine the oil content based on the processed signal.

[0013] Optionally, the display unit is three LED indicator lights or an LCD digital screen; the detection head is a replaceable metal sensing probe; the system also includes a power supply unit for powering the circuit module and the display unit.

[0014] Optionally, the polyester staple fiber oil content detection system also includes: Cloud platform; The circuit module further includes: A digital temperature and humidity sensor is connected to the main control chip; A multi-mode wireless communication chip, connected to the main control chip, is used to collect the digital identity of the polyester staple fiber sample and the ID of the testing personnel. Together with the raw electrical signal data containing time-series information collected by the main control chip and the environmental parameters collected by the digital temperature and humidity sensor, it is used to construct a multi-source data packet and upload it to the cloud platform. The cloud platform is configured to perform the following operations: The multi-source data packets are received and parsed, and mapped to a pre-built digital twin model; the digital twin model integrates a dynamic task topology network based on production batches; When an abnormal oil content event is identified in the digital twin model, an intelligent traceability mechanism is activated to construct a digital trajectory of detection operations associated with the abnormal event; wherein, the digital trajectory of detection operations includes: a high-dimensional feature vector that integrates electrical signal stability index, detection time series distribution map, spatial positioning topology, environmental parameter fluctuation spectrum and historical detection consistency feature matrix. The digital trajectory of the detection operation is analyzed by a multi-source data fusion analysis engine, and the abnormal distribution pattern within the batch is analyzed in combination with the dynamic task topology network. The abnormal distribution pattern analysis within the batch includes: evaluating the correlation and dispersion of abnormal events between different detection nodes within the same production batch. Based on the analysis results of the digital trajectory of the detection operation and the abnormal distribution pattern within the batch, cross-validation is performed to identify non-standard use of the detection pen caused by the unintentional actions of the detection personnel. When the unintentional non-standard usage behavior is identified, the characteristic patterns of the behavior and its impact on batch quality data are located through correlation graph analysis. Based on the aforementioned feature patterns and their scope of influence, a digital correction scheme is generated, which includes personalized operation correction guidelines and intelligent re-inspection paths, and is pushed to the corresponding on-site management terminal in real time.

[0015] In one embodiment, the process of constructing the digital trajectory of the detection operation includes: The original electrical signal data is extracted from the multi-source data packet. Based on the detection start time stamp, signal stabilization time stamp and detection end time stamp, the data subsequence of the stable detection stage is determined. The coefficient of variation of the data subsequence is calculated as an indicator of electrical signal stability. Calculate the duration before and after stabilization, compare the two currently measured duration values ​​with the corresponding historical averages, and classify the time pattern as too fast, normal, or too slow by calculating the deviation between the current value and the historical average and comparing it with the threshold range set based on the standard deviation of historical data, so as to generate a detection time series distribution map. Using the received sample unique identifier as a query condition, a targeted query is initiated to the graph database on which the digital twin model is based. The specific location information of the sample in the factory spatial layout is obtained through graph traversal algorithm, and one-hot encoding technology is used to convert it into a machine-readable feature vector to generate spatial positioning topology. The precise location information of the sample is extracted from the spatial positioning topology, and the raw environmental humidity data associated with the corresponding location during a specified time period before the start of the detection is obtained by querying the time series database. The variance of the time series is calculated as the environmental parameter fluctuation spectrum. Obtain historical detection event data within a preset number of times for the operator, and calculate the mean and variance of its electrical signal stability index and total detection time to form a historical detection consistency feature matrix; The electrical signal stability index, detection time-series distribution map, spatial positioning topology, environmental parameter fluctuation spectrum, and historical detection consistency feature matrix are fused into a high-dimensional feature vector to form the digital trajectory of the detection operation.

[0016] The present invention has achieved the following beneficial effects: It enables rapid on-site detection of oil content in polyester staple fiber, greatly improving production efficiency; by eliminating subjective errors from human experience judgment through objective electrical signal measurement, it provides an immediate and reliable quantitative basis for process adjustment.

[0017] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the method for detecting the oil content of polyester staple fiber in an embodiment of the present invention; Figure 2 This is a schematic diagram of the detection pen in an embodiment of the present invention; Figure 3 This is a schematic diagram of the polyester staple fiber oil content detection system in an embodiment of the present invention; Figure 4 This is a flowchart illustrating the steps involved in configuring and performing operations on a cloud platform, as described in this embodiment of the invention. Figure 5 This is a flowchart illustrating another step in configuring the cloud platform to perform operations in an embodiment of the present invention. Detailed Implementation

[0020] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0021] This application provides a method for detecting the oil content of polyester staple fiber, such as... Figure 1 As shown, it includes: S1. Insert the detection head of the detection pen into the polyester staple fiber sample; S2. Obtain the electrical signal that changes due to the change in fiber oil content through the detection head; S3. Compare the electrical signal with the preset oil content benchmark data; S4. Output the oil content information obtained based on the comparison results.

[0022] The core of this embodiment lies in systematically utilizing the physical properties of the surface oil of polyester fibers to accurately convert its content into measurable electrical parameters, and achieving accurate quantitative or semi-quantitative analysis through a pre-established calibration model. The specific implementation path is as follows: When the metal detection head at the tip of the pen comes into contact with the polyester staple fiber sample, it essentially forms a composite electrical sensor together with the fiber. Its physical basis includes two key mechanisms: First, the dielectric property change mechanism. Since the oil on the fiber acts as a dielectric material, its dielectric constant is significantly higher than that of the fiber itself and air. When the detection head (as an electrode) is inserted into the fiber, the presence of the oil alters the dielectric environment of the electric field between the detection head and the reference ground (which could be the grounded part of the pen body or another conductive path in the fiber). The higher the oil content, the greater the equivalent dielectric constant in the sensing area, directly leading to an increase in the capacitance value sensed by the detection head. Second, the conductivity change mechanism. The antistatic agents and other components in the oil have a certain degree of ionic conductivity. An increase in the oil content creates better conductive paths between the fibers, thereby causing a regular change in the resistance or impedance value sensed by the detection head.

[0023] These minute changes in capacitance or resistance sensed by the detection head are then precisely captured, conditioned, and digitized by a dedicated signal conversion circuit integrated within the pen (such as a capacitance-to-digital converter (CDC) or a precision analog front-end (AFE)), forming a high-precision digital electrical signal. This digital signal is immediately sent to the main control chip for processing. This chip pre-stores a database (calibration curve) of electrical signal-oil content mapping relationships established through numerous rigorous experiments. This database was established by accurately measuring the true oil content of a series of known samples using traditional standard methods (such as Soxhlet extraction) in a laboratory environment, while simultaneously measuring the corresponding electrical signal values ​​using the prototype of this detection pen, and then fitting the data. Upon receiving the real-time signal, the main control chip automatically and accurately maps the electrical signal to an estimated oil content value for the current fiber sample through an efficient lookup table algorithm and interpolation calculation.

[0024] In a preferred embodiment of the detection pen of the present invention, the detection head and the reference ground physically constitute an open-ended coaxial probe structure. This structure consists of a central sensing electrode (corresponding to the detection head) and a coaxially surrounding external reference electrode (corresponding to the reference ground or the grounding portion of the pen body). Its measurement principle is based on the fringing field effect: when the open cross-section of the probe tip contacts the polyester short fiber sample, excitation electric field lines emanate from the central electrode, penetrate a certain volume of the multiphase mixture of fiber, oil, air, and unavoidable ambient moisture near the probe tip, and then return to the external reference electrode.

[0025] The dielectric and conductivity variation mechanisms described in this invention are physically unified and characterized as the complex permittivity of the mixed medium under test. ).in: 1. Real part (Dielectric constant): Reflects the energy storage characteristics (capacitive effect) of a dielectric material; its value is determined by the fiber ( ),Air( ), oil-based oil ( ) and moisture ( The volume fraction of ) is determined together.

[0026] 2. Imaginary part (Dielectric loss): Reflects the energy dissipation characteristics (conductivity effect) of a medium. Its main contribution comes from the ionic conductivity of antistatic agents (such as quaternary ammonium salts and other ionic compounds) contained in the oil, as well as the dipolar relaxation of water molecules.

[0027] However, the sample to be tested is a multiphase, multivariable system composed of fibers, air, moisture, and oil, and its electrical response is also strongly disturbed by the sample's bulk density (i.e., the compaction of the fibers). If capacitance or resistance is measured only at a single frequency, the oil content cannot be uniquely calculated from the strong interference of moisture and density.

[0028] To address this multivariable decoupling challenge, the main control chip of this invention employs electrochemical impedance spectroscopy (EIS) (also known as dielectric spectroscopy). By applying excitation at multiple frequency points (e.g., a frequency sweep from 1 kHz to 10 MHz), the dedicated signal conversion circuit measures and acquires the complete complex permittivity spectrum. Due to the ionic conductivity of the oil ( ) and the dipole relaxation of water ( The frequency response characteristics of the two (which have relaxation peaks at a specific frequency) are quite different, and this spectral data contains sufficient information to distinguish between them.

[0029] At this point, the pre-stored mapping data within the chip is essentially based on physical equations from dielectric mixture theory (such as the Bruggeman-Hanai model). The main control chip uses advanced algorithms to process the measured data... By fitting spectral data to this physical model, multiple parameters such as oil content, water content, and bulk density can be simultaneously decoupled and inverted. This method fundamentally eliminates measurement interference caused by ambient humidity and uneven compaction during sample insertion by the operator.

[0030] In a further preferred embodiment, to address the probe contamination problem caused by oil residue, the probe structure further includes a guard electrode located between the central sensing electrode and the external reference electrode. The circuit employs active shielding technology, meaning that the guard electrode is driven by a buffer amplifier to ensure its potential is always perfectly aligned with the potential of the central sensing electrode (same potential, same phase). Since the potential difference between the sensing electrode and the guard electrode is zero, any oil contaminants bridging these two electrodes cannot generate leakage current. ,because This design makes the probe electrically immune to surface contamination, greatly improving the robustness and reliability of continuous measurements in harsh industrial environments.

[0031] This embodiment enables rapid on-site detection of oil content in polyester staple fibers, greatly improving production efficiency; by eliminating subjective errors from human experience-based judgment through objective electrical signal measurement, it provides an immediate and reliable quantitative basis for process adjustment.

[0032] In one embodiment, the detection head has a surface-treated contact portion, the surface treatment making the contact portion oleophobic or oleophilic; the electrical signal is the capacitance or resistance value of the detection head.

[0033] This embodiment utilizes an oleophobic or oleophilic surface treatment (such as a Teflon coating) on ​​the contact area of ​​the detection head to precisely control the wetting behavior and distribution of the oil on the probe surface. This specific surface characteristic makes the detection head more sensitive to the presence of oil on the fiber surface, generating more significant and regular changes in electrical signals (capacitance / resistance values), thereby improving detection sensitivity and repeatability.

[0034] In one embodiment, the step of comparing the electrical signal with preset oil content benchmark data is performed by the main control chip inside the detection pen, and the main control chip has pre-stored mapping relationship data between the electrical signal and the oil content.

[0035] This embodiment utilizes the main control chip (MCU) inside the detection pen to execute the core judgment logic. The main control chip pre-stores a data table (calibration curve) of the electrical signal-oil content mapping relationship established through a large number of standard samples. After the signal conversion circuit converts the analog signal from the detection head into a digital electrical signal, the main control chip reads the signal in real time and maps it to a specific oil content value or grade through table lookup and calculation.

[0036] In one embodiment, the step of outputting oil content information includes: displaying the oil content level in different colors using multiple LED indicator lights, or displaying the percentage value of oil content using an LCD digital screen.

[0037] This embodiment provides two result output methods. For rapid qualitative judgment, multiple LED indicator lights (such as green, yellow, and red) are used to visually display the oil content level, such as qualified, warning, or unqualified. For scenarios requiring quantitative data, the precise percentage value of oil content (such as 0.15%) is directly displayed on the LCD digital screen.

[0038] This application provides a system for detecting the oil content of polyester staple fiber, comprising: A detection pen, the detection pen being configured to perform the method described in any of the above embodiments.

[0039] This embodiment integrates all the detection steps defined in the aforementioned method claims into a portable detection pen. The core of this system lies in the fact that the detection pen is configured and programmed to automatically execute the complete detection process from fiber contact, signal acquisition, data processing to result output.

[0040] In one embodiment, such as Figure 2 As shown, the detection pen includes: Pen body 100; The detection head 200 is disposed at the front end of the pen body 100; A circuit module housed inside the pen body 100, the circuit module being electrically connected to the detection head 200; The display unit 300 is disposed outside the pen body 100 and is electrically connected to the circuit module.

[0041] This embodiment constructs the basic physical architecture of the detection pen. The pen body 100 serves as the main structure and handheld part, housing and protecting all internal components. The detection head 200, as a sensor, is located at the front end and is responsible for directly interacting with the fibers and generating raw signals. The circuit module acts as the brain, processing and judging signals, while the display unit 300 serves as the human-machine interface, outputting results. The two are electrically connected to form a complete signal chain. The detection pen also has a start button 400 for starting the pen and a reset button 500 for resetting the pen's operation.

[0042] In one embodiment, the circuit module includes: A signal conversion circuit is used to receive and process electrical signals from the detection head; The main control chip is connected to the signal conversion circuit and is used to query the mapping relationship data and determine the oil content based on the processed signal.

[0043] This embodiment defines the internal structure of the circuit module. The signal conversion circuit (such as a capacitor-to-digital converter (CDC) or an analog-to-digital converter (ADC)) is responsible for amplifying, filtering, and converting the weak analog signal sensed by the detection head into a digital signal. The main control chip receives this digital signal, runs its internal program, and executes the oil content determination logic by querying its pre-stored mapping database.

[0044] In one embodiment, the display unit is three LED indicator lights or an LCD digital screen; the detection head is a replaceable metal sensing probe; the system also includes a power supply unit for powering the circuit module and the display unit.

[0045] This embodiment defines preferred solutions for key components. The replaceable metal sensing probe design allows for replacement when the probe wears or becomes contaminated without discarding the entire pen, and can be equipped with dedicated probes for different fiber types. Three LED indicators or an LCD digital screen offer different options in terms of cost and functionality. A power supply unit (such as a button battery) provides the necessary portable power for the entire system.

[0046] In one embodiment, such as Figure 3 As shown, the polyester staple fiber oil content detection system also includes: Cloud platform; The circuit module further includes: A digital temperature and humidity sensor is connected to the main control chip; A multi-mode wireless communication chip, connected to the main control chip, is used to collect the digital identity of the polyester staple fiber sample and the ID of the testing personnel. Together with the raw electrical signal data containing time-series information collected by the main control chip and the environmental parameters collected by the digital temperature and humidity sensor, it is used to construct a multi-source data packet and upload it to the cloud platform. Among them, such as Figure 4 As shown, the cloud platform is configured to perform the following operations: The multi-source data packets are received and parsed, and mapped to a pre-built digital twin model; the digital twin model integrates a dynamic task topology network based on production batches. When an abnormal oil content event is identified in the digital twin model, an intelligent traceability mechanism is activated to construct a digital trajectory of detection operations associated with the abnormal event; wherein, the digital trajectory of detection operations includes: electrical signal stability index, detection time series distribution map, spatial positioning topology, environmental parameter fluctuation spectrum, and historical detection consistency feature matrix; The digital trajectory of the detection operation is analyzed by a multi-source data fusion analysis engine, and the abnormal distribution pattern within the batch is analyzed in combination with the dynamic task topology network. The abnormal distribution pattern analysis within the batch includes: evaluating the correlation and dispersion of abnormal events between different detection nodes within the same production batch. Based on the analysis results of the digital trajectory of the detection operation and the abnormal distribution pattern within the batch, cross-validation is performed to identify non-standard use of the detection pen caused by the unintentional actions of the detection personnel. When the unintentional non-standard usage behavior is identified, the characteristic patterns of the behavior and its impact on batch quality data are located through correlation graph analysis. Based on the aforementioned feature patterns and their scope of influence, a digital correction scheme is generated, which includes personalized operation correction guidelines and intelligent re-inspection paths, and is pushed to the corresponding on-site management terminal in real time.

[0047] Furthermore, such as Figure 5 As shown, the process of constructing the digital trajectory of the detection operation includes: The original electrical signal data is extracted from the multi-source data packet. Based on the detection start time stamp, signal stabilization time stamp and detection end time stamp, the data subsequence of the stable detection stage is determined. The coefficient of variation of the data subsequence is calculated as an indicator of electrical signal stability. Calculate the duration before and after stabilization, compare the two currently measured duration values ​​with the corresponding historical averages, and classify the time pattern as too fast, normal, or too slow by calculating the deviation between the current value and the historical average and comparing it with the threshold range set based on the standard deviation of historical data, so as to generate a detection time series distribution map. Using the received sample unique identifier as a query condition, a targeted query is initiated to the graph database on which the digital twin model is based. The specific location information of the sample in the factory spatial layout is obtained through graph traversal algorithm, and one-hot encoding technology is used to convert it into a machine-readable feature vector to generate spatial positioning topology. The precise location information of the sample is extracted from the spatial positioning topology, and the raw environmental humidity data associated with the corresponding location during a specified time period before the start of the detection is obtained by querying the time series database. The variance of the time series is calculated as the environmental parameter fluctuation spectrum. Obtain historical detection event data within a preset number of times for the operator, and calculate the mean and variance of its electrical signal stability index and total detection time to form a historical detection consistency feature matrix; The electrical signal stability index, detection time-series distribution map, spatial positioning topology, environmental parameter fluctuation spectrum, and historical detection consistency feature matrix are fused into a high-dimensional feature vector to form the digital trajectory of the detection operation.

[0048] The polyester staple fiber oil content detection system described in this embodiment constructs a complete quality control process, from edge data acquisition, cloud-based twin mapping, multi-dimensional feature extraction to hybrid intelligent traceability and closed-loop correction. Its core utilizes a portable testing pen as a high-frequency data acquisition terminal, combined with the powerful graph correlation analysis and machine learning capabilities of the cloud platform, to achieve deep quality traceability from abnormal results to behavioral attribution.

[0049] The system's workflow begins with a highly integrated circuit module inside the detection pen. This module integrates a main control chip (MCU), a multi-mode wireless communication chip, and multiple sensor interfaces. Among them, the multi-mode wireless communication chip, as the core of data scheduling, connects to the main control chip through the communication interface and is responsible for coordinating all sensor data acquisition, packaging, and transmission tasks.

[0050] During system initialization, the multimode wireless communication chip first connects to the enterprise intranet via a factory-deployed wireless access point (AP), and then accesses a standard time server to calibrate its internal real-time clock (RTC) to microsecond-level accuracy using a time synchronization protocol. This ensures that all subsequently acquired sensor data has a unified, high-precision time reference.

[0051] When an inspector initiates a test, the main control chip immediately obtains the test start timestamp with microsecond precision from the RTC, and simultaneously begins polling the signal conversion circuit at a fixed sampling rate via the bus interface. This signal conversion circuit is responsible for converting the analog capacitance signal sensed by the detection head into a high-precision digital quantity in real time.

[0052] During data acquisition, a real-time signal analysis algorithm runs internally within the main control chip. This algorithm uses a rolling average calculation within a preset sample window and continuously monitors the standard deviation of this rolling average. When the standard deviation first falls below a preset stability threshold, the main control chip immediately records the timestamp of this moment as the signal stabilization point. When the operator releases the button, the main control chip records the detection end timestamp. Thus, the system obtains its first key technical parameter: raw electrical signal data containing timing information. This data is acquired by the main control chip collecting raw capacitance value arrays from the signal conversion circuit at a fixed frequency between the start and end of detection, supplemented by three high-precision timestamps: detection start, signal stabilization, and detection end.

[0053] While acquiring electrical signals, the system simultaneously obtains information from other dimensions. The main control chip polls a digital temperature and humidity sensor integrated into the pen's casing via a bus interface, capturing the precise ambient temperature and relative humidity readings when the signal stabilizes, as environmental parameters.

[0054] Simultaneously, the near-field communication (NFC) controller on the main control chip activates the circuit module. According to the NFC standard protocol, the operator must perform the following actions: First, bring the testing pen close to the NFC tag pre-attached to the polyester staple fiber sample packaging. The NFC controller reads its unique identifier, which serves as the digital identity of the polyester staple fiber sample. Next, the operator scans their personal work ID using the same NFC reader. The NFC controller reads and transmits the card's unique identifier, which serves as the testing personnel's ID.

[0055] After completing all data acquisition, the main control chip packages these five technical parameters and compresses them into a compact binary data packet using an efficient serialization scheme. This data packet is then transmitted to the multi-mode wireless communication chip, which, through its Wi-Fi module, uses a secure communication protocol to publish this multi-source data packet to the designated message endpoint of the cloud platform, completing the data acquisition and upload at the edge.

[0056] Upon arrival at the cloud platform, the second phase of processing is immediately triggered. The cloud platform employs an event-driven streaming architecture, receiving data packets through a secure message connector and forwarding them intact to a specific topic on the streaming data platform. This design leverages the platform's persistence and high-throughput capabilities to build a reliable data buffer and distribution bus.

[0057] The construction of the digital twin model is one of the core components of this system. This model is implemented based on graph database technology and predefines a complete ontology structure. Within the graph database, the system establishes various types of entity nodes, including Operator nodes representing inspection personnel, PenDevice nodes representing inspection equipment, ProductionBatch nodes representing production batches, Sample nodes representing the tested samples, and DetectionEvent nodes representing individual inspection events. These nodes are interconnected through carefully designed relationship types, forming a dynamic task topology network integrating production batches, accurately reflecting the actual conditions of the production site.

[0058] A dedicated microservice for model mapping, namely the graph construction service, is used as a consumer of the streaming data platform to subscribe to relevant topics. When a new data packet is received, the service first deserializes it and then calls the same computational logic as the built-in algorithm of the detection pen to derive the final oil content value from the raw electrical signal data. Based on whether this value exceeds the limit, a Boolean anomaly flag is set.

[0059] The mapping process is implemented through transactional graph query operations. The service performs queries in the graph database, using the MERGE command to find existing entity nodes in the graph based on the input sample ID, operator ID, and device ID. Then, it uses the CREATE command to generate a completely new DetectionEvent node, and finally creates a new relationship to connect this new node to the aforementioned entity node. This process of creating new nodes and weaving them into the existing topology is a precise implementation of mapping them to a pre-built digital twin model.

[0060] When the system creates a DetectionEvent node and sets its anomaly flag to true, i.e., when an oil content anomaly event is identified in the digital twin model, the intelligent tracing mechanism is immediately triggered. The map construction service generates an additional anomaly event notification message and pushes this message to another dedicated message topic. A separate digital trajectory generation service subscribes to this anomaly topic and is responsible for constructing the detection operation digital trajectory associated with the anomaly event. This trajectory is a high-dimensional feature vector, and its construction process involves the fine calculation and extraction of five key technical parameters: The method for obtaining the electrical signal stability index is to extract the raw electrical signal data array from the message and locate the data subsequence of the stable detection stage based on the time sequence information. The specific calculation process includes three steps: first, calculate the arithmetic mean of the subsequence to characterize the baseline level of the signal; second, calculate the standard deviation to reflect the dispersion of the data points relative to the mean; and finally, calculate the coefficient of variation (CV), which is calculated as follows: Coefficient of Variation (CV) = (Standard Deviation / Mean) × 100%. This relative metric eliminates the influence of the absolute value of the signal, making the stability between different detection events comparable. For example, when the detected average capacitance is 1500 fF and the standard deviation is 12 fF, the coefficient of variation is calculated as (12 / 1500) × 100% = 0.8%. A higher CV value (e.g., exceeding 1.0%) indicates that the signal still fluctuates significantly during the stable detection stage, which may be caused by operator hand tremors, uneven pressure between the probe and the fiber, or uneven density of the fiber sample itself.

[0061] The acquisition of the detection time-series distribution map specifically includes: the service accurately calculates two key time intervals: the pre-stabilization duration and the post-stabilization duration. The pre-stabilization duration is the time difference between the signal stabilization point and the detection start point, reflecting the time required from the start of detection to the signal reaching a stable state; the post-stabilization duration is the time difference between the detection end point and the signal stabilization point, characterizing the duration for which the operator maintains the detection state after the signal stabilizes. The service then queries the operator's historical detection database to obtain the historical average of the pre-stabilization duration and post-stabilization duration of all recent detection events as an individual benchmark. The system compares and analyzes the two currently measured duration values ​​with the corresponding historical averages. Specifically, it calculates the deviation between the current value and the historical average and compares this deviation with a threshold range set based on the standard deviation of historical data: when the deviation exceeds the positive threshold, it is classified as "too slow"; when it is below the negative threshold, it is classified as "too fast"; and when it is within the threshold range, it is classified as "normal". This classification method based on statistical process control can objectively identify abnormal time control by the operator in specific detections.

[0062] The acquisition of spatial positioning topology relationships specifically includes: the service uses the received sample's unique identifier as a query condition to initiate a targeted query to the graph database upon which the pre-built digital twin model relies. This query, through a graph traversal algorithm, follows predefined relationship paths such as BELONGS_TO to locate physical location nodes such as production lines and silos associated with the sample, thereby accurately obtaining the sample's specific location information in the factory's spatial layout (e.g., "Line 3, Silo A04"). After acquiring the location information, the system uses one-hot encoding technology to convert it into a machine-readable feature vector: first, a complete location dictionary containing all possible location categories is established; then, a zero vector with the same dimension as the dictionary is created; finally, the vector value corresponding to the current location in the dictionary is set to 1, while other locations remain 0. This encoding method preserves the category information of the location while meeting the input data format requirements of machine learning algorithms.

[0063] The acquisition of the environmental parameter fluctuation spectrum specifically includes the following steps: First, the service extracts the precise location information of the sample from the aforementioned spatial positioning topology, using this as a spatial filtering condition for environmental data queries. Next, the service queries a time-series database that stores historical environmental data, requesting raw environmental humidity data associated with the corresponding location within five minutes prior to the start of the detection, forming a complete time-series dataset. Based on this time series, the service calculates the environmental parameter fluctuation spectrum. The specific calculation process is as follows: First, the arithmetic mean of the time series is calculated; then, the sum of squared deviations of each data point from the mean is calculated; finally, the sum of squared deviations is divided by the number of data points to obtain the variance. This variance value quantitatively characterizes the severity of fluctuations in environmental humidity during the critical preparation stage before detection begins. A high variance value indicates unstable environmental conditions, which may interfere with the capacitance detection results.

[0064] The historical detection consistency feature matrix is ​​obtained by querying the service graph database to retrieve the 50 most recent detection event nodes for the operator ID. Two key parameters are extracted from these detection events: the electrical signal stability index (i.e., the previously calculated CV value) and the total detection duration (the difference between the detection end time and the detection start time). Based on these 50 sets of data, four statistics are calculated: the arithmetic mean of the CV values, the variance of the CV values, the arithmetic mean of the total detection duration, and the variance of the total detection duration. These four statistics together constitute a feature matrix characterizing the operator's baseline operating habits. The mean reflects the operator's typical operating level, while the variance reflects the consistency of their operations. This feature matrix quantitatively and comprehensively describes the operator's long-term operating characteristics, providing a reliable benchmark for identifying abnormal behavior.

[0065] Ultimately, this complete high-dimensional feature vector, which integrates micro-operations, space, environment, and individual history, along with the corresponding event ID and the batch ID to which the sample belongs, is packaged by the digital trajectory generation service and pushed to the next analysis queue message topic.

[0066] The multi-source data fusion analysis engine is the intelligent core of this system. It adopts a hybrid analysis architecture, innovatively combining deterministic rule analysis with probabilistic machine learning analysis. This engine consumes analysis queue message topics and first performs batch-based anomaly distribution pattern analysis based on the dynamic task topology network. This is a rule-based macro-analysis process used to determine whether the problem originates from the batch of materials themselves. This analysis includes the evaluation of two key parameters: correlation and dispersion. The specific method for obtaining and calculating the correlation assessment of abnormal events is as follows: The engine uses the batch ID to query the graph database and counts all detected event nodes marked as abnormal in that batch and their corresponding operator information. Then, the engine calculates the correlation index, with the specific formula: Correlation = Maximum number of abnormal operator events / Total number of abnormal events. Here, the denominator represents the total number of abnormal events in the batch, and the numerator represents the number of abnormal events caused by the same operator (i.e., the operator with the most abnormal events). For example, if a batch has 10 abnormal events, and operator A caused 8 of them, then the correlation index is 0.8. This index quantifies the concentration of abnormal events within the operator group; a higher correlation (closer to 1.0) indicates that the problem is more likely to originate from individual factors of a specific operator.

[0067] The specific method for obtaining and calculating the dispersion assessment of abnormal events is as follows: The engine queries the graph database to obtain the total number of operator nodes that tested the batch and the number of operator nodes that generated at least one abnormality. The specific formula for calculating the dispersion index is: Dispersion = Number of operators with abnormalities / Total number of operators. For example, if the total number of operators in the test batch is 10, and 8 of them reported at least one abnormality, then the dispersion is 0.8. This index reflects the widespread distribution of abnormal events among the operator group. High dispersion indicates that the problem is prevalent among different operators, strongly pointing to batch material or systemic environmental problems; low dispersion points to individual operator problems.

[0068] The analytics engine integrates a business rules engine that loads pre-written business rules. After the engine completes the calculation of correlation and dispersion, it passes these two metrics to the rules engine's working memory. The rules engine then performs logical judgments based on a predefined rule base. For example, a typical rule might be set as: if the dispersion is greater than or equal to 0.6, it is determined to be a material problem. This means that when the dispersion metric reaches or exceeds 0.6, this rule is triggered, and the system marks the root cause of the anomaly as a fiber quality problem. Once such a deterministic rule is triggered, the analytics engine immediately adopts its conclusion and outputs the root cause result, while terminating subsequent, more time-consuming machine learning analysis processes to ensure the system's efficiency when handling explicit cases.

[0069] If the results of the macroscopic analysis are ambiguous and no rules are triggered, the system will proceed to the next stage, which is to perform cross-validation based on the analysis results of the digital trajectory of the detection operation and the abnormal distribution pattern within the batch, in order to identify non-standard use of the detection pen caused by the detection personnel's unconscious actions.

[0070] This cross-validation essentially involves launching a probabilistic machine learning model to perform deep analysis of the previously constructed digital trajectories of the detection operations. The construction process of this machine learning model includes data preparation and annotation phases, where the system exports past anomaly events and their corresponding digital trajectory feature vectors from a historical database. A quality control expert team then manually reviews and annotates these events, labeling each anomaly trajectory as a specific category.

[0071] Since non-standard usage behaviors typically constitute only a minority of all anomalies, this leads to a severe class imbalance problem. To address this issue, the system employs an oversampling technique during the model training phase. This algorithm is applied only to the training set, generating new synthetic samples by interpolating minority class samples in the feature space.

[0072] During the model training phase, the system constructs a complete machine learning training pipeline. First, feature standardization is performed using StandardScaler to standardize continuous features with Z-scores, ensuring comparability of features with different scales. Then, an ensemble learning classifier is built, selecting the RandomForest algorithm, setting the number of decision trees to 200, the maximum tree depth to 25 layers, the minimum number of samples required for internal node splitting to 10, and the minimum number of samples for leaf nodes to 5. The pipeline is fitted using a training set balanced by SMOTE, employing a 5-fold cross-validation method. The training set is randomly divided into 5 mutually exclusive subsets; 4 subsets are used for training each time, with the remaining subset used for validation, repeated 5 times to ensure each subset is used as the validation set. The optimization process uses the F1 score as the primary evaluation metric, while simultaneously monitoring the balance between precision and recall, finding the optimal hyperparameter combination through grid search. An early stopping strategy is employed during training: training automatically terminates when the validation set loss function no longer decreases for 10 consecutive epochs to prevent overfitting.

[0073] During the model validation and deployment phase, after training is complete, a reserved test set is used to evaluate the model's real-world performance. Once the model performance meets the standards, this complete pipeline will be serialized into a model file, and a standalone machine learning model service will be built.

[0074] When cross-validation is required, the analytics engine calls the API endpoint of the machine learning model service, sending the feature vector of the currently analyzed digital trajectory. The model service returns a probability value. When this probability value exceeds a preset confidence threshold, the system formally identifies the unintentional non-standard usage behavior.

[0075] When the fusion analytics engine determines that the cause is non-standard usage behavior, it pushes a result message containing the event ID, operator ID, root cause, and key defect characteristics to the last message topic. A digital correction service subscribes to this topic and immediately initiates subsequent processing steps when non-standard usage behavior is identified.

[0076] First, the service execution locates the characteristic patterns of the behavior and their impact on batch quality data through correlation graph analysis. Specifically, based on the key defect features identified in the message, the service performs correlation graph analysis queries in the graph database to find other detection events with similar defect characteristic patterns.

[0077] Secondly, based on this analysis, the service generates a digital correction plan that includes personalized operation correction guidelines and intelligent re-inspection paths. Specifically, the service maintains an internal mapping table between defect features and correction guidelines. When a specific defect feature is detected, it generates a corresponding personalized operation correction guide. Simultaneously, the service uses the queried impact range as the intelligent re-inspection path.

[0078] Finally, this digital correction solution, which includes the correction guidelines and a list of re-inspection paths, is pushed in real-time to the corresponding on-site management terminal via a secure API interface by the digital correction service. Upon receiving the push, the application immediately pops up an alarm, enabling managers to intervene promptly, provide precise one-on-one guidance, and arrange for personnel to re-inspect the samples along the re-inspection path.

[0079] This complete closed loop from data collection to behavior correction continuously optimizes the entire quality control system through continuous data accumulation and analysis, forming a virtuous cycle of continuous improvement.

[0080] The main technical advantages of this system are reflected in the following aspects: At the data acquisition level, multi-sensor fusion and high-precision time-series synchronization ensure the integrity, reliability, and traceability of raw data. At the data processing level, streaming architecture and graph database technology are used to achieve real-time analysis and complex correlation mining of large-scale heterogeneous data. At the intelligent analysis level, rule-based reasoning and machine learning are combined to ensure the efficiency of analyzing routine problems while providing probabilistic analysis capabilities for in-depth insights. At the application level, precise behavior recognition and personalized corrective guidance enable the root-cause resolution of quality problems.

[0081] Of particular note is the system's ability to identify unintentional, non-standard usage behaviors, a capability that traditional quality control methods struggle to achieve. Through multi-dimensional feature analysis of digital trajectories, the system can identify seemingly normal but actually problematic operational behaviors. These subtle issues are often overlooked in traditional methods, yet they are crucial factors affecting the consistency of detection.

[0082] Furthermore, the system's learning and evolutionary capabilities are also worth emphasizing. With accumulated usage time, the system continuously collects new detection data and behavioral samples. This new data can be used to periodically retrain the machine learning model, optimize the feature extraction algorithm, and adjust the judgment threshold. This self-improvement mechanism enables the system to adapt to changes in the production environment, continuously improving its recognition accuracy and practicality.

[0083] In summary, this system, through precise multi-source data acquisition, deep digital twin mapping, intelligent behavior recognition and analysis, and accurate closed-loop correction, constructs a comprehensive, intelligent, and adaptive quality control system, providing strong quality assurance capabilities for polyester staple fiber production.

[0084] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for detecting the oil content of polyester staple fiber, characterized in that, include: Insert the testing head of the testing pen into the polyester staple fiber sample; The detection head acquires electrical signals that change due to variations in fiber oil content. The electrical signal is compared with preset oil content benchmark data; Output the oil content information obtained based on the comparison results; The digital identification of the polyester staple fiber sample and the ID of the testing personnel are collected together with the raw electrical signal data containing time sequence information collected by the main control chip and the environmental parameters collected by the digital temperature and humidity sensor, and then constructed into a multi-source data package and uploaded to the cloud platform. The cloud platform is configured to perform the following operations: receive and parse the multi-source data packets, and map them to a pre-built digital twin model; the digital twin model integrates a dynamic task topology network based on production batches; when an oil content anomaly event is identified in the digital twin model, an intelligent traceability mechanism is activated to construct a detection operation digital trajectory associated with the anomaly event; the detection operation digital trajectory is parsed through a multi-source data fusion analysis engine, and the batch-specific anomaly distribution pattern is analyzed in conjunction with the dynamic task topology network; The batch-level anomaly distribution pattern analysis includes: assessing the correlation and dispersion of abnormal events among different detection nodes within the same production batch; cross-validating the analysis results of the digital trajectory of the detection operation with the batch-level anomaly distribution pattern to identify non-standard use of the detection pen caused by unintentional actions of the inspection personnel; when unintentional non-standard use is identified, locating the characteristic pattern of the behavior and its impact range in the batch quality data through correlation graph analysis; and generating a digital correction scheme containing personalized operation correction guidelines and intelligent re-inspection paths based on the characteristic pattern and its impact range, and pushing it to the corresponding on-site management terminal in real time. The process of constructing the digital trajectory of the detection operation includes: extracting raw electrical signal data from the multi-source data packets; determining a data subsequence for the stable detection stage based on the detection start time stamp, signal stabilization time stamp, and detection end time stamp; calculating the coefficient of variation of this data subsequence as an indicator of electrical signal stability; calculating the duration before and after stabilization; comparing the two currently measured duration values ​​with the corresponding historical averages; classifying the time pattern as too fast, normal, or too slow by calculating the deviation between the current value and the historical average and comparing it with a threshold range set based on the standard deviation of historical data, to generate a detection time series distribution map; using the received sample unique identifier as a query condition, initiating a targeted query to the graph database on which the digital twin model relies; obtaining the specific location information of the sample in the factory spatial layout through a graph traversal algorithm; and using a unique thermal... Encoding techniques are used to transform the data into machine-readable feature vectors to generate spatial positioning topology. Precise sample location information is extracted from these topology. A time-series database is queried to obtain raw environmental humidity data associated with the corresponding location within a specified time period before the start of detection. This raw environmental humidity data is then arranged in chronological order to form a time-series dataset, and the variance of this dataset is calculated as an environmental parameter fluctuation spectrum. Historical detection event data within a preset number of detection attempts are obtained, and the mean and variance of the electrical signal stability index and total detection duration are calculated to form a historical detection consistency feature matrix. The electrical signal stability index, detection time-series distribution map, spatial positioning topology, environmental parameter fluctuation spectrum, and historical detection consistency feature matrix are fused into a high-dimensional feature vector to constitute the digital trajectory of the detection operation.

2. The method for detecting the oil content of polyester staple fiber as described in claim 1, characterized in that, The detection head has a surface-treated contact portion, which makes the contact portion oleophobic or oleophilic; the electrical signal is the capacitance or resistance value of the detection head.

3. The method for detecting the oil content of polyester staple fiber as described in claim 2, characterized in that, The step of comparing the electrical signal with preset oil content benchmark data is executed by the main control chip inside the detection pen. The main control chip has pre-stored mapping relationship data between electrical signals and oil content.

4. The method for detecting the oil content of polyester staple fiber as described in claim 1, characterized in that, The steps for outputting oil content information include: displaying the oil content level in different colors using multiple LED indicators, or displaying the percentage value of oil content on an LCD digital screen.

5. A system for detecting the oil content of polyester staple fiber, characterized in that, include: A detection pen, the detection pen being configured to perform the method as described in any one of claims 1 to 4.

6. The polyester staple fiber oil content detection system as described in claim 5, characterized in that, The detection pen includes: Pen body; The detection head is located at the front end of the pen body; A circuit module housed inside the pen body, the circuit module being electrically connected to the detection head; The display unit is located on the outside of the pen body and is electrically connected to the circuit module.

7. The polyester staple fiber oil content detection system as described in claim 6, characterized in that, The circuit module includes: A signal conversion circuit is used to receive and process electrical signals from the detection head; The main control chip is connected to the signal conversion circuit and is used to query the mapping relationship data and determine the oil content based on the processed signal.

8. The polyester staple fiber oil content detection system as described in claim 6, characterized in that, The display unit is three LED indicator lights or an LCD digital screen; the detection head is a replaceable metal sensing probe; the system also includes a power supply unit for powering the circuit module and the display unit.

Images