A plastic part production line real-time monitoring method and system

Abstract

The present application relates to injection molding process technical field, especially to a kind of plastic part production line real-time monitoring method and system.Method includes: obtaining the benchmark batch historical data of plastic part production line and standardization, construct the benchmark process model containing benchmark centroid and benchmark shape matrix;Real-time acquisition current cycle process parameter vector and standardization, store in time sequence sliding window;Current centroid and current shape matrix in time sequence sliding window are calculated;Current window shape expansion factor and current window position drift factor are obtained, and respectively compared with shape expansion threshold value and position drift threshold value, complete plastic part production line real-time monitoring.The present application constructs benchmark process model, combines sliding window and shape, position factor, judges system state by double threshold value, solves the problem of traditional algorithm time sequence perception deficiency, improves the accuracy and practicality of plastic part production line real-time monitoring, supports early abnormal identification and accurate diagnosis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of injection molding process, and particularly relates to a plastic part production line real-time monitoring method and system. BACKGROUND

[0002] The injection molding of plastic parts is one of the core processes in modern manufacturing industry. The quality of plastic part products, such as electrical appliance housings and precision components, highly depends on the precise coordination of multiple process parameters such as injection pressure, mold temperature, holding time, and cooling time. In the production process, ensuring that the multiple parameters always maintain within the optimal process window is the key to guaranteeing the consistency and qualification rate of mass production.

[0003] Currently, many production lines still rely on traditional monitoring methods, that is, setting fixed upper and lower thresholds for a single process parameter, such as the mold temperature should not exceed a certain set value, to monitor. This monitoring method can only detect the extreme out-of-limit of a single parameter, but cannot identify coordination abnormalities. Coordination abnormalities refer to a situation where although each parameter is within its allowed tolerance range individually, the combination of these parameters deviates from the optimal process setting, which can also lead to batch quality defects such as shrinkage and flash in products.

[0004] In order to solve the problem of detecting coordination abnormalities, the existing technology usually uses unsupervised anomaly detection algorithms based on density, such as the Local Outlier Factor (LOF) algorithm. The LOF algorithm compares the local density of a data point, which is the combination of multiple parameters in a production cycle, with the local density of its neighbor data points, to determine whether it deviates from the normal data cluster.

[0005] However, the traditional LOF algorithm lacks the ability to perceive the temporal context when applied to the continuous and real-time time series data stream of injection molding production lines. It only concerns the spatial neighborhood and ignores the temporal neighborhood, leading to the failure of detecting process drift type abnormalities. For example, when a device slowly drifts away from the baseline window due to slow wear, each new data point is still locally dense compared to its neighbors that are also drifting, causing the traditional LOF algorithm to always misjudge it as normal. Moreover, the traditional LOF algorithm cannot distinguish between different types of abnormalities. Whether it is a transient spike anomaly caused by sudden equipment failure or a gradual drift anomaly caused by slow equipment wear, the traditional algorithm can only classify it as an abnormal point, and cannot provide differentiated information for subsequent fault diagnosis and maintenance decision-making, reducing the practicality of the real-time monitoring system of the plastic part production line. SUMMARY

[0006] In order to solve the technical problems in the prior art that real-time monitoring cannot effectively detect process drift abnormalities due to lack of temporal context, and cannot distinguish between different types of abnormalities, the present application provides a plastic part production line real-time monitoring method and system.

[0007] In a first aspect, the present application provides a plastic part production line real-time monitoring method, which adopts the following technical solution:

[0008] A plastic part production line real-time monitoring method, comprising: obtaining historical data of a plurality of production cycles containing a reference batch as a plastic part production line and performing standardization processing, constructing a reference process model based on the standardized historical data, the reference process model containing a reference centroid and a reference shape matrix; real-time obtaining process parameter vectors of the plastic part production line in the current production cycle and performing standardization processing, storing the standardized process parameter vectors into a preset time sequence sliding window; calculating the current centroid and the current shape matrix of all process parameter vectors in the time sequence sliding window; determining a current window shape inflation factor for characterizing the data dispersion degree within the short-term time sequence sliding window according to the inverse matrix of the current shape matrix and the reference shape matrix; determining a current window position drift factor for characterizing the degree of centroid deviation of the short-term time sequence sliding window from the reference centroid according to the current centroid, the reference centroid and the inverse matrix of the reference shape matrix; setting a shape inflation threshold and a position drift threshold, based on the comparison results of the current window shape inflation factor and the shape inflation threshold and the current window position drift factor and the position drift threshold, diagnosing the state of the plastic part production line as one of normal, transient spike, gradual drift or spike in drift, and completing the real-time monitoring of the plastic part production line.

[0009] The beneficial effects are that by constructing a reference process model containing a reference centroid and a reference shape matrix, the modeling of the normal working condition of the plastic part production line is realized, which can accurately reflect the statistical characteristics and covariance structure of the normal production process, providing a reliable reference for subsequent anomaly detection; by the time sequence sliding window mechanism, the dynamic management of real-time process parameter vectors is realized, which can capture the data feature changes within the short-term time window, providing a data basis for time sequence context awareness; by calculating the current window shape inflation factor and the position drift factor, the quantitative characterization of different types of anomalies is realized, the shape inflation factor reflects the change of data dispersion degree, and the position drift factor reflects the degree of overall deviation; by setting the dual judgment mechanism of the shape inflation threshold and the position drift threshold, the accurate classification of the four states of normal, transient spike, gradual drift and spike in drift is realized, providing differentiated information support for subsequent fault diagnosis and maintenance decision; the present application effectively solves the problem of lack of time sequence awareness of the traditional LOF algorithm, improves the accuracy and practicality of the plastic part production line real-time monitoring system, and provides reliable technical support for early identification and accurate diagnosis of process anomalies.

[0010] Further, the construction of the reference process model comprises: taking the mean of all process parameter vectors of the standardized historical data as the reference centroid; and taking the covariance matrix of all process parameter vectors of the standardized historical data as the reference shape matrix.

[0011] The beneficial effects are that: by taking the mean of the standardized historical data as the reference centroid, the central tendency under normal working conditions of the plastic part production line is accurately reflected; by taking the covariance matrix as the reference shape matrix, the correlation between the process parameters and the discrete characteristics of the data are effectively captured; the construction of the reference centroid ensures accurate representation of the central position under normal working conditions, providing a reliable reference point for the calculation of the position drift factor; and the construction of the reference shape matrix retains the covariance information between the parameters, providing an accurate shape reference for the calculation of the shape expansion factor.

[0012] Further, the standardization processing adopts Z-score standardization.

[0013] Further, the current window shape expansion factor satisfies:

[0014] ; in the formula, is the current production cycle is the corresponding current window shape expansion factor, is the current production cycle is the corresponding current shape matrix, is the inverse matrix of the reference shape matrix, is the trace operation of the matrix.

[0015] The beneficial effects are that: by calculating the trace of the product of the current shape matrix and the inverse matrix of the reference shape matrix, an evaluation index of the shape expansion factor is constructed, which can effectively measure the change in the dispersion degree of the current window data relative to the reference model; the matrix trace operation ensures that the shape expansion factor is a scalar value, facilitating subsequent threshold comparison and state diagnosis; when the current shape matrix is similar to the reference shape matrix, the product is close to the unit matrix, the trace value is close to the parameter dimension, and the shape expansion factor is 1; when the dispersion degree of the current shape matrix is greater than the reference, the trace value increases, and the shape expansion factor is greater than 1; the shape expansion factor can effectively identify the increase in the dispersion degree of the data caused by equipment wear or process parameter drift, providing an important basis for the detection of gradual drift anomalies and improving the accuracy and sensitivity of anomaly detection.

[0016] Further, the current window position drift factor satisfies:

[0017] ; in the formula, is the current production cycle is the corresponding current window position drift factor, for the current production cycle a corresponding current centroid, for the reference centroid, for the inverse of the reference shape matrix, for a vector transposition operation.

[0018] The beneficial effects are that: the position drift factor is constructed by the Mahalanobis distance formula, which realizes scientific evaluation of the degree of deviation of the current window centroid from the reference centroid, fully considers the covariance structure of the reference distribution, and ensures the accuracy and rationality of distance calculation; the introduction of the inverse of the reference shape matrix makes the distance calculation consider the correlation between process parameters and the variability of different parameters, and when the deviation direction is consistent with the high variability direction, the distance is relatively small, and when the deviation direction is consistent with the low variability direction, the distance is relatively large; the position drift factor can effectively identify the overall position change caused by equipment drift or systematic deviation, providing an important basis for distinguishing between instantaneous spikes and gradual drift anomalies, and improving the accuracy and reliability of anomaly type identification.

[0019] Further, the condition for diagnosing as the instantaneous spike includes that the current window shape inflation factor is greater than a shape inflation threshold, and the current window position drift factor is less than or equal to a position drift threshold.

[0020] Further, the condition for diagnosing as the gradual drift includes that the current window shape inflation factor is not greater than a shape inflation threshold, and the current window position drift factor is greater than a position drift threshold.

[0021] Further, the condition for diagnosing as the drift-in-spike includes that the current window shape inflation factor is greater than a shape inflation threshold, and the current window position drift factor is greater than a position drift threshold.

[0022] Further, it further includes adaptively updating the reference process model, including: after a gradual drift anomaly lasts for a set number of cycles, triggering manual confirmation, if it is confirmed to be a benign drift, extracting process parameter vectors of the last multiple cycles as new historical data, and reconstructing the reference process model.

[0023] The beneficial effects are that: by setting the number of cycles to trigger the manual confirmation mechanism, intelligent identification and processing of continuous gradual drift are realized, which can ensure system stability while responding to possible process improvement or equipment adjustment in a timely manner; the manual confirmation link ensures the accuracy and reliability of the reference model update, avoiding false updates caused by real faults, and only when it is confirmed to be a benign drift, the reference model is reconstructed, ensuring the accuracy of the monitoring system; by extracting process parameter vectors of the last multiple cycles as new historical data, adaptive updating of the reference model is realized, which can adapt to the normal evolution of the process and maintain the timeliness and accuracy of the reference model.

[0024] In a second aspect, the present application provides a plastic part production line real-time monitoring system, which adopts the following technical scheme:

[0025] A plastic part production line real-time monitoring system comprises a processor and a memory, and the memory stores computer program instructions, which, when executed by the processor, implement the above-mentioned plastic part production line real-time monitoring method.

[0026] By adopting the above technical scheme, the above-mentioned plastic part production line real-time monitoring method is generated into a computer program and stored in the memory to be loaded and executed by the processor, so that a terminal device is made according to the memory and the processor, and use is facilitated.

[0027] The present application has the following technical effects:

[0028] (1) Break through the limitation of traditional LOF algorithm which only focuses on spatial neighborhood and ignores time sequence context, realize effective identification of slow drift anomaly through time sequence sliding window and position drift factor, build a benchmark process model based on historical stable data, and real-time calculate the deviation degree of the centroid of the current window and the benchmark centroid. When the process parameters are cooperatively and slowly deviated from the benchmark due to equipment wear, the position drift factor will continuously accumulate and rise. Even if a single data point and the adjacent drift data still maintain local density, it can also be accurately determined as a gradual drift through the time sequence characteristics of the continuous deviation of the centroid, solving the missed detection problem of process drift type anomaly of traditional algorithm.

[0029] (2) Abandon the mode of traditional algorithm which only judges anomaly and cannot classify, realize accurate diagnosis of multiple abnormal states through the combination of shape inflation factor and position drift factor in two dimensions, and directly provide differentiated guidance for operation and maintenance through classification diagnosis, such as emergency troubleshooting of instantaneous sharp peak for sudden failure, and gradual drift for planned shutdown maintenance, solving the problem of single abnormal information of traditional method which cannot support accurate decision-making, and improving the practicality of the plastic part real-time monitoring system.

[0030] (3) Through historical benchmark model construction and data standardization processing, the monitoring benchmark is fitted to the real stable working condition of the production line, avoiding misjudgment of normal process fluctuations, such as parameter fluctuations caused by small differences in raw material batches, as anomalies; at the same time, comparison of the current shape matrix and the benchmark shape matrix can more accurately evaluate the data dispersion degree, filter false fluctuations caused by measurement noise, and ensure the reliability of anomaly judgment, solving the problem of traditional algorithm which is easily disturbed and has high false positive rate.

[0031] (4) The time sequence sliding window updates the process parameter vector of the current production cycle in real time, and through real-time calculation of the current centroid and the current shape matrix and comparison with the benchmark, the rapid response of the abnormal is realized. Whether it is a transient peak of sudden damage of the equipment or a gradual drift of slow wear of the equipment, it can be captured and diagnosed in time, avoiding the expansion of the fault due to monitoring lag, such as product batch unqualified caused by drift accumulation, serious damage of the equipment, etc., to meet the continuous and high real-time monitoring requirements of the plastic part production line. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 is a method flow chart of a plastic part production line real-time monitoring method according to an embodiment of the present application.

[0033] Figure 2 is a time sequence diagram of the original process parameters in a plastic part production line real-time monitoring method according to an embodiment of the present application.

[0034] Figure 3 is a time sequence diagram of double-index diagnosis in a plastic part production line real-time monitoring method according to an embodiment of the present application. DETAILED DESCRIPTION

[0035] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0036] The embodiments of the present application disclose a plastic part production line real-time monitoring method, referring to Figure 1 , comprising steps S001-S004:

[0037] S001: Offline construction of benchmark process model.

[0038] In the initial stage of the deployment of the present embodiment, a mathematical model of the benchmark process window needs to be established first.

[0039] (1) Data acquisition and pretreatment: historical production data are called, and at least one production cycle data of a confirmed benchmark batch is selected by the production line manager or the product inspection personnel. The production cycle data of the benchmark batch represents the process parameter combination when the production quality is optimal. It is assumed that each production cycle monitors process parameters, for example, , the parameters are: injection pressure, mold temperature and cooling time, and the data of one production cycle is a dimensional vector , such as a 3-dimensional vector injection pressure Mold temperature Cooling time ; all the process parameter vectors of the reference batches constitute a historical reference set , due to the large difference in the dimension and numerical range of different process parameters, such as pressure unit MPa, temperature ℃, time s, normalization processing is required; this embodiment adopts Z-score standardization, and calculates the mean value and standard deviation of each dimension in each dimension ; the mean vector and the standard deviation vector are calculated; all subsequent data are standardized using the above statistical vectors.

[0040] (2) Construct a reference process model: calculate and store the core statistical characteristics of the standardized , and constitute a reference process model (Base Craftsmanship Model, BCM) , , which contains two core components: reference centroid and reference shape matrix , is the dimensional mean vector of all data points in , according to the definition of Z-score standardization, the mean value of the standardized data is a zero vector, that is , wherein is the vector transposition operation, represents the center position of the reference process window, that is, the reference centroid; is the covariance matrix of , which describes the shape and size of the reference process window, that is, the reference shape matrix, the diagonal elements are the variances of each parameter itself, which are 1 after standardization, and the non-diagonal elements are the covariances between two parameters, such as the cooperative relationship between pressure and temperature; offline store and the normalized parameter .

[0041] (3) Initialize the time sequence sliding window: initialize a first-in first-out short-term time sequence sliding window , the capacity is preset , for example, periods.

[0042] By offline construction of the reference process model, a gold standard is provided for subsequent real-time monitoring of the plastic part production line, and a reference center and normal fluctuation form for evaluating the process state are established.

[0043] S002: Real-time computation of time-series sliding window state features.

[0044] In the real-time monitoring stage of the plastic part production line, for each newly generated production cycle, the following operations are performed:

[0045] (1) Data update: Collect the original data vector of the production cycle , for example, is collected , and the same Z-score standardization is performed on and to obtain , the calculation process is as follows: ; then, is stored in , if is full of 100 cycles, the earliest data is removed, i.e. .

[0046] (2) Calculate the window internal statistics: After the window is filled with at least data points, the current centroid and the current shape matrix of the time-series sliding window are calculated in real time, is the mean vector of data points in , reflecting the average process state of the last cycles; is the covariance matrix of data points in , reflecting the process fluctuation shape of the last cycles.

[0047] (3) Calculate the shape expansion factor : used to evaluate the looseness of the current time-series sliding window relative to the reference process window , a stable process window should have the same internal data shape as the reference shape matrix , the intrusion of a transient spike will cause it to drastically expand, i.e. the variance increases dramatically, the calculation process is as follows:

[0048] ;

[0049] where is the current production cycle​ The corresponding current window shape inflation factor, For the current production cycle The corresponding current form matrix, It is the inverse of the base form matrix. The trace operation is the sum of the elements on the main diagonal. The above formula calculates the generalized variance ratio of the current time-series sliding window shape relative to the baseline shape.

[0050] Calculation example:

[0051] When the production line is stable, see Figure 2 0-200 cycles and Very close, assuming close identity matrix , ,therefore, , such as in In the example See Figure 3 The red curve remained at a low level in the early stage;

[0052] When a transient spike occurs, see Figure 2 The area marked by the red dashed line represents an extreme anomaly, such as... Breaking into the window ,lead to The variance increases dramatically, such as , It will rise instantly, much greater than See Figure 3 Red curve It surges instantly during peak hours;

[0053] When gradual drift occurs, see Figure 2 After 200 cycles, all points within the window move closely together, maintaining a compact internal shape. The changes are minor, therefore, Staying low, still Nearby, see Figure 3 The red curve remained at a low level during the drift phase after 200 cycles.

[0054] (4) Calculate the position drift factor : The metric is used to evaluate the current time series sliding window. center of mass Offset from the reference process window centroid The degree of a gradual drift process, the shape inside the window. Maintain stability, but center of gravity The displacement will continue, the calculation process is as follows:

[0055] ;

[0056] In the formula, is the current production cycle corresponding to the current window position drift factor, is the current production cycle corresponding to the current centroid, is the reference centroid, is the inverse matrix of the reference shape matrix, is the vector transpose operation, is the difference vector between the current time series sliding window centroid and the reference centroid, the calculation result is the square of the Mahalanobis Distance from to Mahalanobis Distance is a normalized distance that takes into account the data shape, i.e. the covariance .

[0057] Calculation example of

[0058] When the production line is stable, see Figure 2 period 0-200, is very close to , i.e. , therefore, will be close to 0, see Figure 3 the orange curve in the early stage remains low;

[0059] When a transient spike occurs, see Figure 2 marked by the red dotted line, although the abnormal point is extreme, is the average of data points, therefore, the abnormal point is smoothed, the change of the centroid is very small, remains low, see Figure 3 the orange curve only has a slight fluctuation at the spike moment;

[0060] When a gradual drift occurs, see Figure 2 after period 200, the data points in the window move as a whole, like continuously moves to , which is obviously far from , like , will continue to rise, see Figure 3The middle orange curve keeps steadily rising after 200 periods.

[0061] By calculating the morphology indicator and the position indicator of the current window in real time , two decoupled feature quantities that can distinguish different types of anomalies are obtained, providing a basis for the subsequent diagnosis step.

[0062] S003: Abnormal diagnosis of the plastic part production line based on double-index decoupling.

[0063] This step realizes the abnormal diagnosis of the plastic part production line by using the two indicators obtained by real-time calculation and .

[0064] (1) Threshold setting: Set the morphology inflation threshold , which can be set based on the data dimension , for example, , i.e. allowing 50% of the morphology inflation; Set the position drift threshold , which can be set based on the statistical characteristics of , for example, the 99.9% quantile of the squared Mahalanobis distance from to , to ensure statistical robustness, such as ; In , Figure 3 corresponds to the red dashed line, i.e. the spike threshold, corresponds to the orange dashed line, i.e. the drift threshold.

[0065] (2) Abnormal diagnosis logic: At each time point, i.e. the current production period , the system compares and with their respective threshold values according to the following logic, see Figure 3 :

[0066] The condition for diagnosing as a transient spike is and , such as and ; The diagnosis logic is: the average position of the window is normal, i.e. is normal, but the internal morphology is suddenly loose, i.e. is too high, which corresponds to a sudden abnormal point breaking into the window, see Figure 3 when the red curve transiently crosses the threshold, while the orange curve is still below the threshold;

[0067] The condition for diagnosing as a gradual drift is and​​ , and ; the diagnostic logic is that the inner shape of the window is compact, i.e. normal, but the average position has deviated from the historical baseline significantly, i.e. high, which corresponds to a slow and concerted movement of the entire process, i.e. a gradual drift, see the stage where the orange curve crosses the threshold in Figure 3 , while the red curve is still below the threshold;

[0068] The condition for diagnosing a spike in drift is and , as in and ; the diagnostic logic is that the average position of the window has deviated, while its inner shape has also become loose, which indicates that the production line has been hit by a transient spike while it was already in a state of drift.

[0069] The condition for diagnosing normality is and , as in and ; the diagnostic logic is that both the shape and the position of the window are within the range allowed by the historical baseline.

[0070] By combining the two decoupled indicators of shape and position, we achieve accurate classification and diagnosis of different physical causes of anomalies, such as transient spikes and gradual drifts, solving the problem that traditional algorithms cannot distinguish between different types of anomalies.

[0071] S004: Perform post-processing and adaptive update.

[0072] (1) Alarm distribution: the system pushes different types of alarms diagnosed to the production line management system or the on-site alarm system, guiding relevant personnel to carry out differentiated treatment: for example, a transient spike may prompt the inspection of sensor or actuator for occasional failure; while a gradual drift may prompt that the equipment may have slow wear and tear, requiring planned maintenance.

[0073] (2) Adaptive baseline model: this step is to solve the problem of false positives caused by static baseline when the process changes, such as changing raw materials or molds, etc.

[0074] Trigger: when the gradual drift alarm lasts for periods, for example, , the system automatically suspends the alarm and sends a process window drift confirmation request to the production line manager or quality inspector.

[0075] Confirmation: the operator checks the current batch, such as the quality of the products in the last periods.

[0076] Decision: malicious drift, operator confirms unqualified, such as shrinkage, flash has appeared, system locks the current non-updated, and continues to alarm, indicating that the production line must be shut down for maintenance; benign drift, operator confirms qualified, which indicates that the process window after the current drift is a new, acceptable baseline process window, such as replacement of a new batch of raw materials or a new mold causing overall translation of process parameters.

[0077] Update: the system triggers a baseline model update, automatically extracts the last period, for example, qualified data as new , completely re-executes all steps in S001 to generate a new .

[0078] Recovery: the system continues to monitor using the new , since the current window is close to the new baseline , recovered to below, the alarm is automatically lifted, and the system completes the self-adaptation to the new process.

[0079] Through the closed loop of alarm distribution and adaptive update, differentiated guidance is provided for the operation and maintenance of the plastic part production line, and at the same time, the real-time monitoring system can actively learn and adapt to benign process changes, ensuring the long-term usability and accuracy of the monitoring system in a dynamic production environment.

[0080] The embodiment of the application also discloses a plastic part production line real-time monitoring system, comprising a processor and a memory, and the memory stores computer program instructions, which realize the plastic part production line real-time monitoring method according to the application when executed by the processor.

[0081] The above system also comprises other components such as a communication bus and a communication interface, which are well known to those skilled in the art, and their settings and functions are known in the art, so they will not be described here.

[0082] The above are preferred embodiments of the application, and are not intended to limit the protection scope of the application, therefore: any equivalent changes made in structure, shape, principle, etc. according to the application should be covered within the protection scope of the application.

Claims

1. A method of real-time monitoring of a plastic part production line, characterized in that, The method comprises the following steps: acquiring historical data of a plurality of production cycles comprising a benchmark batch of a plastic part production line and performing standardization processing, constructing a benchmark process model based on the standardized historical data, the benchmark process model comprising a benchmark centroid and a benchmark shape matrix; acquiring a process parameter vector of the plastic part production line in a current production cycle in real time and performing standardization processing, and storing the standardized process parameter vector in a preset time sequence sliding window; calculating a current centroid and a current shape matrix of all process parameter vectors in the time sequence sliding window; determining a current window shape inflation factor for representing the dispersion degree of the data inside the short-term time sequence sliding window according to the inverse matrix of the current shape matrix and the benchmark shape matrix, satisfying: ； for the current production cycle a corresponding current window shape inflation factor, for the current production cycle a corresponding current shape matrix, an inverse matrix of the reference shape matrix, a trace operation of the matrix; a current window position drift factor is determined for characterizing a degree of deviation of the short-term time-series sliding window centroid from the reference centroid, based on the current centroid, the reference centroid, and the inverse matrix of the reference shape matrix, satisfying: ； for the current production cycle a corresponding current window position drift factor, for the current production cycle a corresponding current centroid, for the reference centroid, is a vector transposition operation; setting a shape inflation threshold and a position drift threshold, and based on the comparison results of the current window shape inflation factor and the shape inflation threshold and the current window position drift factor and the position drift threshold, diagnosing the state of the plastic part production line as one of normal, transient spike, gradual drift, or spike in drift, and completing real-time monitoring of the plastic part production line.

2. The method of claim 1, wherein The method further comprises the following steps for constructing the benchmark process model: taking the mean of all process parameter vectors of the standardized historical data as the benchmark centroid; taking the covariance matrix of all process parameter vectors of the standardized historical data as the benchmark shape matrix.

3. The method of claim 1, wherein The standardization processing adopts Z-score standardization.

4. The method of claim 1, wherein The conditions for diagnosing the transient spike include: the current window shape inflation factor is greater than the shape inflation threshold, and the current window position drift factor is less than or not greater than the position drift threshold.

5. The method of claim 1, wherein The conditions for diagnosing the gradual drift include: the current window shape inflation factor is not greater than the shape inflation threshold, and the current window position drift factor is greater than the position drift threshold.

6. The method of claim 1, wherein The conditions for diagnosing the spike in drift include: the current window shape inflation factor is greater than the shape inflation threshold, and the current window position drift factor is greater than the position drift threshold.

7. The method of claim 1, wherein The method further comprises adaptively updating the benchmark process model, comprising the following steps: after the gradual drift anomaly lasts for a set number of cycles, triggering manual confirmation, and if it is confirmed to be a benign drift, extracting process parameter vectors of the most recent plurality of cycles as new historical data and reconstructing the benchmark process model.

8. A plastic part production line real-time monitoring system, characterized by, The method comprises the following steps: a processor and a memory, the memory storing computer program instructions, when the computer program instructions are executed by the processor, realizing a plastic part production line real-time monitoring method according to any one of claims 1-7.

Images