Real-time monitoring method and system for position deviation of rotary table of multicolor injection molding machine

Abstract

The invention relates to the technical field of injection molding monitoring, and discloses a multicolor injection molding machine turntable position deviation real-time monitoring method and system, and the method comprises the steps: obtaining turntable design parameters, constructing a three-dimensional digital model, and dividing a fan-shaped monitoring region; acquiring no-load motion data to generate a no-load motion feature vector; collecting real-time motion data to generate a real-time motion feature vector; comparing the real-time feature vector with the no-load feature vector to judge whether position deviation occurs or not; if deviation occurs, a diagnosis process is triggered; through multi-dimensional data fusion analysis, real-time accurate monitoring and fault diagnosis of the position deviation of the turntable are realized, and the production quality and efficiency of the multi-color injection molding machine are remarkably improved.

Description

Technical Field

[0001] The invention relates to the technical field of injection molding monitoring, and more specifically to a real-time monitoring method and system for position deviation of a turntable of a multi-color injection molding machine. Background Art

[0002] With the rapid development of modern multi-color injection molding machines, the quality requirements for injection molded products are increasing. In the production process of multi-color injection molding machines, the position accuracy of the turntable directly affects the stability and consistency of product quality. Therefore, real-time monitoring of the position deviation of the turntable of the multi-color injection molding machine and timely discovery and diagnosis of the cause of the deviation are of great significance to ensure the quality of injection molded products.

[0003] The Chinese patent with the authorization announcement number CN205522372U discloses a closed-loop control system for the turntable of an injection molding machine based on encoder signal feedback. The system obtains the position information of the turntable through the encoder, realizes the closed-loop control of the turntable position, and improves the positioning accuracy and reliability of the turntable. However, the system mainly focuses on the control of the turntable position, lacks comprehensive monitoring and analysis of multi-dimensional information such as pressure and vibration during the movement of the turntable, and is difficult to fully evaluate the real-time working conditions of the turntable. The response to position deviation is not timely and accurate enough.

[0004] The Chinese patent application with publication number CN107942941A discloses a method and system for interactive signals between an injection molding machine and an electric turntable. The system realizes two-way communication between the injection molding machine and the electric turntable driver through the CAN bus, and can display the rotation parameters of the turntable and the working parameters of the servo motor in real time on the human-machine interface, providing an intuitive monitoring method for operators. However, the system mainly focuses on the macroscopic working state of the turntable and the servo motor, lacks quantitative analysis and diagnosis of position deviation, and cannot meet the needs of high-precision, intelligent real-time monitoring.

[0005] The Chinese patent application with publication number CN115871185A discloses an injection mold control system and control method. The system adopts temperature detection and feedback control to monitor the mold temperature in real time, calculate the temperature deviation and deviation change rate, and dynamically adjust the on-off time ratio of the relay to achieve precise control of the mold temperature characteristics. However, the system is mainly aimed at the temperature control of the injection mold, and does not involve the monitoring and diagnosis of the turntable position deviation, and cannot be directly applied to the real-time monitoring of the turntable of a multi-color injection molding machine.

[0006] In summary, most of the existing injection molding machine control systems use single-dimensional detection methods and lack the ability to integrate and analyze multi-source heterogeneous data. They are insufficient in real-time and accurate monitoring and intelligent diagnosis of turntable position deviations. Summary of the invention

[0007] In order to overcome the above-mentioned defects of the prior art, the present invention provides a real-time monitoring method and system for the position deviation of the turntable of a multi-color injection molding machine. The method obtains the design parameters of the turntable and the multi-dimensional motion data such as pressure, displacement, vibration, etc. under no-load conditions, and establishes a no-load motion feature vector that reflects the inherent characteristics of the turntable; during the production process, the multi-dimensional motion data of the turntable is collected in real time to generate a real-time motion feature vector; by comparing the real-time feature vector with the no-load feature vector, it is determined whether the turntable has a position deviation, and the intelligent diagnosis process is automatically triggered when a deviation is detected. This method can significantly improve the monitoring accuracy and response speed of the position deviation of the turntable of a multi-color injection molding machine, and provide a strong guarantee for the quality control of injection molded products.

[0008] To achieve the above object, the present invention provides the following technical solutions: The real-time monitoring method of the position deviation of the turntable of a multi-color injection molding machine includes: The design parameters of the turntable of the multi-color injection molding machine are obtained, and a three-dimensional digital model of the turntable is constructed. Based on the three-dimensional digital model of the turntable, the turntable is divided into M sector-shaped monitoring areas, and the no-load motion data of the turntable under the no-load condition is collected to generate a no-load motion feature vector P0 reflecting the inherent characteristics of the turntable; the no-load motion data includes no-load pressure data, no-load displacement data and no-load vibration data; the design parameters include the number of mold cavities, the number of mold cavities in the M sector-shaped monitoring areas is the same, and M is a positive integer; The real-time motion data of the turntable during the injection molding production process is collected, and a real-time motion feature vector M1 is generated based on the real-time motion data of the turntable and the real-time rotation speed N1 of the turntable. Based on the real-time motion feature vector M1 and the no-load motion feature vector P0, it is determined whether the turntable has a position deviation. If it is determined that the turntable has a position deviation, the deviation cause diagnosis process is triggered.

[0009] Further, the collecting of no-load motion data of the turntable under no-load working condition to generate a no-load motion feature vector P0 reflecting the inherent characteristics of the turntable includes: Each sector monitoring area is equipped with T p A pressure sensor and T d A displacement sensor; Install T radially on the turntable r A vibration sensor is installed on the axial direction of the turntable. a A vibration sensor; When the turntable is running without load, the output signals of the pressure sensor, displacement sensor and vibration sensor are collected synchronously to obtain M×T p No-load pressure data, M×T d No-load displacement data and T r +T a No-load vibration data; the no-load vibration data includes no-load radial vibration data and no-load axial vibration data; The time-frequency domain features of the no-load pressure data, no-load displacement data and no-load vibration data are extracted to obtain the dimension D=M×(T p +T d )+T r +T a The no-load motion characteristic vector ;in, is the no-load pressure eigenvector, is the no-load displacement eigenvector, is the no-load radial vibration eigenvector, is the no-load axial vibration eigenvector.

[0010] Furthermore, the real-time motion data includes M×T p dimensional real-time pressure data matrix P1, M×T d Dimensional real-time displacement data matrix D1, T r Real-time radial vibration vector V 1r and T a Dimensional real-time axial vibration vector V 1a ; The real-time motion characteristic vector M1 is generated according to the real-time motion data of the turntable and the real-time rotation speed N1 of the turntable, including: the real-time pressure data matrix P1, the real-time displacement data matrix D1, the real-time radial vibration vector V 1r and real-time axial vibration vector V 1a Perform time-frequency domain feature extraction to obtain real-time motion feature vectors ;in, is the real-time pressure feature vector, is the real-time displacement feature vector, is the real-time radial vibration eigenvector, is the real-time axial vibration eigenvector.

[0011] Further, judging whether the turntable has position deviation according to the real-time motion feature vector M1 and the no-load motion feature vector P0 includes: Calculate the distance E1 between the real-time motion feature vector M1 and the no-load motion feature vector P0; The deviation threshold T1 is preset. If E1≤T1, the turntable position accuracy meets the requirement; if E1>T1, it is determined that a position deviation has occurred, triggering the deviation cause diagnosis process.

[0012] Further, the calculating of the distance E1 between the real-time motion feature vector M1 and the no-load motion feature vector P0 includes: According to the real-time pressure feature vector and the no-load pressure eigenvector Calculate the pressure characteristic distance d p ; According to the real-time displacement feature vector and the no-load displacement eigenvector Calculate the displacement characteristic distance d d ; According to the real-time radial vibration eigenvector and the no-load radial vibration eigenvector Calculate the radial vibration characteristic distance d vr ; According to the real-time axial vibration eigenvector and the no-load axial vibration eigenvector Calculate the axial vibration characteristic distance d va ; Calculate the speed deviation between the real-time speed N1 of the turntable and the rated speed ; The distance E1 is obtained by weighted summation of the pressure characteristic distance, the displacement characteristic distance, the radial vibration characteristic distance, the axial vibration characteristic distance and the rotational speed deviation.

[0013] Furthermore, judging whether the turntable has position deviation according to the real-time motion feature vector M1 and the no-load motion feature vector P0 further includes: If d p >B p , the pressure characteristic is determined to be abnormal, and the pressure abnormality flag bit f is recorded p =1; otherwise, f p =0; B p is a preset pressure threshold; If d d >B d , then the displacement characteristic is determined to be abnormal, and the displacement abnormal flag bit f is recorded d =1; otherwise, f d =0; B d is the preset displacement threshold; If d vr >B vr , then the radial vibration characteristic is determined to be abnormal, and the radial vibration abnormal flag bit f is recorded vr =1; otherwise, f vr =0; B vr is a preset radial vibration threshold; If d va >B va , then the axial vibration characteristic is determined to be abnormal, and the axial vibration abnormal flag position f is recorded va =1; otherwise, f va =0; B va is the preset axial vibration threshold; If |Δn|>B n , the speed is determined to be abnormal, and the abnormal speed flag bit f is recorded n=1; otherwise, f n =0; B n is a preset speed deviation threshold; Calculate the total number of exception flag bits f=f p +f d +f vr +f va +f n , set the position deviation trigger threshold K1, when f ≥ K1, it is determined that the turntable has a position deviation.

[0014] Furthermore, the deviation cause diagnosis process includes: Collecting motion monitoring time series data within a normalized time window TW to form a sample, and summarizing all continuously collected samples to form a sample set {Q}; the motion monitoring time series data includes pressure data, displacement data, vibration data and rotation speed; Extract high-dimensional time-frequency features of each sample in the sample set {Q} to form a high-dimensional feature set {F}; Divide the high-dimensional feature set {F} into K feature subsets {S1}, {S2}, ..., {S K}, each feature subset corresponds to a typical position deviation pattern; For each feature subset {S j} Calculate its sample mean to form the mean vector L of the jth deviation mode j ; For each feature subset {S j} Calculate its sample variance to form the variance vector C of the j-th deviation mode j ; According to the mean vector L j and variance vector C j , construct the deviation pattern feature library {(L1,C1),(L2,C2),...,(L K ,C K )}, each deviation pattern in the deviation pattern feature library (L j ,C j ) corresponds to a known position deviation cause R j , forming a mapping set of deviation patterns and position deviation causes ; 1≤j≤K; When a new sample is obtained, the similarity between the new sample and each deviation pattern is calculated, and the position deviation cause corresponding to the deviation pattern with the greatest similarity is diagnosed as the most likely cause of the current deviation.

[0015] A multi-color injection molding machine turntable position deviation real-time monitoring system, which is used to implement the above-mentioned multi-color injection molding machine turntable position deviation real-time monitoring method, the system comprises: Turntable partitioning module: used to obtain the design parameters of the turntable of the multi-color injection molding machine, build a three-dimensional digital model of the turntable, and divide the turntable into M sector-shaped monitoring areas based on the three-dimensional digital model of the turntable. The design parameters include the number of mold cavities. The number of mold cavities in the M sector-shaped monitoring areas is the same, and M is a positive integer. No-load data acquisition module: used to collect no-load motion data of the turntable under no-load working conditions, and generate a no-load motion feature vector P0 that reflects the inherent characteristics of the turntable; Real-time data acquisition module: used to collect the real-time motion data of the turntable during the injection molding production process, and generate the real-time motion feature vector M1 based on the real-time motion data of the turntable and the real-time rotation speed N1 of the turntable; Position deviation judgment module: used to judge whether the turntable has position deviation based on the real-time motion feature vector M1 and the no-load motion feature vector P0; if it is determined that the turntable has position deviation, the deviation cause diagnosis process is triggered.

[0016] An electronic device comprises a memory, a central processing unit and a computer program stored in the memory and executable on the central processing unit. When the central processing unit executes the computer program, the above-mentioned method for real-time monitoring position deviation of a turntable of a multi-color injection molding machine is implemented.

[0017] A computer-readable storage medium stores a computer program, and when the computer program is executed, the method for real-time monitoring position deviation of a turntable of a multi-color injection molding machine is implemented.

[0018] Compared with the prior art, the present invention has the following beneficial effects: By real-time monitoring of the motion state of the turntable of a multi-color injection molding machine, the present invention can timely detect and diagnose position deviations, significantly improving the precision of injection molding and product quality. At the same time, it reduces downtime and rework caused by position deviations, and improves the overall efficiency of the production line. The present invention adopts multi-source data fusion technology to comprehensively analyze various sensor data such as pressure, displacement and vibration, and can more comprehensively monitor the operating state of the turntable. Through the time-frequency domain feature extraction of multi-dimensional feature vectors, an accurate description of the turntable movement is achieved. When a position deviation is detected, the system automatically triggers the diagnosis process, and quickly locates the cause of the abnormality through high-dimensional feature analysis of sample data. The establishment of a deviation pattern feature library enables different deviation causes to be efficiently identified, providing strong support for maintenance and repair. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 The principle flow chart of the real-time monitoring method of the position deviation of the turntable of a multi-color injection molding machine in the present invention; Figure 2 A flow chart of a method for collecting no-load motion data of a turntable and generating no-load motion feature vectors in a real-time monitoring method for position deviation of a turntable of a multi-color injection molding machine according to the present invention; Figure 3 It is a flow chart of a method for collecting real-time motion data of a turntable during an injection molding production process in a real-time monitoring method for position deviation of a turntable of a multi-color injection molding machine according to the present invention; Figure 4 It is a flow chart of a method for determining whether a position deviation occurs on a turntable in a method for real-time monitoring position deviation of a turntable of a multi-color injection molding machine according to Embodiment 1 of the present invention; Figure 5 A flow chart of a method for diagnosing the cause of deviation in a method for real-time monitoring position deviation of a turntable of a multi-color injection molding machine according to the present invention; Figure 6 It is a principle flow chart of judging whether a position deviation occurs on a turntable in a real-time monitoring method for position deviation of a turntable of a multi-color injection molding machine according to Embodiment 2 of the present invention; Figure 7 It is a functional module diagram of the real-time monitoring system for the position deviation of the turntable of a multi-color injection molding machine in the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] Example 1 See also Figure 1 As shown, this embodiment provides a real-time monitoring method for the position deviation of a turntable of a multi-color injection molding machine, comprising: Step S1000, obtaining design parameters of a turntable of a multi-color injection molding machine, constructing a three-dimensional digital model of the turntable, dividing the turntable into M sector-shaped monitoring areas based on the three-dimensional digital model of the turntable, collecting no-load motion data of the turntable under no-load working conditions, and generating a no-load motion feature vector P0 reflecting the inherent characteristics of the turntable; the no-load motion data includes no-load pressure data, no-load displacement data, and no-load vibration data; M is a positive integer; Furthermore, step S1000 includes: Step S1100, obtaining design parameters of a turntable of a multi-color injection molding machine, constructing a three-dimensional digital model of the turntable, and dividing the turntable into M sector-shaped monitoring areas based on the three-dimensional digital model of the turntable; the design parameters include the diameter of the turntable, the number of mold cavities, and the nominal coordinates of each mold cavity; the number of mold cavities in each sector-shaped monitoring area is the same; Specifically, the design parameters of the turntable, including the turntable diameter, the number of cavities, and the nominal coordinates of each cavity, are accurately obtained by consulting the design drawings of the multi-color injection molding machine and using measuring tools such as vernier calipers and dial indicators. The turntable diameter determines its rotation radius and is an important measure that affects the positioning accuracy; the number of cavities reflects the injection molding capacity and the complexity of the cavity arrangement; and the nominal coordinates of each cavity give the theoretical position of the cavity in the radial and circumferential directions of the turntable, which is an important basis for subsequent deviation calculations. Using the obtained design parameters, a digital three-dimensional solid model of the turntable is constructed in the CAD software. Through parametric modeling technology, the association between model size and design parameters is realized. When the parameters change, the model is automatically updated, which greatly improves the modeling efficiency and the accuracy of parameter transfer. In order to facilitate the implementation of the partition monitoring strategy, the disk model is divided into M sectors. The selection of the M value needs to balance the monitoring accuracy and cost. The larger the M, the finer the partition, and the stronger the "detection power" of the position deviation, but the number of sensors and the amount of data processing also increase accordingly. Combining the symmetry of the turntable structure and the uniformity of the cavity distribution, M can be set to equal fractions of one turntable rotation (such as four equal parts or eight equal parts), which not only ensures the relative equality of the partition area, but also facilitates the symmetrical layout of the sensors. Each sector contains an equal number of cavities, which provides the hardware foundation for accurate positioning and traceability.

[0023] Step S1100 obtains the design parameters of the turntable through the system, overcomes the inefficiency and errors of manual mapping, and lays a data foundation for the subsequent construction of a high-fidelity digital twin. Using professional 3D CAD software and parametric modeling technology, a two-way association between the model and the parameters is achieved, which not only ensures the accuracy of the model, but also improves the flexibility of design changes. By partitioning the turntable into sectors, while covering the entire turntable, it highlights the focus on local position deviations, taking into account both the overall situation and the local situation, and providing an entry point for implementing differentiated partition monitoring strategies. The reasonable selection of the number of partitions achieves a beneficial balance between monitoring accuracy and cost. The partition processing of the model points the way for the optimal layout of sensors and is a key link in achieving reliable monitoring and efficient diagnosis.

[0024] Step S1200, collecting no-load pressure data, no-load displacement data and no-load vibration data of the turntable under no-load working conditions, and generating a no-load motion characteristic vector P0 reflecting the inherent characteristics of the turntable.

[0025] Furthermore, if Figure 2 As shown, step S1200 includes: Step S1210: deploy T p A pressure sensor and T d A displacement sensor; Step S1220, install T in the radial direction of the turntable r A vibration sensor is installed on the axial direction of the turntable. a A vibration sensor; Step S1230: When the turntable is running without load, synchronously collect the output signals of the pressure sensor, displacement sensor and vibration sensor to obtain M×T p No-load pressure data, M×T d No-load displacement data and T r +T a No-load vibration data; the no-load vibration data includes no-load radial vibration data and no-load axial vibration data; Step S1240: extract the time-frequency domain features of the no-load pressure data, no-load displacement data, and no-load vibration data to obtain a dimension of D=M×(T p +T d )+T r +T a The no-load motion characteristic vector ;in, is the no-load pressure eigenvector, is the no-load displacement eigenvector, is the no-load radial vibration eigenvector, is the no-load axial vibration eigenvector.

[0026] Specifically, the pressure sensor is arranged at the entrance of the convex surface of the mold cavity, and the stability of the injection molding process is evaluated by monitoring the change of the gate pressure; the displacement sensor is arranged in the middle of the convex surface, and the model closure is estimated by measuring the normal displacement of the mold cavity. The layout point should be as close to the mold cavity as possible, and avoid interference sources such as gates and ejectors. The specific model and range of the sensor need to match the mold cavity size and injection molding process parameters. Commonly used pressure sensors include strain gauges and piezoresistive types, and displacement sensors include grating scales and laser displacement sensors. In order to associate the local coordinates obtained by the sensor with the global position of the turntable, it is necessary to use a high-precision three-dimensional coordinate measuring instrument or laser tracker to spatially locate the installation point and establish a coordinate transformation matrix between the local coordinate system and the turntable fixed coordinate system. The strategy of partitioned point monitoring is adopted to obtain rich local information while taking into account the system complexity and cost feasibility. The reasonable configuration of pressure and displacement sensors can accurately perceive the dynamic changes of mold cavity force and cavity closure, providing data support for injection molding process stability control and product quality traceability. The key to distributed measurement is to achieve global positioning of measurement points based on spatial coordinate calibration. The introduction of coordinate transformation matrix integrates the local perception of discrete measurement points into the global pose estimation, which is the mathematical basis for achieving integrated measurement and control. The monitoring partition division method not only reduces the computational load but also ensures the diagnosability of faults, which is a beneficial attempt in engineering practice. In the future, with the advancement of intelligent sensing and Internet of Things technologies, partitioned point monitoring is expected to further optimize sensor selection and realize adaptive encryption of measurement points and intelligent data analysis.

[0027] Radial vibration reflects the lateral swing of the turntable in the XY plane of the inertial coordinate system, which is closely related to the rotation accuracy; axial vibration corresponds to the axial movement in the Z direction, which affects the quality of mold clamping. Both types of vibration can be directly measured by the sensor or calculated from the vibration at the turntable bearing seat. Commonly used vibration sensors include accelerometers, velocity sensors and displacement sensors. When selecting, sensitivity, frequency response and measurement range must be considered. No-load acquisition is carried out under the condition that the injection molding machine is running normally but no mold is installed or injection is performed. The purpose is to obtain the ideal motion mode inherent in the turntable and provide a benchmark for subsequent monitoring. When collecting, it is necessary to ensure that the injection molding machine reaches thermal stability and various process parameters such as oil temperature, pressure, and speed are stable at the set values. Feature extraction of the collected time domain signal can unify different physical quantities into signal features that reflect the essence and compress the data dimension. Commonly used time domain features include the mean that reflects the signal level, the variance that measures signal fluctuations, and the root mean square value that takes into account both positive and negative pulses and is sensitive to the position deviation of the turntable; frequency domain features such as the main frequency and modulation frequency reflect the periodicity and frequency stability of the turntable motion, and are also important indicators for evaluating the performance of the turntable. The complementary deployment of radial and axial vibration sensors can capture the complete posture changes of the turntable in three-dimensional space, making up for the coverage blind spots of the pressure displacement measurement points. The no-load acquisition extracts the inherent dynamic behavior of the turntable by isolating production interference, which simplifies the complexity of mathematical modeling and reduces the difficulty of algorithm development. The extraction of time-frequency domain features enables heterogeneous physical quantities to be integrated at the information level in the feature space, which not only compresses the data scale but also highlights the essential properties of the signal. It is a typical method for mining engineering big data.

[0028] Step S1000 systematically realizes the holographic acquisition and characterization of the injection molding machine turntable posture from turntable digital modeling, partition monitoring point layout, to vibration signal perception and no-load benchmark acquisition. The three-dimensional digital twin provides an intuitive and visual platform for the design of measurement and control solutions. The sector-shaped partition monitoring makes full use of the periodic symmetry of the turntable structure, while meeting the fault location requirements and taking into account the measurement cost. The complementary deployment of vibration sensors depicts the complete changes in the turntable posture from a dynamic perspective and makes up for the blind spots of discrete measuring points. No-load feature acquisition and fusion characterization further condense the inherent mode of the turntable and simplify the process of obtaining the benchmark threshold.

[0029] Step S2000, collect the real-time motion data of the turntable during the injection molding production process, generate a real-time motion feature vector M1 based on the real-time motion data of the turntable and the real-time rotation speed N1 of the turntable; determine whether the turntable has a position deviation based on the real-time motion feature vector M1 and the no-load motion feature vector P0; if it is determined that the turntable has a position deviation, collect the motion monitoring time series data within a time window, construct a deviation pattern feature library, diagnose the cause of the deviation, and obtain the diagnosis result of the cause of the deviation.

[0030] Furthermore, step S2000 includes: Step S2100, collecting real-time motion data of the turntable during the injection molding production process; the real-time motion data includes M×T p dimensional real-time pressure data matrix P1, M×T d Dimensional real-time displacement data matrix D1, T r Real-time radial vibration vector V 1r and T a Dimensional real-time axial vibration vector V 1a ; Furthermore, if Figure 3 As shown, step S2100 includes: Step S2110, obtaining the real-time pressure data of the mold cavity in each sector monitoring area, forming M×T p Dimensional real-time pressure data matrix P1; Step S2120, obtaining the real-time displacement data of the mold cavity in each sector monitoring area, forming M×T d Dimensional real-time displacement data matrix D1; Step S2130, obtaining the real-time radial vibration data of the turntable to form T r Real-time radial vibration vector V 1r ; Step S2140, obtaining real-time axial vibration data of the turntable to form T a Dimensional real-time axial vibration vector V 1a .

[0031] Specifically, real-time data acquisition is the basis of deviation monitoring. The real-time pressure data of the mold cavity in each sector-shaped monitoring area is obtained in real time through the pressure sensor array. The real-time normal displacement data of the mold cavity in each sector-shaped monitoring area is obtained in real time through the displacement sensor array. The real-time radial vibration data and real-time axial vibration data of the turntable are obtained in real time through the vibration sensor array. The real-time motion data is preprocessed as necessary, including filtering and denoising, zero drift correction, and outlier removal, to improve the signal-to-noise ratio and consistency of the data. The real-time pressure data matrix P1 reflects the dynamic distribution of the force on the mold cavity in each sector-shaped monitoring area, and the real-time displacement data matrix D1 depicts the real-time trajectory of the normal motion of the mold cavity. The two complement each other and jointly depict the force and motion coupling characteristics of the mold cavity during the injection molding process. Real-time radial vibration vector V 1r and real-time axial vibration vector V 1aThe overall vibration behavior of the turntable is characterized in the horizontal and vertical directions respectively. The sensor sampling frequency can be configured according to the turntable speed and injection cycle to cover the main frequency components of the turntable movement. Data preprocessing helps to suppress external interference and measurement noise and meet the quality requirements of subsequent data analysis. The fusion modeling of local pressure displacement and overall vibration data can fully characterize the dynamic behavior of the turntable and achieve the complementarity of local information and overall information, which is an important manifestation of holographic monitoring.

[0032] Step S2200, generating a real-time motion feature vector M1 according to the real-time motion data of the turntable and the real-time rotation speed N1 of the turntable; For the real-time pressure data matrix P1, real-time displacement data matrix D1, real-time radial vibration vector V 1r and real-time axial vibration vector V 1a Perform time-frequency domain feature extraction to obtain real-time motion feature vectors ;in, is the real-time pressure feature vector, is the real-time displacement feature vector, is the real-time radial vibration eigenvector, is the real-time axial vibration eigenvector.

[0033] Specifically, the rotation speed, as a key parameter affecting the dynamic characteristics of the turntable, needs to be obtained in real time through a dedicated sensor. Common speed sensors include incremental encoders, Hall sensors, etc., which calculate the rotation speed by measuring the number of pulses or rotation angles per unit time. When selecting a speed sensor, indicators such as range, resolution and linearity should be considered. The installation location should be close to the axis of the turntable and electromagnetic shielding should be done well. Incorporating the rotation speed into the real-time motion feature vector can improve the accuracy and reliability of position deviation discrimination. When synchronously collecting data of different physical quantities, timestamp alignment is crucial. Through network synchronization protocols such as PTP and NTP, accurate synchronization between distributed sensors and data processing units can be achieved, and asynchronous data such as pressure, displacement, and vibration can be mapped to a unified time reference. On the basis of data synchronization, the time domain and frequency domain characteristics of each physical quantity are extracted to compress the data size while condensing the key information that characterizes the motion state of the turntable. The general principles of feature selection include: close correlation with position deviation, significant sensitivity to process parameters, and moderate computational complexity. Real-time pressure feature vector You can select first-order statistics such as mean, variance, and displacement characteristics You can select high-order moments such as peak-to-peak value and RMS value, while the frequency domain features can focus on the main peak of the spectrum and its distribution, etc. , The extraction idea of ​​is similar to that of . Feature extraction enables heterogeneous information to be fused and mapped in feature space, simplifies the subsequent state discrimination model, and also provides the possibility for knowledge transfer. Incorporating the instantaneous speed N1 into the real-time motion feature vector can not only correct the random error introduced by speed fluctuation, but also provide a regulating variable for deviation compensation.

[0034] Step S2300, judging whether the turntable has position deviation according to the real-time motion feature vector M1 and the no-load motion feature vector P0; Furthermore, if Figure 4 As shown, step S2300 includes: Step S2310, calculating the distance E1 between the real-time motion feature vector M1 and the idle motion feature vector P0; Step S2320, preset the deviation threshold T1, if E1≤T1, the turntable position accuracy meets the requirement, and return to step S2100 to continue monitoring; In step S2330, if E1>T1, it is determined that position deviation has occurred, triggering the deviation cause diagnosis process of step S2400.

[0035] The calculation of the distance E1 between the real-time motion feature vector M1 and the no-load motion feature vector P0 includes: According to the real-time pressure feature vector and the no-load pressure eigenvector Calculate pressure characteristic distance; According to the real-time displacement feature vector and the no-load displacement eigenvector Calculate the displacement characteristic distance; According to the real-time radial vibration eigenvector and the no-load radial vibration eigenvector Calculate radial vibration characteristic distance; According to the real-time axial vibration eigenvector and the no-load axial vibration eigenvector Calculate the axial vibration characteristic distance; Calculate the speed deviation between the real-time speed N1 of the turntable and the rated speed; The distance E1 is obtained by weighted summation of the pressure characteristic distance, the displacement characteristic distance, the radial vibration characteristic distance, the axial vibration characteristic distance and the rotational speed deviation.

[0036] The specific formula for distance E1 is: in: : Indicates pressure characteristic distance.

[0037] : represents the displacement characteristic distance.

[0038] : Indicates the radial vibration characteristic distance.

[0039] : Indicates the axial vibration characteristic distance.

[0040] : Indicates speed deviation.

[0041] : The distance between the real-time motion feature vector M1 and the no-load motion feature vector P0.

[0042] : The total number of sector monitoring areas.

[0043] : No. The importance weight of the sector monitoring area satisfies The technical personnel in this field make a comprehensive determination through a large number of experiments based on factors such as the location of the mold cavity in the area and the importance of the injection molded product.

[0044] :Indicates the The real-time pressure characteristic vector of a sector-shaped monitoring area during real-time operation.

[0045] :Indicates the The no-load pressure characteristic vector of a sector-shaped monitoring area when it is no-load.

[0046] : Hadamard product (element-by-element multiplication), used to measure the interaction between the real-time pressure characteristics and the no-load pressure characteristics.

[0047] : Infinity norm, that is, the maximum absolute value of each element of the vector, is used to measure the maximum amplitude of pressure deviation.

[0048] :Indicates the The real-time displacement feature vector of a sector-shaped monitoring area during real-time operation.

[0049] :Indicates the The no-load displacement characteristic vector of a sector-shaped monitoring area when it is no-load.

[0050] : Hadamard division (element-by-element division) is used to measure the rate of change of the real-time displacement characteristic relative to the no-load displacement characteristic.

[0051] 1: All-1 vector, minus 1 is used to calculate the relative rate of change.

[0052] The norm, which is the sum of the absolute values ​​of the elements of the vector, is used to measure the cumulative magnitude of the position deviation.

[0053] : The number of displacement sensors, divided by It is to normalize the L1 norm of different dimensions.

[0054] : The global weight factor of the radial vibration feature is determined by technicians in this field through a large number of experiments.

[0055] : The global weight factor of the axial vibration characteristic is determined by a technician in this field through a large number of experiments. and Reflects the overall impact of vibration characteristics on position accuracy.

[0056] : Weight factor of speed deviation.

[0057] : Rated speed.

[0058] : Vector outer product, generating a matrix to measure the interaction of various frequency components of radial vibration.

[0059] The purpose of subtracting the mean is to eliminate the DC component of radial vibration.

[0060] Norm.

[0061] : Calculate the natural logarithm of each element to measure the relative change of energy of each frequency component of axial vibration.

[0062] : The number of axial vibration sensors, divided by It is to normalize the L2 norm of different dimensions.

[0063] This formula treats the deviation contribution of different fan-shaped areas differently to highlight the influence of key areas. The Hadamard product is introduced to consider the interaction effect between real-time pressure and no-load pressure in the pressure feature to more comprehensively characterize the pressure deviation. The Hadamard division is introduced to consider the rate of change of real-time displacement relative to no-load displacement in the displacement feature, rather than the absolute deviation. The vector outer product is introduced to consider the coupling effect between different frequency components in the radial vibration feature to capture the overall change of the spectrum structure. The logarithmic transformation is introduced to convert the absolute energy difference into the relative energy ratio in the axial vibration feature to more sensitively reflect the variation of the axial vibration. Different metrics such as the L1 norm, L2 norm, and infinity norm are used in combination to evaluate the deviation contribution of various features from multiple angles. , , , Prior knowledge is introduced to flexibly adjust the weight of each feature according to the characteristics of the process and equipment. The speed is a key parameter that affects the dynamic characteristics of the turntable. Including it in the distance measurement can more comprehensively describe the actual motion state of the turntable and improve the accuracy of deviation judgment. The speed deviation term is introduced in the formula , improving the accuracy of position deviation judgment.

[0064] Step S2300 calculates the distance E between the real-time motion feature vector M1 and the idle motion feature vector P0 1， The degree to which the turntable motion state deviates from the ideal no-load mode is quantified. This distance metric comprehensively considers multi-source information such as pressure, displacement, and vibration to comprehensively evaluate the dynamic changes in the turntable's posture. Among them, the pressure characteristic distance compares the degree of deviation between the real-time pressure distribution and the ideal pressure distribution, reflecting the uniformity of the force on the mold cavity; the displacement characteristic distance measures the geometric difference between the actual displacement trajectory and the ideal displacement trajectory, and describes the spatial accuracy of the mold cavity motion; the radial and axial vibration characteristic distances evaluate the abnormal degree of the turntable vibration level from the horizontal and vertical directions respectively, capturing the slight changes in the global posture. The introduction of the speed deviation term can suppress the random interference introduced by the speed fluctuation and improve the robustness of the deviation discrimination.

[0065] In the distance calculation, the weighted summation method is adopted to assign different weight coefficients to different physical quantities, highlighting the influence of key features. The pressure feature uses the infinite norm to measure the maximum pressure deviation in each monitoring area, with the worst case as the optimization target; the displacement feature uses the Manhattan distance to characterize the average displacement deviation, with the overall accuracy as the optimization direction; the vibration feature and the speed deviation directly use the Euclidean distance to measure the amplitude difference, and the calculation is simple and intuitive. The nonlinear function E1 is used to integrate the distances of each component, which can suppress the adverse effects of the dimension and order of magnitude differences. The setting of the threshold T1 needs to balance between production efficiency and product quality. If it is too small, it may cause false alarms, and if it is too large, it may miss detection. The deviation levels of different batches can be statistically analyzed in combination with the injection molding process requirements, product accuracy specifications, etc., and the threshold selection can be adaptively optimized. Continuous monitoring helps to grasp the dynamic trend of deviation occurrence and timely warn of deterioration risks. When it is determined that a position deviation occurs, the intelligent diagnosis should be started with minimal delay to shorten the downtime.

[0066] Step S2300 designs a unified distance metric for the multi-source heterogeneous features of the turntable position, so that the discrimination process can be implemented under the same mathematical framework, simplifying the threshold setting and correction process. The introduction of the weighted summation mechanism can flexibly adjust the sensitivity of each component feature to meet different monitoring needs. The nonlinear comprehensive mapping function suppresses the interference of dimension and order of magnitude differences and enhances the applicability of the distance metric. The adaptive optimization of the deviation threshold can further improve the discrimination accuracy and realize the automation and intelligence of the monitoring process. The adoption of this scheme can significantly improve the discovery efficiency and recognition accuracy of the turntable position deviation, timely warn abnormal working conditions, and ensure the quality of injection molding products. The fusion and utilization of multi-source information fully taps the potential value of monitoring data, reveals the inherent connection between different physical quantities, and provides new ideas for comprehensive characterization of the state. The unified definition of the distance metric makes the threshold setting and correction more concise and intuitive, reducing the difficulty of engineering implementation. The introduction of the adaptive mechanism gives the deviation discrimination greater flexibility and robustness, clearing the obstacles for state monitoring under complex working conditions. Looking into the future, with the further development of sensing technology and data analysis, holographic perception, precise diagnosis and intelligent control of the turntable's position will become possible, injecting new impetus into the high-end and intelligent upgrading of injection molding equipment.

[0067] Step S2400, if it is determined that the turntable has a position deviation, collect the motion monitoring time series data within a time window, build a deviation pattern feature library, diagnose the cause of the deviation, and obtain the diagnosis result of the cause of the deviation; the motion monitoring time series data includes pressure data, displacement data, vibration data and rotation speed; Furthermore, if Figure 5 As shown, step S2400 includes: Step S2410, collecting motion monitoring time series data within a normalized time window TW to form a sample, and summarizing all continuously collected samples to form a sample set {Q}; Step S2420, extracting high-dimensional time-frequency features of each sample in the sample set {Q} to form a high-dimensional feature set {F}; Step S2430, divide the high-dimensional feature set {F} into K feature subsets {S1}, {S2}, ..., {S K}, each feature subset corresponds to a typical position deviation pattern; Step S2440, for each feature subset {S j} Calculate its sample mean to form the mean vector L of the jth deviation mode j ; For each feature subset {S j} Calculate its sample variance to form the variance vector C of the j-th deviation mode j ; Step S2450, according to the mean vector L j and variance vector C j , construct the deviation pattern feature library {(L1,C1),(L2,C2),...,(L K ,C K )}, each deviation pattern in the deviation pattern feature library (L j ,C j ) corresponds to a known position deviation cause R j , forming a mapping set of deviation patterns and position deviation causes ; 1≤j≤K; Step S2460, when a new sample is obtained, the similarity between the new sample and each deviation pattern is calculated, and the position deviation cause corresponding to the deviation pattern with the greatest similarity is diagnosed as the most likely cause of the current deviation.

[0068] Specifically, when the position deviation of the turntable is detected, it is necessary to further diagnose the root cause of the deviation so that corrective measures can be taken in time. Due to the complexity and variability of the injection molding production process, the deviation data at a single moment is often difficult to fully reflect the nature of the fault. Therefore, multi-source heterogeneous time series data such as pressure, displacement, vibration, and speed are continuously collected within a normalized time window TW to form a high-dimensional sample. The length of the time window must take into account the representativeness of the data and the computational efficiency, and is usually set to an integer multiple of the injection molding cycle. Normalization is to eliminate the systematic errors between different batches and different equipment, and to facilitate the horizontal comparison of samples. Through continuous collection of sliding windows, a sample set {Q} containing a representative deviation evolution process can be obtained.

[0069] In order to mine the deviation patterns contained in the sample set {Q}, it is necessary to extract a series of high-dimensional time-frequency features that reflect its intrinsic characteristics for each sample to form a high-dimensional feature set {F}. The idea of ​​feature extraction is to characterize the morphological characteristics of a single sample from multiple perspectives and levels from the two perspectives of time domain and frequency domain, and from the two scales of local and overall. For example, the peak value and mean value of the pressure signal can be extracted to reflect its numerical characteristics, and the gradient, kurtosis, etc. can be extracted to reflect its shape characteristics; from the frequency domain perspective, the main frequency, bandwidth, peak value, etc. of the spectrum can be focused on to characterize the periodicity and frequency distribution of the signal; similar feature extraction operators can also be used for displacement and vibration signals. In addition, cross-features between different physical quantities, such as the pressure-displacement correlation coefficient, can reveal the coupling effects between different subsystems. High-dimensional feature extraction makes full use of the multi-source heterogeneity of data and depicts the overall picture of the sample from different aspects, but it also brings the trouble of dimensionality disaster.

[0070] In order to find valuable deviation patterns in massive high-dimensional data, it is necessary to use machine learning ideas to reduce the dimension of the feature set {F} and cluster it. Using clustering algorithms such as K-means and DBSCAN, {F} is divided into K feature subsets {S1}, {S2},..., {S K}, each feature subset corresponds to a typical position deviation pattern. The clustering process can be seen as "feeding" the machine with data, allowing the machine to learn to distinguish between normal and abnormal, and diagnose different abnormal patterns in a fine-grained manner. The key to clustering performance is similarity measurement, which requires comprehensive consideration of the scale differences and noise interference of features in each dimension, and can be achieved through metric functions such as weighted Euclidean distance and Mahalanobis distance. The selection of K value needs to balance between clustering accuracy and real-time performance, and usually follows the elbow rule, that is, when the K value is large enough, the clustering performance improvement is not obvious.

[0071] For each feature subset obtained by clustering {S j}, calculate its sample mean to form the mean vector L j , calculate the sample variance to form the variance vector C j。 Mean vector L j reflects the central position of the jth deviation pattern and is the most representative “portrait” of the pattern; the variance vector C j It describes the discrete degree of the internal samples of the pattern and reflects the evolution range of the pattern. j and variance C j The two-tuple (L j ,C j ) uniquely determines a Gaussian distribution, which can be used for subsequent deviation pattern matching. j ,C j ) to build a deviation pattern feature library {(L1,C1),(L2,C2),...,(L K ,C K)}, which has become the diagnostic knowledge accumulated over a long period of time and laid the foundation for intelligent diagnosis.

[0072] However, to transform the learned deviation pattern into an interpretable diagnostic conclusion, it is necessary to map it with the known causes of position deviation with the help of expert experience. j ,C j ), the expert team can analyze the morphological characteristics of its pressure, displacement, vibration and other signals, and combine the early accumulated failure cases to summarize the most likely cause of the pattern. j , such as core deviation, cavity wear, oil pressure fluctuation, etc. j ,C j ) and R j The mapping relationship is entered into the diagnostic knowledge base to form a mapping set from deviation mode to position deviation cause, such as { , ,…}. The formation of the mapping set enables the diagnosis of deviation patterns to move from experience to knowledge, allowing the machine to have a preliminary ability to analyze the cause of the fault, and realize the transition from data to knowledge.

[0073] When a new deviation sample is obtained, on-site personnel do not need to manually interpret complex high-dimensional data. They only need to call the trained model mentioned above to automatically perform the following steps: ① Extract high-dimensional time-frequency features consistent with the training stage for the new sample; ② Use the similarity function to calculate the similarity between the sample and the K deviation patterns in the feature library; ③ Diagnose the deviation pattern with the largest similarity as the category to which the new sample belongs; ④ Output the cause of the position deviation mapped by the category to form a diagnosis report.

[0074] This method extracts common rules through big data analysis, avoiding the limitations of individual experience; the feature library can be continuously expanded through incremental learning to cope with process changes; the introduction of high-dimensional features enhances the reliability of diagnosis; the combination of offline training and online matching ensures real-time response while identifying new faults; the diagnostic conclusions are traceable and explainable, which is convenient for expert review and iterative optimization.

[0075] For example, a company used this method to monitor the position deviation of a 160-ton injection molding machine. The equipment reported multiple deviation warnings within one day, and collected 200 historical samples and 10 new samples. 100-dimensional time-frequency features were extracted from the historical samples, and 5 typical deviation patterns were formed through K-means clustering, which were interpreted by the expert team as: core offset, nozzle wear, insufficient clamping force, cushion aging and conventional error. The 5 deviation patterns were entered into the feature library, and 10 new samples were diagnosed one by one. The results showed that 6 samples were diagnosed as core offset, 3 were diagnosed as conventional errors, and 1 was diagnosed as nozzle wear. On-site verification found that the core axis had accumulated offset and the nozzle was also slightly worn, which was highly consistent with the diagnosis results. By correcting the deviation in the first time, the scrap rate was avoided from further deterioration.

[0076] In summary, step S2400 forms a set of intelligent diagnosis methods for turntable position deviation based on big data mining by introducing the integration of machine learning and expert knowledge. This method uses a sliding window to continuously collect multi-source heterogeneous data, adaptively extracts high-dimensional time-frequency features, discovers typical deviation patterns through cluster analysis, and then forms a mapping set of patterns and causes with the help of expert experience, and finally realizes real-time diagnosis of new samples through similarity matching. This closed-loop diagnostic process can significantly improve the level of intelligent diagnosis, reduce the labor intensity of on-site personnel, and provide a digital gripper for precise deviation correction. In the future, with the continuous accumulation of samples and the continuous optimization of expert feedback, the diagnostic model will become more mature and stable, and become a tool for injection molding companies to achieve independent control and continuous improvement.

[0077] The calculating of the similarity between the new sample and each deviation pattern includes: in, Represents a new sample With Deviation Mode similarity; Represents the new sample dimensional feature vector; Indicates The mean vector of the deviation modes and variance vector ; Indicates The deviation mode is The mean component on the feature dimension, Indicates The deviation mode is The variance components on the feature dimensions, ; For example, assuming there are two deviation modes, and the feature vector of each deviation mode is 3-dimensional, then: The mean vector of the first deviation mode , the variance vector is ; The mean vector of the second deviation mode , the variance vector is ; in, represents the mean component of the first deviation mode in the second feature dimension, represents the mean component of the second deviation mode in the third feature dimension, represents the variance of the first deviation mode in the second feature dimension, It represents the variance of the second deviation mode in the third feature dimension, and so on.

[0078] is a global scaling parameter used to control the scaling of the overall similarity. The larger it is, the smaller the overall similarity is, and the stronger the ability to distinguish deviation patterns is.

[0079] is a local scaling parameter used to control the distance scaling in a single dimension. The larger it is, the smaller the impact of the numerical differences between different dimensions on the similarity.

[0080] is a sensitive parameter used to control the sensitivity of similarity to distance. The larger the value, the faster the similarity decays with the distance.

[0081] is an exponential function with the base e of the natural logarithm as its base.

[0082] It is obtained by technicians in this field through cross-validation and other methods on the training set to balance the accuracy and generalization performance of similarity calculation. The feature vector of the new sample It is obtained by extracting features from the latest collected motion monitoring time series data within a time window.

[0083] This formula takes into account the impact of numerical differences in different feature dimensions on similarity, and maps the distances of each dimension to the same scale through an exponential function, thus avoiding the problem of individual dimensions dominating the similarity calculation. The importance of different feature dimensions is adaptively weighted. The dimension with larger variance has smaller weight in similarity calculation, thus improving the robustness of similarity calculation. The adjustment of can flexibly control the sensitivity of similarity calculation to distance and balance the ability to distinguish deviation patterns and generalization performance. The value range is , which is convenient for threshold judgment and probability interpretation. With deviation mode When the mean values ​​of each dimension of the new sample differ greatly, the similarity approaches 0; when the mean values ​​of each dimension of the new sample and the deviation pattern differ slightly, the similarity approaches 1.

[0084] In summary, the similarity formula comprehensively considers factors such as the differences between feature dimensions, the internal variance of the deviation pattern, and distance sensitivity, and can more accurately and robustly calculate the similarity between new samples and known deviation patterns, providing a reliable basis for fault diagnosis. Of course, in practical applications, it is also necessary to appropriately adjust and optimize the parameters in the formula according to specific scenarios and data characteristics to achieve the best diagnostic effect.

[0085] Step S2500: output a corresponding correction strategy based on the deviation cause diagnosis result and a pre-calibrated deviation-compensation mapping rule.

[0086] Specifically, compensation control is the last link in the closed-loop management of position deviation. Due to the complex structure of the injection molding machine, which involves multiple disciplines such as hydraulics, machinery, and electronic control, the compensation control methods for position deviations caused by different reasons are often quite different. Through mechanism analysis and expert experience, the causes, symptoms, effects, compensation methods, etc. of various position deviations can be summarized to establish a complete diagnosis-decision-making knowledge base. This knowledge base realizes the rapid mapping of deviation causes to control strategies, which not only includes standard handling procedures for common faults, but also allows experts to continuously enrich and optimize. When the diagnosis module outputs the cause of the position deviation, the corresponding compensation control strategy can be quickly retrieved through the mapping rule base. The dynamic generation and optimization adjustment of the compensation control strategy can significantly improve the fault self-healing ability and production efficiency of the injection molding machine. In the long run, it is also possible to integrate multi-source heterogeneous information such as process parameters, material characteristics, and mold structure, and adopt cutting-edge artificial intelligence technologies such as deep reinforcement learning to achieve global optimization and flexible adaptation of the injection molding process.

[0087] Step S2000 starts with collecting the real-time motion data of the turntable during the injection molding process. The key physical quantities reflecting the dynamic behavior of the turntable are obtained by arranging the pressure and displacement sensors in each sector monitoring area and the radial and axial vibration sensors of the turntable. Among them, the real-time pressure data obtained by the pressure sensor reflects the dynamic distribution of the force on the mold cavity during the injection molding process, and the real-time displacement data obtained by the displacement sensor depicts the real-time trajectory of the normal motion of the mold cavity. The two complement each other and jointly depict the force and motion coupling characteristics of the mold cavity. The vibration data comprehensively characterizes the vibration behavior of the turntable as a whole in the horizontal and vertical directions. After obtaining heterogeneous multimodal data, necessary data preprocessing is required. Filtering and denoising can suppress random interference in the measurement process and improve the signal-to-noise ratio; zero drift correction eliminates the systematic error introduced by the initial value drift of the sensor; and outlier elimination avoids the interference of "wild point" data on subsequent analysis. The real-time pressure, displacement, and vibration data after preprocessing are not only improved in quality, but also greatly improved in consistency, laying a data foundation for subsequent multi-source information fusion.

[0088] After obtaining reliable and consistent real-time motion data, it is necessary to extract key features that reflect the dynamic characteristics of the turntable. Time domain features such as mean and variance can measure the basic statistical properties of the signal, frequency domain features such as main frequency and modulation frequency reveal the periodic characteristics of the signal, and time-frequency domain features such as wavelet coefficients can capture the multi-scale and multi-resolution details of the signal. Feature extraction not only greatly compresses the data size, but more importantly, it condenses the inherent laws and connections hidden in the original data. In order to accurately characterize the actual motion state of the turntable, it is also necessary to obtain the real-time speed of the turntable. The speed not only directly affects the physical and chemical parameters of the injection molding process such as temperature and pressure, but also affects its vibration frequency by changing the dynamic characteristics of the turntable. Incorporating the speed into the real-time motion feature vector can improve the accuracy and reliability of position deviation discrimination. By integrating heterogeneous multimodal features such as pressure, displacement, and vibration and combining them with real-time speed data, a unified real-time motion feature vector M1 is finally generated, which fully characterizes the dynamic behavior of the turntable in the actual injection molding process.

[0089] Based on the distance measurement between the real-time motion feature vector M1 and the no-load reference feature P0, the real-time judgment of the turntable position deviation can be achieved. This distance measurement uses the norm theory and comprehensively considers the difference contribution of multiple source information such as pressure, displacement, and vibration. The core idea is that when the turntable has a position deviation, the measured pressure, displacement, and vibration characteristics must deviate from its ideal no-load reference state, and the degree of deviation is positively correlated with the amplitude of position misalignment. By quantifying the difference between the real-time feature and the reference feature, the degree of deviation of the turntable position can be indirectly evaluated. In order to balance the different sensitivities of various physical quantities to position deviation, the weight w is introduced to differentiate the contribution of each monitoring partition, and the parameter λ is introduced to play a regulating role in the vibration feature distance. The setting of the threshold T1 needs to be based on empirical statistics, taking into account both the detection rate and the false alarm rate. If the measured distance exceeds the threshold, the downstream deviation tracing and cause diagnosis process is triggered.

[0090] Step S2000 monitors the turntable in sector-shaped partitions, ensuring global coverage while taking local accuracy into account, and greatly improving the monitoring resolution; secondly, the complementary perception of pressure, displacement, and vibration realizes the holographic acquisition of the turntable's posture, making up for the blind spot of a single measurement; thirdly, the real-time speed fusion and no-load benchmark comparison adaptively eliminate the random errors introduced by changes in operating conditions; finally, feature extraction realizes the fusion of heterogeneous information at the semantic level, which not only simplifies the deviation discrimination model, but also provides a bridge for operating condition migration.

[0091] For example, taking a certain injection molding workshop as an example, by arranging 64 pressure sensors, 32 displacement sensors and 4 vibration sensors on the turntable of a 32-cavity injection molding machine, a wealth of real-time motion data of the turntable was obtained. After data preprocessing and feature extraction, it was found that the mean pressure and peak-to-peak displacement at the 1# cavity were 12% and 9% higher than the no-load benchmark. Combined with the speed data, it was determined that there was a position deviation of 0.3mm. Source tracing analysis found that the deviation was caused by the wear of the fixed template slider. By replacing the slider and compensating the position of the rear template, the hidden danger was successfully eliminated. This case fully demonstrates the effectiveness of step S2000 in the accurate perception and cause diagnosis of position deviations. The reasonable layout of sensors, the optimal selection of features and the robust design of algorithms are the key to ensuring the actual application effect of this step.

[0092] In general, step S2000 realizes real-time monitoring and early warning of the position deviation of the turntable of the injection molding machine by integrating the perception of multi-source information such as real-time pressure, displacement, vibration, etc., combined with the speed data and no-load feature comparison. The feature-driven deviation measurement avoids the subjectivity of threshold selection, and the deviation traceability with partition refinement also lays a data foundation for accurate diagnosis. In the future, with the development of technologies such as intelligent sensing and edge computing, online monitoring of turntable position deviation is expected to further realize self-optimization of sensor configuration and self-learning of data analysis, and promote the digitalization and intelligent upgrading of injection molding workshops.

[0093] Example 2 This embodiment provides a method for real-time monitoring of position deviation of a turntable of a multi-color injection molding machine based on the first embodiment, including: Step S2300, judging whether the turntable has position deviation according to the real-time motion feature vector M1 and the no-load motion feature vector P0; Furthermore, if Figure 6 As shown, step S2300 includes: Step S2340, calculate the real-time pressure feature vector and the no-load pressure eigenvector The pressure characteristic distance d p , if d p >B p , the pressure characteristic is determined to be abnormal, and the pressure abnormality flag bit f is recorded p =1; otherwise, f p =0; B p is a preset pressure threshold; Step S2350, calculate the real-time displacement feature vector and the no-load displacement eigenvector The displacement characteristic distance d d , if d d >B d , then the displacement characteristic is determined to be abnormal, and the displacement abnormal flag bit f is recorded d =1; otherwise, f d =0; B d is the preset displacement threshold; Step S2360, calculating the real-time radial vibration eigenvector and the no-load radial vibration eigenvector The radial vibration characteristic distance d vr , if d vr >B vr , then the radial vibration characteristic is determined to be abnormal, and the radial vibration abnormal flag bit f is recorded vr =1; otherwise, f vr =0; B vr is a preset radial vibration threshold; Step S2370, calculate the real-time axial vibration characteristic vector and the no-load axial vibration eigenvector The axial vibration characteristic distance d va , if d va >B va , then the axial vibration characteristic is determined to be abnormal, and the axial vibration abnormal flag position f is recorded va =1; otherwise, f va =0; B vais the preset axial vibration threshold; Step S2380, calculating the speed deviation between the real-time speed N1 of the turntable and the rated speed N0 , if |Δn|>B n , the speed is determined to be abnormal, and the abnormal speed flag bit f is recorded n =1; otherwise, f n =0; B n is a preset speed deviation threshold; Step S2390, calculate the total number of abnormal flag bits f=f p +f d +f vr +f va +f n , set the position deviation trigger threshold K1, when f ≥ K1, it is determined that the turntable has a position deviation.

[0094] Specifically, the pressure characteristic distance d p It reflects the degree of deviation of the force distribution of the mold cavity relative to the no-load condition during the injection molding process. Abnormal pressure often indicates defects such as gate blockage and cavity expansion and cracking, which is a key factor affecting product quality. Therefore, the pressure characteristic distance can be used as an important indicator to evaluate the stability of injection molding. Displacement characteristic distance d d It measures the deviation of the normal motion trajectory of the mold cavity relative to the no-load reference. Excessive displacement deviation is often caused by mold wear, increased clearance between guide pins and guide sleeves, etc., which can easily lead to product size deviation. Radial vibration characteristic distance d vr and axial vibration characteristic distance d va The deterioration of the turntable motion posture relative to the ideal state is characterized in both horizontal and vertical directions. Increased vibration will accelerate the fatigue failure of mechanical components and reduce positioning repeatability. The introduction of speed deviation Δn takes into account the impact of production beat fluctuations on deviation monitoring.

[0095] The advantages of thresholding each component distance separately are: More accurate fault tracing. By independently setting thresholds for pressure, displacement, vibration, etc., position deviation can be deconstructed into multiple possible fault modes, such as "pressure abnormality dominated", "displacement abnormality dominated", etc., thereby achieving refined fault tracing and cause diagnosis, which is conducive to the formulation of targeted correction measures.

[0096] The threshold can be adjusted independently. Different physical quantities have different sensitivities to position deviation and different contributions. Independent thresholds allow different physical quantities to be given different weights based on actual engineering experience, and the trigger conditions for suspected deviations can be flexibly set, which not only increases the tolerance for false alarms, but also reduces the risk of missed detections.

[0097] The algorithm is more flexible. The idea of ​​separate judgment can be implemented in serial or parallel according to the needs of the site. Taking the pressure characteristic distance as an example, synchronous judgment can be made at the end of each sampling cycle, or asynchronous judgment can be made after continuous accumulation of multiple sampling cycles to achieve dynamic configuration of monitoring granularity. The comprehensive distance law needs to wait for the data of each physical quantity to be ready at the same time before subsequent calculations can be carried out, and its real-time performance and flexibility are poor.

[0098] Adapt to functional degradation conditions. In actual use, the injection molding machine may experience functional degradation such as partial sensor damage and signal line failure, which may cause some physical quantity data to be unavailable or seriously distorted. In this case, the separate determination method can still make effective decisions based on the available physical quantities, while the comprehensive distance law may be completely ineffective. The separate determination method has stronger fault tolerance and higher system robustness.

[0099] Example 3 This embodiment provides a real-time monitoring system for the position deviation of a turntable of a multi-color injection molding machine based on the embodiment 1. Figure 7 As shown, including: Turntable partitioning module: used to obtain the design parameters of the turntable of the multi-color injection molding machine, build a three-dimensional digital model of the turntable, and divide the turntable into M sector-shaped monitoring areas based on the three-dimensional digital model of the turntable. The design parameters include the number of mold cavities. The number of mold cavities in the M sector-shaped monitoring areas is the same, and M is a positive integer. No-load data acquisition module: used to collect no-load motion data of the turntable under no-load working conditions, and generate a no-load motion characteristic vector P0 reflecting the inherent characteristics of the turntable; the no-load motion data includes no-load pressure data, no-load displacement data and no-load vibration data; Real-time data acquisition module: used to collect the real-time motion data of the turntable during the injection molding production process, and generate the real-time motion feature vector M1 based on the real-time motion data of the turntable and the real-time rotation speed N1 of the turntable; Position deviation judgment module: used to judge whether the turntable has position deviation based on the real-time motion feature vector M1 and the no-load motion feature vector P0; if it is determined that the turntable has position deviation, the deviation cause diagnosis process is triggered.

[0100] In the no-load data acquisition module, the no-load motion data of the turntable under no-load working condition is collected to generate the no-load motion feature vector P0 reflecting the inherent characteristics of the turntable, including: Step S1210: deploy T p A pressure sensor and T d A displacement sensor; Step S1220, install T in the radial direction of the turntable r A vibration sensor is installed on the axial direction of the turntable. a A vibration sensor; Step S1230: When the turntable is running without load, synchronously collect the output signals of the pressure sensor, displacement sensor and vibration sensor to obtain M×T p No-load pressure data, M×T d No-load displacement data and T r +T a No-load vibration data; the no-load vibration data includes no-load radial vibration data and no-load axial vibration data; Step S1240: extract the time-frequency domain features of the no-load pressure data, no-load displacement data, and no-load vibration data to obtain a dimension of D=M×(T p +T d )+T r +T a The no-load motion characteristic vector ;in, is the no-load pressure eigenvector, is the no-load displacement eigenvector, is the no-load radial vibration eigenvector, is the no-load axial vibration eigenvector.

[0101] In the real-time data acquisition module, the real-time motion data includes M×T p dimensional real-time pressure data matrix P1, M×T d Dimensional real-time displacement data matrix D1, T r Real-time radial vibration vector V 1r and T a Dimensional real-time axial vibration vector V 1a ; The real-time motion data of the turntable collected during the injection molding production process includes: Step S2110, obtaining the real-time pressure data of the mold cavity in each sector monitoring area, forming M×T p Dimensional real-time pressure data matrix P1; Step S2120, obtaining the real-time displacement data of the mold cavity in each sector monitoring area, forming M×T d Dimensional real-time displacement data matrix D1; Step S2130, obtaining the real-time radial vibration data of the turntable to form T r Real-time radial vibration vector V 1r ; Step S2140, obtaining real-time axial vibration data of the turntable to form T a Dimensional real-time axial vibration vector V 1a .

[0102] In the real-time data acquisition module, the generation of the real-time motion feature vector M1 includes: the real-time pressure data matrix P1, the real-time displacement data matrix D1, the real-time radial vibration vector V 1r and real-time axial vibration vector V 1a Perform time-frequency domain feature extraction to obtain real-time motion feature vectors ;in, is the real-time pressure feature vector, is the real-time displacement feature vector, is the real-time radial vibration eigenvector, is the real-time axial vibration eigenvector.

[0103] In the position deviation judgment module, judging whether the turntable has position deviation according to the real-time motion feature vector M1 and the no-load motion feature vector P0 includes: Step S2310, calculating the distance E1 between the real-time motion feature vector M1 and the idle motion feature vector P0; Step S2320, preset the deviation threshold T1, if E1≤T1, the turntable position accuracy meets the requirement, and return to step S2100 to continue monitoring; In step S2330, if E1>T1, it is determined that position deviation has occurred, triggering the deviation cause diagnosis process of step S2400.

[0104] The calculation of the distance E1 between the real-time motion feature vector M1 and the no-load motion feature vector P0 includes: According to the real-time pressure feature vector and the no-load pressure eigenvector Calculate pressure characteristic distance; According to the real-time displacement feature vector and the no-load displacement eigenvector Calculate the displacement characteristic distance; According to the real-time radial vibration eigenvector and the no-load radial vibration eigenvector Calculate radial vibration characteristic distance; According to the real-time axial vibration eigenvector and the no-load axial vibration eigenvector Calculate the axial vibration characteristic distance; Calculate the speed deviation between the real-time speed N1 of the turntable and the rated speed; The distance E1 is obtained by weighted summation of the pressure characteristic distance, the displacement characteristic distance, the radial vibration characteristic distance, the axial vibration characteristic distance and the rotational speed deviation.

[0105] In the position deviation judgment module, the deviation cause diagnosis process includes: Step S2410, collecting motion monitoring time series data within a normalized time window TW to form a sample, and summarizing all continuously collected samples to form a sample set {Q}; Step S2420, extracting high-dimensional time-frequency features of each sample in the sample set {Q} to form a high-dimensional feature set {F}; Step S2430, divide the high-dimensional feature set {F} into K feature subsets {S1}, {S2}, ..., {S K}, each feature subset corresponds to a typical position deviation pattern; Step S2440, for each feature subset {S j} Calculate its sample mean to form the mean vector L of the jth deviation mode j ; For each feature subset {S j} Calculate its sample variance to form the variance vector C of the j-th deviation mode j ; Step S2450, according to the mean vector L j and variance vector C j , construct the deviation pattern feature library {(L1,C1),(L2,C2),...,(L K ,C K )}, each deviation pattern in the deviation pattern feature library (L j ,C j ) corresponds to a known position deviation cause R j , forming a mapping set of deviation patterns and position deviation causes ; 1≤j≤K; Step S2460, when a new sample is obtained, the similarity between the new sample and each deviation pattern is calculated, and the position deviation cause corresponding to the deviation pattern with the greatest similarity is diagnosed as the most likely cause of the current deviation.

[0106] Example 4 This embodiment discloses an electronic device, which may include one or more processors and one or more memories, wherein the memories store computer-readable codes, which, when executed by the one or more processors, may execute the above-mentioned method for real-time monitoring of position deviation of a turntable of a multi-color injection molding machine.

[0107] The method or system according to the embodiment of the present application can also be implemented with the help of the architecture of an electronic device. The electronic device may include a bus, one or more CPUs, a read-only memory (ROM), a random access memory (RAM), a communication port connected to a network, an input / output component, a hard disk, etc. A storage device in the electronic device, such as a ROM or a hard disk, can store the real-time monitoring method for the position deviation of the turntable of a multi-color injection molding machine provided in the present application. A real-time monitoring method for the position deviation of a turntable of a multi-color injection molding machine may, for example, include: obtaining the design parameters of the turntable of the multi-color injection molding machine, constructing a three-dimensional digital model of the turntable, dividing the turntable into M sector-shaped monitoring areas based on the three-dimensional digital model of the turntable, collecting no-load motion data of the turntable under no-load conditions, and generating a no-load motion characteristic vector P0 reflecting the inherent characteristics of the turntable; the no-load motion data includes no-load pressure data, no-load displacement data, and no-load vibration data; the design parameters include the number of mold cavities, the number of mold cavities in the M sector-shaped monitoring areas is the same, and M is a positive integer; collecting real-time motion data of the turntable during the injection molding production process, and generating a real-time motion characteristic vector M1 based on the real-time motion data of the turntable and the real-time rotation speed N1 of the turntable; judging whether the turntable has a position deviation based on the real-time motion characteristic vector M1 and the no-load motion characteristic vector P0; if it is determined that the turntable has a position deviation, triggering the deviation cause diagnosis process.

[0108] Furthermore, the electronic device may also include a user interface. Of course, the architecture disclosed in the present invention is only exemplary, and when implementing different devices, one or more components in the electronic device disclosed in the present invention may be omitted according to actual needs.

[0109] Example 5 This embodiment discloses a computer-readable storage medium, on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor, the real-time monitoring method for the position deviation of the turntable of a multi-color injection molding machine according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0110] In addition, according to an embodiment of the present application, the process described above with reference to the flowchart may be implemented as a computer software program. For example, the present application provides a non-temporary machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application, for example: obtaining the design parameters of a turntable of a multi-color injection molding machine, constructing a three-dimensional digital model of the turntable, dividing the turntable into M sector-shaped monitoring areas based on the three-dimensional digital model of the turntable, collecting no-load motion data of the turntable under no-load conditions, and generating a no-load motion feature vector P0 reflecting the inherent characteristics of the turntable; the no-load motion data includes no-load pressure data, no-load displacement data and no-load vibration data; the design parameters include the number of mold cavities, the number of mold cavities in the M sector-shaped monitoring areas is the same, and M is a positive integer; collecting real-time motion data of the turntable during the injection molding production process, and generating a real-time motion feature vector M1 based on the real-time motion data of the turntable and the real-time rotation speed N1 of the turntable; judging whether the turntable has a position deviation based on the real-time motion feature vector M1 and the no-load motion feature vector P0; if it is determined that the turntable has a position deviation, the deviation cause diagnosis process is triggered. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are performed.

[0111] The methods, systems, and devices of the present application may be implemented in many ways. For example, the methods, systems, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers recording media storing programs for executing the method according to the present application.

[0112] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0113] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A real-time monitoring method for position deviation of a turntable of a multi-color injection molding machine, characterized in that: The method comprises: The design parameters of the turntable of the multi-color injection molding machine are obtained, and a three-dimensional digital model of the turntable is constructed. Based on the three-dimensional digital model of the turntable, the turntable is divided into M sector-shaped monitoring areas, and the no-load motion data of the turntable under the no-load condition is collected to generate a no-load motion feature vector P0 reflecting the inherent characteristics of the turntable; the no-load motion data includes no-load pressure data, no-load displacement data and no-load vibration data; the design parameters include the number of mold cavities, the number of mold cavities in the M sector-shaped monitoring areas is the same, and M is a positive integer; The real-time motion data of the turntable during the injection molding production process is collected, and a real-time motion feature vector M1 is generated based on the real-time motion data of the turntable and the real-time rotation speed N1 of the turntable. Based on the real-time motion feature vector M1 and the no-load motion feature vector P0, it is determined whether the turntable has a position deviation. If it is determined that the turntable has a position deviation, the deviation cause diagnosis process is triggered.

2. The method for real-time monitoring of position deviation of a turntable of a multi-color injection molding machine according to claim 1, characterized in that: The collecting of no-load motion data of the turntable under no-load working condition and generating a no-load motion feature vector P0 reflecting the inherent characteristics of the turntable includes: Each sector monitoring area is equipped with T p A pressure sensor and T d A displacement sensor; Install T radially on the turntable r A vibration sensor is installed on the axial direction of the turntable. a A vibration sensor; When the turntable is running without load, the output signals of the pressure sensor, displacement sensor and vibration sensor are collected synchronously to obtain M×T p No-load pressure data, M×T d No-load displacement data and T r +T a No-load vibration data; the no-load vibration data includes no-load radial vibration data and no-load axial vibration data; The time-frequency domain features of no-load pressure data, no-load displacement data and no-load vibration data are extracted, and the dimension is D=M×(T p +T d )+T r +T a The no-load motion characteristic vector ;in, is the no-load pressure eigenvector, is the no-load displacement eigenvector, is the no-load radial vibration eigenvector, is the no-load axial vibration eigenvector.

3. The method for real-time monitoring of position deviation of a turntable of a multi-color injection molding machine according to claim 1, characterized in that: The real-time motion data includes M×T p dimensional real-time pressure data matrix P1, M×T d Dimensional real-time displacement data matrix D1, T r Real-time radial vibration vector V 1r and T a Dimensional real-time axial vibration vector V 1a ; The real-time motion characteristic vector M1 is generated according to the real-time motion data of the turntable and the real-time rotation speed N1 of the turntable, including: the real-time pressure data matrix P1, the real-time displacement data matrix D1, the real-time radial vibration vector V 1r and real-time axial vibration vector V 1a Perform time-frequency domain feature extraction to obtain real-time motion feature vectors ;in, is the real-time pressure feature vector, is the real-time displacement feature vector, is the real-time radial vibration eigenvector, is the real-time axial vibration eigenvector.

4. The method for real-time monitoring of position deviation of a turntable of a multi-color injection molding machine according to claim 1, characterized in that: The step of judging whether the turntable has position deviation according to the real-time motion feature vector M1 and the no-load motion feature vector P0 includes: Calculate the distance E1 between the real-time motion feature vector M1 and the no-load motion feature vector P0; The deviation threshold T1 is preset. If E1≤T1, the turntable position accuracy meets the requirement; if E1>T1, it is determined that a position deviation has occurred, triggering the deviation cause diagnosis process.

5. The method for real-time monitoring of position deviation of a turntable of a multi-color injection molding machine according to claim 4, characterized in that: The calculation of the distance E1 between the real-time motion feature vector M1 and the no-load motion feature vector P0 includes: According to the real-time pressure feature vector and the no-load pressure eigenvector Calculate the pressure characteristic distance d p ; According to the real-time displacement feature vector and the no-load displacement eigenvector Calculate the displacement characteristic distance d d ; According to the real-time radial vibration eigenvector and the no-load radial vibration eigenvector Calculate the radial vibration characteristic distance d vr ; According to the real-time axial vibration eigenvector and the no-load axial vibration eigenvector Calculate the axial vibration characteristic distance d va ; Calculate the speed deviation between the real-time speed N1 of the turntable and the rated speed ; The distance E1 is obtained by weighted summation of the pressure characteristic distance, the displacement characteristic distance, the radial vibration characteristic distance, the axial vibration characteristic distance and the rotational speed deviation.

6. The method for real-time monitoring of position deviation of a turntable of a multi-color injection molding machine according to claim 5, characterized in that: The step of judging whether the turntable has position deviation according to the real-time motion feature vector M1 and the no-load motion feature vector P0 further comprises: If d p >B p , the pressure characteristic is determined to be abnormal, and the pressure abnormality flag bit f is recorded p =1; otherwise, f p =0; B p is a preset pressure threshold; If d d >B d , then the displacement characteristic is determined to be abnormal, and the displacement abnormal flag bit f is recorded d =1; otherwise, f d =0; B d is the preset displacement threshold; If d vr >B vr , then the radial vibration characteristic is determined to be abnormal, and the radial vibration abnormal flag bit f is recorded vr =1; otherwise, f vr =0; B vr is a preset radial vibration threshold; If d va >B va , then the axial vibration characteristic is determined to be abnormal, and the axial vibration abnormal flag position f is recorded va =1; otherwise, f va =0; B va is the preset axial vibration threshold; If |Δn|>B n , the speed is determined to be abnormal, and the abnormal speed flag bit f is recorded n =1; otherwise, f n =0; B n is a preset speed deviation threshold; Calculate the total number of exception flag bits f=f p +f d +f vr +f va +f n , set the position deviation trigger threshold K1, when f ≥ K1, it is determined that the turntable has a position deviation.

7. The method for real-time monitoring of position deviation of a turntable of a multi-color injection molding machine according to claim 1, characterized in that: The deviation cause diagnosis process includes: Collecting motion monitoring time series data within a normalized time window TW to form a sample, and summarizing all continuously collected samples to form a sample set {Q}; the motion monitoring time series data includes pressure data, displacement data, vibration data and rotation speed; Extract high-dimensional time-frequency features of each sample in the sample set {Q} to form a high-dimensional feature set {F}; Divide the high-dimensional feature set {F} into K feature subsets {S1}, {S2}, ..., {S K }, each feature subset corresponds to a typical position deviation pattern; For each feature subset {S j } Calculate its sample mean to form the mean vector L of the jth deviation mode j ; For each feature subset {S j } Calculate its sample variance to form the variance vector C of the j-th deviation mode j ; According to the mean vector L j and variance vector C j , construct the deviation pattern feature library {(L1,C1),(L2,C2),...,(L K ,C K )}, each deviation pattern in the deviation pattern feature library (L j ,C j ) corresponds to a known position deviation cause R j , forming a mapping set of deviation patterns and position deviation causes ; 1≤j≤K; When a new sample is obtained, the similarity between the new sample and each deviation pattern is calculated, and the position deviation cause corresponding to the deviation pattern with the greatest similarity is diagnosed as the most likely cause of the current deviation.

8. A real-time monitoring system for the position deviation of a turntable of a multi-color injection molding machine, which is used to implement the real-time monitoring method for the position deviation of a turntable of a multi-color injection molding machine according to any one of claims 1 to 7, characterized in that: The system comprises: Turntable partitioning module: used to obtain the design parameters of the turntable of the multi-color injection molding machine, build a three-dimensional digital model of the turntable, and divide the turntable into M sector-shaped monitoring areas based on the three-dimensional digital model of the turntable. The design parameters include the number of mold cavities. The number of mold cavities in the M sector-shaped monitoring areas is the same, and M is a positive integer. No-load data acquisition module: used to collect no-load motion data of the turntable under no-load working conditions, and generate a no-load motion feature vector P0 that reflects the inherent characteristics of the turntable; Real-time data acquisition module: used to collect the real-time motion data of the turntable during the injection molding production process, and generate the real-time motion feature vector M1 based on the real-time motion data of the turntable and the real-time rotation speed N1 of the turntable; Position deviation judgment module: used to judge whether the turntable has position deviation based on the real-time motion feature vector M1 and the no-load motion feature vector P0; if it is determined that the turntable has position deviation, the deviation cause diagnosis process is triggered.

9. An electronic device comprising a memory, a central processing unit, and a computer program stored in the memory and executable on the central processing unit, characterized in that: When the central processing unit executes the computer program, the real-time monitoring method for the position deviation of the turntable of a multi-color injection molding machine described in any one of claims 1-7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed, the method for real-time monitoring of position deviation of a turntable of a multi-color injection molding machine according to any one of claims 1 to 7 is implemented.

Images