A PLC control system for centrifuge fault diagnosis

Abstract

The present application relates to the technical field of centrifuge fault diagnosis, and particularly relates to a PLC control system for centrifuge fault diagnosis. The system comprises: a parameter acquisition module that buffers running parameters in real time, and freezes the fault time parameters when a fault is triggered; a fault diagnosis output module that accurately distinguishes cumulative faults and instantaneous faults based on the time sequence distribution characteristics of abnormal events and the parameter degradation trajectory quantitative values; for cumulative faults, an abnormal event screening module screens candidate abnormal events associated with fault parameters; and a fault tracing and positioning module that determines the abnormal transmission direction and energy attenuation gradient based on the characteristic offset vector and the propagation time delay deviation, and realizes accurate positioning of the fault source. The present application realizes accurate differentiation of centrifuge cumulative faults and instantaneous faults and rapid and accurate positioning of the fault source by combining the technical means of parameter buffer freezing, fault type quantitative discrimination, abnormal event correlation screening, and fault source positioning based on the characteristic offset vector and the propagation time delay deviation.

Description

Technical Field

[0001] This invention relates to the field of centrifuge fault diagnosis technology, and in particular to a PLC control system for centrifuge fault diagnosis. Background Technology

[0002] As a core piece of industrial separation equipment, the operational reliability of centrifuges directly impacts production efficiency and safety. Existing centrifuge fault diagnosis systems primarily rely on threshold alarms and simple parameter monitoring. When a fault occurs, they can only indicate out-of-limit information, making it difficult to distinguish between cumulative and transient faults. For cumulative faults, current technology cannot trace the propagation path of abnormal events, making it difficult to accurately locate the source of the fault. This forces maintenance personnel to check each component one by one, resulting in low maintenance efficiency and long downtime. Furthermore, critical operating parameters at the moment of a fault are often overwritten by subsequent data, and the lack of an effective fault data freezing mechanism makes post-fault analysis difficult.

[0003] Chinese Patent Publication No. CN118287272A discloses a centrifuge fault diagnosis device and method. The device includes a microphone, a signal processing module, an edge computing gateway, and a server. The microphone is connected to the centrifuge via a bonding method. The microphone is electrically connected to the signal processing module. The signal processing module is connected to the edge computing gateway via a communication port. The edge computing gateway is connected to the server via wired or wireless means. The method involves the microphone acquiring the acoustic signal emitted by the centrifuge. After processing by the signal processing module, the acoustic signal is transmitted to the edge computing gateway. The edge computing gateway, acting as an intermediate node, further synchronizes the processed signal with the signal transmitted by the centrifuge control system before transmitting them together to the server for fault diagnosis, thus obtaining the diagnostic result. The beneficial effects of this invention are: it eliminates the need for exhaustive fault samples and complex equipment repair, thus simply and effectively improving the accuracy of centrifuge fault diagnosis.

[0004] However, this existing technology has the following problems:

[0005] Relying solely on acoustic signals for fault diagnosis lacks comprehensive acquisition of multi-source operating parameters and time-series analysis of abnormal events. It cannot distinguish between cumulative and instantaneous faults, nor can it trace the fault propagation path or accurately locate the fault source. Summary of the Invention

[0006] Therefore, the present invention provides a PLC control system for centrifuge fault diagnosis, which overcomes the problems of existing technologies that rely solely on acoustic signals for fault diagnosis, lack comprehensive acquisition of multi-source operating parameters and time sequence analysis of abnormal events, cannot distinguish between cumulative faults and instantaneous faults, and are difficult to trace the fault propagation path and accurately locate the fault source.

[0007] To achieve the above objectives, the present invention provides a PLC control system for centrifuge fault diagnosis, comprising:

[0008] The parameter acquisition module collects and caches the centrifuge's operating parameters in real time;

[0009] An anomaly detection module is used to monitor the operating parameters in real time, generate anomaly event records based on the operating parameters, and mark the occurrence time and recovery time of the anomaly events.

[0010] The fault triggering module is used to generate a fault triggering signal when a centrifuge malfunction is detected and the centrifuge stops.

[0011] The parameter freezing and storage module, in response to the fault trigger signal, is used to freeze the running parameters cached in the parameter acquisition module at the time of the fault occurrence and store them as fault shutdown parameters.

[0012] The fault diagnosis output module is used to determine the time-series distribution feature quantization value and parameter degradation trajectory quantization value of the abnormal events based on the abnormal event records and operating parameters within the time window from the normal operating state to the time of the fault occurrence, and to determine the fault type of the centrifuge based on the time-series distribution feature quantization value and the parameter degradation trajectory quantization value, wherein the fault type includes cumulative fault type and instantaneous fault type.

[0013] The abnormal event filtering module, in response to the centrifuge's fault type being a cumulative fault type, extracts the fault feature parameters from the fault shutdown parameters, and filters the abnormal event records within the time window based on the fault feature parameters, filtering out candidate abnormal events that have parameter correlation with the fault shutdown parameters.

[0014] The fault tracing and localization module is used to determine the propagation direction and energy attenuation gradient of the candidate abnormal events based on the feature offset vector of the candidate abnormal events and the propagation delay deviation between the candidate abnormal events, and to locate the fault source based on the propagation direction and energy attenuation gradient of the candidate abnormal events.

[0015] Furthermore, the fault diagnosis output module determines the quantized values ​​of the temporal distribution characteristics of the abnormal event and the quantized values ​​of the parameter degradation trajectory, wherein,

[0016] The fault diagnosis output module arranges the abnormal events within the time window in a temporal sequence, calculates the rate of change of the time interval between adjacent abnormal events, and obtains the quantified value of the temporal distribution characteristics.

[0017] The fault diagnosis output module extracts the operating parameters of the same parameter category from the abnormal event records, constructs a parameter value sequence according to the time sequence of the abnormal occurrence, fits the parameter value sequence using the curve fitting method, calculates the slope of the fitting curve, and obtains the quantified value of the parameter degradation trajectory.

[0018] The same parameter category refers to operating parameters that have the same physical dimensions and the same sensor source.

[0019] Furthermore, the fault diagnosis output module determines the fault type of the centrifuge based on the quantized value of the time-series distribution characteristics and the quantized value of the parameter degradation trajectory, wherein,

[0020] The fault diagnosis output module calculates the fault type discrimination coefficient based on the quantized value of the temporal distribution feature and the quantized value of the parameter degradation trajectory.

[0021] If the fault type discrimination coefficient is greater than or equal to the preset discrimination coefficient, the fault diagnosis output module determines that the fault type of the centrifuge is a cumulative fault type;

[0022] If the fault type discrimination coefficient is less than the preset discrimination coefficient, the fault diagnosis output module determines that the fault type of the centrifuge is an instantaneous fault type.

[0023] Furthermore, the abnormal event filtering module filters out candidate abnormal events that have a parameter correlation with the fault shutdown parameters, wherein,

[0024] The abnormal event screening module extracts the fault feature parameters from the fault shutdown parameters, and calculates the parameter correlation degree between the operating parameters corresponding to each abnormal event within the time window and the fault feature parameters. The parameter correlation degree is obtained by weighted calculation based on parameter waveform similarity and parameter change synchronicity.

[0025] The abnormal event filtering module marks abnormal events whose parameter correlation degree is greater than a preset parameter correlation degree threshold as candidate abnormal events.

[0026] Furthermore, the parameter waveform similarity is determined by calculating the dynamic time warping distance between the operating parameter sequence corresponding to the abnormal event and the fault characteristic parameter sequence, and the parameter change synchronicity is determined by calculating the time offset between the time point of the abnormal event occurrence and the time point of the sudden change in the fault characteristic parameter.

[0027] Furthermore, the fault tracing and localization module determines the propagation direction and energy attenuation gradient of the candidate abnormal event, wherein,

[0028] The fault tracing and localization module determines the feature correlation between candidate abnormal events based on the feature offset vectors of each candidate abnormal event, and constructs the transmission path of the candidate abnormal events according to the feature correlation.

[0029] The fault tracing and localization module determines the transmission direction of the candidate abnormal event along the transmission path based on the change characteristics of the feature offset vector between adjacent candidate abnormal events in the transmission path.

[0030] The fault tracing and localization module calculates the energy attenuation gradient of the candidate abnormal event along the transmission path based on the temporal relationship between adjacent candidate abnormal events in the transmission path and the change in the feature offset vector.

[0031] The feature offset vector is the difference vector between the running parameter value at the time the candidate abnormal event occurs and the preset benchmark value.

[0032] Furthermore, the fault tracing and location module determines the propagation direction of the candidate abnormal event along the propagation path, wherein,

[0033] The fault tracing and localization module calculates the angle between the feature offset directions of the feature offset vectors of two adjacent candidate abnormal events in the transmission path.

[0034] If the included angle of the feature offset direction is less than the preset included angle threshold, it is determined that there is a feature correlation between the two adjacent candidate abnormal events, and the transmission direction is determined according to the sign of the included angle of the feature offset direction.

[0035] Specifically, when the angle between the feature offset directions is positive, it indicates that the feature offset directions tend to be consistent, and the transmission direction is from candidate anomalies with smaller feature offset vector magnitudes to candidate anomalies with larger feature offset vector magnitudes; when the angle between the feature offset directions is negative, it indicates that the feature offset directions tend to be opposite, and the transmission direction is from candidate anomalies with larger feature offset vector magnitudes to candidate anomalies with smaller feature offset vector magnitudes.

[0036] Furthermore, the fault tracing and localization module calculates the energy attenuation gradient of the candidate abnormal event along the transmission path, wherein,

[0037] The fault tracing and localization module obtains the propagation delay deviation between two adjacent candidate abnormal events in the transmission path, and the propagation delay deviation is the difference between the actual propagation delay and the theoretical propagation delay;

[0038] The fault tracing and localization module calculates the magnitude change of the feature offset vectors of two adjacent candidate abnormal events, and determines the energy attenuation gradient between the two adjacent candidate abnormal events based on the ratio of the magnitude change to the propagation delay deviation.

[0039] Furthermore, the fault source tracing and localization module locates the fault source based on the propagation direction and energy attenuation gradient of the candidate abnormal events, wherein,

[0040] The fault tracing and localization module determines the upstream and downstream relationship of each candidate abnormal event in the transmission path according to the transmission direction, and marks the candidate abnormal event located at the upstream of the transmission path as a potential fault source node.

[0041] The fault tracing and localization module calculates the energy attenuation gradient of the potential fault source node as it propagates along the transmission path to the adjacent downstream node.

[0042] If the energy decay gradient is less than a preset decay threshold, then the potential fault source node is determined as the initial fault source.

[0043] If the energy attenuation gradient is greater than or equal to the preset attenuation threshold, the fault tracing and location module traces back upstream along the transmission path and determines the node corresponding to the first time the energy attenuation gradient is less than the preset attenuation threshold as the initial fault source.

[0044] The fault tracing and location module outputs the device component associated with the initial fault source as the fault source.

[0045] Furthermore, the anomaly detection module includes a threshold judgment unit and a rate of change detection unit;

[0046] The threshold judgment unit is used to compare each operating parameter with the corresponding preset threshold range. When any operating parameter exceeds the preset threshold range, a first type of abnormal event record is generated.

[0047] The rate of change detection unit is used to calculate the rate of change of each operating parameter within a unit of time. When the rate of change exceeds a preset rate of change threshold, a second type of abnormal event record is generated.

[0048] The anomaly detection module is also used to merge the corresponding first-type anomaly event record and the second-type anomaly event record into a composite anomaly event record when the same operating parameter triggers the threshold judgment unit and the rate of change detection unit at the same time.

[0049] The first type of abnormal event record, the second type of abnormal event record, and the composite abnormal event record together constitute the abnormal event record.

[0050] Compared with the prior art, the beneficial effects of the present invention are that, based on the time interval change rate of abnormal events, the present invention quantifies the temporal distribution characteristics, which can accurately identify the frequency trend and evolution rhythm of abnormal events. Combined with the fitting of parameter degradation trajectory of the same parameter category, the long-term deterioration trend of parameters is quantified by the curve slope, which realizes the accurate distinction between cumulative faults and instantaneous faults, effectively improving the accuracy and intelligence level of fault diagnosis.

[0051] Furthermore, this invention employs a weighted fusion method to comprehensively calculate the quantized values ​​of temporal distribution features and parameter degradation trajectory into a fault type discrimination coefficient, and compares it with a preset discrimination coefficient. This achieves accurate differentiation between cumulative faults and instantaneous faults, solving the problem that existing technologies struggle to distinguish fault evolution types. It provides a reliable classification basis for subsequent accurate source tracing of cumulative faults, effectively improving the accuracy and intelligence level of fault diagnosis.

[0052] Furthermore, the present invention, through an abnormal event screening module, after determining that a cumulative fault has occurred, extracts fault feature parameters from the fault shutdown parameters and calculates the parameter correlation degree based on the parameter waveform similarity and parameter change synchronicity. Abnormal events with parameter correlation degrees greater than a preset parameter correlation degree threshold are screened as candidate abnormal events, effectively eliminating redundant abnormal events unrelated to the fault. This provides a concise and highly relevant set of candidate abnormal events for subsequent fault tracing, improving the efficiency and accuracy of fault tracing.

[0053] Furthermore, this invention constructs feature associations and transmission paths based on the feature offset vectors of candidate abnormal events through a fault source tracing and localization module. It also uses the angle between the feature offset directions to determine the transmission direction and calculates the energy attenuation gradient using the ratio of the magnitude change to the propagation delay deviation. This achieves accurate identification of the propagation direction of abnormal events and quantitative characterization of the degree of energy attenuation, providing a reliable quantitative basis for fault source localization and improving the accuracy and efficiency of tracing and diagnosing accumulated faults.

[0054] Furthermore, this invention constructs a transmission path topology based on the transmission direction and determines the upstream and downstream relationships through a fault source tracing and localization module. The upstream node is marked as a potential fault source. The fault source is accurately determined by comparing the energy attenuation gradient with a preset attenuation threshold. When the gradient exceeds the threshold, the system backtracks upstream until the node where the attenuation gradient first reaches the threshold is found. This achieves the technical effect of accurately identifying the fault origin from the abnormal event propagation path, significantly shortening the centrifuge fault troubleshooting time and improving maintenance efficiency and equipment availability. Attached Figure Description

[0055] Figure 1 This is a schematic diagram of the PLC control system used for centrifuge fault diagnosis in this embodiment;

[0056] Figure 2This is a flowchart illustrating the process of determining the fault type in the PLC control system used for centrifuge fault diagnosis in this embodiment;

[0057] Figure 3 This is a schematic diagram of the anomaly detection module in the PLC control system used for centrifuge fault diagnosis in this embodiment. Detailed Implementation

[0058] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0059] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0060] Please see Figure 1 As shown, Figure 1 This is a schematic diagram of the PLC control system used for centrifuge fault diagnosis in this embodiment.

[0061] This embodiment provides a PLC control system for centrifuge fault diagnosis, including:

[0062] The parameter acquisition module collects and caches the centrifuge's operating parameters in real time;

[0063] An anomaly detection module, connected to the parameter acquisition module, is used to monitor the operating parameters in real time, generate anomaly event records based on the operating parameters, and mark the occurrence time and recovery time of the anomaly events.

[0064] The fault triggering module, which is connected to the parameter acquisition module and the anomaly detection module, is used to generate a fault triggering signal when a centrifuge malfunctions and stops.

[0065] A parameter freezing and storage module, which is connected to the parameter acquisition module and the fault triggering module, responds to the fault triggering signal and freezes the running parameters cached in the parameter acquisition module at the time of the fault occurrence, and stores them as fault shutdown parameters.

[0066] The fault diagnosis output module, which is connected to the anomaly detection module and the parameter freezing and storage module, is used to determine the time-series distribution characteristic quantization value and parameter degradation trajectory quantization value of the abnormal event based on the abnormal event records and operating parameters within the time window from the normal operating state to the time point of the fault occurrence, and to determine the fault type of the centrifuge based on the time-series distribution characteristic quantization value and the parameter degradation trajectory quantization value, wherein the fault type includes cumulative fault type and instantaneous fault type;

[0067] An abnormal event filtering module, which is connected to the fault diagnosis output module, the abnormal detection module and the parameter freezing and storage module, extracts the fault feature parameters from the fault shutdown parameters in response to the centrifuge's fault type being a cumulative fault type, and filters the abnormal event records within the time window based on the fault feature parameters to filter out candidate abnormal events that have parameter correlation with the fault shutdown parameters.

[0068] The fault tracing and localization module, which is connected to the abnormal event filtering module, is used to determine the propagation direction and energy attenuation gradient of the candidate abnormal events based on the feature offset vector of the candidate abnormal events and the propagation delay deviation between the candidate abnormal events, and to locate the fault source based on the propagation direction and energy attenuation gradient of the candidate abnormal events.

[0069] In this embodiment of the invention, the operating parameters include, but are not limited to, the vibration amplitude and frequency of the centrifuge shaft, bearing temperature, main motor current, drum speed, feed pressure, discharge temperature, lubricating oil pressure and temperature, and casing vibration acceleration. For transient fault types, the transient parameter change characteristics before and after the fault occur are recorded, a transient fault event log is generated, and pushed to the human-machine interface for display, so that maintenance personnel can quickly locate the source of occasional disturbance or external interference factors.

[0070] In this embodiment of the invention, the parameter acquisition module includes multiple sensor interface units, a cache unit, and a timestamp marking unit. The multiple sensor interface units are respectively connected to vibration sensors, temperature sensors, speed sensors, and current sensors installed at different parts of the centrifuge, and are used to acquire the detection signals of each sensor in real time according to a preset sampling frequency, and convert the detection signals into digital operating parameters. The cache unit adopts a first-in-first-out (FIFO) circular storage method, continuously storing the operating parameters according to a preset cache depth. The preset cache depth covers at least the operating parameters within a first preset time period prior to the current moment, ensuring that historical data within a certain period before the fault occurs can be completely preserved. The timestamp marking unit is used to mark the corresponding acquisition timestamp for each set of acquired operating parameters, and synchronously stores the timestamped operating parameters in the cache unit, while simultaneously transmitting the operating parameters to the anomaly detection module for monitoring and analysis in real time. When the fault triggering module generates a fault trigger signal, the parameter acquisition module responds to the fault trigger signal by stopping the data update operation of the cache unit, locking all operating parameters currently stored in the cache unit to form a snapshot of the operating parameters before the fault time, and waiting for the parameter freezing and storage module to read it.

[0071] In this embodiment of the invention, the first preset duration is determined as follows: historical fault data of the centrifuge is collected, and the time interval from the occurrence of the first abnormal event to the shutdown of the fault is counted before each fault. The upper quartile of each time interval is taken as the first preset duration. When historical fault data is lacking, 0.3 times the theoretical failure period is taken as the first preset duration based on the theoretical failure period of the key components of the centrifuge and the propagation characteristics of the abnormal signal. The value range of the first preset duration is from 10 seconds to 300 seconds.

[0072] Specifically, the fault diagnosis output module determines the quantized values ​​of the temporal distribution characteristics of the abnormal event and the quantized values ​​of the parameter degradation trajectory, wherein,

[0073] The fault diagnosis output module arranges the abnormal events within the time window in a temporal sequence, calculates the rate of change of the time interval between adjacent abnormal events, and obtains the quantified value of the temporal distribution characteristics.

[0074] The fault diagnosis output module extracts the operating parameters of the same parameter category from the abnormal event records, constructs a parameter value sequence according to the time sequence of the abnormal occurrence, fits the parameter value sequence using the curve fitting method, calculates the slope of the fitting curve, and obtains the quantified value of the parameter degradation trajectory.

[0075] The same parameter category refers to operating parameters that have the same physical dimensions and the same sensor source.

[0076] In this embodiment of the invention, the fault diagnosis output module first acquires all abnormal event records within a time window from normal operation to the time of fault occurrence, and arranges the abnormal events chronologically according to the order of their occurrence to form an abnormal event time sequence. For the abnormal event time sequence, the fault diagnosis output module sequentially calculates the time interval between two adjacent abnormal events, and calculates the rate of change between adjacent time intervals based on the changing trend of each time interval, outputting the statistical characteristic value of the rate of change as a time sequence distribution feature quantification value. Specifically, the time sequence distribution feature quantification value is used to characterize the density of abnormal event distribution on the time axis and its changing trend. When the time interval between adjacent abnormal events shows a gradually shortening trend, the rate of change is negative, indicating that the frequency of abnormal event occurrence is accelerating; when the time interval shows a gradually increasing trend, the rate of change is positive, indicating that the frequency of abnormal event occurrence is decreasing; when the time interval fluctuates irregularly, the rate of change oscillates around zero. The fault diagnosis output module performs statistical analysis on the rate of change, extracting at least one of its mean, variance, or cumulative change as a time sequence distribution feature quantification value.

[0077] In this embodiment of the invention, it is assumed that before a centrifuge experiences a cumulative "bearing wear" failure, the anomaly detection module records five "vibration amplitude exceeding limits" anomaly events within a 120-second time window. The specific occurrence times are as follows: 1st anomaly: 0 seconds; 2nd anomaly: 30 seconds; 3rd anomaly: 55 seconds; 4th anomaly: 75 seconds; 5th anomaly: 90 seconds; failure shutdown time: 120 seconds; then the time interval between adjacent anomaly events is calculated: Interval 1: 2nd anomaly time - 1st anomaly time = 30 seconds - 0 seconds = 30 seconds; Interval 2: 3rd anomaly time - 2nd anomaly time = 55 seconds - 30 seconds = 25 seconds; Interval 3: 4th anomaly time - 3rd anomaly time = 75 seconds - 55 seconds = 20 seconds; Interval 4: 5th anomaly time - 4th anomaly time = 90 seconds - 75 seconds = 15 seconds; the resulting time interval sequence is: [30 seconds, 25 seconds, 20 seconds, 15 seconds].

[0078] Calculate the rate of change between adjacent time intervals: Rate of change = (next interval - previous interval) / previous interval. Rate of change 1: (25-30) / 30 = -0.167; Rate of change 2: (20-25) / 25 = -0.200; Rate of change 3: (15-20) / 20 = -0.250; The resulting rate of change sequence is: [-0.167, -0.200, -0.250]. Extract the quantized value of the time-series distribution feature, and take the average of all rates of change to obtain the quantized value of the time-series distribution feature: Quantized value of time-series distribution feature = (-0.167 - 0.200 - 0.250) / 3 = -0.206. This negative value indicates that the interval between abnormal events is constantly shortening, that is, the frequency of abnormal events is accelerating. This is a typical precursor characteristic of accumulated faults approaching the time of fault shutdown. If the value is positive, it indicates that the frequency of abnormalities is slowing down; if it is close to 0, it indicates that the abnormality is irregular and tends to be an instantaneous fault.

[0079] The fault diagnosis output module further extracts operating parameters of the same parameter category from the abnormal event records. The same parameter category refers to operating parameters with the same physical dimensions and originating from the same sensor monitoring location, such as vibration amplitude sequences collected by the same vibration sensor, temperature sequences collected by the same temperature sensor, or current sequences collected by the same current sensor. The fault diagnosis output module constructs a parameter value sequence of the operating parameters of the same parameter category according to the chronological order of the occurrence times of each abnormal event. Each parameter value in the parameter value sequence corresponds to an operating parameter value at the time of an abnormal event. The fault diagnosis output module uses a curve fitting method to fit the parameter value sequence. The curve fitting method includes, but is not limited to, linear regression fitting, polynomial fitting, or exponential fitting. After obtaining the fitted curve using the curve fitting method, the slope of the fitted curve within a preset interval is calculated, and the slope is output as the quantized value of the parameter degradation trajectory. The preset interval is determined based on the temporal distribution of abnormal events within the time window. Specifically, it constructs a fitting interval for the parameter degradation trajectory by taking the earliest occurrence time of the abnormal event within the time window as the starting point and the latest occurrence time of the abnormal event or the fault occurrence time as the ending point. In a preferred embodiment, when there is an unrecovered persistent anomaly in the abnormal event record, the preset interval is taken as the starting point of the persistent anomaly and the fault occurrence time point as the ending point.

[0080] This invention quantifies the temporal distribution characteristics based on the rate of change of time intervals of abnormal events, which can accurately identify the frequency trend and evolution rhythm of abnormal events. Combined with the fitting of parameter degradation trajectory of the same parameter category, the long-term deterioration trend of parameters is quantified by the curve slope, which realizes the accurate distinction between cumulative faults and instantaneous faults, effectively improving the accuracy and intelligence level of fault diagnosis.

[0081] Please see Figure 2 As shown, Figure 2 This is a flowchart illustrating the process of determining the fault type in the PLC control system used for centrifuge fault diagnosis in this embodiment.

[0082] Specifically, the fault diagnosis output module determines the fault type of the centrifuge based on the quantized value of the time-series distribution characteristics and the quantized value of the parameter degradation trajectory, wherein,

[0083] The fault diagnosis output module calculates the fault type discrimination coefficient based on the quantized value of the temporal distribution feature and the quantized value of the parameter degradation trajectory.

[0084] If the fault type discrimination coefficient is greater than or equal to the preset discrimination coefficient, the fault diagnosis output module determines that the fault type of the centrifuge is a cumulative fault type;

[0085] If the fault type discrimination coefficient is less than the preset discrimination coefficient, the fault diagnosis output module determines that the fault type of the centrifuge is an instantaneous fault type.

[0086] In this embodiment of the invention, the fault diagnosis output module calculates the fault type discrimination coefficient using a weighted fusion method. The quantized value of the temporal distribution feature corresponds to a first weight coefficient, and the quantized value of the parameter degradation trajectory corresponds to a second weight coefficient. The sum of the first and second weight coefficients is 1. A larger quantized value of the temporal distribution feature indicates a more pronounced accelerating upward trend in the frequency of abnormal events; a larger absolute value of the quantized value of the parameter degradation trajectory indicates a more significant monotonic degradation trend in the operating parameters of the same parameter category. When both are at a high level, the fault type discrimination coefficient approaches its upper limit, indicating that the fault has typical cumulative characteristics. The first and second weighting coefficients can be determined based on historical data statistical analysis, including: collecting historical cumulative fault samples and instantaneous fault samples of centrifuges, calculating the time-series distribution characteristic quantification value and parameter degradation trajectory quantification value of each sample respectively; using principal component analysis or factor analysis to calculate the contribution of the two feature quantities in distinguishing the two types of faults, and normalizing the contribution values ​​as weighting coefficients; or using a logistic regression model to train historical samples, and normalizing the regression coefficients corresponding to the two feature quantities after the model training converges to obtain the first and second weighting coefficients.

[0087] In this embodiment of the invention, the preset discrimination coefficient is a pre-calibrated classification threshold, which is calibrated by: determining the optimal classification boundary through classification model training based on historical fault data of the centrifuge; or setting it after theoretical calculation and on-site debugging according to the equipment type, operating conditions, and sensor configuration parameters of the centrifuge. When the fault type discrimination coefficient is greater than or equal to the preset discrimination coefficient, the fault diagnosis output module determines that the current fault is a cumulative fault type. At this time, the formation of the fault has undergone a gradual evolution process from quantitative change to qualitative change, and there is a traceable abnormal event propagation chain. It is necessary to further call the abnormal event screening module to perform fault source location analysis. When the fault type discrimination coefficient is less than the preset discrimination coefficient, the fault diagnosis output module determines that the current fault is an instantaneous fault type. At this time, the fault is directly caused by a sudden event, and there is no need to perform fault propagation path analysis. The instantaneous fault alarm information and the corresponding over-limit parameters are directly output.

[0088] This invention uses a weighted fusion method to comprehensively calculate the quantized values ​​of temporal distribution features and parameter degradation trajectory into a fault type discrimination coefficient, and compares it with a preset discrimination coefficient. This enables accurate differentiation between cumulative faults and instantaneous faults, solves the problem that existing technologies have difficulty in distinguishing fault evolution types, provides a reliable classification basis for subsequent accurate source tracing of cumulative faults, and effectively improves the accuracy and intelligence level of fault diagnosis.

[0089] Specifically, the abnormal event filtering module filters out candidate abnormal events that are correlated with the fault shutdown parameters, wherein,

[0090] The abnormal event screening module extracts the fault feature parameters from the fault shutdown parameters, and calculates the parameter correlation degree between the operating parameters corresponding to each abnormal event within the time window and the fault feature parameters. The parameter correlation degree is obtained by weighted calculation based on parameter waveform similarity and parameter change synchronicity.

[0091] The abnormal event filtering module marks abnormal events whose parameter correlation degree is greater than a preset parameter correlation degree threshold as candidate abnormal events.

[0092] Specifically, the parameter waveform similarity is determined by calculating the dynamic time warping distance between the operating parameter sequence corresponding to the abnormal event and the fault characteristic parameter sequence, and the parameter change synchronicity is determined by calculating the time offset between the time point of the abnormal event and the time point of the sudden change of the fault characteristic parameter.

[0093] In this embodiment of the invention, after receiving a signal from the fault diagnosis output module indicating a cumulative fault type, the abnormal event filtering module first extracts fault feature parameters from the fault shutdown parameters stored in the parameter freezing and storage module. The fault feature parameters are at least one operating parameter with the largest deviation between the fault occurrence time and the normal operating state baseline value, or at least one operating parameter with the largest deviation from a preset threshold range. The abnormal event filtering module traverses all abnormal event records within the time window and, for each abnormal event, obtains the corresponding operating parameter sequence. The operating parameter sequence is the continuous operating parameter values ​​collected from the time of the abnormal event to the time of the abnormal recovery during the duration of the abnormal event. Simultaneously, the abnormal event filtering module extracts a fault feature parameter sequence of the same parameter category as the abnormal event from the fault shutdown parameters. This fault feature parameter sequence is the continuous operating parameter values ​​within a second preset time period before the time of the fault occurrence.

[0094] In this embodiment of the invention, the abnormal event screening module calculates the parameter waveform similarity between the operating parameter sequence and the fault feature parameter sequence. Specifically, a dynamic time warping algorithm is used to calculate the minimum distance between the two sequences; the smaller the minimum distance, the higher the waveform similarity. Alternatively, the Pearson correlation coefficient is used to calculate the linear correlation between the two sequences; the closer the correlation coefficient is to 1, the higher the waveform similarity. The abnormal event screening module maps the minimum distance or correlation coefficient to a waveform similarity quantification value. The abnormal event screening module also calculates the parameter change synchronicity between the abnormal event and the fault shutdown parameters. Specifically, the time offset between the occurrence time of the abnormal event and the abrupt change time of the fault feature parameter is obtained, where the abrupt change time is the starting moment when the rate of change in the fault feature parameter sequence exceeds a preset abrupt change threshold. The smaller the absolute value of the time offset, the higher the parameter change synchronicity. The abnormal event screening module maps the time offset to a synchronicity quantification value, where the synchronicity quantification value is at its maximum when the time offset is zero, and the synchronicity quantification value monotonically decreases as the absolute value of the time offset increases.

[0095] In this embodiment of the invention, the abnormal event screening module calculates the parameter correlation degree using a weighted fusion method. The parameter waveform similarity corresponds to a third weight coefficient, and the parameter change synchronicity corresponds to a fourth weight coefficient. The sum of the third and fourth weight coefficients is 1. The abnormal event screening module compares the calculated parameter correlation degree with a preset parameter correlation degree threshold. When the parameter correlation degree is greater than the preset threshold, the abnormal event is marked as a candidate abnormal event. All the selected candidate abnormal events constitute a candidate abnormal event set for subsequent analysis by the fault tracing and localization module.

[0096] In this embodiment of the invention, the normal operating state benchmark value is determined by collecting the operating parameters of the centrifuge under stable operating conditions within a preset calibration period and calculating the average value of each operating parameter as the normal operating state benchmark value; or it can be directly set according to the centrifuge's factory calibration parameters. The preset threshold range is determined based on the normal operating state benchmark value combined with the centrifuge's allowable fluctuation range, taking ± a first percentage of the benchmark value as the preset threshold range. The second preset duration is determined based on the average evolution cycle of a centrifuge abnormal event from occurrence to evolution into a fault, taking 0.5 to 1 times this evolution cycle as the extraction duration of the fault characteristic parameter sequence; or it can be dynamically set according to the control system's storage resources and diagnostic response speed requirements. The preset mutation threshold is determined based on the statistical value of the maximum rate of change of operating parameters under normal operating conditions, taking 1.5 to 2 times the statistical maximum value as the preset mutation threshold.

[0097] In this embodiment of the invention, the third and fourth weighting coefficients are determined as follows: based on historical accumulated fault samples, principal component analysis is used to calculate the contribution of two feature quantities in reflecting parameter correlation, and the contribution is normalized to obtain the weighting coefficients; or a logistic regression model is used to train historical samples, and the regression coefficients are normalized to obtain the weighting coefficients; or, based on the signal propagation characteristics of the centrifuge, higher weights are given to parameter waveform similarity or parameter change synchronicity. A preset parameter correlation threshold is set by collecting samples of abnormal events known to be correlated with faults from historical accumulated faults, calculating the parameter correlation of each sample, and taking the minimum or lower quartile of the correlation of each sample as the preset parameter correlation threshold. A preset calibration duration is determined based on the time required for the centrifuge to enter steady-state operation after startup, and is set to 1.2 to 1.5 times this time as the preset calibration duration; or, based on the stability requirements of the sampled data, a sampling duration that can obtain a stable statistical mean is selected. First percentage: determined based on the normal fluctuation range of the centrifuge operating parameters, taking the ratio of the parameter fluctuation range to the benchmark value, and adding a first safety margin; or set according to the allowable deviation range specified in industry standards or equipment manuals.

[0098] This invention uses an abnormal event screening module to extract fault feature parameters from the fault shutdown parameters after determining that the fault is a cumulative fault. The module then calculates the parameter correlation degree based on the similarity of the parameter waveform and the synchronicity of parameter changes. Abnormal events with a parameter correlation degree greater than a preset parameter correlation degree threshold are selected as candidate abnormal events. This effectively eliminates redundant abnormal events that are unrelated to the fault, providing a concise and highly relevant set of candidate abnormal events for subsequent fault tracing, thereby improving the efficiency and accuracy of fault tracing.

[0099] Specifically, the fault tracing and localization module determines the propagation direction and energy attenuation gradient of the candidate abnormal event, wherein,

[0100] The fault tracing and localization module determines the feature correlation between candidate abnormal events based on the feature offset vectors of each candidate abnormal event, and constructs the transmission path of the candidate abnormal events according to the feature correlation.

[0101] The fault tracing and localization module determines the transmission direction of the candidate abnormal event along the transmission path based on the change characteristics of the feature offset vector between adjacent candidate abnormal events in the transmission path.

[0102] The fault tracing and localization module calculates the energy attenuation gradient of the candidate abnormal event along the transmission path based on the temporal relationship between adjacent candidate abnormal events in the transmission path and the change in the feature offset vector.

[0103] The feature offset vector is the difference vector between the running parameter value at the time the candidate abnormal event occurs and the preset benchmark value.

[0104] Specifically, the fault tracing and location module determines the transmission direction of candidate abnormal events along the transmission path, wherein,

[0105] The fault tracing and localization module calculates the angle between the feature offset directions of the feature offset vectors of two adjacent candidate abnormal events in the transmission path.

[0106] If the included angle of the feature offset direction is less than the preset included angle threshold, it is determined that there is a feature correlation between the two adjacent candidate abnormal events, and the transmission direction is determined according to the sign of the included angle of the feature offset direction.

[0107] Specifically, when the angle between the feature offset directions is positive, it indicates that the feature offset directions tend to be consistent, and the transmission direction is from candidate anomalies with smaller feature offset vector magnitudes to candidate anomalies with larger feature offset vector magnitudes; when the angle between the feature offset directions is negative, it indicates that the feature offset directions tend to be opposite, and the transmission direction is from candidate anomalies with larger feature offset vector magnitudes to candidate anomalies with smaller feature offset vector magnitudes.

[0108] Specifically, the fault tracing and localization module calculates the energy attenuation gradient of candidate abnormal events along the propagation path, wherein,

[0109] The fault tracing and localization module obtains the propagation delay deviation between two adjacent candidate abnormal events in the transmission path, and the propagation delay deviation is the difference between the actual propagation delay and the theoretical propagation delay;

[0110] The fault tracing and localization module calculates the magnitude change of the feature offset vectors of two adjacent candidate abnormal events, and determines the energy attenuation gradient between the two adjacent candidate abnormal events based on the ratio of the magnitude change to the propagation delay deviation.

[0111] In this embodiment of the invention, assuming a centrifuge is operating normally, the preset reference values ​​for a certain monitoring point (e.g., the drive motor bearing housing) are as follows: vibration amplitude reference value: 0.05mm, bearing temperature reference value: 65℃; when a candidate abnormal event occurs, the actual operating parameters collected at this monitoring point are as follows: actual vibration amplitude: 0.18mm, actual bearing temperature: 82℃; calculate the offset of each parameter dimension, feature offset = actual operating parameter value − preset reference value; vibration amplitude offset = 0.18mm - 0.05mm = +0.13mm, bearing temperature offset = 82℃ - 65℃ = +17℃; construct a feature offset vector, arranging the offsets of each dimension in a fixed order to form a multi-dimensional vector: feature offset vector = (+0.13mm, +17℃).

[0112] In this embodiment of the invention, after receiving the set of candidate abnormal events output by the abnormal event filtering module, the fault tracing and location module first obtains the feature offset vector corresponding to each candidate abnormal event. The feature offset vector is the difference vector between the operating parameter value at the time of the candidate abnormal event and a preset benchmark value. The preset benchmark value is determined by: collecting the operating parameters of the centrifuge within a preset calibration time under stable operating conditions, and calculating the average value of each operating parameter as the preset benchmark value; or directly setting it according to the centrifuge's factory calibration parameters.

[0113] In this embodiment of the invention, the fault tracing and localization module constructs a feature offset vector matrix based on the feature offset vectors of each candidate abnormal event, and calculates the feature correlation degree between any two candidate abnormal events. The feature correlation degree is calculated as follows: the cosine similarity between the two feature offset vectors is calculated; the closer the cosine similarity is to 1, the more consistent the feature offset directions are; or the Euclidean distance between the two feature offset vectors is calculated; the smaller the Euclidean distance, the closer the amplitude characteristics of the feature offsets are. When the cosine similarity is greater than a preset similarity threshold or the Euclidean distance is less than a preset distance threshold, it is determined that there is a feature correlation between the two candidate abnormal events.

[0114] In this embodiment of the invention, the fault tracing and localization module constructs a transmission path for candidate abnormal events based on the feature association relationships. Specifically, a feature association graph is constructed using each candidate abnormal event as a node and the lines connecting pairs of nodes with feature association relationships as edges. A depth-first search or breadth-first search algorithm is used to start from the node with the smallest feature offset vector modulus, traverse the feature association graph, and extract the connected subgraph containing the most nodes as the main transmission path. Alternatively, each candidate abnormal event is sequentially ordered according to its occurrence time; if a feature association relationship exists between candidate abnormal events at adjacent time points, they are connected to form a temporal transmission chain.

[0115] In this embodiment of the invention, the fault tracing and localization module determines the transmission direction of a candidate abnormal event along the transmission path based on the changing characteristics of the feature offset vectors between adjacent candidate abnormal events in the transmission path. Specifically, it calculates the angle between the feature offset directions of two adjacent candidate abnormal events. When the angle between the feature offset directions is less than a preset angle threshold, it is determined that there is a directional correlation between the two. The transmission direction is determined according to the change in the modulus of the feature offset vectors. When the modulus increases along the path, the transmission direction is from small modulus to large modulus; when the modulus decreases along the path, the transmission direction is from large modulus to small modulus; or the transmission direction is determined according to the chronological order of the occurrence of the abnormal events, with the earlier occurrence pointing to the later occurrence.

[0116] In this embodiment of the invention, the fault tracing and localization module calculates the energy attenuation gradient of a candidate abnormal event along the transmission path based on the temporal relationship between adjacent candidate abnormal events and the change in the feature offset vector. Specifically, the actual propagation delay between two adjacent candidate abnormal events is obtained, where the actual propagation delay is the time difference between the occurrence times of the two abnormal events; the theoretical propagation delay is obtained, which is calculated based on the physical distance between key components of the centrifuge and the signal propagation speed; the difference between the actual propagation delay and the theoretical propagation delay is calculated to obtain the propagation delay deviation. Simultaneously, the magnitude change of the feature offset vector of two adjacent candidate abnormal events is calculated, where the magnitude change is the difference between the magnitude of the feature offset vector of the downstream node and the magnitude of the feature offset vector of the upstream node. The ratio of the magnitude change to the propagation delay deviation is used as the energy attenuation gradient, or the absolute value of the magnitude change is divided by the absolute value of the propagation delay deviation to obtain the energy attenuation gradient.

[0117] In this embodiment of the invention, the preset calibration time is determined based on the time required for the centrifuge to enter steady-state operation after startup, and is taken as 1.2 to 1.5 times that time; or, based on the stability requirements of the sampled data, a sampling time that can obtain a stable statistical mean is taken. The preset similarity threshold is set by collecting candidate abnormal event pairs with known feature correlations from historical accumulated faults, calculating the cosine similarity of the feature offset vectors of each event pair, and taking the minimum value or lower quartile of each similarity as the preset similarity threshold; or, dynamically set according to the accuracy requirements for correlation determination during on-site debugging. The preset distance threshold is set by collecting candidate abnormal event pairs with known feature correlations from historical accumulated faults, calculating the Euclidean distance of the feature offset vectors of each event pair, and taking the maximum value or upper quartile of each distance as the preset distance threshold; or, based on the amplitude distribution range of the feature offset vectors, taking a certain percentage of the amplitude range as the preset distance threshold.

[0118] In this embodiment of the invention, the preset directional angle threshold is determined as follows: based on the physical topology and signal propagation characteristics between key components of the centrifuge, the theoretical phase difference range of operating parameters between different monitoring points under normal operating conditions is calculated, and the upper limit of this range is taken as the preset directional angle threshold; or, abnormal event pairs with known causal relationships are collected from historical accumulated faults, and the statistical distribution of their characteristic offset directional angles is calculated, with the upper quantile of the distribution taken as the preset directional angle threshold. The theoretical propagation delay is calculated based on the physical distance between key components of the centrifuge and the propagation speed of abnormal signals in the equipment structure, wherein the propagation speed is obtained based on the equipment material properties and historical calibration data; or, it is determined based on the statistical mean of the time difference between upstream and downstream abnormal events caused by the same fault source in historical fault data.

[0119] This invention constructs feature associations and transmission paths based on the feature offset vectors of candidate abnormal events through a fault source tracing and localization module. It uses the angle between the feature offset directions to determine the transmission direction and calculates the energy attenuation gradient using the ratio of the magnitude change to the propagation delay deviation. This achieves accurate identification of the propagation direction of abnormal events and quantitative characterization of the degree of energy attenuation, providing a reliable quantitative basis for fault source localization and improving the accuracy and efficiency of tracing and diagnosing accumulated faults.

[0120] Specifically, the fault source tracing and localization module locates the fault source based on the propagation direction and energy attenuation gradient of the candidate abnormal events, wherein,

[0121] The fault tracing and localization module determines the upstream and downstream relationship of each candidate abnormal event in the transmission path according to the transmission direction, and marks the candidate abnormal event located at the upstream of the transmission path as a potential fault source node.

[0122] The fault tracing and localization module calculates the energy attenuation gradient of the potential fault source node as it propagates along the transmission path to the adjacent downstream node.

[0123] If the energy decay gradient is less than a preset decay threshold, then the potential fault source node is determined as the initial fault source.

[0124] If the energy attenuation gradient is greater than or equal to the preset attenuation threshold, the fault tracing and location module traces back upstream along the transmission path and determines the node corresponding to the first time the energy attenuation gradient is less than the preset attenuation threshold as the initial fault source.

[0125] The fault tracing and location module outputs the device component associated with the initial fault source as the fault source.

[0126] In this embodiment of the invention, after determining the propagation direction and energy attenuation gradient of each candidate abnormal event, the fault tracing and localization module constructs a complete propagation path topology based on the propagation direction and determines the upstream and downstream relationships of each candidate abnormal event in the propagation path according to the propagation direction. Specifically, in the propagation path, the node pointed to by the propagation direction is the downstream node, the node from which the propagation direction originates is the upstream node, and the candidate abnormal event located at the beginning of the propagation path and without an upstream node is marked as a potential fault source node. The fault tracing and localization module calculates the energy attenuation gradient of the potential fault source node as it propagates along the propagation path to adjacent downstream nodes. If the potential fault source node has multiple downstream branches, the energy attenuation gradient of each branch is calculated separately, and the minimum value is taken as the representative value of the energy attenuation gradient of that node.

[0127] In this embodiment of the invention, the fault tracing and location module compares the calculated energy attenuation gradient with a preset attenuation threshold. If the energy attenuation gradient is less than the preset attenuation threshold, it indicates that the abnormal energy loss is small when propagating from this node to the downstream node, and this node is close to or is the source of the abnormal energy. Therefore, the potential fault source node is identified as the initial fault source. If the energy attenuation gradient is greater than or equal to the preset attenuation threshold, it indicates that the abnormal energy loss is large when propagating from this node to the downstream node, and this node may not be the source of the abnormality, but rather an intermediate node in the abnormal propagation process. In this case, the fault tracing and location module backtracks upstream along the transmission path, checking the energy attenuation gradient of each upstream node step by step, and identifies the node whose energy attenuation gradient is first less than the preset attenuation threshold as the initial fault source. If the energy attenuation gradient of all nodes is greater than or equal to the preset attenuation threshold when tracing back to the starting point of the transmission path, then the upstreammost node of the transmission path is identified as the initial fault source.

[0128] In this embodiment of the invention, the fault source tracing and localization module outputs the equipment component associated with the initial fault source as the fault source localization result. Specifically, based on the parameter type of the candidate abnormal event corresponding to the initial fault source and the sensor installation location, it is mapped to a specific equipment component of the centrifuge. The equipment component includes, but is not limited to, bearings, shafts, motors, pulleys, feeding mechanisms, discharging mechanisms, or lubrication systems. The method for determining the preset attenuation threshold is as follows: based on the theoretical energy transfer loss coefficient between adjacent monitoring points under normal operating conditions of the centrifuge, combined with a safety margin, the preset attenuation threshold is calculated. The theoretical energy transfer loss coefficient is determined based on the equipment material properties, signal propagation path length, and historical calibration data.

[0129] This invention constructs a transmission path topology based on the transmission direction and determines the upstream and downstream relationships through a fault source tracing and localization module. The upstream node is marked as a potential fault source. The fault source is accurately determined by comparing the energy attenuation gradient with a preset attenuation threshold. When the gradient exceeds the threshold, the system backtracks upstream until the node where the attenuation gradient first reaches the threshold is found. This achieves the technical effect of accurately identifying the fault origin from the abnormal event propagation path, significantly shortening the centrifuge fault diagnosis time and improving maintenance efficiency and equipment availability.

[0130] Please see Figure 3 As shown, Figure 3 This is a schematic diagram of the anomaly detection module in the PLC control system used for centrifuge fault diagnosis in this embodiment.

[0131] Specifically, the anomaly detection module includes a threshold judgment unit and a rate of change detection unit;

[0132] The threshold judgment unit is used to compare each operating parameter with the corresponding preset threshold range. When any operating parameter exceeds the preset threshold range, a first type of abnormal event record is generated.

[0133] The rate of change detection unit is used to calculate the rate of change of each operating parameter within a unit of time. When the rate of change exceeds a preset rate of change threshold, a second type of abnormal event record is generated.

[0134] The anomaly detection module is also used to merge the corresponding first-type anomaly event record and the second-type anomaly event record into a composite anomaly event record when the same operating parameter triggers the threshold judgment unit and the rate of change detection unit at the same time.

[0135] The first type of abnormal event record, the second type of abnormal event record, and the composite abnormal event record together constitute the abnormal event record.

[0136] In this embodiment of the invention, the anomaly detection module receives the operating parameters transmitted by the parameter acquisition module in real time, and performs parallel monitoring and analysis through a threshold judgment unit and a rate of change detection unit. The threshold judgment unit internally stores preset threshold ranges corresponding to each operating parameter. For each operating parameter at each acquisition time, the threshold judgment unit compares it with the corresponding preset threshold range. When the operating parameter is within the preset threshold range, it is determined to be in normal operating condition; when the operating parameter exceeds the preset threshold range, it is determined to be in an abnormal state, and a first type of abnormal event record is generated. The first type of abnormal event record at least includes the abnormal parameter type, the time point of the abnormal occurrence, the abnormal amplitude, the type of exceeding the limit (including upper limit exceeding the limit or lower limit exceeding the limit), and the operating parameter value at the time of the abnormal occurrence.

[0137] The rate of change detection unit is used to calculate the rate of change of each operating parameter per unit time. Specifically, the rate of change detection unit caches continuously collected operating parameters of the same parameter category, calculates the instantaneous rate of change between adjacent sampling points using the difference method, or calculates the average rate of change within a preset time window using the least squares method. When the calculated rate of change exceeds a preset rate of change threshold, it is determined that the parameter has undergone a sudden change anomaly, and a second type of abnormal event record is generated. The second type of abnormal event record includes at least the abnormal parameter type, the time point of the anomaly occurrence, the peak value of the rate of change, the direction of the change trend (including an upward trend or a downward trend), and the operating parameter value at the start time of the anomalous change.

[0138] The anomaly detection module is also used to perform correlation analysis on the monitoring results of the same operating parameter. When both the threshold judgment unit and the rate of change detection unit determine that the same operating parameter is abnormal within the same time window, the anomaly detection module merges the corresponding first-type and second-type abnormal event records to generate a composite anomaly event record. The composite anomaly event record contains all the information fields of the first-type and second-type abnormal event records, and additionally marks the anomaly level. The anomaly level is determined by a weighted sum of the exceedance magnitude and the peak value of the rate of change; the larger the exceedance magnitude and the higher the peak value of the rate of change, the higher the anomaly level. For anomalies that only trigger the threshold judgment unit or only trigger the rate of change detection unit, the first-type or second-type abnormal event record is output as an independent anomaly event record, respectively.

[0139] The anomaly detection module is also used for continuous monitoring of abnormal events. When an abnormal event occurs, it continuously tracks changes in the operating parameter. When the operating parameter recovers to within the preset threshold range and the duration exceeds the preset recovery confirmation time, the abnormal recovery time point is marked, and the duration of the abnormality is calculated. The anomaly detection module associates and stores the anomaly occurrence time point, the anomaly recovery time point, and the anomaly duration in the corresponding anomaly event record, allowing the fault diagnosis output module to exclude the normal operation period after anomaly recovery when determining the time window.

[0140] In this embodiment of the invention, the preset threshold range is defined as follows: Operating parameters of the centrifuge are collected within a preset calibration period under stable operating conditions. The mean and standard deviation of each operating parameter are calculated, and the mean ± k times the standard deviation is taken as the preset threshold range, where k ranges from 2 to 4; or it can be directly set according to the normal fluctuation range specified in the centrifuge's factory calibration parameters; or it can be based on the parameter distribution before the fault occurred in historical fault data, taking the upper and lower limits of the normal distribution range as the preset threshold range. The preset rate of change threshold is determined based on the statistical value of the maximum rate of change of the operating parameters under normal operating conditions. Operating parameters under normal operating conditions are collected within a preset calibration period, and the rate of change of each parameter per unit time is calculated. 1.2 to 1.5 times the statistical maximum value is taken as the preset rate of change threshold; or it can be dynamically set through on-site debugging according to fault sensitivity requirements. The higher the sensitivity, the lower the preset rate of change threshold; or it can be set according to the allowable rate of change range specified in industry standards or equipment manuals.

[0141] In this embodiment of the invention, the preset recovery confirmation time is determined based on the normal fluctuation cycle of the centrifuge operating parameters, taking 2 to 3 times the normal fluctuation cycle as the preset recovery confirmation time to avoid misjudgment caused by instantaneous parameter fluctuations; or it is set according to the response speed requirements of the control system for abnormal recovery. The higher the response speed requirement, the shorter the preset recovery confirmation time. The time window is the time interval from the normal operating state to the time point of the fault occurrence. The normal operating state is determined based on the last abnormal recovery time point in the abnormal event record. When there are no continuously unrecovered abnormal events in the abnormal event record, the last abnormal recovery time point is taken as the starting boundary of the normal operating state.

[0142] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A PLC control system for centrifuge fault diagnosis, characterized in that, include: The parameter acquisition module collects and caches the centrifuge's operating parameters in real time; An anomaly detection module is used to monitor the operating parameters in real time, generate anomaly event records based on the operating parameters, and mark the occurrence time and recovery time of the anomaly events. The fault triggering module is used to generate a fault triggering signal when a centrifuge malfunction is detected and the centrifuge stops. The parameter freezing and storage module, in response to the fault trigger signal, is used to freeze the running parameters cached in the parameter acquisition module at the time of the fault occurrence and store them as fault shutdown parameters. The fault diagnosis output module is used to determine the time-series distribution feature quantization value and parameter degradation trajectory quantization value of the abnormal events based on the abnormal event records and operating parameters within the time window from the normal operating state to the time of the fault occurrence, and to determine the fault type of the centrifuge based on the time-series distribution feature quantization value and the parameter degradation trajectory quantization value, wherein the fault type includes cumulative fault type and instantaneous fault type. The abnormal event filtering module, in response to the centrifuge's fault type being a cumulative fault type, extracts the fault feature parameters from the fault shutdown parameters, and filters the abnormal event records within the time window based on the fault feature parameters, filtering out candidate abnormal events that have parameter correlation with the fault shutdown parameters. The fault tracing and localization module is used to determine the propagation direction and energy attenuation gradient of the candidate abnormal events based on the feature offset vector of the candidate abnormal events and the propagation delay deviation between the candidate abnormal events, and to locate the fault source based on the propagation direction and energy attenuation gradient of the candidate abnormal events. The fault tracing and localization module determines the propagation direction and energy attenuation gradient of the candidate abnormal event, wherein, The fault tracing and localization module determines the feature correlation between candidate abnormal events based on the feature offset vectors of each candidate abnormal event, and constructs the transmission path of the candidate abnormal events according to the feature correlation. The fault tracing and localization module determines the transmission direction of the candidate abnormal event along the transmission path based on the change characteristics of the feature offset vector between adjacent candidate abnormal events in the transmission path. The fault tracing and localization module calculates the energy attenuation gradient of the candidate abnormal event along the transmission path based on the temporal relationship between adjacent candidate abnormal events in the transmission path and the change in the feature offset vector. Wherein, the feature offset vector is the difference vector between the running parameter value at the time of the occurrence of the candidate abnormal event and the preset benchmark value; The fault tracing and localization module determines the transmission direction of the candidate abnormal event along the transmission path, wherein, The fault tracing and localization module calculates the angle between the feature offset directions of the feature offset vectors of two adjacent candidate abnormal events in the transmission path. If the included angle of the feature offset direction is less than the preset included angle threshold, it is determined that there is a feature correlation between the two adjacent candidate abnormal events, and the transmission direction is determined according to the sign of the included angle of the feature offset direction. Wherein, when the included angle of the feature offset direction is positive, it indicates that the feature offset directions tend to be consistent, and the transmission direction is from the candidate anomaly event with the smaller feature offset vector magnitude to the candidate anomaly event with the larger feature offset vector magnitude; when the included angle of the feature offset direction is negative, it indicates that the feature offset directions tend to be opposite, and the transmission direction is from the candidate anomaly event with the larger feature offset vector magnitude to the candidate anomaly event with the smaller feature offset vector magnitude. The fault tracing and localization module calculates the energy attenuation gradient of candidate abnormal events along the propagation path, wherein... The fault tracing and localization module obtains the propagation delay deviation between two adjacent candidate abnormal events in the transmission path, and the propagation delay deviation is the difference between the actual propagation delay and the theoretical propagation delay; The fault tracing and localization module calculates the magnitude change of the feature offset vectors of the two adjacent candidate abnormal events, and determines the energy attenuation gradient between the two adjacent candidate abnormal events based on the ratio of the magnitude change to the propagation delay deviation. The fault source tracing and localization module locates the fault source based on the propagation direction and energy attenuation gradient of the candidate abnormal events, wherein, The fault tracing and localization module determines the upstream and downstream relationship of each candidate abnormal event in the transmission path according to the transmission direction, and marks the candidate abnormal event located at the upstream of the transmission path as a potential fault source node. The fault tracing and localization module calculates the energy attenuation gradient of the potential fault source node as it propagates along the transmission path to the adjacent downstream node. If the energy decay gradient is less than a preset decay threshold, then the potential fault source node is determined as the initial fault source. If the energy attenuation gradient is greater than or equal to the preset attenuation threshold, the fault tracing and location module traces back upstream along the transmission path and determines the node corresponding to the first time the energy attenuation gradient is less than the preset attenuation threshold as the initial fault source. The fault tracing and location module outputs the device component associated with the initial fault source as the fault source.

2. The PLC control system for centrifuge fault diagnosis according to claim 1, characterized in that, The fault diagnosis output module determines the quantized values ​​of the temporal distribution characteristics of the abnormal events and the quantized values ​​of the parameter degradation trajectory, wherein, The fault diagnosis output module arranges the abnormal events within the time window in a temporal sequence, calculates the rate of change of the time interval between adjacent abnormal events, and obtains the quantified value of the temporal distribution characteristics. The fault diagnosis output module extracts the operating parameters of the same parameter category from the abnormal event records, constructs a parameter value sequence according to the time sequence of the abnormal occurrence, fits the parameter value sequence using the curve fitting method, calculates the slope of the fitting curve, and obtains the quantified value of the parameter degradation trajectory. The same parameter category refers to operating parameters that have the same physical dimensions and the same sensor source.

3. The PLC control system for centrifuge fault diagnosis according to claim 1, characterized in that, The fault diagnosis output module determines the fault type of the centrifuge based on the quantized value of the time-series distribution characteristics and the quantized value of the parameter degradation trajectory, wherein, The fault diagnosis output module calculates the fault type discrimination coefficient based on the quantized value of the temporal distribution feature and the quantized value of the parameter degradation trajectory. If the fault type discrimination coefficient is greater than or equal to the preset discrimination coefficient, the fault diagnosis output module determines that the fault type of the centrifuge is a cumulative fault type; If the fault type discrimination coefficient is less than the preset discrimination coefficient, the fault diagnosis output module determines that the fault type of the centrifuge is an instantaneous fault type.

4. The PLC control system for centrifuge fault diagnosis according to claim 1, characterized in that, The abnormal event filtering module filters out candidate abnormal events that are correlated with the fault shutdown parameters, wherein... The abnormal event screening module extracts the fault feature parameters from the fault shutdown parameters, and calculates the parameter correlation degree between the operating parameters corresponding to each abnormal event within the time window and the fault feature parameters. The parameter correlation degree is obtained by weighted calculation based on parameter waveform similarity and parameter change synchronicity. The abnormal event filtering module marks abnormal events whose parameter correlation degree is greater than a preset parameter correlation degree threshold as candidate abnormal events.

5. The PLC control system for centrifuge fault diagnosis according to claim 4, characterized in that, The similarity of the parameter waveforms is determined by calculating the dynamic time warping distance between the sequence of operating parameters corresponding to the abnormal event and the sequence of fault characteristic parameters, and the synchronicity of parameter changes is determined by calculating the time offset between the time point of the abnormal event and the time point of the sudden change of the fault characteristic parameters.

6. The PLC control system for centrifuge fault diagnosis according to claim 1, characterized in that, The anomaly detection module includes a threshold judgment unit and a rate of change detection unit; The threshold judgment unit is used to compare each operating parameter with the corresponding preset threshold range. When any operating parameter exceeds the preset threshold range, a first type of abnormal event record is generated. The rate of change detection unit is used to calculate the rate of change of each operating parameter within a unit of time. When the rate of change exceeds a preset rate of change threshold, a second type of abnormal event record is generated. The anomaly detection module is also used to merge the corresponding first-type anomaly event record and the second-type anomaly event record into a composite anomaly event record when the same operating parameter triggers the threshold judgment unit and the rate of change detection unit at the same time. The first type of abnormal event record, the second type of abnormal event record, and the composite abnormal event record together constitute the abnormal event record.

Images