A rule base driven vibration fault diagnosis method and system for multi-point rotating machinery

Abstract

The application discloses a rule base driven vibration fault diagnosis method and system for multi-measuring-point rotating machinery, and belongs to the technical field of equipment state monitoring and intelligent fault diagnosis. The method comprises the following steps: S1, measuring point modeling and parameter binding; S2, data import and preprocessing; S3, spectrum analysis (FFT); S4, envelope demodulation (ENV) adaptive link; S5, target fault frequency generation; S6, joint hit criterion; S7, rule evidence mapping and screening; S8, AST rule evaluation and interpretable output; and S9, multi-measuring-point fusion and sorting. The application can avoid evidence confusion between different measuring points, different parts or different data files, and improve the traceability of a diagnosis conclusion.

Description

Technical Field

[0001] This invention relates to a rule-based vibration fault diagnosis method and system for multi-measurement rotating machinery, belonging to the field of equipment condition monitoring and intelligent fault diagnosis technology. Background Technology

[0002] Many existing systems suffer from the problem of "disconnect between signal analysis results and rule base labels": although frequency components are extracted, they cannot be stably mapped to rule conclusions; at the same time, it is common to use only frequency difference or only peak value as a single indicator for judgment, resulting in a high false hit rate for weak peaks.

[0003] From the perspective of existing technologies, common solutions typically treat vibration signal processing, equipment ledgers, rule base configuration, and diagnostic reports as relatively independent functional modules: the signal processing module outputs the spectral peak value, the rule base module configures the rule conditions separately, and the reporting module manually concatenates the conclusions. This approach lacks unified evidence fields and measurement point identification, making it difficult to reliably trace the same frequency peak value back to a specific measurement point, specific location, and specific rule.

[0004] Other solutions, while capable of calculating bearing fault characteristic frequencies or gear meshing frequencies, often rely solely on frequency deviations for hit determination, failing to consider factors such as peak relative noise levels, energy proportions, octave or sideband corroboration, and differences in FFT / ENV sources. When on-site operating conditions change, speed fluctuates, or weak peaks are submerged by noise, false hits, missed hits, or insufficient rule interpretation can easily occur.

[0005] In multi-point diagnostic scenarios, existing technologies lack a unified interactive link to merge evidence from individual test points into device-level conclusions. Different test points may hit different rules, and if only the results of each rule are displayed one by one, it is impossible to form a primary fault, secondary fault, and evidence ranking that conforms to engineering decision-making habits.

[0006] In multi-measurement scenarios, existing solutions lack a unified fusion and sorting mechanism, which can easily lead to conflicting conclusions from different measurement points. In addition, the rule label text contains mixed terms across devices (such as pump feature words mixed in the fan scenario), which reduces the reliability of diagnosis. Summary of the Invention

[0007] The purpose of this invention is to solve the problem that existing technologies cannot form primary faults, secondary faults, and evidence ranking that conform to engineering decision-making habits, and to provide a rule-based vibration fault diagnosis method and system for multi-measurement rotating machinery; To achieve the above objectives / to solve the above technical problems, the present invention is implemented using the following technical solution: On the one hand: A rule-based vibration fault diagnosis method for multi-measurement rotating machinery, the method comprising: S1: Bind each measuring point of the equipment to its corresponding part, sampling parameters, speed parameters, bearing parameters and / or gear parameters, and generate a unique measuring point identifier; S2: Import the vibration time-domain signal bound to the measuring point identifier, and perform mean removal, trend removal, unit conversion, filtering and / or noise reduction processing on the vibration time-domain signal to obtain the pre-processed signal; S3: Perform spectrum analysis on the preprocessed signal to extract the peak frequency, peak amplitude and local noise floor, and form a spectrum candidate peak set P_fft; S4: Perform envelope demodulation on the preprocessed signal, and generate an envelope candidate peak set P_env and corresponding source identifiers in the order of bandpass ENV priority, direct-pass ENV backoff, and FFT proxy backoff; S5: Based on the bound bearing parameters, gear parameters, and speed parameters, generate a target frequency set F_target consisting of the basic fault frequency, the equipment-specific fault frequency, and their harmonics or sidebands; S6: For each target frequency in the target frequency set F_target, search for the optimal candidate peak in the spectrum candidate peak set P_fft and the envelope candidate peak set P_env, and determine the frequency matching table using the joint hit criterion composed of frequency difference tolerance, peak signal-to-noise ratio PTN, peak energy ratio and harmonic evidence. S7: The frequency matching table is filtered by rules according to the device name and key parts, and sorted by soft constraints according to the monitoring parts and monitoring locations to map into rule evidence vectors; S8: Based on the rule evidence vector, perform AND, OR, NOT and / or weighted node evaluation on the AST logic tree in the rule base, and output the hit rule, frequency matching degree, number of evidence items, unsatisfied atoms, unsupported atoms and interpretation chain; S9: The diagnostic conclusions of multiple measurement points are merged and sorted according to the frequency matching degree first, and the severity first when the frequency matching degree is the same, to form a comprehensive equipment-level diagnostic conclusion and a list of conclusions for each measurement point.

[0008] Optionally, step S1 specifically includes: Construct a hierarchical relationship at the equipment level, including equipment name, key components, monitoring components, and monitoring locations; Each measurement point is bound to sampling parameters, which include at least the sampling frequency fs, sampling duration, channels, and rotational speed rpm; Each measuring point is bound to bearing parameters, which include at least the number of rolling elements n, the rolling element diameter d, the pitch circle diameter D, and the contact angle θ. Depending on the equipment type, the gear parameters can be selected for binding. The gear parameters include at least the number of teeth z1 / z2, the transmission ratio, the number of stages, and the meshing stage relationship. The measurement point name is generated by concatenating the measurement point number, monitoring location, monitoring position, and data import status, and the measurement point status is saved to ensure a one-to-one correspondence between measurement points, locations, parameters, and data.

[0009] Optionally, step S2 specifically includes: Import vibration time-domain signals in xlsx, csv, txt and / or dat formats, and bind the vibration time-domain signals with the corresponding measurement point identifiers; Perform mean removal, trend removal, and unit conversion on the imported signal; Perform high-pass filtering, low-pass filtering, notch filtering and / or noise reduction processing according to preset processing parameters; The preprocessing parameters are persisted so that the processing status of the corresponding measurement points can be restored when refreshed or re-imported.

[0010] Optionally, step S5 specifically includes: The rotational frequency fr is generated based on the rotational speed. Generate the outer ring failure frequency (BPFO), inner ring failure frequency (BPFI), rolling element failure frequency (BSF), and cage failure frequency (FTF) based on the bearing parameters. Perform 2x to 5x frequency multiplication on fr, BPFO, BPFI, BSF, and FTF; The receiver uses a device-specific frequency and performs frequency multiplication, wherein the device-specific frequency includes at least the blade passage frequency (BPF). When the equipment name, key parts, or monitoring parts contain the keyword "gearbox", the gear meshing frequency GMF and at least one of the following sideband frequencies, GMF+fr, GMF-fr, 2GMF+fr, and 2GMF-fr, are generated based on the gear parameters.

[0011] Optionally, step S6 specifically includes: Frequency difference criterion: Calculate the deviation between the peak frequency f_peak and the target frequency f_target. When |f_peak-f_target|<=max(minimum tolerance Hz, tolerance ratio * f_target), the frequency difference condition is met. PTN criterion: Calculate the peak signal-to-noise ratio (PTN) = When PTN >= PTN_min, the peak signal-to-noise ratio condition is met; Energy criterion: Calculate the peak energy percentage. When the peak energy percentage >= E_min(feature, source), the energy condition is met. Frequency harmonic evidence: When the main criterion is in a critical state, if the number of hits in the 2x to 5x harmonics is >= k, then it is determined to be a harmonic support hit; The frequency matching table includes at least the target frequency, peak frequency, peak amplitude, deviation, PTN, peak energy percentage, harmonic hit, source, and hit status.

[0012] Optionally, step S3 specifically includes: The spectrum is calculated for each measurement point in the spectrum channel, and the peak frequency, peak amplitude and local noise floor are extracted to form a spectrum candidate peak set P_fft.

[0013] Optionally, step S4 specifically includes: First, execute the bandpass ENV in the envelope channel. If the bandpass ENV is not informative enough, fall back to the passthrough ENV. If the passthrough ENV is still not informative enough, then enable FFT proxy evidence.

[0014] Optionally, the source identifier is used to distinguish whether the candidate peak comes from FFT, bandpass ENV, direct-pass ENV, or FFT proxy evidence; the source identifier is retained synchronously in subsequent frequency hit determination and rule evaluation.

[0015] Optionally, the rule evidence vector includes at least a feature key, a measured value, a threshold, whether the condition is met, and its source. The equipment name and key components are used as the main filtering criteria for rules to determine the range of candidate rules. The monitoring location and monitoring position are used as soft constraints for ranking, rather than as hard elimination conditions.

[0016] The second aspect discloses a rule-based vibration fault diagnosis system for multi-measurement-point rotating machinery, the system comprising: The measurement point modeling module is used to bind each measurement point of the equipment with its corresponding part, sampling parameters, speed parameters, bearing parameters and / or gear parameters, and generate a unique measurement point identifier; The data preprocessing module is used to import the vibration time-domain signal bound to the measuring point identifier, and perform mean removal, trend removal, unit conversion, filtering and / or noise reduction processing on the vibration time-domain signal to obtain the preprocessed signal. The spectrum feature analysis module is used to perform spectrum analysis on the preprocessed signal, extract the peak frequency, peak amplitude and local noise floor, and form a spectrum candidate peak set P_fft; The envelope feature analysis module is used to perform envelope demodulation on the preprocessed signal and generate an envelope candidate peak set P_env and corresponding source identifiers in the order of bandpass ENV priority, direct-pass ENV backoff, and FFT proxy backoff. The target frequency generation module is used to generate a target frequency set F_target consisting of the basic fault frequency, the equipment-specific fault frequency, and their multiples or sidebands, based on the bound bearing parameters, gear parameters, and speed parameters. The joint hit determination module is used to search for the optimal candidate peak in the spectrum candidate peak set P_fft and the envelope candidate peak set P_env for each target frequency in the target frequency set F_target. The frequency matching table is determined by the joint hit criteria composed of frequency difference tolerance, peak signal-to-noise ratio PTN, peak energy ratio and harmonic evidence. The rule-based evidence mapping module is used to perform rule-based primary filtering of the frequency matching table according to the device name and key parts, and to perform soft-constraint sorting and mapping according to the monitoring parts and monitoring locations into rule-based evidence vectors. The rule evaluation module is used to perform AND, OR, NOT and / or weighted node evaluation on the AST logic tree in the rule base based on the rule evidence vector, and output the hit rules, frequency matching degree, number of evidence items, unsatisfied atoms, unsupported atoms and interpretation chain; The fusion output module is used to merge and sort the diagnostic conclusions of multiple measurement points according to the priority of frequency matching degree and the severity priority when the matching degree is the same, so as to form a comprehensive diagnostic conclusion at the equipment level and a list of conclusions for each measurement point.

[0017] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: This invention uses measurement point identification throughout the entire process, which can avoid confusion of evidence between different measurement points, different locations, or different data files, and improve the traceability of diagnostic conclusions.

[0018] By combining frequency difference, PTN, energy ratio, and octave / sideband evidence into a joint hit criterion, this invention can improve hit stability in scenarios with weak peaks, noise peaks, and speed fluctuations, and reduce the risk of misdiagnosis based solely on a single frequency deviation.

[0019] By structuring the frequency matching results into rule evidence vectors and performing AST evaluation, this invention can output the reasons for hits and the reasons for misses, making the diagnostic conclusions auditable, facilitating review by maintenance personnel, and also facilitating subsequent iterations of the rule base.

[0020] Through a multi-measurement point fusion and sorting mechanism, this invention can convert local evidence from multiple measurement points into equipment-level primary and secondary conclusions, reducing the workload of manually comparing multiple measurement point reports and improving the efficiency of engineering decision-making. Attached Figure Description

[0021] Figure 1 The diagram shown is a schematic representation of the overall system architecture of this invention. Figure 2 The diagram shown is a flowchart of the signal processing and feature extraction process of this invention. Figure 3 The diagram shown is a flowchart of the frequency hit determination process of this invention. Figure 4The diagram shown is a flowchart illustrating the rule evaluation and interpretation process of this invention. Figure 5 The diagram shown is a flowchart of the multi-point fusion diagnosis process of this invention. Figure 6 The diagram shown is a schematic diagram of the page interaction logic of this invention. Detailed Implementation

[0022] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0023] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are used only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0024] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0025] Example 1

[0026] like Figures 1-6 As shown, a rule-based vibration fault diagnosis method for multi-measurement rotating machinery is disclosed, the method comprising: Step S1: Measurement point modeling and parameter binding: Establish a four-level structure at the equipment level, including equipment name, key parts, monitoring parts, and monitoring locations. Bind each measurement point to its corresponding part, sampling parameters, bearing parameters, and / or gear parameters, and generate a unique measurement point identifier. Step S2: Data import and preprocessing: Import the vibration time domain signal bound to the measuring point identifier, and perform mean removal, trend removal, unit conversion, filtering and / or noise reduction processing on the vibration time domain signal to obtain the preprocessed signal; Step S3: Spectrum Analysis: Perform spectrum analysis on the preprocessed signal to extract peak frequency, peak amplitude and local noise floor, forming a spectrum candidate peak set P_fft; Step S4: Envelope demodulation adaptive analysis: Perform envelope demodulation on the preprocessed signal, and generate an envelope candidate peak set P_env and corresponding source identifiers in the order of bandpass ENV priority, direct-pass ENV backoff, and FFT proxy backoff; Step S5: Target Fault Frequency Generation: Based on the bearing parameters, gear parameters, speed parameters, and equipment-specific frequencies bound in S1, generate a target frequency set F_target consisting of the base fault frequency, the equipment-specific fault frequency, and their multiples or sidebands. Step S6: Joint Hit Decision: For each target frequency in F_target, search for the optimal candidate peak in P_fft and P_env, and determine the frequency matching table using the joint hit criteria composed of frequency difference tolerance, peak signal-to-noise ratio PTN, peak energy ratio and harmonic evidence. Step S7: Rule Evidence Mapping and Filtering: Map the frequency matching table to rule evidence vectors, perform main rule filtering according to device name and key parts, and sort by soft constraints according to monitoring parts and monitoring locations; Step S8: AST rule evaluation and interpretation chain generation: Based on the rule evidence vector, perform AND, OR, NOT and / or weighted node evaluation on the AST logic tree in the rule base, and output the hit rules, matching degree, number of evidence entries, unsatisfied atoms, unsupported atoms and interpretation chain; Step S9: Multi-point fusion and sorting: The diagnostic conclusions of multiple measurement points are fused and sorted according to the matching degree priority and the severity priority when the matching degree is the same, to form a comprehensive equipment-level diagnostic conclusion and a list of sub-measurement point conclusions; Step S10: Output and Human-Machine Interaction: Output device name, diagnostic location, fault type, fault severity, hit rules, sub-measurement point results and frequency matching table, and support measurement point switching, spectrum zooming, cursor reading and characteristic frequency display / hiding.

[0027] In the specific implementation of this embodiment, step S1 specifically includes: Construct a hierarchical relationship at the equipment level, including equipment name, key components, monitoring components, and monitoring locations; Each measurement point is bound to sampling parameters, which include at least the sampling frequency fs, sampling duration, channels, and rotational speed rpm; Each measuring point is bound to bearing parameters, which include at least the number of rolling elements n, the rolling element diameter d, the pitch circle diameter D, and the contact angle θ. Depending on the equipment type, the gear parameters can be selected for binding. The gear parameters include at least the number of teeth z1 / z2, the transmission ratio, the number of stages, and the meshing stage relationship. The measurement point name is generated by concatenating the measurement point number, monitoring location, monitoring position, and data import status, and the measurement point status is saved to ensure a one-to-one correspondence between measurement points, locations, parameters, and data.

[0028] In this embodiment, step S2: data import and preprocessing: The time-domain signal imported in this step is the vibration signal corresponding to the monitoring position of the measurement point defined in step S1. It is bound to the measurement point identifier during import. The preprocessed signal is used as the common input for S3 and S4. The specific processing process is as follows: 1) Import time-domain signals such as xlsx / csv / txt / dat; 2) Perform mean removal, trend removal, and unit conversion; 3) Perform optional filtering (high-pass, low-pass, notch filtering) and optional noise reduction; 4) Persist preprocessing parameters and automatically restore them after refreshing.

[0029] In this embodiment, step S3, Spectrum Analysis (FFT), uses the preprocessed time-domain signal output from S2 as input to calculate the spectrum for each measurement point and form an FFT candidate peak set P_fft. Specifically: The spectrum is calculated for each measurement point, and the default display is 0~500Hz with zoom / cursor reading support. Output peak frequency, peak amplitude, and low local noise; Form the FFT candidate peak set P_fft.

[0030] In the specific implementation of this embodiment, step S4: Envelope Demodulation (ENV) Adaptive Link: This step also takes the preprocessed time-domain signal output from S2 as input, and adaptively switches between bandpass ENV, direct-pass ENV and FFT proxy evidence according to the demodulation information.

[0031] First execute the bandpass ENV; If the information provided by the pass-through ENV is insufficient, then fall back to the pass-through ENV; If direct access to ENV is still insufficient, then FFT proxy evidence will be enabled; The candidate peak set P_env and the "source identifier" (FFT / ENV) are formed.

[0032] In the specific implementation process of this embodiment, step S5: Target fault frequency generation: This step takes the rotational speed, bearing parameters and / or gear parameters bound in S1, and combines them with the equipment-specific frequency settings to generate a set of target frequencies to be matched, F_target, for S6 to search in the candidate peak set of S3 / S4.

[0033] Basic bearing frequencies: fr, BPFO, BPFI, BSF, FTF; Multiplier extension: 2x~5x; Equipment-specific frequencies: such as blade passing frequency (BPF), gear meshing frequency (GMF) and its sidebands; Generate the target frequency set F_target.

[0034] In the specific implementation of this embodiment, step S6: Joint hit criterion: For each f∈F_target, search for the optimal candidate in P_fft and P_env, and output at least: target frequency, peak frequency, peak amplitude, deviation, PTN, peak energy percentage, harmonic hit, and source.

[0035] in: Deviation criterion: |f_peak-f_target|<=max(minimum tolerance Hz, tolerance ratio × f_target); PTN Criterion: PTN = The requirement is that PTN >= PTN_min; Energy criterion: Peak energy percentage >= E_min(feature, source); Frequency doubling evidence: When the main criterion is critical, if the number of hits in 2x~5x is >=k, then it can be determined as "harmonic support hit".

[0036] In the specific implementation of this embodiment, step S7: rule-based evidence mapping and filtering: The hit results (i.e., the frequency matching table fields output by S6, including target frequency, peak frequency, peak amplitude, deviation, PTN, peak energy percentage, harmonic hit and source) are mapped to evidence vectors (feature key, measured value, threshold, whether it is satisfied, source); The main filter for rules uses "equipment name + key parts"; The monitoring location and monitoring position are only used as soft constraints for ranking, and are not used for hard elimination.

[0037] In the specific implementation process of this embodiment, step S8: AST rule evaluation and interpretable output: This step takes the evidence vector generated in S7 and the filtered candidate rule set as input, performs logical evaluation on each rule atom, and generates an interpretation chain that can be referenced by the comprehensive conclusion in S9.

[0038] Performs comparison evaluation on regular atoms, supporting AND / OR / NOT and weighted nodes; Output the hit rules, matching degree, number of evidence items, unsatisfied atoms, and unsupported atoms; Generate rule explanation text for use in the overall conclusion.

[0039] In the specific implementation process of this embodiment, step S9: multi-measurement point fusion and sorting: this step takes the rule hit, matching degree, severity and rule explanation text output by S8 as input, and performs device-level fusion and sorting of the conclusions of multiple measurement points.

[0040] First, sort by rule matching degree; In case of a tie, sort by severity (extreme > severe > moderate > mild); A list of overall equipment-level conclusions and sub-test point conclusions will be generated.

[0041] In the specific implementation of this embodiment, step S10: Output and Human-Computer Interaction: Output "equipment name, diagnostic location, fault type, fault severity, hit rule, and test point results"; The graph supports zooming, measurement point switching, and feature frequency visibility. The frequency matching table displays key fields: target frequency, peak frequency, peak value, deviation, PTN, energy percentage, harmonics, source, and hit status.

[0042] This embodiment organizes its technical solution around "data object interaction" and "evidence chain interaction." Data objects include a measurement point model, a preprocessed time-domain signal, an FFT candidate peak set P_fft, an ENV candidate peak set P_env, a target frequency set F_target, a frequency matching table, rule evidence vectors, rule evaluation results, and a comprehensive diagnostic conclusion. Each data object is transmitted in a fixed order, and the measurement point identifier, device level, and source field are retained.

[0043] Interactive Measurement Point Modeling: In S1, the user or system first establishes a hierarchical relationship between equipment name, key components, monitoring parts, and monitoring locations, and then binds sampling frequency, sampling duration, channel type, rotational speed, bearing parameters, and / or gear parameters to each measurement point. This measurement point identification is used throughout subsequent signal import, feature extraction, rule matching, and conclusion output, ensuring that every piece of evidence can be traced back to the specific monitoring location.

[0044] Signal processing interaction: In S2 to S4, the imported time-domain signal must be bound to the measurement point identifier defined in S1; the preprocessed signal is simultaneously input into the FFT spectrum analysis and ENV envelope demodulation links. The FFT outputs P_fft, and the ENV outputs P_env and the source identifier; when the envelope information is insufficient, the system backtracks in the order of bandpass ENV, direct-pass ENV, and FFT proxy evidence, thereby ensuring that subsequent frequency matching can still obtain interpretable evidence sources.

[0045] Interaction between target frequency and candidate peak: In S5, the system calculates fr, BPFO, BPFI, BSF, FTF, GMF, BPF and their harmonics or sidebands based on the rotational speed, bearing parameters and / or gear parameters bound in S1, and generates F_target; In S6, the system searches for the optimal candidate peak in P_fft and P_env for each target frequency in F_target, and outputs the target frequency, peak frequency, peak amplitude, deviation, PTN, peak energy percentage, harmonic hit, source and hit status.

[0046] Rule-based evidence interaction: In S7, the frequency matching table is converted into a rule-based evidence vector. The evidence vector includes at least the feature key, measured value, threshold, comparison method, whether it is satisfied, source, measurement point identifier, and explanatory text fragment; rule filtering first performs primary filtering by device name and key parts, and then sorts by monitoring parts and monitoring locations using soft constraints to avoid missed diagnoses due to differences in field naming.

[0047] Rule evaluation and interpretation interaction: In S8, the system parses the rule configuration into an AST logic tree, performs comparison evaluation on each rule atom, and supports AND, OR, NOT, weights, thresholds, and optional evidence nodes. The evaluation result not only indicates whether the rule is matched, but also outputs matched atoms, unsatisfied atoms, unsupported atoms, matching degree, and rule interpretation text, providing citationable reasons for the S9 synthesis conclusion.

[0048] Fusion Output Interaction: In S9 to S10, the system aggregates the hit results of similar rules by equipment name, key parts and fault type, and performs fusion sorting based on matching degree, severity, number of covered test points and reliability of evidence source. Finally, it outputs the equipment-level comprehensive conclusion, sub-test point conclusion, hit rule, diagnosis location, fault type, fault degree and evidence interpretation chain.

[0049] The specific implementation example of a single-stage dynamically adjustable fan with a dual-measuring-point bearing failure is as follows: Step 1. Modeling Configuration: Equipment Name = Single-Stage Dynamic Adjustment Fan; Measuring point 1 = Fan free end - rolling bearing - non-driving end; Measuring point 2 = Fan drive end - rolling bearing - drive end; Sampling parameters: fs=6400Hz, rpm=600 or 1500 (acceleration according to operating conditions), channel=acceleration.

[0050] Step 2. Bearing parameter entry: Measurement point 1 and measurement point 2 can share bearing parameters. If the models are the same, they can be reused with one click; if they are different, they should be entered separately.

[0051] Step 3. Signal processing: Preprocessing, FFT and ENV are performed at the two measurement points respectively; the system records each feature from FFT or ENV.

[0052] Step 4. Feature Hit: Measurement point 1 hits the BPFI / FTF area in ENV and the octave corroborates the findings. Combined with the loosening-related features, we obtain "moderate loosening". Measurement point 2 shows a high PTN hit in BPFI within ENV, which is corroborated by the presence of 2x in FFT, indicating "severe wear".

[0053] Step 5. Rule evaluation: Use "single-stage dynamic adjustment fan + fan" as the main filtering range, and use the monitoring parts / locations as soft constraints for sorting to avoid missing valid rules due to slight deviations in the modeling fields.

[0054] Step 6. Output: Conclusion of measurement points: Measurement point 1 - Fan free end - Rolling bearing - Data import = Fan free end rolling bearing is moderately loose; Measurement point 2 - Fan drive end - Rolling bearing - Data import = Severe wear of the rolling bearing at the fan drive end.

[0055] Overall conclusion: First sort by matching degree, then sort by severity for those with the same score, and output the device-level primary conclusion and secondary conclusion.

[0056] In this embodiment, the time-domain signals of the two measurement points are respectively bound to the measurement point identifier formed by S1. The system calculates P_fft and P_env for the two measurement points respectively, and then generates a frequency matching table and a rule evidence vector respectively; when measurement point 1 and measurement point 2 both point to the same critical part but have different fault degrees, the system determines the equipment-level main conclusion according to the matching degree and severity, while retaining their respective evidence interpretations in the sub-measurement point conclusions.

[0057] This embodiment is used to support subordinate protection points such as multi-measurement point modeling, measurement point parameter reuse, independent calculation of sub-measurement point evidence, equipment-level fusion sorting, and sub-measurement point conclusion output.

[0058] Triggered at a device-specific frequency: Taking the blade passage frequency BPF as a specific example, the details are as follows: Step 1. Enter BPF=400Hz in "Equipment-Specific Fault Frequency Setting" and automatically extend 2x / 3x harmonics; Step 2. After importing the test signal dominated by BPF only, the FFT shows a main peak at 400Hz, and ENV can be used as supplementary evidence; Step 3. Add a BPF row to the frequency matching table to display the target frequency, peak frequency, deviation, PTN, energy percentage, and source; Step 4. If the rule label is written as "guide blade passes through frequency harmonics" and the rule name is written as "wind turbine blade wear", the system performs semantic alignment through keyword mapping; Step 5. After evaluating the rule, output the corresponding level of "wind turbine blade wear" and provide BPF evidence entries in the explanation; Step 6. When the selected monitoring site does not completely match the rule words, it still participates in the matching within the key site range, so as not to cause missed diagnosis due to hard filtering.

[0059] In this embodiment, the BPF, as a device-specific frequency, is included in F_target and uses the same candidate peak search and joint hit criterion as the base bearing frequency. If the BPF forms the main peak in the FFT while the ENV only serves as supplementary evidence, the system marks the source field as FFT or FFT+ENV in the frequency matching table and explains the source of the BPF evidence in the rule explanation text.

[0060] This embodiment is used to support device-specific frequency settings, automatic frequency multiplication, keyword semantic mapping, and protection points that can still be evaluated even when the rule label and rule name are not completely consistent.

[0061] Taking the gearbox scenario as an example, the specific implementation is as follows: Step 1. When the equipment name / critical part / monitored part contains the keyword "gearbox", the system enables mandatory field verification for gear parameters; Step 2. After inputting the number of teeth z1 / z2, transmission ratio, and number of stages, the system will automatically calculate the GMF and sideband frequency (e.g., GMF±fr, 2GMF±fr). Step 3. Identify the GMF main peak in FFT and the sideband structure in ENV, and use the combined criteria to determine the hit. Step 4. Map evidence against rules in the rule base such as "poor gear meshing, gear wear, gearbox noise"; Step 5. When outputting, distinguish between "bearing-related fault evidence" and "gear-related fault evidence", and give the primary fault type and secondary fault type to avoid confusion.

[0062] In this embodiment, gearbox parameters trigger GMF and sideband frequency calculations. The system can identify the GMF main peak in the FFT and the modulation sideband in the ENV, and write them into the evidence vector as evidence from different sources. When bearing-related evidence and gear-related evidence exist simultaneously, the system clusters them according to fault type and outputs the order of primary and secondary faults.

[0063] This embodiment is used to support protection points such as mandatory verification of gear parameters, generation of GMF and sidebands, hierarchical output of bearing evidence and gear evidence, and sorting of primary and secondary faults.

[0064] Example 2 discloses a rule-based vibration fault diagnosis system for multi-measurement rotating machinery, the system comprising: The measurement point modeling module is used to establish a four-level structure at the equipment level, including equipment name, key parts, monitoring parts, and monitoring locations. It binds each measurement point to its corresponding part, sampling parameters, speed parameters, bearing parameters, and / or gear parameters, and generates a unique measurement point identifier. The data preprocessing module is used to import the vibration time-domain signal bound to the measuring point identifier, and perform mean removal, trend removal, unit conversion, filtering and / or noise reduction processing on the vibration time-domain signal to obtain the preprocessed signal. The spectrum feature analysis module is used to perform spectrum analysis on the preprocessed signal, extract the peak frequency, peak amplitude and local noise floor, and form a spectrum candidate peak set P_fft; The envelope feature analysis module is used to perform envelope demodulation on the preprocessed signal and generate an envelope candidate peak set P_env and corresponding source identifiers in the order of bandpass ENV priority, direct-pass ENV backoff, and FFT proxy backoff. The target frequency generation module is used to generate a target frequency set F_target consisting of the basic fault frequency, the equipment-specific fault frequency, and their multiples or sidebands, based on the bound bearing parameters, gear parameters, and speed parameters. The joint hit determination module is used to search for the optimal candidate peak in the spectrum candidate peak set P_fft and the envelope candidate peak set P_env for each target frequency in the target frequency set F_target. The frequency matching table is determined by the joint hit criteria composed of frequency difference tolerance, peak signal-to-noise ratio PTN, peak energy ratio and harmonic evidence. The rule-based evidence mapping module is used to perform rule-based primary filtering of the frequency matching table according to the device name and key parts, and to perform soft-constraint sorting and mapping according to the monitoring parts and monitoring locations into rule-based evidence vectors. The rule evaluation module is used to perform AND, OR, NOT and / or weighted node evaluation on the AST logic tree in the rule base based on the rule evidence vector, and output the hit rules, frequency matching degree, number of evidence items, unsatisfied atoms, unsupported atoms and interpretation chain; The fusion output module is used to merge and sort the diagnostic conclusions of multiple measurement points according to the priority of frequency matching degree and the severity priority when the matching degree is the same, so as to form a comprehensive diagnostic conclusion at the equipment level and a list of conclusions for each measurement point.

[0065] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A rule-based vibration fault diagnosis method for multi-measurement rotating machinery, characterized in that, include: S1: Bind each measuring point of the equipment to its corresponding part, sampling parameters, speed parameters, bearing parameters and / or gear parameters, and generate a unique measuring point identifier; S2: Import the vibration time-domain signal bound to the measuring point identifier, perform preprocessing on the vibration time-domain signal, and obtain the preprocessed signal; S3: Perform spectrum analysis on the preprocessed signal to extract peak frequency, peak amplitude and local noise floor, and form a spectrum candidate peak set P_fft; S4: Perform envelope demodulation on the preprocessed signal and generate a set of candidate envelope peaks P_env and corresponding source identifiers in a preset order; S5: Based on the parameters of the bound bearings, gears, and rotational speed, generate a target frequency set F_target consisting of the basic fault frequency, the equipment-specific fault frequency, and their harmonics or sidebands; S6: For each target frequency in the target frequency set F_target, search for the optimal candidate peak in the spectrum candidate peak set P_fft and the envelope candidate peak set P_env, and determine the frequency matching table using the joint hit criterion composed of frequency difference tolerance, peak signal-to-noise ratio PTN, peak energy ratio and harmonic evidence. S7: Perform rule-based primary filtering on the frequency matching table according to equipment name and key parts, and sort and map it into rule evidence vectors according to monitoring parts and monitoring locations using soft constraints; S8: Based on the rule evidence vector, perform AND, OR, NOT and / or weighted node evaluation on the AST logic tree in the rule base, and output the hit rule, frequency matching degree, number of evidence items, unsatisfied atoms, unsupported atoms and interpretation chain; S9: The diagnostic conclusions of multiple measurement points are merged and sorted according to the frequency matching degree first, and the severity first when the frequency matching degree is the same, to form a comprehensive equipment-level diagnostic conclusion and a list of conclusions for each measurement point.

2. The rule-based vibration fault diagnosis method for multi-measurement rotating machinery according to claim 1, characterized in that, Step S1 specifically includes: Construct a hierarchical relationship at the equipment level, including equipment name, key components, monitoring components, and monitoring locations; Each measurement point is bound to sampling parameters, which include at least the sampling frequency fs, sampling duration, channels, and rotational speed rpm; Each measuring point is bound to bearing parameters, which include at least the number of rolling elements n, the rolling element diameter d, the pitch circle diameter D, and the contact angle θ. Depending on the equipment type, the gear parameters can be selected and bound, and the gear parameters include at least the number of teeth z1 / z2, the transmission ratio, the number of stages, and the meshing stage relationship; The measurement point name is generated by concatenating the measurement point number, monitoring location, monitoring position, and data import status, and the measurement point status is saved to ensure a one-to-one correspondence between measurement points, locations, parameters, and data.

3. The rule-based vibration fault diagnosis method for multi-measurement rotating machinery according to claim 1, characterized in that, Step S2 specifically includes: Import vibration time-domain signals in xlsx, csv, txt and / or dat formats, and bind the vibration time-domain signals with the corresponding measurement point identifiers; Perform mean removal, trend removal, and unit conversion on the imported signal; Perform high-pass filtering, low-pass filtering, notch filtering and / or noise reduction processing according to preset processing parameters; The preprocessing parameters are persisted so that the processing status of the corresponding measurement points can be restored when refreshed or re-imported.

4. The rule-based vibration fault diagnosis method for multi-measurement rotating machinery according to claim 1, characterized in that, Step S5 specifically includes: The rotational frequency fr is generated based on the rotational speed. Generate the outer ring failure frequency (BPFO), inner ring failure frequency (BPFI), rolling element failure frequency (BSF), and cage failure frequency (FTF) based on the bearing parameters. Perform 2x to 5x frequency multiplication on fr, BPFO, BPFI, BSF, and FTF; The receiver uses a device-specific frequency and performs frequency multiplication, wherein the device-specific frequency includes at least the blade passage frequency (BPF). When the equipment name, key parts, or monitoring parts contain the keyword "gearbox", the gear meshing frequency GMF and at least one of the following sideband frequencies, GMF+fr, GMF-fr, 2GMF+fr, and 2GMF-fr, are generated based on the gear parameters.

5. The rule-based vibration fault diagnosis method for multi-measurement rotating machinery according to claim 1, characterized in that, Step S6 specifically includes: Frequency difference criterion: Calculate the deviation between the peak frequency f_peak and the target frequency f_target. When |f_peak-f_target|<=max(minimum tolerance Hz, tolerance ratio * f_target), the frequency difference condition is met. PTN criterion: Calculate the peak signal-to-noise ratio (PTN) = When PTN >= PTN_min, the peak signal-to-noise ratio condition is met; Energy criterion: Calculate the peak energy percentage. When the peak energy percentage >= E_min(feature, source), the energy condition is met. Frequency harmonic evidence: When the main criterion is in a critical state, if the number of hits in the 2x to 5x harmonics is >= k, then it is determined to be a harmonic support hit; The frequency matching table includes at least the target frequency, peak frequency, peak amplitude, deviation, PTN, peak energy percentage, harmonic hit, source, and hit status.

6. The rule-based vibration fault diagnosis method for multi-measurement rotating machinery according to claim 1, characterized in that, Step S3 specifically includes: The spectrum is calculated for each measurement point in the spectrum channel, and the peak frequency, peak amplitude and local noise floor are extracted to form a spectrum candidate peak set P_fft.

7. The rule-based vibration fault diagnosis method for multi-measurement rotating machinery according to claim 1, characterized in that, Step S4 specifically includes: In the envelope channel, the bandpass ENV is executed first. If the bandpass ENV is not informative enough, it falls back to the passthrough ENV. If the passthrough ENV is still not informative enough, FFT proxy evidence is enabled.

8. The rule-based vibration fault diagnosis method for multi-measurement rotating machinery according to claim 1, characterized in that, The source identifier is used to distinguish candidate peaks from FFT, bandpass ENV, direct-pass ENV, or FFT proxy evidence; the source identifier is retained synchronously in subsequent frequency hit determination and rule evaluation.

9. The rule-based vibration fault diagnosis method for multi-measurement rotating machinery according to claim 1, characterized in that, The rule-based evidence vector includes at least the feature key, measured value, threshold, whether it is satisfied, and its source; The equipment name and key components are used as the main filtering criteria for rules to determine the range of candidate rules. The monitoring location and monitoring position are used as soft constraints for ranking, rather than as hard elimination conditions.

10. A rule-based vibration fault diagnosis system for multi-measurement rotating machinery, characterized in that, The system includes: The measurement point modeling module is used to bind each measurement point of the equipment with its corresponding part, sampling parameters, speed parameters, bearing parameters and / or gear parameters, and generate a unique measurement point identifier; The data preprocessing module is used to import the vibration time-domain signal bound to the measuring point identifier, and perform mean removal, trend removal, unit conversion, filtering and / or noise reduction processing on the vibration time-domain signal to obtain the preprocessed signal. The spectrum feature analysis module is used to perform spectrum analysis on the preprocessed signal, extract the peak frequency, peak amplitude and local noise floor, and form a spectrum candidate peak set P_fft; The envelope feature analysis module is used to perform envelope demodulation on the preprocessed signal and generate an envelope candidate peak set P_env and corresponding source identifiers in the order of bandpass ENV priority, direct-pass ENV backoff, and FFT proxy backoff. The target frequency generation module is used to generate a target frequency set F_target consisting of the basic fault frequency, the equipment-specific fault frequency, and their multiples or sidebands, based on the bound bearing parameters, gear parameters, and speed parameters. The joint hit determination module is used to search for the optimal candidate peak in the spectrum candidate peak set P_fft and the envelope candidate peak set P_env for each target frequency in the target frequency set F_target. The frequency matching table is determined by the joint hit criteria composed of frequency difference tolerance, peak signal-to-noise ratio PTN, peak energy ratio and harmonic evidence. The rule-based evidence mapping module is used to perform rule-based primary filtering of the frequency matching table according to the device name and key parts, and to perform soft-constraint sorting and mapping according to the monitoring parts and monitoring locations into rule-based evidence vectors. The rule evaluation module is used to perform AND, OR, NOT and / or weighted node evaluation on the AST logic tree in the rule base based on the rule evidence vector, and output the hit rules, frequency matching degree, number of evidence items, unsatisfied atoms, unsupported atoms and interpretation chain; The fusion output module is used to merge and sort the diagnostic conclusions of multiple measurement points according to the priority of frequency matching degree and the severity priority when the matching degree is the same, so as to form a comprehensive diagnostic conclusion at the equipment level and a list of conclusions for each measurement point.

Images