Hydraulic gear pump fault prediction system based on multi-parameter fusion
The hydraulic gear pump fault prediction system, which integrates multiple parameters, solves the problems of false alarms and missed alarms in the prior art by using the serial decision logic of timing and amplitude matching degree, and realizes accurate early warning and rapid location of hydraulic gear pump faults.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- TAIZHOU JINFUJIA MASCH CO LTD
- Filing Date
- 2026-02-04
- Publication Date
- 2026-06-05
Smart Images
Figure CN122153565A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of fault prediction technology, specifically a fault prediction system for hydraulic gear pumps based on multi-parameter fusion. Background Technology
[0002] As the core power component of hydraulic transmission systems, hydraulic gear pumps are widely used in engineering machinery, industrial production equipment and other fields. Their operating status directly affects the stability and reliability of the entire system.
[0003] Current hydraulic gear pump fault prediction technology mainly relies on single parameter or simple multi-parameter superposition analysis, such as relying solely on vibration threshold or oil contamination degree for judgment, which has problems such as large interference, low sensitivity, and monitoring lag.
[0004] Existing technologies fail to consider the coordinated changes of parameters under different fault modes, and also fail to distinguish the characteristics of stable and impacted operating states of equipment. This leads to false alarms and missed alarms in early fault identification, making it difficult to meet the requirements for accurate early warning. Summary of the Invention
[0005] The purpose of this invention is to provide a hydraulic gear pump fault prediction system based on multi-parameter fusion to solve the problems mentioned in the background art.
[0006] A hydraulic gear pump fault prediction system based on multi-parameter fusion includes: The data acquisition module is used to acquire the real-time monitoring parameter set of the target hydraulic gear pump. The real-time monitoring parameter set includes at least vibration parameters, pressure parameters, and oil wear parameters. The knowledge base module stores a preset fault knowledge model, which is used to determine the key parameter set and benchmark coordinated change mode corresponding to the current fault mode to be diagnosed based on the real-time monitoring parameter set. Among them, the benchmark coordinated change pattern includes the change timing relationship and magnitude range of each parameter in the key parameter group when the failure mode occurs; The collaborative analysis module is used to calculate the temporal conformity and amplitude matching degree of key parameter groups based on real-time monitoring parameter sets and benchmark collaborative change patterns. The cascaded decision module is used to take timing compliance as the primary constraint and amplitude matching as the verification constraint, and calculate the actual synergy index of the key parameter group through cascaded decision logic. The cascaded decision logic requires that timing compliance first reach the first threshold before initiating the judgment on whether amplitude matching has reached the second threshold. The early warning output module is used to generate and output diagnostic early warning information for the fault mode when the actual coordination index exceeds the preset diagnostic threshold.
[0007] Specifically, the system effectively integrates multi-dimensional parameter information through a sequential decision logic of "time sequence first, amplitude verification," thereby avoiding false alarms caused by instantaneous anomalies or random fluctuations of a single parameter, thus improving the reliability and accuracy of fault prediction.
[0008] In some possible implementations, the specific methods by which the collaborative analysis module calculates timing compliance include: Identify the time points when each parameter in the key parameter group experiences anomalies; The identified time point sequence is compared with the expected time series defined in the changing time series relationship; If the identified time series is completely consistent with the expected time series in order, it is determined to be a strong temporal match; If the identified time series does not match the expected time series, a time series deep analysis process is executed. The time series deep analysis process includes: When the time sequence of some parameters is consistent with the expected time series, the determination is made based on the number and criticality of the consistent parameters; When the time sequence of all parameters is inconsistent with the expected time series, perform local time series pattern matching analysis.
[0009] This hierarchical and layered timing compliance analysis method enhances the fault tolerance capability for complex operating conditions and incomplete fault signals. Even when some parameters are disturbed or the initial characteristics of the fault are not obvious, it can effectively capture and evaluate potential fault timing patterns, thereby improving the robustness of the system.
[0010] In some possible implementations, local temporal pattern matching analysis includes: From the key parameter set, select a subset of parameters whose time point order is consistent with the expected time series; Determine whether a subset of parameters constitutes a complete causal transmission chain defined in a changing temporal relationship; If so, it is determined to be a weak timing match, and the parameter subset is identified as a valid fault symptom group. At the same time, the parameters not included in the parameter subset are marked as disturbed parameters.
[0011] This step, by identifying valid fault symptom clusters and interfered parameters, can isolate interference signals and focus on the core parameter combinations that truly reflect the causal chain of the fault, providing a clear and reliable data foundation for subsequent accurate diagnosis and interference source analysis.
[0012] In some possible implementations, the collaborative analysis module is also used to perform interference source inference: When the disturbed parameter is identified through the time series deep analysis process, based on the fault mode corresponding to the effective fault symptom group, as well as the type and expected performance of the disturbed parameter, and combined with the system composition and signal flow knowledge of the hydraulic gear pump, the potential interference source that caused the disturbed parameter to not perform as expected is deduced. Potential sources of interference are sent to the early warning output module as auxiliary diagnostic information.
[0013] In some possible implementations, the specific methods by which the collaborative analysis module calculates the magnitude matching degree include: Calculate the magnitude of abnormal changes for each parameter in the key parameter group; The abnormal changes in each parameter are compared according to the expected proportional relationship defined in the range of amplitude proportions; When the abnormal changes of each parameter fall within the corresponding range defined by the amplitude ratio range, it is determined to be a strong amplitude match; when the abnormal changes of the core parameter fall within the corresponding range defined by the amplitude ratio range, while the abnormal changes of the auxiliary parameter deviate from the corresponding range defined by the amplitude ratio range but the direction of change is the same, it is determined to be a weak amplitude match.
[0014] In some possible implementations, the serial decision logic of the serial decision module specifically includes: When the timing conformity is strong, the actual coordination index is directly determined by the level of amplitude matching. When the timing match is weak, the actual coordination index is only judged to exceed the diagnostic threshold when the amplitude match is strong.
[0015] Specifically, this decision-making logic reflects the principle of prioritizing temporal reliability. When the temporal characteristics are highly reliable, the system's decision-making is more sensitive to changes in magnitude; when the reliability of the temporal characteristics is moderate, then the evidence regarding magnitude must be very sufficient. This dynamic weighting strategy further optimizes the system's risk control capabilities under different confidence levels.
[0016] In some possible implementations, the execution of the cascaded decision module also depends on fault development stage markers; The fault development stage marker is dynamically updated based on the timing compliance determination result and is used to retrieve the refined amplitude ratio range corresponding to the current stage from the knowledge base module; the refined amplitude ratio range has a stricter allowable deviation than the initial amplitude ratio range.
[0017] This invention achieves adaptive and refined adjustment of fault warning thresholds by introducing the concept of "fault development stages" and dynamically applying more stringent stage-based criteria. Raising the judgment criteria at stages where fault characteristics are clearly defined can effectively filter out minor early fluctuations, prevent unreasonably inflated warning levels, and make warning information more practically instructive.
[0018] In some possible implementations, the knowledge base module stores at least two baseline collaborative change patterns under different operating states for the same fault mode. The system also includes a working condition identification module, which is used to identify the current operating status of the equipment based on a real-time monitoring parameter set, and the knowledge base module automatically selects a benchmark collaborative change mode that matches the current operating status of the equipment.
[0019] In some possible implementations, different operating states include at least a stable operating state and an impact-induced operating state; In the baseline coordinated change mode corresponding to the impacted operating state, the upper limit of the amplitude ratio range is preset to K times the upper limit corresponding to the stable operating state, where K is a constant greater than 1. The cascaded decision module only calculates and judges the amplitude matching degree after the timing compliance reaches the strong compliance level.
[0020] The amplitude tolerance (K times) has been specifically relaxed for impact conditions, but by mandating "strong timing conformity" as a prerequisite for activation, a clever balance between sensitivity and noise immunity is achieved. This ensures that under impact conditions, only signals with highly questionable timing patterns will enter the amplitude verification stage, thus avoiding frequent false alarms caused by impact noise while guaranteeing the timely detection of real faults.
[0021] In the system provided in this application, the time series compliance is a quantitative or qualitative evaluation result derived from a comparative analysis of the real-time monitoring parameter sequence and the expected time series relationship. Based on the accuracy of the matching, it can be divided into different levels, for example: Strong match: This means that the identified time series is completely consistent with the expected time series in terms of sequence.
[0022] Weak coincidence: This refers to a situation where, although a strong coincidence is not achieved, a subset of parameters constituting a complete causal transmission chain can be identified through in-depth analysis (such as local timing pattern matching), indicating that the timing pattern of the fault is partially revealed.
[0023] Other intermediate states: Based on the number of consistent parameters, their criticality, etc., other levels such as moderate compliance can be defined to more finely mark different stages of fault development.
[0024] The above classification aims to more accurately describe the significance of fault characteristics and provide a basis for subsequent cascade decisions and stage marking.
[0025] The technical solutions provided by the embodiments of this disclosure have at least the following beneficial effects: 1. Differentiated amplitude ratio ranges are set for two operating states: stable and impacted. The upper limit of the amplitude ratio range under impacted operation is set to K times that under stable operation. At the same time, more stringent and refined amplitude ratio ranges are configured for different fault development stages. This effectively solves the problem of judgment bias caused by the failure of traditional technology to consider the differences between equipment operating conditions and fault stages, and improves the adaptability and accuracy of fault diagnosis. 2. The amplitude matching degree is only calculated and judged when the temporal characteristics perfectly match the fault evolution pattern. In cases where the temporal characteristics do not strongly match, the decision-making process is directly simplified. This design avoids the risk of false alarms caused by instantaneous parameter fluctuations and operating condition interference from the root, while reducing meaningless calculation operations, making the system response more agile and the decision more reliable. 3. By identifying the affected parameters and based on valid fault symptom clusters, fault modes, system configuration, and signal flow knowledge, potential interference sources can be deduced. This effectively distinguishes whether parameter anomalies are caused by the evolution of actual faults or by external interference. This not only avoids misjudgment but also outputs potential interference sources as auxiliary diagnostic information, providing on-site maintenance personnel with clear directions for fault investigation and shortening fault location time. 4. When the timing does not fully meet expectations, it can uncover effective local timing features or assess the consistency of some parameters, thereby identifying potential fault symptom clusters and marking the disturbed parameters separately. This ensures the sensitivity of early fault detection and provides a data foundation for subsequent interference source analysis and accurate diagnosis. Attached Figure Description
[0026] Figure 1 This is a schematic diagram of the system framework structure of the present invention. Detailed Implementation
[0027] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0028] Please see Figure 1 This application provides a hydraulic gear pump fault prediction system based on multi-parameter fusion, including: The data acquisition module is used to acquire the real-time monitoring parameter set and working status identifier of the target hydraulic gear pump. The real-time monitoring parameter set includes at least vibration parameters, pressure parameters, and oil wear parameters. Different working states include at least stable operation state and impact operation state. The working status identifier is synchronously transmitted from the equipment control system to the data acquisition module and sent to subsequent modules along with the real-time monitoring parameter set as the basis for calling the benchmark coordinated change mode.
[0029] Specifically, the main body of the data acquisition module is the distributed data acquisition unit, which integrates multiple sensors and data transmission modules.
[0030] The data acquisition module collects vibration, pressure, and oil wear parameters of the hydraulic gear pump through an accelerometer, a pressure sensor, and an online oil monitoring sensor, respectively. It also ensures that the timestamps of all collected data are aligned through a built-in time synchronization module, forming a real-time monitoring parameter set with consistent timing.
[0031] It should be noted that, in order to ensure the consistency of multi-parameter timing, the data acquisition module has a built-in time synchronization module, which uses GPS timing or NTP protocol to align the timestamps of each sensor data and control time synchronization errors.
[0032] The collected real-time monitoring parameter set and working status identifier are transmitted to the system processing unit through the industrial Ethernet interface. The transmission protocol is ModbusTCP, which controls the data transmission delay and sets up a data caching mechanism.
[0033] Furthermore, the data acquisition module also has a parameter self-checking function. If abnormal sensor data or communication failure is detected, it will immediately generate a fault prompt message and upload it to the system backend, so that maintenance personnel can handle it in a timely manner.
[0034] The knowledge base module stores a preset fault knowledge model, which is used to determine the key parameter set and benchmark coordinated change mode corresponding to the current fault mode to be diagnosed and the working state based on the real-time monitoring parameter set and working status identifier. The benchmark coordinated change mode includes the temporal relationship of change of each parameter in the key parameter group when the fault mode occurs and the initial amplitude ratio range. In addition, the knowledge base module pre-stores the refined amplitude ratio range corresponding to each stage of the fault development. The refined amplitude ratio range has a stricter allowable deviation than the initial amplitude ratio range. In the benchmark coordinated change mode corresponding to the impact operation state, the upper limit of the amplitude ratio range is pre-set to K times the upper limit value corresponding to the stable operation state, where K is a constant greater than 1.
[0035] It should be noted that the value range of parameter K is set to 1.2-2.0. This range is based on statistical analysis of the variation law of failure parameter amplitude under typical impact conditions of hydraulic gear pumps, and can effectively cover common industrial scenarios from light impact to medium and strong impact.
[0036] Specifically, the knowledge base module adopts a distributed storage architecture, using a structured knowledge graph combined with a rule base to store and retrieve fault knowledge models. The fault knowledge models are built upon experimental and field operation data of hydraulic gear pumps under various typical operating conditions (e.g., stable operation and impact operation), generated through statistical analysis and machine learning training. Therefore, the constructed knowledge base pre-stores benchmark collaborative change patterns corresponding to different equipment operating states for the same fault mode, ensuring that the model and decision logic have inherent adaptability to different operating states.
[0037] It should be understood that the training of the fault knowledge model is based on labeled fault experimental data. Taking gear tooth surface wear as an example, the training dataset should contain at least the following information: Baseline data under normal operating conditions: each parameter was sampled continuously for at least 100 hours; Fault injection experimental data: records of at least 20 complete fault evolution processes; Parameter alignment: The timestamp alignment error for all parameters is ≤10ms; Fault tags: Fault start time and development stage markers accurate to the second; By statistically analyzing the parameter coordination change patterns in the above data, the temporal relationships and amplitude ratio ranges described in the examples are extracted to form a quantifiable and verifiable benchmark coordination change pattern.
[0038] The fault modes cover four core faults: gear tooth surface wear, bearing damage, seal leakage, and oil trapping. Each fault mode corresponds to a unique fault identification code, and each fault mode is divided into different fault development stages. Each stage is configured with a corresponding refined amplitude ratio range, which has a stricter allowable deviation than the initial amplitude ratio range.
[0039] Example illustration: Taking gear tooth surface wear failure as an example, a specific benchmark cooperative change mode can be defined as follows: Key parameter set: peak vibration acceleration, pressure pulsation amplitude, and ferrographic particle concentration.
[0040] The time sequence of changes is as follows: from the abnormal start of the peak vibration acceleration to (lag 30-120 seconds) the abnormal start of the pressure pulsation amplitude, and then to (lag 5-15 minutes) the abnormal start of the ferrographic particle concentration.
[0041] Initial amplitude ratio range under stable operation: Assuming the abnormal amplitude of the peak vibration acceleration is reference 1, the corresponding range for pressure pulsation amplitude is [0.6, 0.9], and the corresponding range for ferrographic particle concentration is [0.3, 0.5].
[0042] Adjustment under impact operation (K=1.5): According to the adjustment rule of the upper limit of the amplitude ratio range (upper limit of stable operation × K), the upper limit of the pressure pulsation amplitude range is expanded from 0.9 to 0.9×1.5=1.35, and the upper limit of the ferrographic particle concentration range is expanded from 0.5 to 0.5×1.5=0.75.
[0043] As should be understood, an example of constructing a fault knowledge model is as follows: To enable those skilled in the art to construct the fault knowledge model necessary for realizing the present invention, the following uses two typical faults, gear tooth surface wear and bearing rolling element spalling, as examples to fully illustrate the process of determining their baseline cooperative change mode.
[0044] For the target model of hydraulic gear pump, a fault injection experiment was conducted on a test bench to obtain complete data from normal state to failure.
[0045] Gear tooth surface wear test: Simulated pitting defects are machined on the tooth surface of the test gear, and the test is repeated at least 8 times.
[0046] Bearing rolling element spalling test: Install a bearing with pre-existing fatigue cracks and perform at least 8 repeated tests.
[0047] Each experiment involved synchronous and continuous acquisition of all monitoring parameters, including vibration, pressure, and oil levels, and precise recording of the initial moment when the fault was introduced artificially.
[0048] Determine the key parameter set (taking gear tooth surface wear as an example): Steps: Collect commonly used features of three types of parameters—vibration, pressure, and oil—as a candidate set. Use the random forest classification algorithm, labeling the experimental data as "normal," "early failure," and "severe failure," to train the model and calculate the importance score of each parameter feature for classification.
[0049] Results: The parameter with the highest importance score from each category was selected to form a key parameter group. Based on the analysis of experimental data, the key parameter group for "gear tooth surface wear" was determined to be: peak vibration acceleration, pressure pulsation amplitude, and ferrographic abrasive particle concentration.
[0050] Similarly, the key parameter set for "bearing rolling element spalling" determined using the same method is: effective vibration value, vibration peak factor, and oil moisture content.
[0051] Determine the temporal relationship of the changes: Steps: For the key parameter sets of data in each experimental case, a sliding window analysis of variance combined with the 3σ criterion is used to identify the precise time point at which each parameter first shows a persistent abnormal deviation.
[0052] Statistics: Calculate the time delay between the outlier of the next parameter and the outlier of the previous parameter in all valid experimental cases.
[0053] Results: The distribution of these delay times was statistically analyzed, and the time interval in which 80% of the experimental cases fell was taken as the time sequence relationship of the fault.
[0054] According to the statistical data of gear tooth surface wear test, the time sequence relationship is described as follows: after the abnormal start of the peak vibration acceleration, after 40 to 110 seconds, the abnormal start of the pressure pulsation amplitude occurs; after the abnormal start of the pressure pulsation amplitude, after 5 to 15 minutes, the abnormal start of the ferrographic abrasive concentration occurs.
[0055] According to the statistical data of the bearing rolling element spalling test, its time sequence relationship is described as follows: after the abnormal start of the vibration effective value, the vibration peak factor shows an abnormal start after 1 to 4 minutes; after the abnormal start of the vibration peak factor, the oil moisture content shows an abnormal start after 10 to 25 minutes.
[0056] Determine the range of amplitude proportions: Steps: After each parameter starts to show abnormality, take the monitoring data for the next 10 minutes and calculate the average change relative to the normal baseline mean.
[0057] Calculation and Statistics: Using the magnitude of change of the first abnormal parameter as a baseline, calculate the ratio of the magnitudes of change of other parameters to it. Statistically analyze the distribution of these ratios across all experimental cases.
[0058] Result: The middle 90% of these ratios is taken as the range of magnitude of the fault under stable operating conditions.
[0059] According to the statistical data of gear tooth surface wear test, the amplitude ratio range is as follows: the ratio of pressure pulsation amplitude change to vibration acceleration peak value change is between 0.58 and 0.88; the ratio of ferrographic abrasive particle concentration change to vibration acceleration peak value change is between 0.25 and 0.48.
[0060] According to the statistical data of bearing rolling element spalling test, the amplitude ratio range is as follows: the ratio of the change in vibration peak factor to the change in vibration effective value is between 1.8 and 3.2; the ratio of the change in oil moisture content to the change in vibration effective value is between 0.08 and 0.18.
[0061] It should be understood that the construction of the "baseline coordinated change mode" for other failure modes such as "seal leakage" and "oil trapping" mentioned above follows the same method and process as described in this chapter.
[0062] It should be noted that the aforementioned "bearing damage" encompasses common failure modes including rolling element spalling and inner / outer ring spalling. This embodiment uses "rolling element spalling" as a typical example to fully reveal a general method for constructing a "baseline co-variation mode" for any specific bearing failure mode (i.e., obtaining data through targeted experiments and following the aforementioned feature screening, time-series statistics, and amplitude statistics process). Those skilled in the art can use this general method to construct corresponding rules for other specific modes under the broad category of "bearing damage" without any inventive effort.
[0063] The determination of key parameter groups is not fixed, but rather based on the specific dynamic screening of fault modes. Moreover, the screening rules are not related to the working state of the hydraulic gear pump, ensuring the consistency of core parameters for fault judgment under different states.
[0064] In one implementation of this step, a feature importance ranking algorithm is used. A random forest model is used to calculate the contribution weight of each monitoring parameter to different failure modes. Parameters with higher contribution weights to the failure modes are selected to form a key parameter group. The specific weight thresholds are determined through feature importance analysis. For example, the key parameter group corresponding to gear tooth surface wear includes peak vibration acceleration, pressure pulsation amplitude, and ferrographic particle concentration; the key parameter group corresponding to bearing damage includes effective vibration value, vibration peak factor, and moisture content.
[0065] In another approach, an initial set of key parameters can be preset based on expert experience, and then optimized iteratively using on-site data to remove parameters with lower contribution weights, ensuring the relevance and effectiveness of the key parameter set.
[0066] The temporal relationship of the baseline coordinated change mode is determined through time series correlation analysis. The dynamic time warping algorithm is used to calculate the time delay of parameter changes. The setting of the temporal relationship is independent of the working state of the hydraulic gear pump, ensuring that the judgment standard of the temporal characteristics of fault evolution under different states is consistent. For example, in the gear tooth surface wear fault, the time when the ferrographic particle concentration begins to rise lags behind the time when the vibration acceleration peak appears abnormal, forming a specific temporal relationship. The initial amplitude ratio range under stable operating conditions is determined by statistically analyzing the ratio of the changes in the amplitude of each parameter during the fault development process. For example, for different fault modes (such as gear tooth surface wear), there is a specific proportional relationship between the changes in the amplitude of their key parameters, and this proportional relationship will be adaptively adjusted according to whether the equipment is in a stable or impacted operating state (such as increasing the upper limit of the proportional range by K times under impacted conditions).
[0067] It should be understood that the baseline collaborative change model is not static. As system operation data accumulates, it can be dynamically updated using incremental learning algorithms (such as online random forests and incremental clustering algorithms). Update triggering conditions include: the number of new failure cases reaching a preset threshold (e.g., 10 cases) or the system operation time accumulating to a certain period (e.g., 3 months). Before updating, data verification and expert review are required to ensure the reliability of the new model. The update process should simultaneously cover the amplitude ratio range of the two working states to make the model more closely reflect actual operating conditions.
[0068] Furthermore, the knowledge base module supports online updates. As the system accumulates new fault case data, the fault knowledge model can be iteratively optimized through manual review or automatic learning. The update cycle can be set as needed, and historical versions are retained after each update, supporting rollback operations.
[0069] From the perspective of practical application scenarios, if a new gear scuffing failure mode is added, the corresponding key parameter set can be trained and generated by supplementing the experimental data and field data of this failure, such as vibration kurtosis, pressure fluctuation frequency, oil viscosity change rate and benchmark coordinated change mode. At the same time, the different development stages of the failure can be divided and the corresponding refined amplitude ratio range can be configured. The upper limit value of the amplitude ratio range can be set for stable operation and impact operation respectively. After being entered into the knowledge base module, the failure mode can be extended and compatible.
[0070] The collaborative analysis module calculates the timing compliance of key parameter groups based on real-time monitoring parameter sets, operational status indicators, and baseline collaborative change patterns. The system introduces the concept of fault development stages; after determining the timing compliance, this module dynamically generates a fault development stage marker based on the determination result. This marker is not only sent to the serial decision module to provide a basis for its decision execution; More importantly, it directly determines which judgment criteria the system retrieves from the knowledge base. Specifically, it retrieves a more stringent amplitude ratio range corresponding to that particular development stage, one that allows for a stricter deviation than the initial amplitude ratio range, as the basis for subsequent amplitude matching degree calculations. The collaborative analysis module, according to the instructions of the serial decision module, retrieves the aforementioned determined amplitude ratio range to calculate the amplitude matching degree.
[0071] Specifically, the timing compliance is calculated with reference to the timing relationship in the baseline cooperative change mode and is implemented using the timing alignment algorithm. The criteria for determining timing compliance are not related to the working state of the hydraulic gear pump, ensuring the consistency of fault timing characteristic determination under different states.
[0072] First, the abnormal start time point of each parameter in the key parameter group is extracted. The abnormal start time point is determined by the 3σ criterion, where σ is the standard deviation of the parameter’s historical data under normal operating conditions. The abnormal judgment threshold is usually set to ±3σ. If the data sampling frequency is high, sliding window variance analysis can be used to assist in the judgment. The recommended window length is 10 to 50 sampling points. This criterion is widely used in industrial monitoring to identify effective abnormal signals from background noise.
[0073] The specific methods by which the collaborative analysis module calculates the time series compliance include: identifying the time points when each parameter in the key parameter group experiences anomalies; It should be understood that, in addition to the core 3σ criterion, the identification process also incorporates the slope of parameter change trend to assist in the judgment. When the slope of parameter change exceeds the preset abnormal trend threshold, it is marked as a potential abnormal starting point in advance. This threshold is set based on the fluctuation characteristics of normal operating parameters and is subsequently confirmed in conjunction with the 3σ criterion. This design can improve the sensitivity of early abnormal time point identification and avoid the lag caused by a single criterion.
[0074] The identified time point sequence is compared with the expected time series defined in the changing time series relationship; It should be noted that the sequence alignment algorithm is used to calculate the order matching degree of the two during the comparison. The order matching degree = number of parameter pairs with consistent order / total number of parameter pairs. The total number of parameter pairs is the number of pairwise combinations of parameters in the key parameter group. This formula can quantify the degree of fit of the sequence order and provide a quantitative basis for subsequent judgment.
[0075] If the identified time series is completely consistent with the expected time series in order, it is determined to be a strong temporal match; If the identified time series is inconsistent with the expected time series, then the time series deep analysis process is executed; Understandably, this design takes into account that fault evolution may be affected by interference in actual working conditions, leading to sequence deviation. Through in-depth analysis, it avoids direct judgment as non-compliance and improves the fault identification tolerance.
[0076] The time series in-depth analysis process includes: when the time sequence of some parameters is consistent with the expected time series, a determination is made based on the number and criticality of the consistent parameters; As an example of an achievable decision rule: First, count the total number of parameters N in the key parameter group; Calculate the number M of parameters whose time point sequence matches the expectation; Obtain the pre-stored criticality weight Wi for each parameter from the knowledge base module; The criticality weighted sum of the consistency parameters is calculated as S = Σ(Wi of the consistency parameters).
[0077] Then the following judgment rule is applied: if M / N≥2 / 3 and S≥0.7, it is judged as "moderately satisfactory"; If M / N≥1 / 3 and S≥0.4, then it is determined to be a weak fit; Otherwise, it is judged as non-compliant. The above thresholds (2 / 3, 0.7, 1 / 3, 0.4) can be adjusted according to specific application scenarios.
[0078] Specifically, firstly, the proportion of consistent parameters is calculated as the number of consistent parameters / the total number of parameters in the key parameter group. At the same time, based on the preset parameter keyness weights in the knowledge base module, the weighted sum of the keyness of consistent parameters is calculated as Σ(keyness weight of consistent parameter i × 1). This weight allocation is determined based on the priority of the impact of parameter keyness on fault diagnosis.
[0079] When the time sequence of all parameters is inconsistent with the expected time series, perform local time series pattern matching analysis. An example of local temporal pattern matching analysis is as follows: For each failure mode, a “core causal chain” is predefined in its baseline co-variation mode (e.g., gear wear is “vibration to pressure”).
[0080] During the analysis, a subset of parameters whose time sequence matches the expectation is selected from the key parameter group, and it is determined whether the subset fully contains and conforms to the order of the "core causal chain".
[0081] If yes, it is determined as "weak timing match (partial match)" and the subset is identified as a valid fault symptom group, while the remaining parameters are marked as disturbed parameters; if no, it is determined as "non-match".
[0082] Furthermore, this analysis first divides the time-point sequence into multiple continuous local time-series segments, then calculates the matching degree between each local segment and the corresponding segment of the expected sequence. The local matching degree is calculated as 1 - (dynamic time warping distance / maximum allowable distance). The maximum allowable distance is a certain proportion of the baseline time-series span of the segment. Finally, the overall time-series compliance degree St2 is obtained by weighting and summing the segments according to their weights. The segment weights are set according to the criticality of the included parameters. This design can uncover effective local time-series features and avoid ignoring potential fault signals due to overall sequence disorder.
[0083] Meanwhile, the collaborative analysis module dynamically updates the fault development stage markers based on the timing conformity determination results. When the timing conformity is strong, it is marked as the mid-stage of fault development; when the timing conformity is moderate, it is marked as the early stage of fault development; when the timing conformity is weak, it is marked as the potential budding stage of fault; and after local timing pattern matching analysis, it is marked as the stage where fault characteristics are unclear. Different markers correspond to different development stages of the fault, providing a staged basis for subsequent amplitude matching degree calculation and series decision.
[0084] It should be noted that the collaborative analysis module adopts a high-performance parallel computing architecture, which can quickly complete the calculation of timing compliance and amplitude matching degree, thus meeting the system's real-time requirements.
[0085] An outlier removal mechanism is set up during the calculation process. When there is a significant abnormal jump in the data of a certain parameter, linear interpolation is used to repair the data and ensure the accuracy of the calculation results.
[0086] Let's assume another scenario: if a parameter experiences a data jump due to a momentary sensor malfunction, the linear interpolation method can repair it based on the average of the 10 sampling points before and after that parameter, thus avoiding the impact of a single abnormal data point on the overall calculation result.
[0087] The core of the cascaded decision module's decision mechanism lies in executing a strict condition-triggered cascaded decision logic. This module is configured to first verify whether the timing compliance meets the condition that serves as the dominant constraint; The module will only trigger subsequent processes if and only if the primary condition of strong compliance is met. This involves calling the corresponding amplitude ratio range based on the working status identifier and the fault development stage marker, and initiating the calculation and judgment of the amplitude matching degree. The fault development stage marker, working status identifier, timing compliance, and amplitude matching degree together serve as the execution basis for this cascaded decision logic, used to calculate the actual coordination index of the key parameter group. The serial decision logic requires that the timing compliance level reaches a strong compliance level before initiating the calculation and judgment of the amplitude matching degree based on the corresponding amplitude ratio range according to the working status indicator. The serial decision logic of the serial decision module specifically includes: When the timing conformity is strong, the actual coordination index is directly determined by the level of amplitude matching. When the timing conformity is weakly consistent, the actual coordination index is only judged to exceed the diagnostic threshold if the amplitude conformity is strong. If the timing conformity does not reach the strong conformity level, the amplitude conformity calculation is not initiated, and the judgment is executed directly according to the corresponding rules.
[0088] Specifically, after receiving the timing compliance judgment result, fault development stage marker, and working status identifier sent by the collaborative analysis module, the serial decision module first performs a timing compliance level judgment. Only when the timing compliance reaches a strong compliance level will it send an amplitude matching degree calculation instruction to the collaborative analysis module. At the same time, based on the working status identifier and fault development stage marker, it retrieves the refined amplitude ratio range corresponding to the current working state and fault development stage from the knowledge base module and sends it to the collaborative analysis module as the basis for its amplitude matching degree calculation. In the refined amplitude ratio range corresponding to the impacted operating state, the upper limit of the amplitude ratio range is K times the upper limit value corresponding to the stable operating state, ensuring that the amplitude matching degree calculation is adapted to the actual operating state of the equipment.
[0089] After receiving the amplitude matching degree calculation instruction and the corresponding refined amplitude ratio range, the collaborative analysis module immediately calculates the amplitude matching degree of the key parameter group based on the refined amplitude ratio range. The specific method includes: calculating the abnormal change amplitude of each parameter in the key parameter group. It should be understood that the abnormal change amplitude of the parameter is the absolute value of the difference between the actual monitored value after the abnormal start time point and the average value of the parameter during the normal operating range. Before calculation, the original monitoring data should be smoothed. The moving average method is used to smooth the monitoring data. The size of the smoothing window is set according to the sampling frequency and the duration of typical noise. It is generally recommended that the window length be 5 to 15 sampling points. If the sampling frequency is 1kHz, the window length can be set to 10 sampling points (i.e. 10ms) to balance the filtering effect and response speed, so as to ensure effective noise filtering without masking fault characteristics. This conforms to the conventional design principles of industrial equipment parameter preprocessing, avoids amplitude calculation errors caused by instantaneous data fluctuations, ensures the authenticity of abnormal change amplitudes, and the smoothing processing rules are unrelated to the equipment's working status.
[0090] The abnormal variation amplitudes of each parameter are compared with the expected proportional relationship defined in the refined amplitude ratio range corresponding to the current working state and fault development stage. It should be noted that the abnormal variation amplitudes of each parameter need to be normalized before comparison. After normalization, the values all fall within the range of 0 to 1 to eliminate the influence of the difference in the units of different parameters on the proportional relationship comparison. The normalization formula is: Actual variation amplitude normalized i = (Actual variation amplitude i - Normal minimum variation amplitude i) / (Fault maximum variation amplitude i - Normal minimum variation amplitude i). Then, the parameter-by-parameter comparison is performed based on the normalized values and the expected proportional relationship of the refined amplitude ratio range. The accuracy of amplitude matching is improved by using stricter allowable deviations. Under the impact operation state, the upper limit value is increased, and the adaptability of the normalized value range is stronger.
[0091] When the abnormal changes in each parameter all fall within the corresponding range defined by the refined amplitude ratio range, it is determined to be a strong amplitude match; When the abnormal change in the core parameter falls within the corresponding range defined by the refinement range ratio, while the abnormal change in the auxiliary parameter deviates from the corresponding range defined by the refinement range ratio but the direction of change is the same, it is judged as a weak amplitude match. After the collaborative analysis module completes the amplitude matching degree calculation, it sends the result to the serial decision module in real time as the basis for the actual collaborative index calculation.
[0092] After receiving the amplitude matching degree calculation result, the cascaded decision module first verifies whether the refined amplitude ratio range used in the result matches the current working state and fault development stage, ensuring the consistency of the decision basis. Then, it executes the corresponding cascaded decision logic according to the timing compliance level, amplitude matching degree level, and fault development stage marker. The first threshold (timing compliance threshold) and the second threshold (amplitude matching threshold) are determined through receiver operating characteristic (ROC) curve analysis. Specifically, based on historical fault datasets, the true positive rate and false positive rate of timing compliance and amplitude matching degree under different thresholds are calculated for different fault modes. The threshold that maximizes the Youden index is selected as the optimal threshold, and the threshold setting is independent of the equipment's working state.
[0093] As an optional example: Analysis of 500 sets of historical data (including 200 failure cases and 300 normal cases) regarding gear tooth surface wear faults yielded the following results: The first threshold for time sequence compliance is calculated based on the order matching degree (number of parameter pairs with consistent order / total number of parameter pairs) as defined above, and its threshold is set to ≥0.85. The second threshold for amplitude matching is calculated based on the deviation between the abnormal change amplitude after normalization and the expected proportion, and its threshold is set to ≤0.25. Actual collaborative index diagnostic threshold (Level 3 warning): ≥0.7 (corresponding to a failure probability >85%) The above thresholds can be fine-tuned according to the specific equipment model and operating conditions, but the adjustment range shall not exceed ±20% of the initial value.
[0094] The sequential decision module executes decision logic based on the timing compliance level: only when the timing compliance is determined to be strong compliance will the calculation and judgment of the amplitude matching degree be initiated, and the actual coordination index be determined based on the amplitude matching result (strong matching or weak matching). If the timing compliance is moderate compliance, weak compliance, or partial compliance, it is directly determined that the actual coordination index has not exceeded the diagnostic threshold, no warning output is issued, the system switches to continuous monitoring mode and records the relevant status; The execution process of the serial decision logic adopts a strict condition triggering mechanism. First, it receives the timing compliance judgment result, the fault development stage mark and the working status mark. The first step is to determine whether the timing compliance has reached the strong compliance level. If so, it calls the corresponding refined amplitude ratio range and sends the amplitude matching degree calculation instruction. After receiving the amplitude matching degree result, it performs mapping assignment. If not, the decision rule of no amplitude matching is executed directly based on the timing compliance level and fault development stage marker, without initiating any amplitude-related calculations and judgments.
[0095] Meanwhile, the serial decision module is equipped with a threshold adjustment interface and a K-value calibration interface, allowing users to calibrate the K-value between 1.2 and 2.0 according to the operating characteristics of different models of hydraulic gear pumps. The adjusted and calibrated parameters are automatically synchronized to the global parameter configuration of the system. Furthermore, the K-value can be calibrated in segments according to different impact intensity levels, improving the adaptability to different impact conditions.
[0096] The module also has a built-in judgment log recording function, which records in detail the working status identifier, K value, fault development stage marker, timing compliance level, amplitude matching level (if any), actual coordination index and judgment result of each judgment. The log is retained for a set duration. By adopting the above method, the rigor, flexibility and traceability of the judgment logic can be achieved, meeting the fault judgment needs of different application scenarios and different working states.
[0097] The early warning output module is used to generate and output diagnostic early warning information for the fault mode when the actual coordination index exceeds the preset diagnostic threshold.
[0098] Specifically, the diagnostic thresholds are set by classifying the severity of the fault and adopting a three-level early warning mechanism. The diagnostic thresholds for each level are set based on the correlation analysis between the severity of the hydraulic gear pump fault and the coordination index, corresponding to three fault states: early warning, mid-term warning, and late emergency. It should be noted that the above threshold settings are based on the correlation analysis between the severity of hydraulic gear pump failures and the synergy index. This threshold division is highly compatible with the failure handling needs of industrial sites, making it easier for maintenance personnel to take targeted measures according to the warning level. The first-level warning corresponds to early minor failures, the second-level warning corresponds to mid-stage development failures, and the third-level warning corresponds to late-stage severe failures. In addition, the warning information will simultaneously mark the failure development stage and the working status, providing more accurate staged and status-based references for on-site maintenance.
[0099] It should be noted that when the actual coordination index exceeds the corresponding diagnostic threshold, the early warning output module first generates standardized diagnostic early warning information. This information includes the fault mode identifier, fault occurrence time, operating status identifier, K-value, fault development stage marker, timing compliance level of key parameter groups, amplitude matching level, actual coordination index, early warning level, and recommended handling measures. For example, for a Level 1 early warning marked as mid-fault development and operating in an impact-induced state, the recommended handling measures are to immediately reduce equipment load, exit the impact-induced operating state, and increase monitoring frequency. For a Level 2 warning, marked as mid-stage of fault development, and operating in a stable state, the recommended handling measures are to shut down the machine to inspect key components, locate the fault, and perform preliminary handling. When a Level 3 warning is issued and the system is marked as being in the middle stage of a fault development, regardless of the operating status, the recommended course of action is to immediately shut down the machine, replace the faulty component, and complete a comprehensive overhaul.
[0100] Furthermore, the diagnostic warning information may also include a link to a parameter change trend chart, making it easier for technicians to view detailed data.
[0101] In one implementation, the early warning information is output in two ways: local output and remote output. Local output is achieved through the system's audible and visual alarm and LCD display. Different early warning levels correspond to different audible and visual signals: Level 1 early warning is indicated by a flashing green indicator light and a low-frequency buzzer; Level 2 early warning is indicated by a solid yellow indicator light and a medium-frequency buzzer; and Level 3 early warning is indicated by a solid red indicator light and a high-frequency buzzer. The LCD display will also simultaneously display the working status indicator and the fault development stage marker. Remote output is achieved through an industrial IoT platform, which supports sending early warning information to designated managers' mobile apps and email addresses. It can also connect to the equipment management system to automatically generate maintenance work orders. The work orders simultaneously carry working status identifiers, K-values, and fault development stage markers, making it easier for maintenance personnel to prepare maintenance plans in advance based on the equipment's working status.
[0102] In another implementation, the output channels can be expanded according to user needs, such as sending warning text messages through an SMS gateway or connecting to the large screen display system of the factory monitoring center.
[0103] It should be understood that the early warning output module is equipped with a false alarm suppression mechanism. It will only officially output early warning information when the actual coordination index calculated multiple times in succession exceeds the diagnostic threshold, in order to avoid false alarms caused by a single abnormal data.
[0104] Meanwhile, it supports the warning cancellation function. When the calculation results are lower than the diagnostic threshold for multiple consecutive times, the warning status will be automatically cancelled. If the timing compliance does not reach the strong compliance level, the current warning status will be automatically cancelled and a warning cancellation notification will be generated. The notification will simultaneously mark the latest working status and fault development stage.
[0105] This invention constructs a benchmark collaborative change mode that links three dimensions: fault mode, fault development stage, and equipment operating status. It sets differentiated amplitude ratio ranges for two operating states: stable and impacted. The upper limit of the amplitude ratio range under impacted operating state is set to K times that of stable operating state. At the same time, it configures more stringent and refined amplitude ratio ranges for different fault development stages. This solves the problem of judgment bias caused by the difference between equipment operating conditions and fault stages in traditional technology, and achieves accurate matching of fault characteristics in all scenarios. The innovative design features a serial decision logic that is strongly time-matched and triggers only. The calculation and judgment of the amplitude matching degree are only initiated when the time characteristics completely match the fault evolution pattern. In non-time-matched states, the decision process is directly simplified. This not only avoids the risk of false alarms caused by instantaneous parameter fluctuations and operating condition interference from the root, but also reduces meaningless calculation operations, which greatly improves the system's fault decision efficiency and anti-interference capability. By combining fault development stage markers with equipment operating status indicators, a two-dimensional precise early warning system was constructed. The three-level early warning level is deeply linked to the development stages of fault initiation, early stage, and middle stage, as well as the actual operating status of the equipment. It is accompanied by targeted operation and maintenance suggestions, replacing the traditional one-size-fits-all early warning method, thus making operation and maintenance more instructive and realizing precise control of faults throughout the entire lifecycle from initiation monitoring to late-stage handling.
[0106] As one embodiment of the present invention, the collaborative analysis module can also provide an interference source auxiliary prompt function: When the interference parameter is identified through the time series deep analysis process, the system queries a preset "Common Interference-Phenomenon Mapping Table" based on the type of the interference parameter and the deviation characteristics between its actual performance and expected performance, thereby generating potential interference source prompt information.
[0107] The "Common Interference-Phenomenon Mapping Table" is based on engineering experience summarizing common interferences in hydraulic systems, and its exemplary entries are as follows: Types of Interference Parameters Observational bias characteristics Potential sources / causes of interference Suggestions for Assisting Investigation Vibration parameters The abnormal peak was earlier and shorter than expected. External instantaneous mechanical shock (such as sudden load change) Check the condition of the load equipment and coupling. Pressure parameters The abnormal lag and fluctuations were far less than expected. Pressure sensor with slow response or excessive damping Check if the sensor's front pressure tube is blocked. Oil wear parameters No abnormal changes Online oil level monitoring sensor sampling failure Check the oil sensor sampling port and pipeline (General) All parameter timings are completely disordered. Data acquisition system clock synchronization failure Check the time synchronization module and network When generating the prompt message, the system sends the item with the highest matching degree (including "potential interference source / cause" and "auxiliary troubleshooting suggestions") to the early warning output module as auxiliary diagnostic information. It should be noted that this prompt message is for reference only and is intended to provide troubleshooting ideas for operations and maintenance personnel; its accuracy should not be used as the basis for core system fault diagnosis decisions.
[0108] For ease of understanding, taking gear scuffing failure as an example, its key parameter set can be set as: vibration kurtosis, pressure fluctuation frequency, and oil viscosity change rate. Under stable operating conditions, the timing relationship is: vibration kurtosis anomaly precedes pressure fluctuation frequency anomaly, and oil viscosity change rate lags behind the former two. Within the amplitude ratio range, the abnormal amplitude ratio of vibration kurtosis to pressure fluctuation frequency is typically 1:0.5~1.2. Under impact conditions, the upper limit of this ratio can be relaxed to 1.5 times (K=1.5). After detecting parameter anomalies that conform to this timing relationship, the collaborative analysis module initiates amplitude matching judgment. If the matching is successful, it outputs gear scuffing warning information.
[0109] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A hydraulic gear pump fault prediction system based on multi-parameter fusion, characterized in that, include: The data acquisition module is used to acquire the real-time monitoring parameter set of the target hydraulic gear pump. The real-time monitoring parameter set includes at least vibration parameters, pressure parameters, and oil wear parameters. The knowledge base module stores a preset fault knowledge model, which is used to determine the key parameter group and benchmark coordinated change mode corresponding to the current fault mode to be diagnosed based on the real-time monitoring parameter set; wherein, the benchmark coordinated change mode includes the change sequence relationship and amplitude ratio range followed by each parameter in the key parameter group when the fault mode occurs. The collaborative analysis module is used to calculate the temporal conformity and amplitude matching degree of key parameter groups based on real-time monitoring parameter sets and benchmark collaborative change patterns. The cascaded decision module is used to take timing compliance as the primary constraint and amplitude matching as the verification constraint, and calculate the actual synergy index of the key parameter group through cascaded decision logic. The cascaded decision logic requires that timing compliance first reach the first threshold before initiating the judgment on whether amplitude matching has reached the second threshold. The early warning output module is used to generate and output diagnostic early warning information for the fault mode when the actual coordination index exceeds the preset diagnostic threshold.
2. The hydraulic gear pump fault prediction system based on multi-parameter fusion according to claim 1, characterized in that, The specific methods used by the collaborative analysis module to calculate timing compliance include: Identify the time points when each parameter in the key parameter group experiences anomalies; The identified time point sequence is compared with the expected time series defined in the changing time series relationship; If the identified time series is completely consistent with the expected time series in order, it is determined to be a strong temporal match; If the identified time series does not match the expected time series, a time series deep analysis process is executed. The time series deep analysis process includes: When the time sequence of some parameters is consistent with the expected time series, the determination is made based on the number and criticality of the consistent parameters; When the time sequence of all parameters is inconsistent with the expected time series, perform local time series pattern matching analysis.
3. The hydraulic gear pump fault prediction system based on multi-parameter fusion according to claim 2, characterized in that, Local temporal pattern matching analysis includes: From the key parameter set, select a subset of parameters whose time point order is consistent with the expected time series; Determine whether a subset of parameters constitutes a complete causal transmission chain defined in a changing temporal relationship; If so, it is determined to be a weak timing match, and the parameter subset is identified as a valid fault symptom group. At the same time, the parameters not included in the parameter subset are marked as disturbed parameters.
4. The hydraulic gear pump fault prediction system based on multi-parameter fusion according to claim 3, characterized in that, The collaborative analysis module is also used to perform interference source inference: When the disturbed parameter is identified through the time series deep analysis process, based on the fault mode corresponding to the effective fault symptom group, as well as the type and expected performance of the disturbed parameter, and combined with the system composition and signal flow knowledge of the hydraulic gear pump, the potential interference source that caused the disturbed parameter to not perform as expected is deduced. Potential sources of interference are sent to the early warning output module as auxiliary diagnostic information.
5. The hydraulic gear pump fault prediction system based on multi-parameter fusion according to claim 1, characterized in that, The specific methods used by the collaborative analysis module to calculate the amplitude matching degree include: Calculate the magnitude of abnormal changes for each parameter in the key parameter group; The abnormal changes in each parameter are compared according to the expected proportional relationship defined in the range of amplitude proportions; When the abnormal changes of each parameter fall within the corresponding range defined by the amplitude ratio range, it is determined to be a strong amplitude match; when the abnormal changes of the core parameter fall within the corresponding range defined by the amplitude ratio range, while the abnormal changes of the auxiliary parameter deviate from the corresponding range defined by the amplitude ratio range but the direction of change is the same, it is determined to be a weak amplitude match.
6. The hydraulic gear pump fault prediction system based on multi-parameter fusion according to claim 1, characterized in that, The serial decision logic of the serial decision module specifically includes: When the timing conformity is strong, the actual coordination index is directly determined by the level of amplitude matching. When the timing match is weak, the actual coordination index is only judged to exceed the diagnostic threshold when the amplitude match is strong.
7. The hydraulic gear pump fault prediction system based on multi-parameter fusion according to claim 6, characterized in that, The execution of the cascaded decision module also depends on the fault development stage markers; The fault development stage marker is dynamically updated based on the timing compliance determination result and is used to retrieve the detailed amplitude ratio range corresponding to the current stage from the knowledge base module; The refined amplitude ratio range has a stricter allowable deviation than the initial amplitude ratio range.
8. The hydraulic gear pump fault prediction system based on multi-parameter fusion according to claim 1, characterized in that, In the knowledge base module, at least two benchmark collaborative change patterns under different operating states are stored for the same fault mode; The system also includes a working condition identification module, which is used to identify the current operating status of the equipment based on a real-time monitoring parameter set, and the knowledge base module automatically selects a benchmark collaborative change mode that matches the current operating status of the equipment.
9. The hydraulic gear pump fault prediction system based on multi-parameter fusion according to claim 8, characterized in that, Different operating states include at least a stable operating state and an operating state under impact; In the baseline coordinated change mode corresponding to the impacted operating state, the upper limit of the amplitude ratio range is preset to K times the upper limit corresponding to the stable operating state, where K is a constant greater than 1. The cascaded decision module only calculates and judges the amplitude matching degree after the timing compliance reaches the strong compliance level.