Electromagnetic valve multi-parameter fusion fault diagnosis model

Abstract

The invention relates to the technical field of equipment state monitoring and fault diagnosis, and discloses an electromagnetic valve multi-parameter fusion fault diagnosis model which comprises the steps that multi-physical field signals such as driving voltage, coil current, an external leakage magnetic field and valve body vibration are synchronously collected; based on a signal physical connotation, three core energy conversion links of electromagnetic conversion, magnetic force conversion and force feedback are defined; and the stability of the energy conversion process is quantified in real time by calculating the collaborative entropy of each link. On the basis, a real-time entropy increasing sequence is constructed and compared with a health benchmark, and sensitive fault detection is achieved. When a fault occurs, the system performs causal analysis on the entropy increase sequence of each link by using the transfer entropy, constructs a weighted directed fault causal relationship graph, and realizes accurate positioning of a fault initial entropy source and clear tracing of a fault propagation path by analyzing the weighted out-degree and in-degree difference of each node in the graph. According to the invention, tracing from phenomena to roots is realized, so that the diagnosis accuracy is improved.

Description

Technical Field

[0001] This invention relates to the field of equipment condition monitoring and fault diagnosis technology, specifically to a multi-parameter fusion fault diagnosis model for solenoid valves. Background Technology

[0002] Solenoid valves are key actuators that convert electrical energy into magnetic energy, and then from magnetic energy into mechanical energy to control the flow of fluids. They play a vital role in precision control systems in aerospace, intelligent manufacturing, and chemical industries. Their reliability directly affects the safety and stable operation of the entire system. However, a solenoid valve is a typical complex system involving the tight coupling of multiple physical fields such as electricity, magnetism, force, and heat. Its internal energy conversion process exhibits significant transient and nonlinear characteristics, leading to extremely complex failure mechanisms.

[0003] Existing solenoid valve diagnostic technologies typically focus on analyzing the characteristics of a single physical signal (such as coil current or valve body vibration). While this approach can detect faults to some extent, the limited information available makes it difficult to delve into the physical nature of the fault. When a subsystem malfunctions, its effects are rapidly transmitted through close coupling and reflected in signals from other physical fields. This results in overlapping representations of multiple fault modes on a single signal, making it impossible to effectively distinguish between the root cause and accompanying phenomena. For example, a simple current anomaly reflects both the state of the electrical drive system and the reaction force of the mechanical components. Therefore, both electrical drive malfunctions and mechanical jamming of the valve core can cause current signal anomalies, and this single signal alone cannot differentiate their underlying causes.

[0004] Although some studies have attempted to integrate multiple signals for analysis, most remain at the level of feature splicing and pattern recognition at the data level. These diagnostic methods, which rely on data model-based decosively approaches, suffer from a disconnect between their internal logic and physical processes, lacking clear physical interpretability and failing to reveal the inherent logic and causal chains of fault propagation across different physical domains. Therefore, when solenoid valves malfunction, existing technologies generally face the problem of being unable to accurately locate the root cause of the fault or clearly trace the fault propagation path, thus limiting the efficiency of fault repair and the reliability of the system. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a multi-parameter fusion fault diagnosis model for solenoid valves. This model solves the problem that existing diagnostic methods typically rely on a single physical signal, making it difficult to comprehensively capture the state information of solenoid valves under the coupling of multiple physical fields. Furthermore, most methods can only identify the fault state and lack the technical means to effectively trace the root cause and propagation path of the fault.

[0006] To address the aforementioned technical problems, this invention provides a multi-parameter fusion fault diagnosis model for solenoid valves.

[0007] The system includes:

[0008] A multi-physics field signal synchronous acquisition module is provided, the input end of which is connected to the measuring point of the solenoid valve, and is used to synchronously acquire the solenoid valve's drive voltage signal, coil current signal, external leakage magnetic field signal, and valve body vibration acceleration signal.

[0009] A link-coordinated entropy decoupling module is included, with its input connected to the output of a multi-physics signal synchronous acquisition module. This module receives the acquired signals and defines electromagnetic conversion links, magnetic conversion links, and force feedback links by pairing the acquired signals. For each link, the module calculates the link-coordinated entropy, which serves as a quantification of energy conversion order, thereby obtaining the time series of electromagnetic conversion entropy, magnetic conversion entropy, and force feedback entropy.

[0010] A fault causal chain localization module is included, with its input connected to the output of the link cooperative entropy decoupling module. This module constructs an entropy increase sequence based on the time series of obtained electromagnetic conversion entropy, magnetic force conversion entropy, and force feedback entropy. It then constructs a fault causal relationship graph based on the quantization results of the directed information flow between the entropy increase sequences using transfer entropy. Finally, the module locates the initial entropy source and propagation path of the fault according to the topology of the fault causal relationship graph and outputs a diagnostic conclusion.

[0011] Preferably, the specific steps for the link cooperative entropy decoupling module to calculate the link cooperative entropy include:

[0012] Mode decomposition is performed on the two signals constituting each of the links to obtain eigenmode function components of the same order;

[0013] Based on the intrinsic mode function components, the instantaneous coordination index characterizing phase synchronization and amplitude correlation is calculated;

[0014] The link collaboration entropy is obtained based on the probability distribution of the instantaneous collaboration index within a time window.

[0015] Preferably, the mode decomposition employs the Hilbert-Huang transform. The Hilbert-Huang transform is used for adaptive decomposition of non-stationary and nonlinear signals. For two signals x1(t) and x2(t) within the link, after obtaining the eigenmode function components of the same order through Hilbert-Huang transform decomposition, a Hilbert transform is performed on them to obtain the instantaneous amplitude A of each eigenmode function component. i (t) and instantaneous phase φ i (t).

[0016] Preferably, the instantaneous coordination index C I (t) is determined by the instantaneous phase coherence C P (t) and the degree of coordination with instantaneous amplitude C a (t) Obtained by weighted fusion. The weighted fusion is used to quantitatively characterize the coupling degree of two signals in the energy evolution within the link. Its calculation formula is:

[0017] C I (t)=w·C p (t)+(1-w)·C a (t);

[0018] in:

[0019] C I (t) is the instantaneous coordination index at time t;

[0020] C P (t) represents the instantaneous phase coherence at time t;

[0021] C a (t) represents the instantaneous amplitude coherence at time t;

[0022] w is the weighting coefficient, and its value ranges from [0,1].

[0023] Preferably, the instantaneous phase coordination degree C P (t) is obtained by calculating the cosine of the instantaneous phase difference between the intrinsic mode function components of the two signals. The calculation formula is:

[0024] C p (t)=cos[φ1(t)-φ2(t)];

[0025] in:

[0026] C P (t) represents the instantaneous phase coherence at time t;

[0027] φ1(t) is the instantaneous phase of the eigenmode function component of the first signal at time t;

[0028] φ2(t) is the instantaneous phase of the eigenmode function component of the second signal at time t.

[0029] The instantaneous amplitude coordination degree C a (t) is obtained by calculating the normalized product of the instantaneous amplitudes of the intrinsic mode function components of the two signals. The calculation formula is:

[0030]

[0031] in:

[0032] Ca (t) represents the instantaneous amplitude coherence at time t;

[0033] A1(t) is the instantaneous amplitude of the eigenmode function component of the first signal at time t;

[0034] A2(t) is the instantaneous amplitude of the eigenmode function component of the second signal at time t;

[0035] <·> indicates time averaging;

[0036] The normalization factor is the product of the root mean square of the instantaneous amplitudes of the intrinsic mode function components of the two signals.

[0037] Preferably, the link cooperative entropy is calculated by measuring the instantaneous cooperative index C. I The Shannon entropy of the probability distribution (t) within the time window is obtained. C I The range of values ​​for (t) [-1, 1] is divided into N intervals, and C is statistically analyzed. I (t) The probability p of falling into the i-th interval i Then the formula for calculating the link cooperative entropy is:

[0038]

[0039] in:

[0040] LSE stands for Link Co-entropy;

[0041] N is the total number of intervals dividing the range of values ​​for the instantaneous coordination index, and is a positive integer;

[0042] i is the range index, from 1 to N;

[0043] p i The probability that the instantaneous collaborative index falls into the i-th interval;

[0044] ln(p i ) represents the probability p i The natural logarithm of .

[0045] Preferably, the fault causal chain localization module constructs the entropy increase sequence by subtracting the value of the electromagnetic conversion entropy, the magnetic force conversion entropy, and the force feedback entropy at each moment in the time series from a predetermined health state benchmark entropy model, thereby obtaining the entropy increase sequence.

[0046] Preferably, the fault causal chain localization module locates the initial entropy source by calculating the weighted out-degree D of each node in the topology of the fault causal relationship graph. out with weighted in-degree D in , will have the largest (Dout -D in The nodes with the difference are determined as the initial entropy sources.

[0047] Preferably, the fault causal chain localization module uses transitive entropy quantization to quantize the directed information flow between the entropy-increasing sequences.

[0048] Preferably, the diagnostic conclusion is a structured report, which includes the fault status assessment, the root cause of the fault, and the fault propagation path.

[0049] Preferably, the fault causal chain localization module is further configured to compare the entropy increase sequence with a predetermined fault alarm threshold, and trigger a fault diagnosis process when the entropy increase sequence exceeds the fault alarm threshold.

[0050] This invention provides a multi-parameter fusion fault diagnosis model for solenoid valves. It has the following beneficial effects:

[0051] 1. This invention defines multi-physics field signal pairing as an energy conversion link and calculates the link cooperative entropy, which can quantitatively describe the coupling degree between different physical processes inside the solenoid valve and the orderliness of energy conversion, providing deeper state information for fault diagnosis.

[0052] 2. By constructing an entropy-increasing sequence and using the propagation entropy to establish a fault causal relationship graph, this invention can realize the location of fault phenomena and the tracing of the propagation path from fault phenomena to fault root causes, providing more in-depth diagnostic conclusions than traditional fault classification methods.

[0053] 3. This invention uses Hilbert-Huang transform to process the signal, enabling it to adaptively analyze the non-stationary and nonlinear signals generated during the operation of the solenoid valve, thereby improving the applicability and analytical accuracy of the model. Attached Figure Description

[0054] Figure 1 This is a block diagram of the overall structure of the multi-parameter fusion fault diagnosis model for a solenoid valve according to an embodiment of the present invention.

[0055] Figure 2 This is a schematic diagram of the arrangement of multi-physics field signal acquisition sensors according to an embodiment of the present invention;

[0056] Figure 3 This is a schematic diagram comparing the waveforms before and after signal preprocessing in an embodiment of the present invention;

[0057] Figure 4 This is a schematic diagram illustrating the energy conversion link definition in an embodiment of the present invention;

[0058] Figure 5 This is a flowchart of the link collaborative entropy calculation according to an embodiment of the present invention;

[0059] Figure 6 This is a schematic diagram illustrating the construction of an entropy increase sequence and fault triggering in an embodiment of the present invention. Detailed Implementation

[0060] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0061] See attached document Figure 1 , Figure 1 This is a block diagram of the overall structure of the multi-parameter fusion fault diagnosis model for solenoid valves according to an embodiment of the present invention.

[0062] The system provided by the present invention may include a multi-physics field signal synchronous acquisition module 100 and a data processing unit 200.

[0063] In this embodiment, the system monitors a solenoid valve under test. A voltage and current acquisition unit is connected in series in the drive circuit of the solenoid valve under test to acquire its drive voltage signal and coil current signal. A Hall sensor is attached to a specific position on the valve body shell of the solenoid valve under test to acquire its external leakage magnetic field signal. An accelerometer is installed on the valve cover of the solenoid valve under test to acquire its valve body vibration acceleration signal. The signal output terminals of the voltage and current acquisition unit, the Hall sensor, and the accelerometer are all electrically connected to the analog signal input terminal of the multi-physics field signal synchronous acquisition module 100.

[0064] The multi-physics field signal synchronous acquisition module 100 can specifically be a multi-channel synchronous data acquisition card. The multi-physics field signal synchronous acquisition module 100 receives analog signals from various sensors and converts them into digital signal sequences through an internal analog-to-digital converter. The multi-physics field signal synchronous acquisition module 100 ensures that the acquisition of drive voltage signals, coil current signals, external leakage magnetic field signals, and valve body vibration acceleration signals is synchronized in time by using an internal shared clock or by receiving external trigger signals.

[0065] The data processing unit 200 can specifically be a host computer, such as an industrial control computer or a general-purpose computer. The input terminal of the data processing unit 200 is connected to the digital signal output terminal of the multi-physics field signal synchronous acquisition module 100 through a data bus (such as a PCIe or USB interface) to receive the synchronously acquired digital signal sequence.

[0066] The data processing unit 200 is internally configured with a processor and a memory, the memory of which stores computer program instructions. When the processor executes the computer program instructions, it implements the functions of the following modules: a link cooperative entropy decoupling module 210 and a fault causal chain localization module 220.

[0067] The output of the data processing unit 200 is connected to a display unit for outputting and displaying the structured diagnostic report generated by the fault cause-effect chain localization module 220.

[0068] During system operation, the multi-physics field signal synchronous acquisition module 100 continuously and synchronously acquires the driving voltage signal, coil current signal, external leakage magnetic field signal and valve body vibration acceleration signal of the solenoid valve under test, and transmits the acquired digital signal sequence to the data processing unit 200 in real time.

[0069] After receiving the data, the data processing unit 200 processes it using the link cooperative entropy decoupling module 210. The link cooperative entropy decoupling module 210 first pairs the four signal sequences to define the electromagnetic conversion link, the magnetic force conversion link, and the force feedback link. Then, for each link, the link cooperative entropy decoupling module 210 performs a series of calculations to obtain the link cooperative entropy, which serves as a quantitative indicator of the ordered energy conversion. Finally, it outputs three time series: electromagnetic conversion entropy, magnetic force conversion entropy, and force feedback entropy.

[0070] The fault causal chain localization module 220 receives three entropy time series output by the link cooperative entropy decoupling module 210. The fault causal chain localization module 220 first subtracts the real-time entropy sequence from the pre-determined health state benchmark entropy model to construct an entropy increase sequence.

[0071] Subsequently, the fault causal chain localization module 220 uses transfer entropy to quantify the directed information flow between entropy-increasing sequences and constructs a fault causal relationship graph based on the quantization results.

[0072] Finally, the fault causal chain localization module 220 analyzes the topological structure of the fault causal relationship graph, locates the initial entropy source of the fault and the specific propagation path, and generates a structured report containing diagnostic conclusions based on this, which is transmitted to the display unit.

[0073] See attached document Figure 2 , Figure 2 This is a schematic diagram of the arrangement of multi-physics field signal acquisition sensors according to an embodiment of the present invention.

[0074] In one specific embodiment of the present invention, the multiphysics signal synchronous acquisition module 100 acquires four signals. To fully capture the transient dynamic characteristics of the solenoid valve during its switching action, this embodiment sets the signal sampling frequency to 50kHz and the sampling precision to 16 bits.

[0075] The acquisition of the drive voltage signal and the coil current signal is achieved through a voltage and current acquisition unit. The voltage acquisition channel of this unit is connected in parallel to the power supply line of the solenoid valve under test using a high-impedance differential input method; its current acquisition channel uses a Hall effect current sensor, which is located on a wire of the drive circuit to measure the coil current in a non-invasive manner.

[0076] The external leakage magnetic field signal is acquired using a high-sensitivity Hall sensor. To capture the magnetic field changes related to the valve core's movement position, the Hall sensor is fixed to the solenoid valve body housing, with its sensing surface facing the midpoint of the valve core's stroke, and its sensing axis kept parallel to the valve core's movement direction.

[0077] The valve body vibration acceleration signal is acquired using a piezoelectric accelerometer. To capture the mechanical impact response generated when the valve core opens and closes, the accelerometer is securely mounted at the center of the solenoid valve cover via threads or a magnetic base.

[0078] To achieve synchronization of the four signals, all analog-to-digital conversion channels within the multi-physics signal synchronization acquisition module 100 are driven by a unified internal clock signal. This configuration ensures that the sampling of the four signals is completed simultaneously at each sampling point, thereby guaranteeing the consistency of the four sets of digital signal sequences in terms of time reference.

[0079] See attached document Figure 3 , Figure 3 This is a schematic diagram comparing the waveforms before and after signal preprocessing in an embodiment of the present invention.

[0080] After the multi-physics field signal synchronous acquisition module 100 transmits the four digital signal sequences to the data processing unit 200, the data processing unit 200 first performs preprocessing operations on these signal sequences. The purpose of this preprocessing operation is to filter out high-frequency noise that is mixed in during signal acquisition and transmission and is unrelated to the physical process of the solenoid valve itself.

[0081] In this embodiment, preprocessing is achieved by applying a digital filter to each digital signal sequence. Specifically, a fourth-order Butterworth low-pass filter is used to process the sequences of drive voltage signal, coil current signal, external leakage magnetic field signal, and valve body vibration acceleration signal.

[0082] The filter's cutoff frequency is set to 10kHz. This cutoff frequency is set to preserve the effective signal frequency components generated by the solenoid valve during its dynamic behaviors such as switching actions and valve core impacts, while attenuating electromagnetic interference and sensor white noise above this frequency range.

[0083] After digital filtering, four signal sequences with high-frequency noise removed are obtained. These four clean signal sequences are stored in the memory of the data processing unit 200 and used as input data for subsequent analysis by the link cooperative entropy decoupling module 210.

[0084] See attached document Figure 4 , Figure 4 This is a schematic diagram illustrating the energy conversion link definition in an embodiment of the present invention.

[0085] In the data processing unit 200, the link cooperative entropy decoupling module 210 receives four signal sequences: preprocessed drive voltage, coil current, external leakage magnetic field, and valve body vibration acceleration. The first step of the link cooperative entropy decoupling module 210 is to define the energy conversion links characterizing different physical processes inside the solenoid valve by specifically pairing these four signal sequences.

[0086] In this embodiment, three energy conversion links are defined.

[0087] The first link is the electromagnetic conversion link, which consists of a drive voltage signal and a coil current signal. This signal pair characterizes the process by which the solenoid valve coil converts input electrical energy into magnetic energy. The drive voltage serves as the input excitation, and the coil current serves as the system response. The dynamic relationship between the two directly reflects the electrical characteristics of the electromagnetic circuit, such as changes in the coil's resistance, inductance, and back electromotive force.

[0088] The second link is the magnetic force conversion link, which consists of a coil current signal and an external leakage magnetic field signal. This signal pair characterizes the process by which the coil current generates a magnetic field, thereby forming the magnetic force that drives the valve core. The coil current is the direct source of the magnetic field, while the external leakage magnetic field signal indirectly reflects the change in magnetic flux in the internal main magnetic circuit. The relationship between the two contains physical information about the magnetic circuit, such as the influence of magnetic reluctance, magnetic saturation, and air gap changes on the magnetic field strength.

[0089] The third link is the force feedback link, which consists of an external leakage magnetic field signal and a valve body vibration acceleration signal. This signal characterizes the process of magnetically driving the movement of mechanical components such as the valve core, generating mechanical impact and vibration. The external leakage magnetic field signal is related to the real-time position and motion state of the valve core, while the vibration acceleration signal is the mechanical system's response when the valve core collides with the valve seat, limit block, etc. The relationship between the two reflects the dynamic characteristics of the mechanical system, such as friction and wear of moving parts and spring stiffness.

[0090] Based on the above definition, the link cooperative entropy decoupling module 210 deconstructs the original multidimensional signal space into three two-dimensional analysis subspaces with clear physical meaning. These three defined energy conversion links form the basis for subsequent link cooperative entropy calculations.

[0091] See attached document Figure 5 , Figure 5 This is a flowchart illustrating the link collaborative entropy calculation process according to an embodiment of the present invention.

[0092] After defining three energy conversion links, the link cooperative entropy decoupling module 210 calculates the link cooperative entropy for the two signal sequences constituting each link, namely the link input signal x1(t) and the link output signal x2(t). This calculation process in this embodiment includes the following steps:

[0093] First, Hilbert-Huang transforms are applied to x1(t) and x2(t) respectively. Specifically, mode decomposition is used to adaptively decompose the non-stationary, nonlinear original signal sequence into a finite number of intrinsic mode function (EMF) components. Each EMF component represents the oscillation mode of the signal at different time scales.

[0094] Then, from the set of intrinsic mode function (IMF) components obtained from each signal decomposition, IMF components of the same order are selected for subsequent analysis. In this embodiment, for both x1(t) and x2(t), the IMF component containing the main energy of the signal is selected for subsequent calculation. Subsequently, Hilbert transforms are performed on the two selected IMF components to obtain their instantaneous amplitude sequences A1(t), A2(t) and instantaneous phase sequences φ1(t), φ2(t).

[0095] Based on the obtained instantaneous parameters, the instantaneous coordination index C is calculated. I (t). This index is derived from the instantaneous phase coherence C. P (t) and the degree of coordination with instantaneous amplitude C a (t) Weighted fusion is used to obtain the instantaneous phase coherence degree C. P The formula for calculating (t) is:

[0096] C p (t)=cos[φ1(t)-φ2(t)];

[0097] in:

[0098] C P (t) represents the instantaneous phase coherence at time t;

[0099] φ1(t) is the instantaneous phase of the eigenmode function component of the first signal at time t;

[0100] φ2(t) is the instantaneous phase of the eigenmode function component of the second signal at time t.

[0101] Instantaneous amplitude coherence C a The formula for calculating (t) is:

[0102]

[0103] in:

[0104] C a (t) represents the instantaneous amplitude coherence at time t;

[0105] A1(t) is the instantaneous amplitude of the eigenmode function component of the first signal at time t;

[0106] A2(t) is the instantaneous amplitude of the eigenmode function component of the second signal at time t;

[0107] <·> indicates time averaging;

[0108] The product of the root mean square of the instantaneous amplitudes of the intrinsic mode function components of the two signals is used as the normalization factor.

[0109] Instantaneous Coordination Index C I The formula for calculating (t) is:

[0110] C I (t)=w·C p (t)+(1-w)·C a (t);

[0111] C I (t) is the instantaneous coordination index at time t;

[0112] C P (t) represents the instantaneous phase coherence at time t;

[0113] C a (t) represents the instantaneous amplitude coherence at time t;

[0114] w is a weighting coefficient, and in this embodiment, the value of w is set to 0.5.

[0115] Finally, the link coordination entropy is calculated. Within a preset time window (e.g., a complete solenoid valve switching cycle), the instantaneous coordination index C is calculated. I The range of values ​​for (t) [-1, 1] is divided into N equal intervals (in this embodiment, N = 100). By statistically analyzing C... I The probability p that the value of the (t) sequence falls into the i-th interval. i The probability distribution is constructed. The link collaborative entropy is calculated based on this probability distribution, and its formula is:

[0116]

[0117] LSE stands for Link Co-entropy;

[0118] N is the total number of intervals dividing the instantaneous coordination index range, which is 100 in this embodiment;

[0119] i is the range index, from 1 to 100;

[0120] p i The probability that the instantaneous collaborative index falls into the i-th interval;

[0121] ln(p i ) represents the probability p i The natural logarithm of .

[0122] The link cooperative entropy decoupling module 210 repeats the above complete calculation steps for the electromagnetic conversion link, the magnetic force conversion link, and the force feedback link respectively, and finally outputs three independent link cooperative entropy sequences that vary with time: electromagnetic conversion entropy, magnetic force conversion entropy, and force feedback entropy.

[0123] A health status baseline entropy model is pre-stored in the memory of the data processing unit 200. This model serves as a quantitative benchmark for subsequent calculations of the entropy increase sequence, and its establishment process is completed before the system is put into real-time diagnosis.

[0124] The establishment of this model first requires using a solenoid valve of the same model that has been confirmed to be in a fault-free and healthy state as a reference component. This reference solenoid valve is installed on a test platform and operated under a set of preset typical operating conditions that cover its main operating range.

[0125] In this embodiment, the typical operating conditions include multiple different combinations of working pressure levels and switching frequencies. For example, three pressure levels (80%, 100%, and 120% of the rated pressure) and two switching frequencies (low frequency and high frequency) are set, constituting a total of six independent operating conditions.

[0126] For each operating condition, the multi-physics signal synchronous acquisition module 100 continuously acquires data for a predetermined number (e.g., 500) complete switching cycles. This yields multiple sets of raw multi-physics signal datasets representing the health status.

[0127] Subsequently, the data processing unit 200 performs the complete link collaborative entropy calculation steps described in Part III for each collected health status dataset. This process generates a time series of electromagnetic conversion entropy, magnetic force conversion entropy, and force feedback entropy for each health condition.

[0128] Finally, a final health state baseline entropy model is constructed by statistically processing the health entropy time series generated under all operating conditions. Specifically, for each type of link entropy (e.g., electromagnetic conversion entropy), the time series of that entropy obtained under all operating conditions are time-aligned with the rising edge of the drive voltage signal as the reference. Then, the average value of all entropy values ​​at each time point is calculated on the aligned time axis.

[0129] By performing the above averaging process on the three types of link entropy respectively, three benchmark entropy time series are finally obtained, which can represent the average orderliness level of this type of solenoid valve under various health conditions. These three benchmark time series together constitute the health state benchmark entropy model that is permanently stored.

[0130] See attached document Figure 6 , Figure 6 This is a schematic diagram illustrating the construction of an entropy increase sequence and fault triggering in an embodiment of the present invention.

[0131] During real-time system monitoring, the fault causal chain localization module 220 receives three real-time time series output by the link cooperative entropy decoupling module 210: electromagnetic conversion entropy, magnetic force conversion entropy, and force feedback entropy. Simultaneously, the fault causal chain localization module 220 retrieves a pre-established health state baseline entropy model from memory.

[0132] The fault causal chain localization module 220 calculates the difference between each real-time entropy sequence and the corresponding baseline entropy sequence in the baseline model point by point, thereby constructing three independent entropy increase sequences. The calculation formula is as follows:

[0133] ΔS link (t)=S real-time,link (t)-S baseline,link (t);

[0134] in:

[0135] ΔS link (t) represents the entropy increase of a certain link at time t;

[0136] S real-time,link (t) represents the real-time link cooperative entropy value of the link at time t;

[0137] S baseline,link (t) represents the baseline entropy value of the link at time t in the health status baseline entropy model.

[0138] The calculations were applied to the three links of electromagnetic conversion, magnetic conversion, and force feedback, respectively, to obtain three entropy increase time series.

[0139] The data processing unit 200 also stores a fault alarm threshold in its memory. In this embodiment, this threshold is determined based on multiple sets of health status datasets used when establishing the health status baseline entropy model. Specifically, for each time point, the standard deviation σ of the link collaborative entropy values ​​of all healthy operating cycles at that point is calculated. The fault alarm threshold at that time point is set as the baseline entropy value at that point plus three times the standard deviation, i.e., S. baseline,link (t)+3σ link (t).

[0140] The fault causal chain localization module 220 continuously compares the value of the collaborative entropy sequence of each real-time link with the corresponding fault alarm threshold.

[0141] When the system detects that the value of the real-time link cooperative entropy exceeds the fault alarm threshold at its corresponding time point, i.e., S real-time,link (t)>S baseline,link (t)+3σ link When (t) occurs, the system determines that the solenoid valve is malfunctioning and triggers the subsequent fault causality analysis process. If the real-time entropy values ​​of all links do not exceed the alarm threshold, the system continues to maintain monitoring and does not perform subsequent diagnostics.

[0142] Once the fault triggering condition is met, the fault causal chain localization module 220 immediately initiates the causal analysis process. The fault causal chain localization module 220 uses three entropy increase sequences within a fixed time window prior to the triggering time as analysis data.

[0143] The core function of the fault causal chain localization module 220 is to use transitive entropy to quantify the intensity of the directed information flow between each pair of these three entropy-increasing sequences. For two discrete-time sequences X = {x1, x2, ..., x...} n} and Y = {y1, y2, ..., y n The transfer entropy T from sequence X to sequence Y X→Y The calculation formula is:

[0144]

[0145] in:

[0146] y n+1 The state value of time series Y at time point n+1;

[0147] Let be the k-dimensional historical state vector of the time series Y at time point n, i.e.

[0148] Let X be the l-dimensional historical state vector of the time series X at time point n, i.e.

[0149] For y n+1 , and The joint probability distribution of the three;

[0150] Given the historical state of Y The historical state of X Under the given conditions, the future state of Y is y n+1 The conditional probability;

[0151] For only the known historical state of Y Under the given conditions, the future state of Y is y n+1 The conditional probability;

[0152] It is a likelihood ratio that quantifies the degree to which the historical state of sequence X provides new information for predicting the future state of sequence Y, given that the history of sequence Y itself is known.

[0153] This represents the amount of local information transferred from X to Y under a specific combination of states, expressed in bits.

[0154] Therefore, the transfer entropy T X→Y It is the mathematical expectation of the local information transfer amount under all possible state combinations, used to measure the average information flow intensity from sequence X to sequence Y.

[0155] The joint probability and conditional probability are estimated by comparing and analyzing the frequency statistics of the discretized entropy-increasing sequence data within the time window.

[0156] In this embodiment, the dimensions k and l of the historical state vector are both set to 1.

[0157] The fault causal chain localization module 220 pairs the electromagnetic conversion entropy increase sequence, the magnetic force conversion entropy increase sequence, and the force feedback entropy increase sequence in pairs to calculate the transfer entropy values ​​in all six directions (e.g., from electromagnetic to magnetic force, from magnetic force to electromagnetic force, etc.).

[0158] Based on the calculated six transit entropy values, the fault causal chain localization module 220 constructs a fault causal relationship graph. This graph is a weighted directed graph containing three nodes, representing the electromagnetic conversion link, the magnetic force conversion link, and the force feedback link, respectively. The weight of the directed edge between nodes is the calculated transit entropy value in the corresponding direction.

[0159] Subsequently, the fault causal chain localization module 220 analyzes the topological structure of the fault causal relationship graph to locate the initial entropy source of the fault. Specifically, it calculates the weighted out-degree D of each node in the graph. out with weighted in-degree D in The weighted out-degree D of a node out The weighted in-degree D of a node is defined as the sum of the weights of all directed edges originating from that node. in Defined as the sum of the weights of all directed edges pointing to that node.

[0160] By comparing the (D) of each node out -D in The node with the largest positive difference is identified as the initial entropy source in this fault event.

[0161] After identifying the initial entropy source node, the fault causal chain localization module 220 determines the main propagation path of the fault by finding the directed edge with the maximum weight originating from that source node. The node pointed to by this maximum weight edge is the next link in the fault propagation.

[0162] After the fault causal chain localization module 220 completes the analysis of the fault causal relationship graph, it integrates the analysis results into a structured diagnostic report. This report is a data structure containing multiple predefined fields.

[0163] In this embodiment, the structured diagnostic report includes the following fields:

[0164] A unique identifier field for a tested solenoid valve is used to record the specific device where a diagnostic event occurred;

[0165] A timestamp field for a diagnostic event, used to record the precise time when the fault was confirmed;

[0166] A current status field, whose value is set to fault, to indicate that the system has deviated from a healthy state;

[0167] A root cause field, the value of which is set to the name of the initial entropy source node identified in the causal analysis, such as an electromagnetic conversion link;

[0168] A fault propagation path field, the value of which is an ordered list that records the names of the nodes traversed along the directed edge with the maximum propagation entropy weight, starting from the initial entropy source node.

[0169] The report also includes quantitative data to support the diagnostic conclusions. Specifically, it includes a peak entropy increase field for each link, recording the maximum value reached by each of the three entropy increase sequences within the analysis time window. Additionally, it includes a causal relationship transfer entropy matrix field, which stores all six transfer entropy values ​​calculated in the causal analysis, characterizing the pairwise information flow intensity between the three entropy increase sequences.

[0170] After generating the structured diagnostic report, the fault causal chain localization module 220 sends it to the display unit via the data interface. The display unit receives the data structure and presents the contents of each field to the operator in a visual form, or stores it as a record in the system operation log database.

Claims

1. A multi-parameter fusion fault diagnosis model for solenoid valves, characterized in that, The system includes: The multi-physics field signal synchronous acquisition module is used to synchronously acquire the solenoid valve's drive voltage signal, coil current signal, external leakage magnetic field signal, and valve body vibration acceleration signal. The link cooperative entropy decoupling module is used to receive the signals collected by the multi-physics field signal synchronous acquisition module, define electromagnetic conversion links, magnetic force conversion links and force feedback links by pairing the collected signals, and calculate the link cooperative entropy as a quantitative index of energy conversion order for each link, thereby obtaining the time series of electromagnetic conversion entropy, magnetic force conversion entropy and force feedback entropy. The fault causal chain localization module is used to construct an entropy increase sequence based on the time series of electromagnetic conversion entropy, magnetic force conversion entropy and force feedback entropy obtained by the link collaborative entropy decoupling module, and construct a fault causal relationship graph based on the quantification result of the directed information flow between the entropy increase sequences using transfer entropy. Finally, based on the topological structure of the fault causal relationship graph, the initial entropy source and propagation path of the fault are located, thereby outputting a diagnostic conclusion.

2. The multi-parameter fusion fault diagnosis model for a solenoid valve according to claim 1, characterized in that, The link cooperative entropy decoupling module calculates the link cooperative entropy through the following steps: First, mode decomposition is performed on the two signals within the link to obtain the eigenmode function components of the same order; Then, based on the intrinsic mode function components, the instantaneous coordination index characterizing phase synchronization and amplitude correlation is calculated; Finally, the link collaboration entropy is calculated based on the probability distribution of the instantaneous collaboration index within a time window.

3. The multi-parameter fusion fault diagnosis model for a solenoid valve according to claim 2, characterized in that, The instantaneous coordination index is obtained by weighted fusion of instantaneous phase coordination degree and instantaneous amplitude coordination degree. The weighted fusion is used to quantitatively characterize the degree of coupling between two signals in the energy evolution within the link.

4. The multi-parameter fusion fault diagnosis model for a solenoid valve according to claim 3, characterized in that, The instantaneous phase coherence is obtained by calculating the cosine of the instantaneous phase difference between the intrinsic mode function components of the two signals, and the instantaneous amplitude coherence is obtained by calculating the normalized product of the instantaneous amplitudes of the intrinsic mode function components of the two signals.

5. The multi-parameter fusion fault diagnosis model for a solenoid valve according to claim 1, characterized in that, The fault causal chain localization module constructs the entropy increase sequence in the following manner: The entropy increase sequence is obtained by subtracting the value of the electromagnetic conversion entropy, magnetic force conversion entropy, and force feedback entropy at each moment in the time series from a predetermined health state baseline entropy model.

6. The multi-parameter fusion fault diagnosis model for a solenoid valve according to claim 1, characterized in that, The fault causal chain localization module locates the initial entropy source in the following manner: In the topology of the fault causal relationship graph, the node with the largest difference between the weighted out-degree and the weighted in-degree is determined as the initial entropy source.

7. The multi-parameter fusion fault diagnosis model for a solenoid valve according to claim 2, characterized in that, The mode decomposition employs the Hilbert-Huang transform, which is used for adaptive decomposition of non-stationary and nonlinear signals.

8. The multi-parameter fusion fault diagnosis model for a solenoid valve according to claim 1, characterized in that, The fault causal chain localization module uses transfer entropy quantification to quantify the directed information flow between the time series of the electromagnetic conversion entropy, the magnetic conversion entropy, and the force feedback entropy.

9. The multi-parameter fusion fault diagnosis model for a solenoid valve according to claim 1, characterized in that, The diagnostic conclusion is a structured report, which includes the fault status assessment, the root cause of the fault, and the fault propagation path.

10. The multi-parameter fusion fault diagnosis model for a solenoid valve according to claim 5, characterized in that, The fault causal chain localization module is further configured to: compare the entropy increase sequence with a preset fault alarm threshold, and trigger a fault diagnosis process when the entropy increase sequence exceeds the fault alarm threshold.

Images