Industrial equipment fault early warning method and system based on pulse neural network

Abstract

The invention belongs to the technical field of data processing. The invention provides an industrial equipment fault early warning method and system based on a pulse neural network. The method comprises the following steps: acquiring a vibration time sequence signal and a temperature time sequence signal of industrial equipment; inputting the vibration time sequence signal and the temperature time sequence signal into a pulse neural network optimized by adopting a quantum annealing algorithm to obtain an output state of each neuron; calculating a connection weight matrix of each neuron of the spiking neural network according to the output state of each neuron, and determining global energy according to the output state of each neuron and the connection weight matrix; when the global energy is lower than a first set threshold value and the holding time is greater than a second set threshold value, determining that the industrial equipment has a fault and generating a fault early warning signal; according to the invention, the quantum annealing algorithm in the quantum computing field is utilized, the processing capability of the pulse neural network on the long-range dependence of the time sequence signal of the industrial equipment is improved, and the fault early warning precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an industrial equipment fault early warning method and system based on a pulse neural network. Background Art

[0002] The statements in this section merely provide background art related to the present invention and do not necessarily constitute prior art.

[0003] Neurons in a spiking neural network (SNN) communicate in continuous time using short pulses called spikes. This contrasts sharply with the way neurons in an artificial neural network (ANN) exchange real-valued signals in discrete time. However, in industrial equipment failure warning scenarios, spiking neural networks have significant shortcomings when handling long-range dependent tasks, as shown in the following: (1) The timing signals of industrial equipment have long-range dependence characteristics. Faults of equipment such as motors and fans (such as bearing wear and stator winding aging) are often related to small abnormal signals from several hours to several days ago (such as the long-period correlation between early small cracks in bearings and later vibration amplitudes). Existing pulse neural networks are difficult to capture such characteristics across long time scales, resulting in a decrease in the accuracy of early fault prediction; (2) Industrial field controllers (such as ARM Cortex-M series, PLC modules) have limited computing power and rely on field buses for power supply. Traditional artificial neural networks cannot be deployed due to their high energy consumption (power consumption of a single inference cycle ≥50mW). Although pulse neural networks without quantum optimization have low energy consumption, they lack long-range dependence performance and cannot meet the reliability requirements of fault warning. Summary of the Invention

[0004] In order to address the shortcomings of the existing technology, the present invention provides an industrial equipment fault warning method and system based on a pulse neural network. By utilizing the quantum annealing algorithm in the field of quantum computing, the pulse neural network's processing capability for the long-range dependence of industrial equipment timing signals is improved, while retaining the low energy consumption advantage of the pulse neural network, adapting to the deployment requirements of industrial edge equipment, and improving the accuracy of fault warning.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides an industrial equipment fault early warning method based on a pulse neural network.

[0006] A method for early warning of industrial equipment failure based on a pulse neural network includes the following steps: Obtain vibration timing signals and temperature timing signals of industrial equipment; The vibration timing signal and the temperature timing signal are input into the pulse neural network optimized by the quantum annealing algorithm to obtain the output state of each neuron; Calculate the connection weight matrix of each neuron of the spiking neural network according to the output state of each neuron, and determine the global energy according to the output state of each neuron and the connection weight matrix; When the global energy is lower than a first set threshold and the maintenance time is greater than a second set threshold, it is determined that the industrial equipment has a fault and a fault warning signal is generated.

[0007] In one implementation of the first aspect of the present invention, a quantum annealing algorithm is used to train a pulse neural network, and the total energy is set as the weighted sum of the global energy and the amplitude-controllable kinetic energy term. By adjusting the coefficient of the kinetic energy term, the objective function continuously jumps from the local minimum to the global minimum, and the latest training obtained is obtained. .

[0008] As a further limitation of the first aspect of the present invention, when training the pulse neural network, the connection weight matrix Data element for: ,in, Representative failure mode Next The output state of a neuron, Representative failure mode Next The output state of a neuron.

[0009] As a further limitation of the first aspect of the present invention, determining the global energy according to the output state of each neuron and the connection weight matrix includes: ,in, Representative The output state of a neuron, No. The output state of a neuron.

[0010] As a further limitation of the first aspect of the present invention, the total energy is a weighted sum of the global energy and the amplitude-controllable kinetic energy term, including: ,in, is the kinetic energy term with controllable amplitude, and are coefficients respectively.

[0011] As a further limitation of the first aspect of the present invention, each failure mode during training Input the same number of vibration timing signals and temperature timing signals.

[0012] As a further limitation of the first aspect of the present invention, the failure modes include: motor bearing failure, stator failure, fan impeller imbalance, pump body cavitation and gearbox wear.

[0013] In an implementation of the first aspect of the present invention, both the vibration time series signal and the temperature time series signal are pulse data.

[0014] In a second aspect, the present invention provides an industrial equipment failure early warning system based on a pulse neural network.

[0015] An industrial equipment fault early warning system based on a pulse neural network, comprising: The time series data acquisition unit is configured to: acquire a vibration time series signal and a temperature time series signal of the industrial equipment; The time series data processing unit is configured to: input the vibration time series signal and the temperature time series signal into the pulse neural network optimized by the quantum annealing algorithm to obtain the output state of each neuron; A global energy calculation unit is configured to: calculate a connection weight matrix of each neuron of the spiking neural network according to the output state of each neuron, and determine a global energy according to the output state of each neuron and the connection weight matrix; The warning signal generating unit is configured to: when the global energy is lower than a first set threshold and the maintenance time is greater than a second set threshold, determine that there is a fault in the industrial equipment and generate a fault warning signal.

[0016] In a third aspect, the present invention provides a computer device comprising: a processor and a computer-readable storage medium; a processor adapted to execute a computer program; A computer-readable storage medium having a computer program stored therein, wherein the computer program, when executed by the processor, implements the industrial equipment fault early warning method based on a pulse neural network as described in the first aspect of the present invention.

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention significantly improves the ability of pulse neural networks to handle long-range dependencies in industrial equipment timing signals by introducing a quantum annealing algorithm to optimize them. Traditional neural networks often struggle to capture correlations between distant time points when processing long time series data due to vanishing gradients or model complexity limitations. The quantum annealing algorithm dynamically adjusts the kinetic energy coefficient in the objective function, enabling the network training process to "leap" from local optimal solutions, thereby more efficiently mining deep time series patterns in vibration and temperature data. This optimization mechanism not only retains the natural event-driven nature of pulse neural networks, but also enhances the network's modeling accuracy for the long-term evolution of equipment states through the global search capability of quantum annealing, providing more reliable technical support for fault feature extraction in complex industrial scenarios.

[0018] The quantum annealing-optimized pulse neural network of the present invention has significant advantages in terms of energy consumption and deployment adaptability. The pulse neural network itself is based on sparse pulse coding, and its calculation process is only activated when the neuron triggers a pulse. It naturally meets the requirements of industrial edge devices for low power consumption and real-time performance. The introduction of the quantum annealing algorithm does not increase the complexity of the network. On the contrary, it further reduces the computational overhead of the training phase by optimizing the convergence efficiency of the connection weight matrix. This lightweight design enables the optimized network to be directly deployed on resource-constrained industrial field equipment without relying on high-performance computing resources in the cloud. It not only reduces the system deployment cost, but also improves the real-time response capability of fault warning, providing an efficient solution for edge intelligent applications in industrial Internet of Things scenarios.

[0019] The collaborative optimization mechanism of the global energy and kinetic energy terms of the present invention directly improves the accuracy and reliability of fault warning. By weightedly combining the global energy (reflecting the network's stable assessment of the current equipment status) with the kinetic energy term of controllable amplitude (driving the network out of the local optimum), the system can more accurately distinguish between normal fluctuations and fault precursors in dynamic industrial environments. When the global energy is continuously lower than the set threshold, it indicates that the network has stably converged to the optimal representation of the actual state of the equipment. The fault warning signal triggered at this time can effectively avoid false alarms caused by transient noise or short-term anomalies. This judgment logic based on energy evolution, combined with the network parameters optimized by quantum annealing, significantly improves the sensitivity and specificity of fault warning, providing a more accurate risk warning method for the safe operation of industrial equipment.

[0020] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0022] Figure 1 A schematic flow chart of an industrial equipment fault early warning method based on a spiking neural network provided as an exemplary embodiment of the present invention; Figure 2 A schematic diagram of a neuron provided for an exemplary embodiment of the present invention; Figure 3 A schematic diagram of a principle of an industrial equipment failure early warning system based on a spiking neural network provided by an exemplary embodiment of the present invention; Figure 4 A schematic diagram of a computer device is provided for an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0023] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0024] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0025] Industrial equipment fault warning is one of the core applications of the Industrial Internet of Things. According to industry statistics, industrial losses caused by unplanned equipment downtime exceed 100 billion yuan annually. Therefore, real-time modeling and anomaly identification of time-series signals such as equipment vibration, temperature, and current are crucial. As described in the background technology, traditional technical paths have the following problems: Artificial neural networks (ANN): For example, Transformer-based fault prediction models can capture long-range dependencies, but the power consumption of a single inference cycle is ≥50mW. They need to rely on the cloud or high-performance edge servers and cannot be deployed on field controllers with limited computing power (such as PLCs and embedded modules). Unoptimized spiking neural networks (SNNs) have low energy consumption (≤20mW), but due to insufficient long-range dependency processing capabilities, the capture rate of fault precursor signals spanning hours is low, resulting in delayed warnings or false alarms. The method of optimizing the SNN connection matrix through classical algorithms relies on thermal fluctuations and is prone to falling into local optimality. In the high-dimensional energy landscape of industrial signals (such as vibration, temperature, and current multi-dimensional signals), the fault feature capture is insufficiently complete, and increasing the number of iterations will increase energy consumption and surpass the SNN advantage.

[0026] In summary, the current industrial scene has a clear demand for a fault warning solution with "edge deployment + long-range dependence + low energy consumption", which cannot be met by existing technologies at the same time. In view of this, this implementation proposes an industrial equipment fault warning method based on a pulse neural network. The pulse neural network using quantum annealing technology achieves the dual goals of "low energy consumption + long-range dependence and high performance" in the industrial equipment fault warning scenario. Specifically, Figure 1 As shown, the following process is included: S101: Acquire vibration time series signals and temperature time series signals of industrial equipment; S102: Inputting the vibration timing signal and the temperature timing signal into the pulse neural network optimized by the quantum annealing algorithm to obtain the output state of each neuron; S103: Calculating a connection weight matrix of each neuron of the spiking neural network according to the output state of each neuron, and determining global energy according to the output state of each neuron and the connection weight matrix; S104: When the global energy is lower than the first set threshold and the maintenance time is greater than the second set threshold, it is determined that the industrial equipment has a fault and a fault warning signal is generated.

[0027] In this implementation, the neuron model of the pulse neural network is as follows: Figure 2 As shown, each neuron receives input signals from other neurons. arrive , each neuron contains a weight arrive and bias ,in, arrive controls the importance of each input signal, and Controls the ease with which neurons are activated. Inside the neuron, the received arrive Multiply by the corresponding weight arrive , added together and added a bias , and we get the result: (1); Then, through the activation function After processing, the output state of the neuron is finally obtained.

[0028] More specifically, taking 256 input signals from industrial sensors as an example, the neuron receives 256 input signals from industrial sensors (denoted as to , corresponding to vibration timing signals and temperature timing signals), each neuron contains weights (denoted as to ) and bias (denoted as ), the parameter definitions and industrial scene adaptation are as follows: Weight to It is used to control the importance of each input signal, and the value range is [-1,1]. Among them, the weights corresponding to the 128 vibration signals (such as to ) is set to 0.6-1.0 (because vibration signals are highly correlated with faults), and the weights corresponding to the 128 temperature signals (such as to ) The absolute value is set to 0.3-0.5 (the temperature signal is mainly used to assist in fault diagnosis); Bias It is used to control the ease of neuron activation, with a value range of [0.8, 1.2]. In high-noise environments at industrial sites (such as workshop dust, electromagnetic interference), the bias Increase to 1.0-1.2 to reduce invalid pulses triggered by noise; Input signal to They are all pulse signals with values ​​of 0 or 1, where "1" indicates that the sensor detects an abnormal event (such as the vibration peak value exceeds the normal threshold of the equipment by 1.2 times, or the temperature exceeds the normal operating range of the equipment), and "0" indicates no abnormal event; The number of input channels is fixed at 256, including 128 vibration event signals and 128 temperature event signals, which meets the needs of multi-dimensional acquisition of industrial sensors.

[0029] The internal computation process of a neuron is divided into two steps: Step 1: Input weighted accumulation, to Multiply by the corresponding weights to , after accumulating all the product results, add the bias , and get the intermediate result , the calculation formula is shown in formula (1).

[0030] Step 2: Activation function processing, using the ReLU-like pulse activation function adapted to industrial scenarios. (This threshold is set to adapt to the industrial noise environment), the neuron outputs 1 (generates a pulse, indicating that a potential fault feature has been detected); otherwise, it outputs 0, and finally the output signal of the neuron is obtained.

[0031] In this implementation, when multiple neurons form a network layer and multiple network layers further form a complex SNN, the connection strength between neurons is determined by the weight matrix (denoted as ) definition, matrix Elements in Representative The neuron pairs The influence strength of each neuron ( 、 Each corresponds to a processing neuron for 256 input signals). To ensure the stability and fault feature capture capability of SNN in industrial scenarios, the matrix There are two core constraints to follow: Symmetrical connection ( ): This means that the effect of neurons on n is the same as the effect on n. This condition is not just for convenience; it is the mathematical premise that ensures that a global energy function (Lyapunov function) can be defined for the network and that the network dynamics always evolve in the direction of decreasing energy.

[0032] No self-join ( ): A neuron is not connected to itself. This rule is also crucial because non-zero self-connectivity ( ) can cause individual neurons to become trapped in a “self-locked” stable state, preventing the network as a whole from exploring and reaching a meaningful collective stable state.

[0033] The value of is the cumulative correlation between the states of and across all failure modes. A large positive value means that in most memories, and tend to be in the same state, thus forming an excitatory connection between them. Conversely, a large negative value indicates that they tend to be in different states, forming an inhibitory connection. Through this simple local rule, the global structural information of all failure modes is cleverly encoded into the network's connection weights.

[0034] Assume that you want to store Failure Mode , Defined as: (2); In formula (2) Will State Mapping If a specific failure mode In , the states of neurons and are the same (both 1 or both 0), then the value of is 1; if their states are different, the value is -1.

[0035] The value of is the accumulation of all the state correlations in all the failure modes. is a large positive number (such as ≥0.6), indicating that the neuron and In most fault modes, the states are the same (e.g., in a certain type of fault, vibration anomaly and temperature anomaly appear simultaneously), and the two form an excitatory connection; if is a large negative number (such as less than or equal to -0.6), indicating that the neuron and In most fault modes, the states are opposite (for example, in a certain type of fault, only vibration is abnormal and temperature does not change), and the two form an inhibitory connection. Through this calculation rule, the global characteristics of the five types of faults (including the temporal correlation of fault signals and multi-dimensional collaborative characteristics) can be fully encoded into the connection matrix. .

[0036] For a symmetric connection matrix For a network with a global energy function, we can define it as the target optimization function of the system: (3); in, is the global energy.

[0037] Neural network training requires finding the global minimum of this function. The industrial scenario definitions of each parameter are as follows: Total number of neurons: 512: The SNN uses a two-hidden layer design, with each layer containing 256 neurons, for a total of 512 neurons, to accommodate the feature expansion requirements of 256 input signals. Representative The state of each neuron is converted from the original pulse signal to -1 or 1 through mapping; Global Energy The value range and fault judgment: The value range of is [-100,50]. In the industrial equipment failure warning scenario, when When , it is determined that the SNN has stably captured the fault characteristics; The core goal of SNN training is to find global capabilities The global minimum of is obtained, ensuring the stable capture of long-range dependent fault signals of industrial equipment.

[0038] In this implementation, quantum annealing technology is used to obtain a global optimal solution for long-range correlations. Training a spiking neural network on a quantum machine and employing quantum computing are key approaches to solving complex problems that are difficult for classical computers to handle. Quantum annealing is an optimization method based on the quantum tunneling effect. Its core mechanism is to adjust the system's Hamiltonian to allow the quantum state to evolve from an initial high-energy state to a low-energy state solution to the target problem. Unlike classical annealing, which relies on thermal fluctuations, quantum annealing utilizes quantum tunneling to penetrate energy barriers, thereby more efficiently avoiding local optimal solutions. Quantum tunneling is a non-local effect that has the potential to bypass local optimal solutions and find the global optimal solution, i.e., the global energy minimum. A major advantage of quantum annealing is that, under ideal conditions, it can guarantee the global minimum of the objective function. Traditional methods such as Monte Carlo, simulated annealing, genetic algorithms, and machine learning-based methods cannot guarantee this and are prone to falling into local minima in complex or high-dimensional energy landscapes.

[0039] The entire quantum annealing process can be used Equation description: (4); in, is the time-dependent wave function of the pulse neural network system, which is used to describe the quantized expression of industrial fault characteristics; i is the imaginary unit ( ); is the reduced Planck constant (the actual value is approximately equal to , a normalized value of 1.0 can be used to simplify calculations in industrial scenarios); The unit is set to milliseconds and is fixed at 1000ms in industrial scenarios (to adapt to the time scale of equipment fault signals); for The total Hamiltonian of the system at each moment controls the evolution of the quantum state.

[0040] The quantum annealing algorithm model consists of two parts, one of which is the global energy , which is the objective function to be optimized. The second part is kinetic energy, which is obtained by introducing a kinetic energy term with controllable amplitude. , that is, regulating the system temperature. As a perturbation of the system, its role is to make the objective function continuously jump from the local minimum to the global minimum. The total Hamiltonian (that is, the total energy) is: (5); in, Represents the potential energy coefficient, which is dynamically adjusted over time. The initial value is 1.0 and decreases to 0.1 at a rate of 0.001 / ms (adjustment is completed within 1000ms). Its function is to reduce the system's dependence on initial noise; Represents the kinetic energy coefficient, which is dynamically adjusted over time. The initial value is 0.1 and increases to 1.0 at a rate of 0.001 / ms (adjustment completed within 1000ms). Its function is to break through the local energy barrier through the quantum kinetic energy driving system.

[0041] Kinetic energy term with controllable amplitude The calculation method is: (6); in, For the The quantum momentum of each neuron follows the normal distribution N(0,1) and is used to simulate the random fluctuation characteristics of industrial signals, which are constantly changing. , so that the system escapes from the original local minimum point, the potential energy changes, then the corresponding connection matrix It also dynamically adjusts, and the neural network system composed of neurons cools to the ground state, that is, the lowest energy state. Since it is a quantum system, the system has the probability of jumping to a distant place with lower potential energy, so it can capture more stable long-distance correlations.

[0042] The biggest advantage of pulse neural networks is that they can make full use of information based on spatiotemporal events. The pulse neural network of the present invention uses quantum annealing technology, which can not only consume less energy than artificial neural networks, but also achieve a huge leap in performance in capturing long-range semantic dependencies.

[0043] In this implementation, more specifically, the collaborative training process of quantum annealing and SNN is divided into four steps, including: Step 1: Collect historical data of industrial equipment operating normally under five types of fault conditions (collect 1,000 sets of time series signals for each type of fault, each containing 256 input features). Initialize the SNN connection matrix T using the Hebbian rule. At the same time, set the initial parameters of quantum annealing, i.e., A(t)=1.0, B(t)=0.1, and start the quantum annealing process. Step 2: Within 1000ms, the potential energy term coefficient A(t) is simultaneously reduced and the kinetic energy term coefficient B(t) is increased. This uses the quantum tunneling effect to explore the high-dimensional energy landscape of industrial signals and break through local energy minima caused by noise (such as false vibration signals caused by electromagnetic interference in the workshop). Step 3: Global Energy As quantum evolution changes in real time, the connection matrix T is dynamically adjusted synchronously. When the long-range characteristics of a certain type of fault are detected (such as changes in the vibration period caused by bearing wear), the connection weights of the corresponding neurons in T are automatically strengthened (such as from 0.4 to 0.8); Step 4: When Down to 0.1, When it rises to 1.0, quantum evolution stops and the system cools to the ground state (the lowest energy state). The connection matrix at this time is As the optimal matrix, SNN can stably capture the long-range dependency characteristics of equipment failures; Step 5: The optimal matrix By solidifying it into an industrial edge controller (such as ARM Cortex-M4), the on-site deployment of the SNN model is completed. There is no need to perform quantum computing on site. Only the optimal matrix needs to be loaded to perform inference, which greatly reduces the on-site computing power requirements.

[0044] More specifically, this implementation proposes an industrial equipment fault warning system whose hardware architecture is divided into four layers. The core components and functions of each layer are as follows: Sensor layer: The core components are the neuromorphic vibration sensor (model InivationDVS240) and the digital temperature sensor (model DS18B20). The vibration sensor uses a 1kHz sampling frequency to detect the device's vibration peak in real time, outputting a pulse (X=1) when the peak exceeds 1.2 times the device's rated vibration. The temperature sensor uses a 100ms sampling interval and outputs a pulse (X=1) when the detected temperature exceeds the device's normal operating temperature (e.g., the upper limit of a motor's normal operating temperature is 80°C). Preprocessing layer: The core component is the filtering module built into the industrial edge controller (ARM Cortex-M4). This module uses a sliding window filtering algorithm with a window size of 5 frames (each frame is 10ms long), which effectively removes isolated noise pulses. It also fuses the pulses corresponding to the vibration and temperature signals into 256 input features, which are then transmitted to the SNN computation layer. SNN computing layer: The core component is the SNN inference module built into the edge controller. This module loads the quantum-optimized connection matrix. , performs neuron accumulation calculation ( ) and the activation function, and finally output the global energy ; Early warning layer: The core components are on-site sound and light alarms and remote 5G modules. When this state lasts for 10 frames (total duration 100ms), the sound and light alarm triggers an on-site alarm; at the same time, the 4G module will send the fault characteristics (including global energy and pulse sequences) are uploaded to the industrial cloud platform to achieve remote early warning and data retention.

[0045] Based on the above hierarchical division, the specific workflow is as follows: The sensor collects equipment vibration and temperature signals in real time. The preprocessing layer generates 256 pulse features per frame every 10ms. After removing noise through sliding window filtering, the feature data is input into the SNN calculation layer. The SNN computing layer loads the optimal connection matrix T and calculates the system energy corresponding to each frame feature ,like , then mark the frame as a "potential fault frame"; If 10 frames of "potential fault frames" (total duration 100ms) are detected continuously, the system will determine that the equipment has a fault precursor and immediately trigger the on-site alarm and remote data upload of the early warning layer; The industrial cloud platform collects new fault data generated on site every month (if any), and uses cloud-based quantum computing equipment to re-optimize the connection matrix through quantum annealing. , and then push the updated matrix to the edge controller through the 5G module to realize model iteration and ensure long-term warning accuracy.

[0046] Optionally, in some other implementations, to address the problem of traditional fault warning "only identifying faults but not determining modes", pre-trained fault baseline features and real-time signals are combined to determine fault modes through the dual dimensions of "real-time energy matching + neuron state distribution verification". The core is to use the optimal connection matrix after quantum optimization and the output state of each neuron to construct a precise matching logic of "real-time features-baseline features".

[0047] When the following criteria are met, it is determined to be the kth type of failure mode: Criterion 1: (7); Criterion 2: (8); in, (9); Represents the global energy of the SNN when the device is running in real time; Represents the pre-trained baseline energy of the k-th type of failure mode; represents the energy matching threshold, with a value of 5; represents the total number of SNN neurons, which is fixed at 512; Representative The real-time state of each neuron (value -1 or 1) is calculated based on the real-time sensor signal The value is quantified. Representative Failure mode The typical state of a neuron; represents the state distribution matching threshold, with a value of 0.2; Represents the optimal connection matrix after quantum optimization.

[0048] Optionally, in some other implementations, to address the problem that traditional SNN neuron state quantization (simple mapping 0 / 1→-1 / 1) cannot accurately capture the weak precursors of industrial equipment failures, the fault sensitivity thresholds of industrial sensors (such as vibration exceeding 1.2 times the rated value, temperature exceeding 80°C) are integrated into the neurons. Value calculation and state mapping, combined with the global energy constraints of quantum annealing, ensure that the neuron state can accurately reflect the "precursor of device abnormality" rather than simply the presence or absence of a signal. Specifically, the output state of the neuron is set to: (10); in, (11); Representative The final state of the neuron (-1 / 1) is used for the subsequent connection matrix calculation and fault identification; represents the weighted accumulation result of the i-th neuron, reflecting the abnormality of the industrial signal; Represents the weight of the neuron to the vibration signal; Represents the weight of the neuron to the temperature signal; Representative Vibration input signal; Representative 1. Temperature input signal; represents neuronal bias; represents the industrial fault sensitive activation threshold; Represents the global energy of SNN; represents the energy threshold of the fault precursor; Optionally, in some other implementations, a fixed evolution rate (potential energy coefficient) is used for conventional quantum annealing. , kinetic energy coefficient The rate is constant), which cannot adapt to the "long-term, gradual" characteristics of industrial equipment failure precursors (such as bearing wear precursors lasting for several hours). The evolution rate of quantum annealing is associated with the actual duration of industrial failure precursors, and dynamically adjusted through the timing adaptive factor. 、 The rate of change of , ensures that quantum tunneling can fully capture the long-range dependence of the fault and avoids the omission of long-term features caused by fixed rate. Specifically, the time-adaptive quantum annealing coefficient evolution formula is given, including: (12); (13); Among them, the timing adaptive factor for: (14); represent The initial value of represent The target value, represent The initial value of , t represents the current time of quantum annealing, represents the total quantum annealing time, Represents the actual duration of precursors to industrial equipment failures, Represents the standard prefault duration.

[0049] Figure 3 An industrial equipment fault early warning system based on a pulse neural network is shown, comprising: The time series data acquisition unit is configured to: acquire a vibration time series signal and a temperature time series signal of the industrial equipment; The time series data processing unit is configured to: input the vibration time series signal and the temperature time series signal into the pulse neural network optimized by the quantum annealing algorithm to obtain the output state of each neuron; A global energy calculation unit is configured to: calculate a connection weight matrix of each neuron of the spiking neural network according to the output state of each neuron, and determine a global energy according to the output state of each neuron and the connection weight matrix; The warning signal generating unit is configured to: when the global energy is lower than a first set threshold and the maintenance time is greater than a second set threshold, determine that there is a fault in the industrial equipment and generate a fault warning signal.

[0050] It is understandable that each of the above-mentioned units can be separately or completely combined into one or several other units to form a unit, or one (or some) of the units can be further divided into multiple functionally smaller units to form a unit, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application. The above-mentioned units are divided based on logical functions. In actual applications, the functions of one unit can also be implemented by multiple units, or the functions of multiple units can be implemented by one unit. In other embodiments of the present application, the system may also include other units. In actual applications, these functions can also be implemented with the assistance of other units and can be implemented by the collaboration of multiple units.

[0051] According to another embodiment of the present application, the system described in this embodiment can be constructed by running a computer program (including program code) capable of executing the steps involved in the corresponding method of the present invention on a general-purpose computing device such as a computer, which includes processing elements and storage elements such as a central processing unit (CPU), random access memory (RAM), and read-only memory (ROM). The computer program can be recorded on, for example, a computer-readable recording medium, and loaded into the above-mentioned computing device through the computer-readable recording medium and run therein.

[0052] Figure 4 A computer device is shown, which includes a processor 401, a communication interface 402, and a computer-readable storage medium 403. The processor 401, the communication interface 402, and the computer-readable storage medium 403 may be connected via a bus or other means.

[0053] Among them, the communication interface 402 is used to receive and send data, the computer-readable storage medium 403 can be stored in the memory of the electronic device, the computer-readable storage medium 403 is used to store computer programs, the computer programs include program instructions, and the processor 401 is used to execute the program instructions stored in the computer-readable storage medium 403.

[0054] The processor 401 is the computing core and control core of the electronic device, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions to implement corresponding method processes or corresponding functions.

[0055] The processor 401 is configured to perform the following process: Obtain vibration timing signals and temperature timing signals of industrial equipment; The vibration timing signal and the temperature timing signal are input into the pulse neural network optimized by the quantum annealing algorithm to obtain the output state of each neuron; Calculate the connection weight matrix of each neuron of the spiking neural network according to the output state of each neuron, and determine the global energy according to the output state of each neuron and the connection weight matrix; When the global energy is lower than a first set threshold and the maintenance time is greater than a second set threshold, it is determined that the industrial equipment has a fault and a fault warning signal is generated.

[0056] Those skilled in the art will appreciate that the units and algorithmic steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional technical personnel may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0057] The above embodiments can be implemented in whole or in part using software, hardware, firmware, or any combination thereof. When implemented using software, they can be implemented in whole or in part in the form of a computer program product. A computer program product comprises one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions according to the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted via a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data processing device such as a server or data center that integrates one or more available media. Available media can include magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives).

[0058] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A method for early warning of industrial equipment failure based on pulse neural network, characterized in that: The following processes are included: Obtain vibration timing signals and temperature timing signals of industrial equipment; The vibration timing signal and the temperature timing signal are input into the pulse neural network optimized by the quantum annealing algorithm to obtain the output state of each neuron; Calculate the connection weight matrix of each neuron of the spiking neural network according to the output state of each neuron, and determine the global energy according to the output state of each neuron and the connection weight matrix; When the global energy is lower than a first set threshold and the maintenance time is greater than a second set threshold, it is determined that the industrial equipment has a fault and a fault warning signal is generated.

2. The industrial equipment failure early warning method based on pulse neural network according to claim 1 is characterized in that: The quantum annealing algorithm is used to train the pulse neural network. The total energy is set as the weighted sum of the global energy and the kinetic energy term with controllable amplitude. By adjusting the coefficient of the kinetic energy term, the objective function is continuously transitioned from the local minimum to the global minimum, and the latest training result is obtained. .

3. The industrial equipment failure early warning method based on pulse neural network according to claim 2, characterized in that: When training a pulse neural network, the connection weight matrix Data element for: ,in, Representative failure mode Next The output state of a neuron, Representative failure mode Next The output state of a neuron.

4. The industrial equipment failure early warning method based on pulse neural network according to claim 2, characterized in that: The global energy is determined based on the output state of each neuron and the connection weight matrix, including: ,in, Representative The output state of a neuron, No. The output state of a neuron.

5. The industrial equipment failure early warning method based on pulse neural network according to claim 4 is characterized in that: The total energy is the weighted sum of the global energy and the amplitude-controllable kinetic energy term, including: ,in, is the kinetic energy term with controllable amplitude, and are coefficients respectively.

6. The industrial equipment failure early warning method based on pulse neural network according to claim 5, characterized in that: Each failure mode during training Input the same number of vibration timing signals and temperature timing signals.

7. The industrial equipment failure early warning method based on pulse neural network according to claim 2, characterized in that: Failure modes include: motor bearing failure, stator failure, fan impeller imbalance, pump cavitation, and gearbox wear.

8. The industrial equipment failure early warning method based on a pulse neural network according to any one of claims 1 to 7, characterized in that: Both the vibration timing signal and the temperature timing signal are pulse data.

9. An industrial equipment fault early warning system based on pulse neural network, characterized in that: include: The time series data acquisition unit is configured to: acquire a vibration time series signal and a temperature time series signal of the industrial equipment; The time series data processing unit is configured to: input the vibration time series signal and the temperature time series signal into the pulse neural network optimized by the quantum annealing algorithm to obtain the output state of each neuron; A global energy calculation unit is configured to: calculate a connection weight matrix of each neuron of the spiking neural network according to the output state of each neuron, and determine a global energy according to the output state of each neuron and the connection weight matrix; The warning signal generating unit is configured to: when the global energy is lower than a first set threshold and the maintenance time is greater than a second set threshold, determine that there is a fault in the industrial equipment and generate a fault warning signal.

10. A computer device, characterized in that: include: a processor and a computer-readable storage medium; a processor adapted to execute a computer program; A computer-readable storage medium having a computer program stored therein, wherein the computer program, when executed by the processor, implements the industrial equipment fault early warning method based on a pulse neural network as described in any one of claims 1 to 8.

Images