Intelligent fault diagnosis method and system for cattle bone powder production line

Abstract

The invention discloses an intelligent fault diagnosis method and system for a cattle bone powder production line, and relates to the related technical field of production fault diagnosis, and the method comprises the steps: collecting equipment operation data and technological parameters of the cattle bone powder production line in real time through a multi-mode sensor, and obtaining a real-time monitoring data set; transmitting the real-time monitoring data set to an edge computing node for preprocessing; inputting the standard fusion data into the cloud digital twin, and dynamically simulating the operation state of the production line to obtain virtual-real mapped production line health state simulation data; analyzing through a hierarchical diagnosis engine to obtain a diagnosis result; and triggering a dynamic response according to a diagnosis result, and performing closed-loop intelligent operation and maintenance of the cattle bone powder production line. The technical problems that in the prior art, fault diagnosis is difficult to conduct on the cattle bone powder production line accurately in real time, the fault diagnosis result lags behind, and the production efficiency and the product quality are affected are solved, and the technical effect of improving the operation stability, the product quality and the production efficiency of the cattle bone powder production line is achieved.

Description

Technical Field

[0001] This application relates to the technical field of production fault diagnosis, and specifically to an intelligent fault diagnosis method and system for a bovine bone meal production line. Background Art

[0002] In the modern food processing industry, bovine bone meal, as an important raw material, is widely used in fields such as feed and nutritional supplements. The efficient and stable operation of the bovine bone meal production line is crucial for ensuring product quality and improving production efficiency. The early fault diagnosis of the bovine bone meal production line mainly relied on manual experience. Operators judged the operating state of the equipment by means of hearing, vision, and touch. This method was not only inefficient, but also the diagnosis results were extremely vulnerable to subjective factors, making it impossible to detect potential faults in a timely and accurate manner, resulting in frequent shutdowns of the production line for maintenance, seriously affecting the production progress and product quality, and being difficult to meet the actual production needs, thus affecting the overall stability, reliability, and intelligent level of the bovine bone meal production line.

[0003] Therefore, in the current related technologies, there are technical problems such as it being difficult to conduct real-time and accurate fault diagnosis on the bovine bone meal production line, resulting in lagging fault diagnosis results and affecting production efficiency and product quality. Summary of the Invention

[0004] This application provides an intelligent fault diagnosis method and system for a bovine bone meal production line, which solves the technical problems in the prior art that it is difficult to conduct real-time and accurate fault diagnosis on the bovine bone meal production line, resulting in lagging fault diagnosis results and affecting production efficiency and product quality, and achieves the technical effect of improving the operating stability, product quality, and production efficiency of the bovine bone meal production line.

[0005] This application provides an intelligent fault diagnosis method for a bovine bone meal production line. The method includes: real-time collecting equipment operation data and process parameters of the bovine bone meal production line through multi-modal sensors to obtain a multi-source heterogeneous real-time monitoring data set; transmitting the real-time monitoring data set to an edge computing node for preprocessing to obtain standard fusion data; inputting the standard fusion data into a cloud digital twin to dynamically simulate the operating state of the production line and obtain virtual-real mapped production line health state simulation data; analyzing the production line health state simulation data through a hierarchical diagnosis engine to obtain a diagnosis result; and triggering a dynamic response according to the diagnosis result to conduct closed-loop intelligent operation and maintenance of the bovine bone meal production line.

[0006] The present application also provides an intelligent fault diagnosis system for a bovine bone powder production line. The system includes: a real-time monitoring data set acquisition module, configured to collect device operation data and process parameters of the bovine bone powder production line in real time through multi-modal sensors, so as to obtain a multi-source heterogeneous real-time monitoring data set; a standard fusion data acquisition module, configured to transmit the real-time monitoring data set to an edge computing node for preprocessing to obtain standard fusion data; a production line dynamic simulation module, configured to input the standard fusion data into a cloud digital twin to dynamically simulate the operation state of the production line, so as to obtain production line health state simulation data with virtual-real mapping; a diagnosis result acquisition module, configured to analyze the production line health state simulation data through a hierarchical diagnosis engine to obtain a diagnosis result; and a dynamic response module, configured to trigger a dynamic response according to the diagnosis result to perform closed-loop intelligent operation and maintenance of the bovine bone powder production line.

[0007] It is intended to propose an intelligent fault diagnosis method and system for a bovine bone powder production line through the present application. Device operation data and process parameters of the bovine bone powder production line are collected in real time through multi-modal sensors to obtain a real-time monitoring data set; the real-time monitoring data set is transmitted to an edge computing node for preprocessing; the standard fusion data is input into a cloud digital twin to dynamically simulate the operation state of the production line, so as to obtain production line health state simulation data with virtual-real mapping; through analysis by a hierarchical diagnosis engine, a diagnosis result is obtained; and a dynamic response is triggered according to the diagnosis result to perform closed-loop intelligent operation and maintenance of the bovine bone powder production line. The technical problem in the prior art that it is difficult to perform fault diagnosis on a bovine bone powder production line in real time and accurately, resulting in a lag in the fault diagnosis result and affecting production efficiency and product quality, is solved, and the technical effects of improving the operation stability, product quality and production efficiency of the bovine bone powder production line are achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the accompanying drawings of the embodiments of the present disclosure will be briefly introduced below. Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the operations before or below do not necessarily need to be executed precisely in sequence. On the contrary, according to the need, they can be executed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or several operations can be removed from these processes.

[0009] Figure 1 It is a schematic flow chart of an intelligent fault diagnosis method for a bovine bone powder production line provided by an embodiment of the present application.

[0010] Figure 2 It is a schematic structural diagram of an intelligent fault diagnosis system for a bovine bone powder production line provided by an embodiment of the present application.

[0011] Description of the attached drawing reference numerals: Real-time monitoring data set acquisition module 10, standard fusion data acquisition module 20, production line dynamic simulation module 30, diagnosis result acquisition module 40, dynamic response module 50. Detailed implementation manners

[0012] The above description is only an overview of the technical solution of the present application. In order to be able to understand the technical means of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present application more obvious and understandable, the following specifically gives the detailed implementation manners of the present application.

[0013] In order to make the purpose, technical solution and advantages of the present application clearer, the present application will be further described in detail below with reference to the attached drawings. The described embodiments should not be regarded as limitations of the present application. All other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0014] In the following description, "some embodiments" are involved, which describe a subset of all possible embodiments. However, it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict. The terms "first\second" involved are only used to distinguish similar objects and do not represent a specific order for the objects. The terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or server including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or modules not clearly listed or inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application.

[0015] The embodiment of the present application provides an intelligent fault diagnosis method for a bovine bone meal production line, as Figure 1 shown. The method includes:

[0016] Step S100, collecting the equipment operation data and process parameters of the bovine bone meal production line in real time through multi-modal sensors to obtain a multi-source heterogeneous real-time monitoring data set.

[0017] Preferably, various sensors are used to collect the equipment operation data and process parameters of the bovine bone powder production line in real time. Specifically, the equipment operation data includes the monitoring of the operation status of key equipment and the monitoring of equipment operation environment parameters. The production of bovine bone powder requires the collaborative work of key equipment such as crushers, grinders, and dryers. Vibration sensors are installed on the outer shell of the crusher to collect the vibration signals generated during equipment operation in real time, intuitively reflecting the operation status of components such as the bearings and rotors of the crusher; motor current sensors monitor the motor current of the grinder, and the current fluctuation can reflect the load change of the grinder. If the material feeding is uneven or the grinding components are worn, the motor current will change accordingly; temperature and humidity sensors are deployed in the production workshop and key parts of the equipment to monitor the temperature and humidity of the equipment operation environment. For the dryer, too high a temperature may lead to a decline in the quality of bovine bone powder, and abnormal humidity will affect the performance of the equipment's electrical components; gas sensors are used to detect the concentration of harmful gases in the workshop. For example, bovine bone may produce peculiar smells or harmful gases during the processing, and timely monitoring can ensure the safety of the production environment and avoid damage such as corrosion to the equipment.

[0018] Preferably, the process parameters include the monitoring of material-related parameters and the monitoring of production process parameters. Specifically, a mass flowmeter is installed in the material conveying pipeline to accurately measure the flow rate of bovine bone raw materials entering the production line; moisture sensors are used to detect the moisture content of bovine bone raw materials and semi-finished products at each production stage. In the drying link, the drying temperature and time are adjusted in real time according to the feedback of the moisture sensor to ensure that the moisture content of bovine bone powder meets the standards; pressure sensors are installed in equipment such as cooking tanks and reaction kettles to monitor the internal pressure. When cooking bovine bone, the pressure control directly affects the bone powder boiling effect and the retention of nutritional components, and the pH value sensor monitors the acidity and alkalinity of the reaction system in real time. These equipment operation data and process parameters collected by multi-modal sensors have diverse sources (different types of sensors, different equipment and process links), and different data formats and natures (such as vibration signals being analog data and equipment start-stop states being digital data), jointly constituting a multi-source heterogeneous real-time monitoring data set.

[0019] Furthermore, step S100 further includes step S110, which monitors the turbidity of the cleaning liquid through a turbidity sensor and monitors the pH value of the cleaning liquid through a pH sensor to obtain a real-time cleaning monitoring data set; step S120, which collects the mechanical vibration signal of the crusher through a vibration sensor and monitors the motor load through a current sensor to obtain a real-time crushing monitoring data set; step S130, which monitors the temperature distribution in the cooking tank through an infrared thermal imager, collects steam pressure data through a pressure sensor, and detects the defatting rate of the aggregate through an on-line fat content detector to obtain a real-time cooking monitoring data set; step S140, which monitors the humidity of the hot air through a humidity sensor and scans the inner wall of the dryer through an infrared thermal imager to obtain a real-time drying monitoring data set; step S150, which collects the noise of the pulverizer through an acoustic sensor and detects the fineness of the bone meal through a laser particle size analyzer to obtain a real-time pulverizing monitoring data set; step S160, which monitors the temperature in the sterilization tank through a temperature sensor array and / or monitors the irradiation dose through an ultraviolet intensity sensor to obtain a real-time sterilization monitoring data set; step S170, which monitors the net content deviation of each package of bone meal through a weight sensor and identifies foreign matter intrusion through a metal detector to obtain a real-time packaging monitoring data set; step S180, which integrates the real-time cleaning monitoring data set, the real-time crushing monitoring data set, the real-time cooking monitoring data set, the real-time drying monitoring data set, the real-time sterilization monitoring data set, and the real-time packaging monitoring data set to obtain a multi-source heterogeneous real-time monitoring data set.

[0020] Preferably, in the cleaning process at the early stage of bone meal production, a turbidity sensor is used to monitor the turbidity of the cleaning liquid. The turbidity of the cleaning liquid can reflect the cleaning condition of the impurities on the surface of the bovine bone. The higher the turbidity, the more impurities there are, and the worse the cleaning effect may be. A pH sensor monitors the pH value of the cleaning liquid. An appropriate pH value helps to improve the cleaning efficiency and prevent corrosion of the equipment, thereby constituting a real-time cleaning monitoring data set and reflecting the key information of the cleaning process. The crusher is a device for initially crushing bovine bones. A vibration sensor collects the mechanical vibration signal of the crusher. When faults such as component loosening and wear occur, the vibration signal will change abnormally. A current sensor monitors the motor load. The change in the motor load is related to the working state of the crusher. For example, excessive feeding or jamming of the crushing components will cause an increase in the motor load. Combining them forms a real-time crushing monitoring data set, which is used to evaluate the operating state and performance of the crusher.

[0021] Preferably, in the cooking process, a thermal imager monitors the temperature distribution in the cooking tank to ensure that the cattle bones are heated evenly during the cooking process. Different temperature distributions may affect the degreasing and softening effects of the cattle bones. The pressure sensor collects steam pressure data. The steam pressure has an important impact on the efficiency and quality of cooking. The appropriate pressure can ensure the smooth progress of the cooking process. The fat content online detector detects the degreasing rate of the aggregates, and the degreasing rate is used to measure the cooking effect. These data together constitute a real-time cooking monitoring data set, which comprehensively reflects the process parameters and product quality related information of the cooking process. During the drying process, the humidity sensor monitors the hot air humidity, which affects the drying speed and final moisture content of the cattle bone meal. The thermal imager scans the inner wall of the dryer to detect whether the inner wall temperature is uniform, avoiding the situation where local overheating or overcooling causes uneven drying of the cattle bone meal. These two data constitute a real-time drying monitoring data set, which is used to control and optimize the drying process.

[0022] Preferably, the grinder further grinds the dried cattle bones into fine powder, and the acoustic sensor collects the grinder noise. The noise is stable during normal operation, and the noise will change when a fault occurs or the grinding effect is not good; the laser particle size analyzer detects the fineness of bone powder, where the fineness of bone powder is an important quality indicator of the product. Different application scenarios have different requirements for the fineness of bone powder. These two data constitute a real-time grinding monitoring data set, which is used to evaluate the working status of the grinder and product quality. In the sterilization process, the temperature in the sterilization tank is monitored by a temperature sensor array to ensure that all parts of the tank meet the temperature conditions required for sterilization and ensure the sterilization effect; or the irradiation dose is monitored by an ultraviolet intensity sensor. If ultraviolet sterilization is used, the irradiation dose directly affects the effect and efficiency of sterilization; these two types of collected data constitute a real-time sterilization monitoring data set to ensure the safety of the product.

[0023] Preferably, during the packaging stage, the weight sensor monitors the net content deviation of each package of cattle bone meal to ensure that the product meets the weight standard and avoids the situation of underweight or overpackaging; the metal detector identifies foreign matter mixed in to prevent metal and other impurities from entering the product, thereby improving product quality and safety; these two data form a real-time packaging monitoring data set for monitoring the quality control of the packaging link. Finally, the real-time monitoring data sets collected from each production link are integrated to form a multi-source heterogeneous real-time monitoring data set, which contains rich information from each link in the cattle bone meal production process.

[0024] Step S200: transmit the real-time monitoring data set to the edge computing node for preprocessing to obtain standard fusion data.

[0025] Preferably, the real-time monitoring data set is transmitted to the edge computing node for preprocessing to obtain standard fusion data, which can make the original multi-source heterogeneous data more regular and orderly. Specifically, the real-time monitoring data set collected by the multimodal sensor is transmitted from each sensor node to the edge computing node through a wired network (such as Ethernet) or a wireless network (such as Wi-Fi, Bluetooth, ZigBee, etc.), wherein the edge computing node is generally a computing device deployed near the production line, which is used to perform preliminary processing of the data near the data source. Preprocessing the real-time monitoring data set at the edge computing node may include data cleaning, data normalization, and data dimensionality reduction.

[0026] Preferably, the data collected by the sensor may have problems such as noise, missing values, and outliers. Data cleaning is to remove interfering information and improve data quality. For example, due to sensor failure or external interference, the turbidity data of the cleaning liquid collected by the turbidity sensor may have outliers that are obviously deviated from the normal range. By setting a reasonable threshold range, the outliers are removed; for missing values caused by communication failures, interpolation methods (such as linear interpolation, spline interpolation, etc.) can be used to supplement. Data normalization is to use minimum-maximum normalization, Z-score normalization, etc. to unify the data to have the same dimension and value range. Data dimensionality reduction is to use principal component analysis (PCA), linear discriminant analysis (LDA), etc. to reduce the number of data features while retaining the main information of the data. For example, in the collected equipment operation data, there may be multiple highly correlated features. Data dimensionality reduction is used to remove redundant features and improve the efficiency of data processing. After completing the preprocessing, data from different sources and of different types are fused to explore the correlation and complementarity between the data, so as to more comprehensively and accurately reflect the operating status of the production line, and finally obtain standard fused data, which can more effectively support dynamic simulation and fault diagnosis analysis.

[0027] Furthermore, step S200 also includes step S210, performing time alignment processing on the time series data to obtain first preprocessed data; step S220, performing signal-to-noise ratio evaluation on the time series signal, and performing adaptive noise reduction processing on the signal with a signal-to-noise ratio lower than a preset signal-to-noise ratio threshold to obtain second preprocessed data; step S230, performing feature extraction on the image and spectral data to obtain third preprocessed data; step S240, performing Z-score standardization processing on the numerical data to obtain fourth preprocessed data; step S250, establishing a spatiotemporal topological association between the first preprocessed data, the second preprocessed data, the third preprocessed data and the fourth preprocessed data to obtain spatiotemporal topological association data; step S260, standardizing and encapsulating the first preprocessed data, the second preprocessed data, the third preprocessed data, the fourth preprocessed data and the spatiotemporal topological association data to obtain standard fusion data.

[0028] Preferably, for the time-series data collected by different sensors in the bovine bone meal production line, due to differences in factors such as sampling frequency and start time, their timestamps may not be consistent. Time alignment processing is to calibrate the time-series data in the time dimension so that the data from different data sources have a corresponding relationship at the same time point. For example, a vibration sensor collects vibration data of a crusher at a frequency of 10 times per second, while a temperature sensor collects equipment temperature data at a frequency of 5 times per second. Through time alignment processing, these two types of data are matched at the same time point to obtain the first preprocessed data. The signal-to-noise ratio reflects the proportional relationship between the useful information and noise in the signal. Evaluating the signal-to-noise ratio of a time-series signal is to calculate the signal-to-noise ratio of each time-series signal and compare it with a preset signal-to-noise ratio threshold. For signals with a signal-to-noise ratio lower than the threshold, it indicates that the noise interference is large and adaptive noise reduction processing is required to improve the signal quality. For example, for the noise signal of a pulverizer collected by an acoustic sensor, if there is other interfering noise in the environment, it may cause the signal-to-noise ratio of this signal to decrease. It is processed through an adaptive noise reduction algorithm (such as wavelet noise reduction) to remove the noise interference and obtain a signal with higher quality, thereby obtaining the second preprocessed data.

[0029] Preferably, feature extraction is performed on the image and spectral data, that is, representative and discriminative features are extracted from the data to reduce the data volume and highlight the key information. Specifically, for the temperature distribution image inside a cooking tank collected by a thermal imager, the boundaries of temperature changes, the shapes of high-temperature regions, etc. are extracted through image processing algorithms (such as edge detection, texture analysis, etc.); for the spectral data of aggregates detected by an on-line fat content detector, the characteristic peaks related to the fat content are extracted; these extracted features are combined to form the third preprocessed data. Numerical data may come from different sensors and have different dimensions and value ranges. Z-score standardization processing is performed on it to convert the data into a standard normal distribution with a mean of 0 and a standard deviation of 1, eliminating the influence of dimensions. For example, the net content data of each package of bovine bone meal collected by a weight sensor and the steam pressure data collected by a pressure sensor have very different dimensions and value ranges. Through Z-score standardization processing, they are unified to the same scale to obtain the fourth preprocessed data.

[0030] Preferably, the first, second, third, and fourth preprocessed data are respectively from different types of sensors and different production links, and there may be spatio-temporal topological relationships. Establish the spatio-temporal topological associations among the first preprocessed data, the second preprocessed data, the third preprocessed data, and the fourth preprocessed data, that is, mine the internal connections between these data in terms of time and space. For example, the sequential association in time between the vibration data of a certain device and the temperature data of adjacent devices, or the distribution relationship in space of sensor data at different positions. The change in the vibration signal of the crusher may affect the feeding situation of the grinder after a period of time, resulting in a change in the motor current of the grinder. By analyzing the chronological order and causal relationship of the data, establish spatio-temporal topological associations to obtain spatio-temporal topological association data. Finally, standardize and package the first preprocessed data, the second preprocessed data, the third preprocessed data, the fourth preprocessed data, and the spatio-temporal topological association data. For example, package them in JSON format, and each data item has a clear identifier and meaning, and finally obtain standard fusion data.

[0031] Further, step S250 further includes step S251, extracting production nodes from the first preprocessed data, the second preprocessed data, the third preprocessed data, and the fourth preprocessed data to obtain multiple production nodes; step S252, based on the multiple production nodes, spatially align the preprocessed data of the same production node to obtain spatial topological association data; step S253, according to the production node flow chart, obtain the timing relationship of the production nodes, and establish the topological association of the preprocessed data according to the technological timing based on the timing relationship and the multiple production nodes to obtain time topological association data; step S254, integrate the spatial topological association data and the time topological association data to obtain spatio-temporal topological association data.

[0032] Preferably, the production nodes are the links in the production process of bone meal. For example, in the production of bone meal, cleaning, crushing, steaming, etc. Extracting production nodes from the first, second, third, and fourth preprocessed data is to identify the data subsets corresponding to each production link from them, classify them into different production nodes, and then obtain multiple production nodes. For example, from the preprocessed data, extract the turbidity, pH, etc. data related to the cleaning link and classify them into the cleaning node; classify the vibration, current, etc. data of the crusher into the crushing node. Then, based on the extracted multiple production nodes, spatially align the preprocessed data of different sensors within the same production node, that is, establish an accurate association for the data in the spatial dimension. For example, in the steaming tank node, the temperature distribution image data collected by the thermal imager and the steam pressure data collected by the pressure sensor are both related to the spatial position. Through spatial alignment, clarify the pressure value corresponding to each position in the image, so as to obtain the spatial topological relationship of the data within this production node; and then perform spatial alignment on all production nodes to finally obtain spatial topological association data.

[0033] Preferably, the production node flow chart describes the sequence and logical relationship of each production node in the production process of bone meal. According to the production node flow chart, the timing relationship of production nodes is obtained, and then based on the timing relationship and multiple production nodes, the preprocessing data of different production nodes are topologically associated in the time sequence of process execution, so as to reflect the change and mutual influence of data over time in the production process, and thus obtain time-topological association data; finally, the space-topological association data and time-topological association data are integrated to obtain spatio-temporal topological association data, which can more comprehensively and accurately describe the operation status of the bone meal production line and the mutual relationship between each link.

[0034] Step S300: Input the standard fusion data into the cloud digital twin to dynamically simulate the operation status of the production line and obtain the production line health status simulation data of virtual-real mapping.

[0035] Preferably, input the standard fusion data into the cloud digital twin to dynamically simulate the operation status of the production line. Among them, the cloud digital twin is a digital twin model based on cloud computing and is an accurate mapping of the bone meal production line in the virtual space. It not only includes the physical structure information of the production line, such as the layout and connection method of equipment, but also includes the logical information of the production process, such as process flow, control strategy, etc.; through real-time data interaction with the actual production line, it can reflect the operation status of the production line in real time. Input the standard fusion data into the cloud digital twin to dynamically simulate the operation status of the production line, specifically including equipment behavior simulation, process simulation, and module cooperation simulation. Then, according to the operation data of the equipment, such as the rotation speed and vibration frequency of the motor, the actual operation of the equipment is simulated. For example, for a crusher, by inputting data such as motor current and vibration signal, the simulation model can accurately present the crushing efficiency and component wear of the crusher; combined with process parameters, such as temperature, pressure, humidity, etc., the physical changes in the production process are simulated. Taking the cooking link as an example, according to data such as steam pressure and temperature distribution, the simulation model can predict the degreasing effect and nutrient retention of bones; considering the mutual influence between each equipment and link in the production line, the coordinated operation of the entire production system is simulated. Through dynamic simulation, the cloud digital twin generates virtual data corresponding to the operation status of the actual production line, that is, the production line health status simulation data of virtual-real mapping, which can timely reflect the latest situation of the production line, predict the future state of the production line, and discover potential fault hazards in advance.

[0036] Further, step S300 further includes step S310 of establishing a first cloud digital twin of the bovine bone meal production line based on physical rules; step S320 of constructing a production fault tree for bovine bone meal according to the historical fault data of the bovine bone meal production line; step S330 of training the fault response logic of the first cloud digital twin according to the production fault tree for bovine bone meal to obtain a second cloud digital twin; and step S340 of inputting the standard fusion data into the second cloud digital twin to dynamically simulate the operation state of the production line and obtain the production line health state simulation data with virtual-real mapping.

[0037] Preferably, the physical rules refer to the scientific principles followed in the production process of bovine bone meal, such as mechanical principles, thermal principles, chemical principles, etc. Based on these physical rules, a virtual bovine bone meal production line model, i.e., the first cloud digital twin, is constructed in the cloud to accurately reflect the physical characteristics, operation mechanisms, etc. of each device in the production line. For example, for a crusher, according to mechanical principles, considering its motor power, crushing chamber structure, blade shape, etc., a model that can accurately simulate its crushing process is established; for a cooking tank, based on thermal principles, the steam heating process, heat transfer method, and the effect on bovine bone are simulated.

[0038] Preferably, according to the historical fault data of the bovine bone meal production line, where the historical fault data records various fault situations that occurred during the past operation of the bovine bone meal production line, including information such as the time, location, phenomenon, and cause of the fault. Specifically, a fault tree is a graphical logical model that starts with an event that is not desired to occur (top event, such as production line shutdown, unqualified product quality, etc.), and through layer-by-layer analysis of the direct causes (intermediate events) and basic causes (bottom events, such as equipment component damage, operation errors, etc.) that lead to this event, a tree-like structure is constructed, that is, a production fault tree for bovine bone meal is constructed. For example, taking "the crusher cannot work properly" as the top event, by analyzing the historical fault data, possible intermediate events are found, such as motor failure, conveyor belt breakage, etc., and then the bottom events that lead to the intermediate events are further found, such as motor overheating, conveyor belt aging, etc., thus constructing a sub-fault tree for crusher faults. Integrating the sub-fault trees of each device and link, the production fault tree for bovine bone meal is obtained.

[0039] Preferably, the constructed fault tree of bone meal production is used to train the fault response logic of the first cloud digital twin. By integrating various fault scenarios and causal relationships in the fault tree into the digital twin model, the digital twin can identify possible faults during the simulation process and make corresponding responses according to the type and severity of the faults. For example, when it is detected during the simulation process that the current of the crusher motor increases abnormally (which may indicate a motor fault), the second cloud digital twin can, according to the trained fault response logic, simulate the possible measures that the production line may take, such as automatic shutdown, alarm, adjustment of the operating parameters of adjacent equipment, etc.

[0040] Preferably, the standard fusion data is input into the optimized second cloud digital twin. The digital twin dynamically adjusts its own state according to the input data, conducts real-time simulation of the operation of the production line. Then, the digital twin presents a highly consistent operation state with the actual production line in the virtual space, so as to obtain the simulation data of the production line health state with virtual-real mapping, reflecting the health status of the production line at each moment, including whether the equipment is operating normally, whether the process parameters meet the requirements, etc. For example, according to the real-time input vibration data of the crusher, motor current data, and temperature and pressure data of the cooking tank, etc., the second cloud digital twin can dynamically simulate the operation state of the entire production line and generate simulation data of the production line health state such as the wear degree of the equipment, product quality indicators, potential fault risks, etc.

[0041] Further, step S320 further includes step S321, determining a set of key fault modes according to the historical fault data of the bone meal production line; step S322, extracting a first fault mode from the set of key fault modes, where the first fault mode is any one in the set of key fault modes; step S323, for the first fault mode, analyzing and extracting equipment-level fault causes from the historical equipment log data to obtain an equipment fault set, and constructing a first-level branch based on the equipment fault set; step S324, associating the historical process parameter data based on the equipment fault set to obtain a process anomaly set, and constructing a second-level branch based on the process anomaly set; step S325, analyzing the historical interaction record data of the equipment fault set and the process anomaly set to obtain a composite fault set, and constructing a third-level branch based on the composite fault set; step S326, inputting the first-level branch, the second-level branch, and the third-level branch into a Bayesian network, obtaining the occurrence probability and propagation weight of each fault cause node, and constructing a dynamically optimized fault tree of bone meal production based on the occurrence probability and the propagation weight.

[0042] Preferably, by analyzing the historical fault data of the bovine bone powder production line, the frequency of faults occurring, the degree of impact of faults on production, etc. are statistically analyzed, and fault modes that have a greater impact on the operation of the production line, a higher occurrence probability, or may lead to serious consequences are screened out to form a set of key fault modes; then a fault mode is randomly selected from them and defined as the first fault mode. For example, in the set of key fault modes, there are modes such as "breaker blade breakage", "cooking tank temperature out of control", "dryer blockage", etc., and "breaker blade breakage" is randomly selected as the first fault mode; then for the selected first fault mode, the equipment-level reasons that may cause this fault are searched for and analyzed from the historical data of the equipment log. Among them, the equipment log usually records information such as the operating parameters, status changes, and maintenance records of the equipment. By analyzing, the specific equipment fault reasons are determined to form an equipment fault set. Then, based on the equipment fault set, the first-level branches of the fault tree are constructed, representing the equipment-related factors that directly lead to the first fault mode.

[0043] Preferably, based on the equipment fault set, the historical process parameter data is further associated. Among them, the historical process parameter data records various parameters in the production process, such as temperature, pressure, flow rate, time, etc. The process anomalies related to the equipment faults are analyzed and determined to form a process anomaly set, and then the second-level branches of the fault tree are constructed, representing the process-level factors related to the equipment faults. For example, for the equipment fault set of "breaker blade breakage", by associating the historical process parameter data, it is found that when the feeding speed is too fast (process anomaly), it will increase the load of the breaker and accelerate the wear of the blade, thus making the blade more likely to break; or when the material humidity does not meet the requirements, it may also affect the crushing effect and cause additional damage to the blade. Analyze the historical data of the interaction records of the equipment fault set and the process anomaly set, that is, search for the situations where equipment faults and process anomalies occur simultaneously or affect each other, determine the compound fault reasons, form a compound fault set, and based on the compound fault set, construct the third-level branches of the fault tree to reflect the impact of the interaction between equipment faults and process anomalies on the occurrence of faults.

[0044] Preferably, input the constructed primary branches, secondary branches, and tertiary branches into the Bayesian network, and calculate the occurrence probability and propagation weight of each fault cause node based on historical data. Among them, the occurrence probability represents the likelihood of each fault cause occurring, and the propagation weight reflects the intensity of the mutual influence and propagation between fault causes. Based on the occurrence probability and propagation weight, optimize the fault tree so that it can more accurately reflect the likelihood of fault occurrence and the relationship between fault causes, thereby constructing a dynamically optimized fault tree for beef bone meal production. For example, in the Bayesian network, the calculated occurrence probability of the equipment fault cause of "unqualified blade material" is 0.2, and the occurrence probability of the process anomaly cause of "too fast feeding speed" is 0.3, and the propagation weight is 0.6, indicating that when "unqualified blade material" occurs, "too fast feeding speed" will increase the probability of "breaker blade fracture" by 0.6 times. Then, according to the occurrence probability and weight information, adjust and optimize the fault tree to make it more consistent with the actual fault occurrence situation in production.

[0045] Step S400, analyze the production line health status simulation data through a hierarchical diagnosis engine to obtain a diagnosis result.

[0046] Preferably, the hierarchical diagnosis engine is a tool for fault diagnosis or status assessment of complex systems. Usually, based on the hierarchical idea, the diagnosis problem is decomposed into multiple levels for processing. Each level is responsible for analyzing different granularity information to more comprehensively and accurately judge the system status. Specifically, use the hierarchical diagnosis engine to gradually analyze the production line health status simulation data. It may first analyze the overall operation performance indicators of the entire production line at the macroscopic level, such as whether the production efficiency and output meet the expectations, etc.; then go deep into the intermediate level to check the operation status of each production unit or key equipment to see if there are abnormal operation parameters or fault signs in the equipment; finally, at the microscopic level, analyze specific process parameters, sensor data, etc. to determine whether there are potential minor problems or hidden dangers. After a comprehensive analysis of the production line health status simulation data by the hierarchical diagnosis engine, a diagnosis result regarding the production line health status is generated, which may include an assessment of the current state of the production line, such as whether it is operating normally, whether there are faults, and the severity of the faults; it may also include a warning of potential problems, indicating which parts may have faults in the future and need to take preventive measures in advance; in addition, the diagnosis result may also provide suggestions or solutions, such as which equipment needs to be repaired and which process parameters need to be adjusted, so that production personnel can take corresponding measures according to the diagnosis result to maintain the normal operation of the production line, improve production efficiency and product quality.

[0047] Further, step S400 further includes step S410 of constructing a first diagnostic engine based on a preset threshold matching rule; step S420 of modeling and predicting the performance degradation trend of the equipment on the bovine bone meal production line to construct a second diagnostic engine; step S430 of mining association rules for multi-equipment collaborative faults to construct a third diagnostic engine; step S440 of constructing a fourth diagnostic engine through causal reasoning based on the fault causality; step S450 of connecting the first diagnostic engine, the second diagnostic engine, the third diagnostic engine, and the fourth diagnostic engine in parallel to obtain a hierarchical diagnostic engine; and step S460 of analyzing the production line health status simulation data through the hierarchical diagnostic engine to obtain a diagnostic result.

[0048] Preferably, according to the normal operating parameter range of the equipment on the bovine bone meal production line, the thresholds of key parameters are preset in advance, and then the first diagnostic engine is constructed. By comparing and matching the real-time collected equipment operating parameters with the preset thresholds, when the parameters exceed the threshold range, it is determined that the equipment may have a fault or abnormality. For example, the normal operating current range of the crusher motor is 20 - 30 amperes. When the first diagnostic engine detects that the real-time current is 35 amperes, exceeding the preset upper threshold, a warning is issued, indicating possible problems such as motor overload. Using historical data and appropriate mathematical models (such as time series analysis, machine learning algorithms, etc.), model the change trend of the performance indicators (such as the wear degree and accuracy decline of the equipment) of the equipment on the bovine bone meal production line over time to construct a second diagnostic engine. Predict the future performance state of the equipment through the model, and detect signs of equipment performance degradation in advance so as to take maintenance measures in time to avoid sudden equipment failures. Mine association rules for multi-equipment collaborative faults to construct a third diagnostic engine. By mining the association relationships between different equipment faults in the historical fault data, find out those relevant fault patterns. For example, through the analysis of a large amount of historical data, the third diagnostic engine finds that when the temperature control of the cooking tank is abnormal, in 30% of the cases, it will cause the pressure of the concentration equipment to be unstable. Based on the association rules, when the temperature of the cooking tank is detected to be abnormal again, the third diagnostic engine will focus on monitoring the state of the concentration equipment and issue a warning of possible faults.

[0049] Preferably, through in-depth understanding of historical fault data and production processes, the causal chains between different faults are determined, and a fourth diagnostic engine is constructed to perform reasoning based on the causal relationships between faults. When a certain fault is detected, it can infer the possible root causes and further predict other faults that may be triggered. For example, when a fault of premature wear of the crusher blade is detected, the fourth diagnostic engine discovers through causal reasoning that it is caused by excessive impurities in the feed (causal relationship). At the same time, it can also predict that due to blade wear, the subsequent grinding effect may be affected and even cause an overload fault in the grinder, thus issuing relevant early warnings in advance. The first, second, third, and fourth diagnostic engines are connected in parallel to form a hierarchical diagnostic engine. Each diagnostic engine is responsible for analyzing the production line health status simulation data from different perspectives (threshold matching, performance degradation prediction, multi-device collaborative fault correlation, fault causal reasoning), and integrating the analysis results of each diagnostic engine to obtain a comprehensive diagnostic result regarding the production line health status, which may include whether the equipment is operating normally, what faults or abnormalities exist, what the causes of the faults are, potential future problems, and corresponding handling suggestions, etc., thereby improving the accuracy and reliability of fault diagnosis.

[0050] Further, step S420 further includes step S421, using an LSTM network to learn the temporal degradation characteristics of vibration signals and acoustic signals, and establishing a tool wear prediction model for the crusher; step S422, fusing the tool wear prediction model and the tool physical model to perform tool remaining life prediction and obtain the predicted remaining life of the tool; step S423, setting an adaptive tool life warning threshold based on the production load, and constructing a second diagnostic engine according to the adaptive tool life warning threshold and the predicted remaining life of the tool.

[0051] Preferably, the long short-term memory network (LSTM) is a special type of recurrent neural network (RNN) that can process and remember long-term sequence information and is suitable for analyzing data that changes over time, such as vibration signals and acoustic signals. During the operation of the crusher, the tool gradually wears with the usage time, and its vibration signals and acoustic signals will change accordingly. The LSTM network can capture the changing patterns of these signals over time, i.e., the temporal degradation characteristics, through learning a large amount of historical vibration signal and acoustic signal data. For example, as the tool wears, the frequency of the vibration signal may shift, and the intensity and frequency distribution of the acoustic signal will also change. The LSTM network automatically learns the relationship between these changes and tool wear, thereby establishing a tool wear prediction model for the crusher. Using the current and past vibration signals and acoustic signals as inputs, it outputs an estimation of the current wear degree of the tool and a prediction of the wear trend in the future for a certain period of time.

[0052] Preferably, the physical model of the cutting tool is a model established based on physical knowledge such as the material properties, working principle, and force-bearing conditions of the cutting tool, which is used to describe the wear process and life law of the cutting tool under normal working conditions. Then, the established cutting tool wear prediction model is integrated with the physical model of the cutting tool to capture the influence of various complex factors on the cutting tool wear in actual production and more accurately predict the remaining life of the cutting tool. The production load refers to the amount of work tasks undertaken by the pulverizer at different production stages, which affects the wear rate of the cutting tool. The greater the production load, the faster the cutting tool wears. Then, an adaptive cutting tool life warning threshold is set based on the production load, that is, according to the current production load situation, the warning standard for the cutting tool life is dynamically adjusted. When the production load is high, the cutting tool wears faster, and the warning threshold will be correspondingly reduced to issue a warning earlier to remind the staff to replace the cutting tool in time; while when the production load is low, the warning threshold can be appropriately increased; finally, according to the adaptive cutting tool life warning threshold and the predicted remaining life of the cutting tool, a second diagnostic engine is constructed to monitor the state of the cutting tool in real time. When the predicted remaining life of the cutting tool approaches or is lower than the adaptive cutting tool life warning threshold, a warning signal is issued to prompt the staff that the cutting tool is about to reach the service life and needs to be replaced or maintained, thereby realizing the dynamic monitoring and diagnosis of the cutting tool state, avoiding affecting production efficiency, product quality, and even causing equipment failures due to excessive wear of the cutting tool, and ensuring the stable operation of the bovine bone meal production line.

[0053] Further, step S400 further includes step A1 of determining the key control nodes and non-key control nodes of the production line based on the historical failure data of the bovine bone meal production line; step A2 of using real-time transmission for the sensor data of the key control nodes and batch transmission for the sensor data of the non-key control nodes; step A3 of establishing a refined simulation model of the key control nodes and a simplified simulation model of the non-key control nodes in the cloud digital twin; step A4 of triggering a first-level response mechanism for the diagnostic results of the key control nodes and a second-level response mechanism for the diagnostic results of the non-key control nodes.

[0054] Preferably, based on the historical fault data of the beef bone powder production line, data transmission and response mechanism strategies are formulated for control nodes of different importance levels to improve the fault diagnosis and handling efficiency of the beef bone powder production line and rationally utilize resources. Specifically, based on the historical fault data of the beef bone powder production line, by analyzing the frequency of fault occurrence and the degree of impact of the fault on production (such as whether it causes production stagnation, serious decline in product quality, severe damage to equipment, etc.), the control nodes in the production line are divided into key control nodes and non-key control nodes. Among them, key control nodes refer to production nodes whose faults will have a serious impact on the output, quality, equipment safety, etc. of beef bone powder production. For example, the motor control node of the crusher. If the motor fails, the production line may stop the crushing operation. Non-key control nodes refer to production nodes with relatively small impact on production. Even if a fault occurs, it will not immediately lead to production stagnation or seriously affect product quality. For example, the control node of auxiliary equipment (workshop lighting equipment), whose fault generally will not directly affect the production process of beef bone powder.

[0055] Preferably, according to the criticality of the control nodes, different sensor data transmission methods are adopted to balance the real-time nature of data transmission and resource consumption. Specifically, the change in the state of key control nodes has a significant impact on production, and the latest data needs to be obtained in a timely manner. The sensor data of key control nodes is transmitted in real time so as to quickly discover potential fault hazards. For example, the current sensor data of the crusher motor is transmitted in real time to timely understand the load condition of the motor. Once the current abnormally increases, measures are immediately taken to avoid motor damage. The sensor data of non-key control nodes is transmitted in batches. Non-key control nodes have a relatively small impact on production and do not require real-time data acquisition. Transmitting data in batches can reduce the frequency of data transmission, reduce the occupancy of network bandwidth and the pressure of data processing. For example, the light intensity sensor data of workshop lighting equipment can be transmitted in a package every once in a while (such as one hour).

[0056] Preferably, in the cloud digital twin, simulation models of different complexity levels are established for control nodes of different criticality levels, including a refined simulation model for key control nodes and a simplified simulation model for non-key control nodes. Among them, the refined simulation model considers more factors and details and can more accurately simulate the operating state of key control nodes. For example, for the key control node of the cooking tank, the refined simulation model considers the dynamic changes of steam flow, pressure, temperature, as well as the physical and chemical changes of beef bones during the cooking process, etc., in order to more accurately predict possible faults. The simplified simulation model ignores some factors with relatively small overall impact to reduce the complexity and computational amount of the model. For example, for the control node of workshop ventilation equipment, the simplified model may only consider the simple relationship between ventilation volume and motor power and does not consider complex factors such as air flow distribution in the air duct.

[0057] Preferably, according to the criticality of the control nodes, different levels of response mechanisms are set for the diagnostic results to achieve targeted fault handling. Specifically, for the diagnostic results of critical control nodes, a first-level response mechanism is triggered. That is, when the diagnostic results of critical control nodes show the existence of faults or potential faults, the first-level response mechanism is triggered. The first-level response mechanism usually has a relatively high priority and will immediately notify relevant technicians for handling. It may also automatically take some emergency measures, such as automatic shutdown, starting standby equipment, etc., to avoid the expansion of faults and reduce the impact on production. For the diagnostic results of non-critical control nodes, a second-level response mechanism is triggered. When the diagnostic results of non-critical control nodes show faults, the second-level response mechanism is triggered. The priority of the second-level response mechanism is relatively low, and personnel may be arranged to handle it within a certain period (such as one working day), or some relatively mild measures may be taken, such as adjusting operating parameters, etc., to maintain the normal operation of production.

[0058] Step S500, trigger a dynamic response according to the diagnostic result to perform closed-loop intelligent operation and maintenance of the bone meal production line.

[0059] Preferably, according to different diagnostic results, corresponding dynamic responses are automatically triggered. Specifically, if a serious fault is diagnosed in the equipment, the dynamic response may be to immediately issue an alarm, notify the maintenance personnel for emergency repair, and automatically stop the operation of the relevant equipment. If the performance of the equipment gradually degrades, such as tool wear, the dynamic response may be to arrange a tool replacement plan in advance according to the predicted remaining life of the tool and the adaptive tool life warning threshold, and at the same time adjust the production plan to minimize the impact on production. And perform closed-loop intelligent operation and maintenance of the bone meal production line, including data collection, analysis and diagnosis, dynamic response, and effect feedback, etc. That is, after the measures are taken in the dynamic response, continuously monitor the operation status of the production line, collect relevant data, evaluate the effect of the operation and maintenance measures. If the expected effect is not achieved, further analyze the reasons, adjust the operation and maintenance strategy, and optimize again to achieve the reasonable arrangement of the production plan and the maintenance plan, reduce manual intervention, improve the operation and maintenance efficiency and accuracy, reduce production costs, and ensure the stable operation of the bone meal production line.

[0060] In the above text, reference is made to Figure 1 described in detail the intelligent fault diagnosis method of the bone meal production line according to the embodiments of the present invention. Next, reference will be made to Figure 2 describe the intelligent fault diagnosis system of the bone meal production line according to the embodiments of the present invention.

[0061] The intelligent fault diagnosis system for the bovine bone powder production line according to the embodiments of the present invention is used to solve the technical problems existing in the prior art, such as the difficulty in diagnosing faults in the bovine bone powder production line in real time and accurately, resulting in lagging fault diagnosis results and affecting production efficiency and product quality, and achieves the technical effects of improving the operation stability, product quality and production efficiency of the bovine bone powder production line. As Figure 2 shown, the intelligent fault diagnosis system for the bovine bone powder production line includes: a real-time monitoring data set acquisition module 10, a standard fusion data acquisition module 20, a production line dynamic simulation module 30, a diagnosis result acquisition module 40, and a dynamic response module 50.

[0062] The real-time monitoring data set acquisition module 10 is used to collect the equipment operation data and process parameters of the bovine bone powder production line in real time through multi-modal sensors, and obtain a multi-source heterogeneous real-time monitoring data set; the standard fusion data acquisition module 20 is used to transmit the real-time monitoring data set to an edge computing node for preprocessing to obtain standard fusion data; the production line dynamic simulation module 30 is used to input the standard fusion data into a cloud digital twin to dynamically simulate the operation state of the production line and obtain production line health state simulation data with virtual-real mapping; the diagnosis result acquisition module 40 is used to analyze the production line health state simulation data through a hierarchical diagnosis engine to obtain a diagnosis result; the dynamic response module 50 is used to trigger a dynamic response according to the diagnosis result and perform closed-loop intelligent operation and maintenance of the bovine bone powder production line.

[0063] Next, the specific configuration of the real-time monitoring data set acquisition module 10 will be described in detail. The real-time monitoring data set acquisition module 10 further includes: monitoring the turbidity of the cleaning liquid through a turbidity sensor and the pH value of the cleaning liquid through a pH sensor to obtain a real-time cleaning monitoring data set; collecting the mechanical vibration signal of the crusher through a vibration sensor and monitoring the motor load through a current sensor to obtain a real-time crushing monitoring data set; monitoring the temperature distribution in the cooking tank through an infrared thermal imager, collecting steam pressure data through a pressure sensor, and detecting the defatting rate of the bone material through an on-line fat content detector to obtain a real-time cooking monitoring data set; monitoring the hot air humidity through a humidity sensor and scanning the inner wall of the dryer through an infrared thermal imager to obtain a real-time drying monitoring data set; collecting the noise of the grinder through an acoustic sensor and detecting the fineness of the bone powder through a laser particle size analyzer to obtain a real-time grinding monitoring data set; monitoring the temperature in the sterilization tank through a temperature sensor array and / or monitoring the irradiation dose through an ultraviolet intensity sensor to obtain a real-time sterilization monitoring data set; monitoring the net content deviation of each pack of bovine bone powder through a weight sensor and identifying foreign matter intrusion through a metal detector to obtain a real-time packaging monitoring data set; integrating the real-time cleaning monitoring data set, the real-time crushing monitoring data set, the real-time cooking monitoring data set, the real-time drying monitoring data set, the real-time sterilization monitoring data set and the real-time packaging monitoring data set to obtain a multi-source heterogeneous real-time monitoring data set.

[0064] Next, the specific configuration of the real-time monitoring data set acquisition module 10 will be further described in detail. The real-time monitoring data set acquisition module 10 further includes: determining the key control nodes and non-key control nodes of the production line based on the historical fault data of the bovine bone meal production line; performing real-time transmission on the sensor data of the key control nodes and batch transmission on the sensor data of the non-key control nodes; establishing a refined simulation model of the key control nodes and a simplified simulation model of the non-key control nodes in the cloud digital twin; triggering a first-level response mechanism for the diagnostic results of the key control nodes and a second-level response mechanism for the diagnostic results of the non-key control nodes.

[0065] Next, the specific configuration of the standard fusion data acquisition module 20 will be described in detail. The standard fusion data acquisition module 20 further includes: performing time alignment processing on the time series data to obtain first preprocessed data; evaluating the signal-to-noise ratio of the time series signal, and performing adaptive noise reduction processing on the signal with a signal-to-noise ratio lower than the preset signal-to-noise ratio threshold to obtain second preprocessed data; extracting features from the image and spectral data to obtain third preprocessed data; performing Z-score standardization processing on the numerical data to obtain fourth preprocessed data; establishing a spatio-temporal topological association between the first preprocessed data, the second preprocessed data, the third preprocessed data, and the fourth preprocessed data to obtain spatio-temporal topological association data; performing standard encapsulation on the first preprocessed data, the second preprocessed data, the third preprocessed data, the fourth preprocessed data, and the spatio-temporal topological association data to obtain standard fusion data.

[0066] Next, the specific configuration of the standard fusion data acquisition module 20 will be further described in detail. The standard fusion data acquisition module 20 further includes: extracting production nodes from the first preprocessed data, the second preprocessed data, the third preprocessed data, and the fourth preprocessed data to obtain a plurality of production nodes; based on the plurality of production nodes, performing spatial alignment on the preprocessed data of the same production node to obtain spatial topological association data; according to the production node flow chart, obtaining the time series relationship of the production nodes, and establishing a topological association of the preprocessed data according to the time series relationship and the plurality of production nodes in the process time series to obtain time topological association data; integrating the spatial topological association data and the time topological association data to obtain spatio-temporal topological association data.

[0067] Next, the specific configuration of the production line dynamic simulation module 30 will be described in detail. The production line dynamic simulation module 30 further includes: establishing a first cloud digital twin of the bovine bone meal production line based on physical rules; constructing a bovine bone meal production fault tree according to the historical fault data of the bovine bone meal production line; training the fault response logic of the first cloud digital twin according to the bovine bone meal production fault tree to obtain a second cloud digital twin; inputting the standard fusion data into the second cloud digital twin, dynamically simulating the operation state of the production line, and obtaining production line health state simulation data with virtual-real mapping.

[0068] Next, the specific configuration of the production line dynamic simulation module 30 will be further described in detail. The production line dynamic simulation module 30 further includes: determining a set of key fault modes according to the historical fault data of the bovine bone meal production line; extracting a first fault mode from the set of key fault modes, where the first fault mode is any one in the set of key fault modes; for the first fault mode, analyzing and extracting device-level fault causes from the historical device log data to obtain a device fault set, and constructing a first-level branch based on the device fault set; based on the device fault set, associating the historical process parameter data to obtain a process anomaly set, and constructing a second-level branch based on the process anomaly set; analyzing the historical interaction record data of the device fault set and the process anomaly set to obtain a composite fault set, and constructing a third-level branch based on the composite fault set; inputting the first-level branch, the second-level branch, and the third-level branch into a Bayesian network, obtaining the occurrence probability and propagation weight of each fault cause node, and constructing a dynamically optimized bovine bone meal production fault tree based on the occurrence probability and the propagation weight.

[0069] Next, the specific configuration of the diagnosis result obtaining module 40 will be described in detail. The diagnosis result obtaining module 40 further includes: constructing a first diagnosis engine based on a preset threshold matching rule; modeling and predicting the performance degradation trend of the bovine bone meal production line equipment to construct a second diagnosis engine; mining association rules for multi-device collaborative faults to construct a third diagnosis engine; constructing a fourth diagnosis engine through causal reasoning according to the fault causal relationship; connecting the first diagnosis engine, the second diagnosis engine, the third diagnosis engine, and the fourth diagnosis engine in parallel to obtain a hierarchical diagnosis engine; analyzing the production line health state simulation data through the hierarchical diagnosis engine to obtain a diagnosis result.

[0070] Next, the specific configuration of the diagnosis result obtaining module 40 will be described in detail. The diagnosis result obtaining module 40 further includes: learning the temporal degradation characteristics of vibration signals and acoustic signals by using an LSTM network, and establishing a tool wear prediction model for the crusher; fusing the tool wear prediction model and the tool physical model to perform tool remaining life prediction and obtain the predicted remaining life of the tool; setting an adaptive tool life warning threshold based on the production load, and constructing a second diagnostic engine according to the adaptive tool life warning threshold and the predicted remaining life of the tool.

[0071] The intelligent fault diagnosis system for the bovine bone meal production line provided by the embodiments of the present invention can execute the intelligent fault diagnosis method for the bovine bone meal production line provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0072] Although various references are made to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or the server. The included units and modules are only divided according to the functional logic, but are not limited to the above division as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of mutual distinction and do not limit the protection scope of the present invention.

[0073] The above specific embodiments do not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. An intelligent fault diagnosis method for a beef bone powder production line, characterized in that The method includes: Real-time collecting the equipment operation data and process parameters of the bovine bone powder production line through multi-modal sensors to obtain a multi-source heterogeneous real-time monitoring data set; Transmitting the real-time monitoring data set to an edge computing node for preprocessing to obtain standard fusion data; Inputting the standard fusion data into a cloud digital twin to dynamically simulate the operation state of the production line and obtain production line health state simulation data with virtual-real mapping; Analyzing the production line health state simulation data through a hierarchical diagnosis engine to obtain a diagnosis result; Triggering a dynamic response according to the diagnosis result to perform closed-loop intelligent operation and maintenance of the bovine bone powder production line.

2. The intelligent fault diagnosis method for the bovine bone meal production line according to claim 1, wherein Real-time collecting the equipment operation data and process parameters of the bovine bone powder production line through multi-modal sensors to obtain a multi-source heterogeneous real-time monitoring data set, including: Monitoring the turbidity of the cleaning liquid through a turbidity sensor and monitoring the acidity and alkalinity of the cleaning liquid through a pH sensor to obtain a real-time cleaning monitoring data set; Collecting the mechanical vibration signal of the crusher through a vibration sensor and monitoring the motor load through a current sensor to obtain a real-time crushing monitoring data set; Monitoring the temperature distribution in the cooking tank through an infrared thermal imager, collecting steam pressure data through a pressure sensor, and detecting the defatting rate of the bone material through an on-line fat content detector to obtain a real-time cooking monitoring data set; Monitoring the hot air humidity through a humidity sensor and scanning the inner wall of the dryer through an infrared thermal imager to obtain a real-time drying monitoring data set; Collecting the noise of the pulverizer through an acoustic sensor and detecting the fineness of the bone powder through a laser particle size analyzer to obtain a real-time pulverizing monitoring data set; Monitoring the temperature in the sterilization tank through a temperature sensor array and / or monitoring the irradiation dose through an ultraviolet intensity sensor to obtain a real-time sterilization monitoring data set; Monitoring the net content deviation of each package of bovine bone powder through a weight sensor and identifying foreign matter mixing through a metal detector to obtain a real-time packaging monitoring data set; Integrating the real-time cleaning monitoring data set, the real-time crushing monitoring data set, the real-time cooking monitoring data set, the real-time drying monitoring data set, the real-time sterilization monitoring data set, and the real-time packaging monitoring data set to obtain a multi-source heterogeneous real-time monitoring data set.

3. The intelligent fault diagnosis method for the bovine bone meal production line according to claim 2, wherein It also includes: Determining the key control nodes and non-key control nodes of the production line based on the historical fault data of the bovine bone powder production line; Performing real-time transmission on the sensor data of the key control nodes and performing batch transmission on the sensor data of the non-key control nodes; Establishing a refined simulation model of the key control nodes and a simplified simulation model of the non-key control nodes in the cloud digital twin; Triggering a first-level response mechanism for the diagnosis results of the key control nodes and triggering a second-level response mechanism for the diagnosis results of the non-key control nodes.

4. The intelligent fault diagnosis method for the bovine bone powder production line according to claim 1, wherein Transmitting the real-time monitoring data set to an edge computing node for preprocessing to obtain standard fusion data, including: Performing time alignment processing on the time series data to obtain first preprocessing data; Evaluating the signal-to-noise ratio of the time series signal, and performing adaptive noise reduction processing on the signal with a signal-to-noise ratio lower than a preset signal-to-noise ratio threshold to obtain second preprocessing data; Performing feature extraction on the image and spectral data to obtain third preprocessing data; Perform Z-score normalization on the numerical data to obtain the fourth preprocessed data; Establish the spatio-temporal topological association among the first preprocessed data, the second preprocessed data, the third preprocessed data, and the fourth preprocessed data to obtain spatio-temporal topological association data; Perform standardized encapsulation on the first preprocessed data, the second preprocessed data, the third preprocessed data, the fourth preprocessed data, and the spatio-temporal topological association data to obtain standard fusion data.

5. The intelligent fault diagnosis method for the bovine bone meal production line according to claim 4, wherein Establish the spatio-temporal topological association among the first preprocessed data, the second preprocessed data, the third preprocessed data, and the fourth preprocessed data to obtain spatio-temporal topological association data, including: Extract production nodes from the first preprocessed data, the second preprocessed data, the third preprocessed data, and the fourth preprocessed data to obtain multiple production nodes; Based on the multiple production nodes, align the preprocessed data of the same production node in space to obtain spatial topological association data; According to the production node flow chart, obtain the timing relationship of the production nodes. According to the timing relationship and the multiple production nodes, establish the topological association of the preprocessed data according to the process timing to obtain time topological association data; Integrate the spatial topological association data and the time topological association data to obtain spatio-temporal topological association data.

6. The intelligent fault diagnosis method for the bovine bone meal production line according to claim 1, characterized in that Input the standard fusion data into the cloud digital twin to dynamically simulate the operation state of the production line and obtain the virtual-real mapped production line health state simulation data, including: Based on physical rules, establish the first cloud digital twin of the bone meal production line; Construct a bone meal production fault tree according to the historical fault data of the bone meal production line; Train the fault response logic of the first cloud digital twin according to the bone meal production fault tree to obtain the second cloud digital twin; Input the standard fusion data into the second cloud digital twin to dynamically simulate the operation state of the production line and obtain the virtual-real mapped production line health state simulation data.

7. The intelligent fault diagnosis method for the bovine bone meal production line according to claim 6, wherein Construct a bone meal production fault tree according to the historical fault data of the bone meal production line, including: Determine the set of key fault modes according to the historical fault data of the bone meal production line; Extract the first fault mode from the set of key fault modes, and the first fault mode is any one in the set of key fault modes; For the first fault mode, analyze and extract the equipment-level fault causes from the historical equipment log data to obtain an equipment fault set, and construct a first-level branch based on the equipment fault set; Based on the equipment fault set, associate the historical process parameter data to obtain a process anomaly set, and construct a second-level branch based on the process anomaly set; Analyze the historical interaction record data of the equipment fault set and the process anomaly set to obtain a composite fault set, and construct a third-level branch based on the composite fault set; Input the first-level branch, the second-level branch, and the third-level branch into the Bayesian network, obtain the occurrence probability and propagation weight of each fault cause node, and construct a dynamically optimized bone meal production fault tree based on the occurrence probability and the propagation weight.

8. The intelligent fault diagnosis method for the bovine bone meal production line according to claim 1, characterized in that, Analyze the production line health state simulation data through a hierarchical diagnosis engine to obtain a diagnosis result, including: Construct a first diagnostic engine based on a preset threshold matching rule; Model and predict the performance degradation trend of the equipment on the bovine bone meal production line, and construct a second diagnostic engine; Mine association rules for multi-device collaborative faults, and construct a third diagnostic engine; Construct a fourth diagnostic engine through causal reasoning based on the fault causal relationship; Connect the first diagnostic engine, the second diagnostic engine, the third diagnostic engine, and the fourth diagnostic engine in parallel to obtain a hierarchical diagnostic engine; Analyze the production line health status simulation data through the hierarchical diagnostic engine to obtain a diagnostic result.

9. The intelligent fault diagnosis method for the bovine bone meal production line according to claim 8, characterized in that Model and predict the performance degradation trend of the equipment on the bovine bone meal production line, and construct a second diagnostic engine, including: Use the LSTM network to learn the temporal degradation characteristics of vibration signals and acoustic signals, and establish a tool wear prediction model for the crusher; Fuse the tool wear prediction model and the tool physical model to predict the remaining tool life and obtain the predicted remaining tool life; Set an adaptive tool life warning threshold based on the production load, and construct a second diagnostic engine according to the adaptive tool life warning threshold and the predicted remaining tool life.

10. Intelligent fault diagnosis system for beef bone powder production line, characterized in that The system is used to implement the intelligent fault diagnosis method for the bovine bone meal production line according to any one of claims 1 to 9, and the system includes: A real-time monitoring data set acquisition module for collecting the equipment operation data and process parameters of the bovine bone meal production line in real time through multi-modal sensors to obtain a multi-source heterogeneous real-time monitoring data set; A standard fusion data acquisition module for transmitting the real-time monitoring data set to an edge computing node for preprocessing to obtain standard fusion data; A production line dynamic simulation module for inputting the standard fusion data into a cloud digital twin to dynamically simulate the operation state of the production line and obtain production line health status simulation data with virtual-real mapping; A diagnostic result acquisition module for analyzing the production line health status simulation data through a hierarchical diagnostic engine to obtain a diagnostic result; A dynamic response module for triggering a dynamic response according to the diagnostic result and performing closed-loop intelligent operation and maintenance of the bovine bone meal production line.

Images