A real-time fault detection method, device, equipment and medium

Abstract

The present disclosure provides a method, apparatus, device and medium for real-time fault detection, the method comprising: obtaining the operating status data of at least one steel plant equipment and performing standardization processing to generate standardized data, dividing the standardized data into standardized data of four operating stages according to the operating stage: raw material preparation, smelting processing, forming processing, and shutdown; constructing four data queues on a cache device, calculating the data flow fluctuation coefficient based on the standardized data cache inflow rate of the four data queues and the remaining capacity of the cache device, and adjusting the cache strategy of the four data queues based on the data flow fluctuation coefficient; inputting the standardized data of the four data queues into the corresponding equipment fault judgment model for processing to obtain fault data, the equipment fault judgment model comprising a first rule layer and a second rule layer. The effect of real-time fault detection is improved by performing caching in stages and dual judgment of static thresholds and dynamic thresholds.

Description

Technical Field

[0001] The present disclosure relates to the technical field of fault detection, and in particular to a method, apparatus, device and medium for real-time fault detection. Background Art

[0002] In modern steelmaking, the proper operation of equipment is crucial for ensuring production efficiency and product quality. Traditional equipment maintenance methods often rely on periodic inspections and manual judgment, making it difficult to predict and address equipment failures in a timely manner, leading to increased equipment downtime and rising maintenance costs.

[0003] With the development of big data applications, the collection, transmission, and control of data content are becoming increasingly important for more intelligent big data processing. However, existing technologies are unable to process real-time data due to the large number of equipment types and the large amount of data generated. Furthermore, equipment failure types are complex. Traditional processing systems rely solely on single parameter thresholds for judgment, resulting in poor fault prediction results and an inability to accurately and efficiently detect abnormal data in the real-time data of steel mill equipment. Summary of the Invention

[0004] The present disclosure proposes a real-time fault detection method, device, equipment and medium to solve the problems in the existing technology such as the inability to perform real-time data processing on steel plant equipment and the inability to provide accurate and efficient early warning.

[0005] In a first aspect, the present disclosure provides a real-time fault detection method, the method comprising:

[0006] Obtaining operating status data of at least one steel mill equipment, standardizing the operating status data to generate standardized data with aligned timestamps, and dividing the standardized data into standardized data of four operating stages: raw material preparation, smelting processing, forming processing, and shutdown;

[0007] Establishing four data queues on the cache device for receiving standardized data of the four operating stages, calculating data flow fluctuation coefficients of the four data queues according to the standardized data cache inflow rates of the four data queues and the remaining capacity of the cache device, and adjusting the cache strategies of the four data queues according to the data flow fluctuation coefficients;

[0008] Inputting the standardized data of the four data queues into corresponding equipment fault judgment models for processing to obtain fault data; wherein different data queues correspond to different equipment fault judgment models, and the equipment fault judgment models include a first rule layer and a second rule layer, the first rule layer is used to perform static threshold judgment on the standardized data, and the second rule layer is used to perform dynamic threshold judgment on the standardized data;

[0009] Send early warning information of steel plant equipment failure based on the fault data.

[0010] According to a real-time fault detection method provided by the present disclosure, the standardized data of the four data queues are respectively input into the corresponding equipment fault judgment model for processing to obtain fault data, specifically including: for any data queue, obtaining multi-source data corresponding to the equipment model from the timestamp-aligned standardized data according to the equipment model; inputting the multi-source data into the first rule layer for static judgment of the parameter threshold, and outputting a first fault detection result, the first fault detection result including: unlabeled multi-source data and / or first-level labeled fault data; inputting the multi-source data into the second rule layer for dynamic judgment of the parameter threshold, and outputting a second fault detection result, the second fault detection result including: unlabeled multi-source data and / or second-level labeled fault data; determining the fault data corresponding to the multi-source data according to the first fault detection result and the second fault detection result, the fault data including fault type and label information.

[0011] According to a real-time fault detection method provided by the present disclosure, the first rule layer includes static judgment rules in the form of multi-level key values ​​constructed according to equipment type, equipment model, fault type, and parameter combination; wherein, the determination of the threshold value of each parameter in the parameter combination of the first rule layer includes at least one of the following: determining the first static threshold value of the parameter according to the basic threshold value set in the equipment manufacturing specification and the years of use of the equipment; or determining the second static threshold value of the parameter according to the historical fault data of the equipment and the health of the equipment; or determining the third static threshold value based on the first static threshold value and the second static threshold value.

[0012] According to a real-time fault detection method provided by the present disclosure, the second rule layer includes dynamic judgment rules in the form of multi-level key values ​​constructed according to the device type, device model, fault type, and parameter combination. The determination of the threshold value of each parameter in the parameter combination of the second rule layer includes: obtaining the threshold value of each parameter in the parameter combination determined by the first rule layer; calculating the coupling between each parameter in the parameter combination based on the historical fault data of the device to obtain a parameter association weight matrix; and calculating the dynamic threshold value of each parameter based on the parameter association matrix and the threshold value of each parameter.

[0013] According to a real-time fault detection method provided by the present disclosure, the data flow fluctuation coefficients of the four data queues are respectively calculated based on the standardized data cache inflow rate of the four data queues and the remaining capacity of the cache device, including: using a flow monitoring sensor at the data transmission interface of the cache device to monitor the cache inflow rate of the standardized data of the four data queues; using a capacity monitor integrated inside the cache device to monitor the remaining capacity of the cache device, the remaining capacity of the cache device includes the remaining capacity of each data queue; and calculating the data flow fluctuation coefficient of each data queue based on the cache inflow rate and remaining capacity of the standardized data of each data queue.

[0014] According to a real-time fault detection method provided by the present disclosure, the method of calculating the data flow fluctuation coefficient of each data queue based on the cache inflow rate and remaining capacity of the standardized data of each data queue includes: calculating the data flow fluctuation coefficient of each data queue according to a first formula, wherein the first formula is:

[0015]

[0016] Among them, the data inflow rate standard deviation is the standard deviation of the data inflow rate within the preset time period, the data inflow rate average is the average inflow rate of the data inflow rate within the preset time period, α is the capacity weight coefficient, "1-remaining capacity percentage" is the cache capacity adjustment factor, and the remaining capacity percentage is the ratio between the remaining capacity of each data queue and the allocated capacity.

[0017] According to a real-time fault detection method provided by the present disclosure, the caching strategy of the four data queues is adjusted according to the data flow fluctuation coefficient, including: classifying the data cache inflow of each data queue into a stable state, a violent fluctuation state, and an extreme fluctuation state according to the data flow fluctuation coefficient and the fluctuation threshold of each data queue; when the data cache inflow of any data queue is in an extreme fluctuation state, temporarily storing the standardized data to be cached in the data queue to a cloud cache device using a high-speed data transmission link; when the data cache inflow of any data queue is in a violent fluctuation state and there is at least one data queue among the four data queues with a stable data cache inflow, increasing the allocation amount of the data queue with a violent fluctuation state while reducing the allocation amount of the data queue with a stable data cache inflow; when the data cache inflow of any data queue is in a violent fluctuation state and there is no data queue among the four data queues with a stable data cache inflow, temporarily storing the standardized data to be cached in the data queue to a cloud cache device using a high-speed data transmission link; when the data cache inflow of any data queue is in a stable state, clearing data that has not been accessed for a long time and sorting the storage fragments of the local cache device.

[0018] In a second aspect, the present disclosure further provides a real-time fault detection device, comprising: a data acquisition module for acquiring operating status data of at least one steel plant equipment, and performing standardization processing on the operating status data to generate standardized data with aligned timestamps, wherein the standardized data is divided into standardized data of four operating stages according to the operating stage: raw material preparation, smelting processing, forming processing, and shutdown;

[0019] a data caching module, configured to construct four data queues on a cache device for receiving standardized data of the four operating stages, calculate data flow fluctuation coefficients of the four data queues according to standardized data cache inflow rates of the four data queues and remaining capacity of the cache device, and adjust cache strategies of the four data queues according to the data flow fluctuation coefficients;

[0020] a processing module, configured to input the standardized data of the four data queues into corresponding equipment fault judgment models for processing to obtain fault data; wherein different data queues correspond to different equipment fault judgment models, and the equipment fault judgment models include a first rule layer and a second rule layer, wherein the first rule layer is configured to perform static threshold judgment on the standardized data, and the second rule layer is configured to perform dynamic threshold judgment on the standardized data;

[0021] The early warning module is used to send early warning information of steel plant equipment failure based on the fault data.

[0022] In a third aspect, the present disclosure further provides an electronic device, comprising: a processor and a memory, wherein the memory stores machine-readable instructions executable by the processor, and the processor is used to read the machine-readable instructions from the memory and execute the machine-readable instructions to implement the real-time fault detection method provided in an embodiment of the present disclosure.

[0023] In a fourth aspect, the present disclosure further provides a computer-readable storage medium storing a computer program, wherein the computer program is used to execute the real-time fault detection method provided in an embodiment of the present disclosure.

[0024] In summary, the present disclosure provides a real-time fault detection method, apparatus, equipment and medium, which improve the data quality of the operation status data by standardizing the real-time operation status data and aligning the timestamps; effectively improves the effect of equipment fault detection by dividing the standardized data into operation stages and setting different fault judgment models for different operation stages; ensures the stability of real-time data processing by adaptively caching the data of each data queue on the cache device; realizes real-time, accurate and efficient detection of various equipment faults by dual judgment of static thresholds and dynamic thresholds on standardized data of different operation stages, and at the same time realizes early warning of faults by dynamic threshold judgment, and effectively improves the fault detection effect by giving early warning before the fault occurs. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the present disclosure or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0026] Figure 1 This is a flow chart of a real-time fault detection method provided by the present disclosure;

[0027] Figure 2 This is a flowchart of a processing method of a fault judgment model provided by the present disclosure;

[0028] Figure 3 This is a structural diagram of a real-time fault detection device provided by the present disclosure;

[0029] Figure 4 It is a structural diagram of an electronic device provided by the present disclosure. DETAILED DESCRIPTION

[0030] To make the objectives, technical solutions, and advantages of this disclosure more clear, the technical solutions of this disclosure will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only part of the embodiments of this disclosure, not all of them. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments of this disclosure without creative effort shall fall within the scope of protection of this disclosure.

[0031] Figure 1 This is a flow chart of a real-time fault detection method provided by the present disclosure. Figure 1 , the method comprising:

[0032] Step S11, obtain the operating status data of at least one steel plant equipment, and standardize the operating status data to generate standardized data with time stamp alignment, and divide the standardized data into standardized data of four operating stages according to the operating stage: raw material preparation, smelting processing, forming processing, and shutdown.

[0033] Specifically, it is understood that the steel mill equipment refers to equipment used in the steel production process, such as ironmaking equipment such as blast furnaces and hot blast furnaces, steelmaking equipment such as converters and electric furnaces, and rolling mills and other steelmaking equipment. The operating status data refers to equipment operating data collected in real time from various steel mill equipment. To ensure the safe and stable operation of steel mill equipment and the smooth progress of the production process, it is necessary to obtain equipment operating status data in real time so as to provide timely warnings of equipment failures and take appropriate measures. The real-time operating status data collected includes, but is not limited to, temperature, vibration, pressure, current, sound, and equipment operating parameters.

[0034] Among them, temperature is a key factor affecting the performance of steel mill equipment. Since steel mill equipment is in a high-temperature environment for a long time, abnormal temperature may indicate wear of equipment parts. For example, high temperature will aggravate the thermal expansion of bearings and even cause abnormal vibration of the room temperature of the shaft system. Therefore, real-time monitoring of equipment temperature changes can provide timely early warning of equipment failures.

[0035] Vibration is an important basis for equipment fault diagnosis. For example, real-time monitoring of vibration data of key parts such as bearings and gears of steel plant equipment can timely detect abnormal vibration in key parts and provide timely warnings.

[0036] Pressure stability is an important basis for the safe operation of equipment. Overpressure may cause explosion, and underpressure may cause the equipment to stop working and thus damage the equipment. At the same time, pressure can also reflect whether the equipment's seal has failed and whether there are problems such as leakage in the equipment.

[0037] The current is closely related to the load condition of the equipment. Abnormal current may indicate a short circuit or overload fault in the equipment. The sound can also reflect the operating status of the equipment to a certain extent. When the equipment is operating normally, the sound is relatively stable. When abnormal noise occurs, such as impact sound, friction sound, etc., it may indicate that internal parts are loose or faulty.

[0038] In some embodiments, data collection of parameters such as temperature, vibration, pressure, current, and sound of steel mill equipment requires the use of sensors. In this case, data collection can be performed using a distributed sensor array, that is, deploying one or more of temperature sensors, vibration sensors, pressure sensors, vibration sensors, and current sensors at key points on each piece of equipment within the steel mill environment. For example, temperature sensors can be installed on the walls of the blast furnace, the furnace body of the converter, and the mold of the continuous caster to monitor temperature changes in the equipment. Pressure sensors can be installed in the cooling system of the blast furnace and the cooling system of the continuous caster to monitor cooling water pressure. Pressure sensors can be installed in the oxygen lance system of the converter to monitor oxygen pressure. Vibration sensors can be installed at the taphole and slag outlet of the blast furnace, the tilting mechanism of the converter, and the straightening machine of the continuous caster to monitor the vibration status of the equipment.

[0039] Specifically, obtaining the operating status data of at least one steel plant equipment includes: obtaining the operating status data of at least one steel plant equipment collected in real time by a distributed sensor array.

[0040] In other embodiments, some steel mill equipment has built-in monitoring instruments, and some parameters of the steel mill equipment, such as temperature and vibration, can be collected in real time through the built-in monitoring instruments.

[0041] Specifically, obtaining the operating status data of at least one steel plant equipment further includes: obtaining the operating status data of the steel plant equipment collected in real time by a built-in monitoring instrument of at least one steel plant equipment.

[0042] It should be noted that the sensors deployed in the steel mill can also include other types of sensors, such as acceleration sensors and velocity sensors. The specific sensors deployed can be adjusted according to the actual data collection needs, and this disclosure does not impose any special restrictions on this. In addition, dual sensors of the same type can be deployed at key points in some equipment to avoid single points of failure.

[0043] Specifically, it is also understandable that due to the numerous and complex equipment in steel mills, these devices continuously generate a large amount of data during operation, including equipment operating parameters, process parameters, and real-time monitoring data collected by sensors. To promptly predict and address equipment failures, it is often necessary to monitor multi-source data about at least one device, collected in real time by multiple sensors. The parameters collected by different sensors may have different formats. Furthermore, the complex production environment of steel mills is subject to interference factors such as high temperature, dust, and vibration, and the equipment data may also contain a certain amount of noise. Therefore, after obtaining the real-time operating status data, it is necessary to perform standardization processing such as data cleaning and format conversion.

[0044] In some embodiments, the operating status data is standardized to generate standardized data with timestamp alignment, including: performing data cleaning on the operating status data to obtain denoised data; performing format conversion on the denoised data to obtain standard format data; and generating standardized data with timestamp alignment based on the acquisition time of the standard format data.

[0045] Among them, the data cleaning can be one of the commonly used denoising methods in the field of big data processing, and the embodiments of the present disclosure are not limited to this. The noise in the operating status data collected in real time is removed by data cleaning. The format conversion refers to the integration and normalization of the denoised data after cleaning to generate a unified data format to ensure that the same type of data maintains a unified data format. The acquisition time refers to the time node when each sensor acquires the corresponding parameters after the various parameters generated by the steel plant equipment are sequentially generated. Since the operating status data of the steel plant equipment has obvious time series characteristics, the standard format data is further time-aligned to facilitate the subsequent prediction and processing of equipment failures.

[0046] Specifically, it is also understandable that, because the equipment has different response sensitivities to parameters at different stages, in order to more accurately detect equipment failures, the real-time collected operating status data is further divided into operating status data of multiple operating stages according to the equipment process flow.

[0047] In some embodiments, the operating status data collected in real time is further divided into operating status data of multiple operating stages according to the equipment process flow. The standardized data with time stamp alignment after standardization can be divided into standardized data of four operating stages: raw material preparation, smelting processing, forming processing, and shutdown. Quality monitoring or other operating stages can also be added according to the actual operating environment of the steel plant. The operating stages can also be further subdivided according to the actual operating environment of the steel plant. The embodiments of the present disclosure do not make special limitations on this.

[0048] The operation phase of raw material preparation can be pre-processing operations such as crushing, screening, and mixing of raw materials, such as crushing large pieces of raw materials into suitable particle sizes with a crusher and screening the crushed materials with a vibrating screen; it can also be the startup and preheating phase of the equipment before operation. Real-time fault detection during the operation phase of raw material preparation can be detection of crusher motor failure, crusher overload, crusher bearing damage, loosening of crusher anchor bolts, loosening of vibrating screen box components, and damage to vibrating screen mesh; it can also be for equipment in the smelting or forming processing phase, by detecting parameters such as temperature changes, lubrication, and sealing to determine whether the equipment has failed during the startup and preheating phase, such as thermal stress cracks in the blast furnace hearth and furnace wall, cracks in the converter lining, etc.

[0049] The operational phase of smelting refers to the conversion of raw materials into molten steel through processes such as high-temperature smelting. For example, a blast furnace reduces iron ore to molten iron, a hot blast furnace provides hot air for the blast furnace, and a converter or electric furnace further refines the molten iron. Real-time fault detection during the operational phase of raw material preparation can include detection of blast furnace hearth burn-through, blast furnace hot blast furnace failure, converter tilt mechanism failure, and converter lining erosion.

[0050] The operational phase of the forming process involves processing the smelted molten steel through various forming equipment into steel products of varying shapes and specifications to meet the needs of various users. Common forming equipment includes continuous casting machines and rolling mills. The operational phase of the forming process can involve the continuous casting machine solidifying the molten steel into ingots, or the rolling mill rolling the ingots into various steel products such as plates, pipes, and wire. Real-time fault detection during the operational phase of the forming process can include detecting faults in the continuous casting machine's mold, wear on the continuous casting machine's segment rollers, faults in the rolling mill's rolls, and faults in the rolling mill's motor.

[0051] The shutdown operation phase refers to when the equipment is shut down. Since the status of each component will gradually return to its initial state when the equipment is shut down, the changes in data during this process can also reflect some performance and potential problems of the equipment. For real-time fault detection during the shutdown operation phase, for example, the temperature of key parts such as the tuyere of the blast furnace and the lining of the converter should drop according to a certain pattern when the equipment is shut down, which may indicate an equipment failure. For another example, when the rolling mill is shut down, the vibration sensor detects that the vibration amplitude always remains at a high level and is accompanied by abnormal low-frequency vibration. After detection, it was found that the roller bearing was damaged, resulting in unstable rotation of the roller and causing vibration.

[0052] In the above method, the data quality, data accuracy and reliability of the collected operating status data are improved by standardizing and aligning the timestamps of the operating status data collected in real time; the standardized data are divided into operating stages according to the process flow of the steel plant equipment to achieve group processing of the standardized data. On the one hand, the difficulty of real-time data processing is reduced; on the other hand, since the parameter types and thresholds that need to be referenced for equipment fault detection in different operating stages are different, dividing the standardized data by stages can effectively improve the effect of equipment fault detection.

[0053] Step S12: construct four data queues on the cache device for receiving standardized data of the four operating stages, calculate the data flow fluctuation coefficients of the four data queues according to the standardized data cache inflow rates of the four data queues and the remaining capacity of the cache device, and adjust the cache strategies of the four data queues according to the data flow fluctuation coefficients.

[0054] The cache device is a local cache device that can receive standardized data collected in real time even when the network is disconnected or weak. A connection channel is established between the cache device and each steel plant device and each sensor. The connection channel can be a wired connection channel or a wireless connection channel. The operating status data collected by each steel plant device or each sensor is transmitted to the cache device through the connection channel for caching. This avoids directly sending large amounts of standardized data to and from fault diagnosis, reduces the pressure of data reading and data processing for fault diagnosis, improves response speed, and maintains the stability of data transmission and processing.

[0055] The initial lengths of the four data queues can be the same, and the initial length of each data queue is calculated according to the equal division of the cache device capacity; the initial lengths of the four data queues can also be different, and different weight values ​​are assigned to each data queue according to the operation stage of the standardized data cached by each data queue, and then the initial length of each data queue is calculated according to the cache device capacity and the weight value of each data queue. For example, since the amount of real-time data collected in the shutdown operation stage is relatively small and the amount of real-time data collected in the smelting and forming operation stages is relatively large, when assigning weight values, the weight of the data queue used to cache the standardized data of each operation stage can be in the order of smelting > forming > raw material preparation > shutdown.

[0056] Specifically, it is understandable that the amount of data collected in real time to monitor the operating conditions of each steel plant equipment at different operating stages is closely related to the operating stage of the steel plant equipment. For example, the amount of data in each stage will change accordingly during the initial startup of the blast furnace, the stage of smelting raw materials into molten iron, and the stage of transferring the molten iron. Especially in the molten iron transfer stage, when the molten iron flows out, multiple sensors such as temperature sensors and flow sensors need to work simultaneously at high frequency, and the amount of data may reach nearly 20 times the usual amount. For example, when collecting vibration data of a rolling mill, the sampling frequency may be 100Hz when the rolling mill is in normal operation, while the sampling frequency may reach 10kHz during high-speed rolling. Therefore, in order to better ensure that the cache device can quickly, accurately, and stably cache the real-time collected operating status data of the steel plant equipment, the data flow fluctuation coefficient can be calculated during caching, and the cache strategy of each data queue can be adaptively adjusted based on the data flow fluctuation coefficient.

[0057] In some embodiments, calculating the data flow fluctuation coefficients of the four data queues respectively based on the normalized data cache inflow rates of the four data queues and the remaining capacity of the cache device includes:

[0058] Step S121 : Using a flow monitoring sensor at a data transmission interface of a cache device, a cache inflow rate of standardized data of four data queues is monitored.

[0059] Step S122: Using a capacity monitor integrated in the cache device to monitor the remaining capacity of the cache device, the remaining capacity of the cache device includes the remaining capacity of each data queue.

[0060] Step S123 , calculating the data flow fluctuation coefficient of each data queue according to the buffer inflow rate and remaining capacity of the standardized data of each data queue.

[0061] Specifically, steps S121 and S122 monitor the standardized data cache inflow rate and remaining capacity for each data queue. When constructing the four data queues on the cache device, the initial length of each data queue is specified. Based on the initial length, the allocated capacity of each data queue can be determined. Based on the monitored amount of data actually cached in each data queue, the remaining capacity of each data queue can be calculated. Then, the data flow fluctuation coefficient of each data queue is calculated according to the first formula, which is:

[0062]

[0063] Among them, the standard deviation of the data inflow rate is the standard deviation of the data inflow rate within the preset time period, and the average data inflow rate is the average inflow rate of the data inflow rate within the preset time period. The preset time period is the past few hours, the past day, or the past week. It can be set according to actual conditions, and this disclosure does not limit this.

[0064] α is the capacity weight coefficient, which is used to control the influence of cache device capacity on the overall fluctuation coefficient. The value range of α is 0~1. Generally, the initial value of α can be set to 0.5.

[0065] "1 - Remaining Capacity Percentage" is the cache capacity adjustment factor, which is the ratio of each data queue's remaining capacity to its allocated capacity. When a data queue has a large remaining capacity, the remaining capacity percentage approaches 1, meaning the cache capacity adjustment factor is close to 0. The calculated data traffic fluctuation coefficient is then completely based on fluctuations within the preset time period. When a data queue has a small remaining capacity, an increase in the cache capacity adjustment factor further increases the calculated data traffic fluctuation coefficient, making the coefficient more sensitive to data traffic fluctuations.

[0066] In some embodiments, adjusting the cache strategies of the four data queues according to the data traffic fluctuation coefficient includes:

[0067] Step S124 , classifying the data buffer inflow of each data queue into a stable state, a violent fluctuation state, and an extreme fluctuation state according to the data flow fluctuation coefficient and the fluctuation threshold of each data queue.

[0068] In which, the fluctuation threshold includes at least a first fluctuation threshold and a second fluctuation threshold, and the fluctuation threshold of each data queue is different, that is, each data queue is separately provided with two fluctuation thresholds; for any data queue, when the data flow fluctuation coefficient is less than or equal to the second threshold, the data cache inflow of the data queue is classified as a stable state; when the data flow fluctuation coefficient is greater than the second threshold and less than the first threshold, the data cache inflow of the data queue is classified as a violent fluctuation state; when the data flow fluctuation coefficient is greater than or equal to the first threshold, the data cache inflow of the data queue is classified as an extreme fluctuation state.

[0069] Step S125 , when the data cache inflow of any data queue is in an extremely fluctuating state, the standardized data to be cached in the data queue is temporarily stored in a cloud cache device using a high-speed data transmission link.

[0070] In step S126, when the data cache inflow of any data queue is in a state of violent fluctuation and there is at least one data queue in the four data queues in which the data cache inflow is in a stable state, the allocation amount of the data queue in which the data cache inflow is in a state of violent fluctuation is increased while the allocation amount of the data queue in which the data cache inflow is in a stable state is reduced.

[0071] Specifically, when there is a queue in the four data queues whose data cache inflow is in a state of violent fluctuation, the remaining capacity can be readjusted with the data queue in a stable state. For example, if there is a data queue in a state of violent fluctuation, the capacity of the data queue in the state of violent fluctuation is expanded according to the data queue in a stable state, thereby reducing the cache pressure of the data queue. If there are multiple data queues in a state of violent fluctuation, the capacity expansion is allocated to the multiple data queues in the state of violent fluctuation according to the data queue in a stable state. The specific expansion capacity of each data queue can be allocated according to the data cache inflow fluctuation coefficient of the multiple data queues in the state of violent fluctuation, or according to the remaining capacity of each of the multiple data queues in the state of violent fluctuation. The embodiments of the present disclosure do not limit this.

[0072] Step S127: When the data cache inflow of any data queue is in a state of violent fluctuation and there is no data queue in the four data queues with a stable data cache inflow, the standardized data to be cached in the data queue is temporarily stored in the cloud cache device using a high-speed data transmission link;

[0073] Specifically, when there are drastic fluctuations in the four data queues and there is no data cache flowing into the data queue in a stable state to provide expansion conditions, it is also necessary to temporarily cache it in the cloud to ensure that the received collected data can be cached in real time, thereby ensuring the timeliness of the data.

[0074] Step S128 : When the data cache flow of any data queue is in a stable state, data that has not been accessed for a long time is cleared and storage fragments of the local cache device are sorted.

[0075] Specifically, when the data cache inflow of the data queue is relatively stable, the cache device may be cleared to clear more space for allocation to each data queue.

[0076] In the above method, by caching standardized data of the four operating stages in the cache device, it is ensured that all real-time collected data can be cached in the cache device, ensuring the timeliness of the data obtained during fault diagnosis. At the same time, the access pressure on each sensor is reduced, and real-time data is directly read from the cache device for fault diagnosis, thereby improving the stability of real-time data processing. By combining the cache capacity adjustment factor to calculate the data flow fluctuation coefficient, the volatility of the data cache inflow rate and the ability of the current cache device to withstand fluctuations can be quickly understood, and the data flow fluctuation coefficient of each data queue can be further used to achieve adaptive adjustment of the data queue cache strategy. By adopting cache adjustment strategies such as temporarily storing the cache inflow status in the cloud for extremely volatile conditions, expanding the cache inflow status for drastically fluctuating conditions, and clearing some data when the cache inflow status is relatively stable, the normal reception and processing of data is guaranteed, overload of the local cache device is avoided, cache space is released, and the utilization rate of the cache device is improved, further ensuring the stability of real-time data processing.

[0077] Step S13: inputting the standardized data of the four data queues into corresponding equipment fault judgment models for processing to obtain fault data.

[0078] Among them, different data queues correspond to different equipment fault judgment models and the equipment fault judgment model includes a first rule layer and a second rule layer. The first rule layer is used to perform static threshold judgment on standardized data, and the second rule layer is used to perform dynamic threshold judgment on standardized data.

[0079] Specifically, it can be understood that the parameter judgment rules for whether the equipment has failed are different in different operating stages. Therefore, before processing the standardized data of the cache device, it is necessary to pre-train the corresponding equipment fault judgment models for the four operating stages of raw material preparation, smelting processing, forming processing, and shutdown. That is, different data queues correspond to different equipment fault judgment models.

[0080] Specifically, it is also understood that due to the wide variety of steel mill equipment, there are also many corresponding fault types. Fault detection for steel mill equipment can be performed by comparing the threshold values ​​of a single parameter to detect equipment failures. For example, bearing damage in a crusher requires monitoring whether the vibration amplitude increases significantly or abnormal vibration frequency occurs. Spring breakage in a vibrating screen requires monitoring whether the vibrating screen amplitude is excessive. Mold failure in a continuous casting machine requires monitoring the mold vibration frequency. Fault detection for steel mill equipment can also be performed through combined analysis of multiple parameters. For example, abnormal blast furnace conditions can be determined by combining factors such as increased furnace top temperature and fluctuations in furnace static pressure. Tuyere burn-through in a blast furnace requires monitoring changes in the temperature difference of the tuyere cooling water and fluctuations in furnace top pressure. Lance nodules in a converter require monitoring of lance current and fan blowing time. Whether single-parameter or multi-parameter fault detection is used, both rely on real-time detection data to determine whether a fault has occurred when it occurs, and cannot provide advance prediction of equipment failures. Therefore, in order to quickly and accurately predict equipment failures through real-time collected operating status data, an equipment fault judgment model consisting of a first rule layer and a second rule layer is constructed. On the one hand, at least two parameter thresholds can be set for the fault judgment rules of each fault type to improve the fault detection accuracy; on the other hand, a static threshold for real-time detection and a dynamic threshold for early prediction can be set. While performing real-time fault detection, faults can also be detected in advance, further improving the detection effect.

[0081] In some embodiments, the first rule layer includes static judgment rules in the form of multi-level key values ​​constructed based on device type, device model, fault type, and parameter combination. The threshold value of each parameter in the parameter combination of the first rule layer is a static threshold value, which can be a unilateral threshold value or a bilateral threshold value. The present disclosure provides three methods for the specific determination of the static threshold value. Among them, the determination of the threshold value of each parameter in the parameter combination of the first rule layer includes at least one of the following:

[0082] a1, the first static threshold of the parameter is determined based on the basic threshold set in the equipment manufacturing specification and the age of the equipment; or, a2, the second static threshold of the parameter is determined based on the historical failure data of the equipment and the health of the equipment; or, a3, the third static threshold is determined based on the first static threshold and the second static threshold.

[0083] Specifically, for the first method a1 of determining the threshold of the parameters of the first rule layer, since equipment is more likely to malfunction after being used for too long, it is necessary to determine the depreciation coefficient based on the equipment's service life, and then calculate the product of the corresponding threshold determined according to the fault type given when the equipment leaves the factory and the depreciation coefficient of the equipment as a static threshold, which can perform fault detection more accurately.

[0084] Specifically, for the second method a2 for determining the threshold value of the parameters of the first rule layer, when the frequency of device failure is high, the device is more likely to fail again, while for devices that have never failed, the probability of failure is relatively low. Therefore, it is possible to consider determining the health of the device based on the number of times the device has failed in the past. For example, the health of the device is set to 1 when no failure has occurred, and the health of the device is set to 0.5 when the number of failures is greater than 10. This disclosure does not make any special restrictions on this.

[0085] Specifically, for the third method a3 for determining the threshold of the parameters of the first rule layer, the thresholds determined by the above two methods are combined to comprehensively determine the parameter threshold of the corresponding fault type to enhance the credibility of the threshold. For example, the average value of the first static threshold and the second static threshold is directly calculated as the third static threshold, or the weighted average value of the first static threshold and the second static threshold is calculated as the third static threshold.

[0086] In some embodiments, the second rule layer includes dynamic judgment rules in the form of multi-level key values ​​constructed based on device type, device model, fault type, and parameter combination. The threshold of each parameter in the parameter combination of the second rule layer is a dynamic threshold. The dynamic threshold is of the same type as the static threshold. When the static threshold is a unilateral threshold, the dynamic threshold is also a unilateral threshold but the threshold value is different. When the static threshold is a bilateral threshold, the dynamic threshold is also a bilateral threshold but the upper and lower limits of the bilateral threshold are different. Determining the threshold of each parameter in the parameter combination of the second rule layer includes:

[0087] Step b1: Obtain the threshold value of each parameter in the parameter combination determined by the first rule layer.

[0088] Specifically, when determining the thresholds of the parameters of the second rule layer for the first fault type, the thresholds of the parameters of the first rule layer determined for the first fault type according to at least one of the above methods a1, a2 or a3 are first obtained.

[0089] Step b2: Calculate the coupling between the parameters in the parameter combination based on the historical fault data of the equipment to obtain a parameter association weight matrix.

[0090] Specifically, due to the strong coupling between the operating status data of steel mill equipment, for example, oxygen lance nodules can cause oxygen pressure fluctuations, which in turn affect the reaction in the furnace, which in turn causes furnace temperature changes, which in turn exacerbate nodules. Therefore, for each fault type detection, a comprehensive analysis is performed by selecting multiple parameters to avoid misjudging equipment faults. Moreover, in the period before the fault occurs, the parameters corresponding to the fault may also have the same numerical combination. Therefore, in order to more accurately detect faults, the embodiment of the present disclosure calculates the dynamic threshold value for early warning corresponding to the fault by analyzing the coupling of parameters, which is used to achieve early warning of faults and improve the fault detection effect.

[0091] Assume that the static thresholds of the parameters of the first fault type obtained in step b1 are: the static threshold of the current parameter is [TA1, TA2], and the static threshold of the temperature parameter is [TW1, TW2]. Then, the coupling between the parameters in the parameter combination is calculated based on the historical fault data of the device. That is, the coupling between the current parameter A and the temperature parameter B is calculated based on the historical fault data of the device. For example, the first fault type that occurred is obtained from the historical fault data, and time and space alignment is performed based on the time of the first fault type. The time series data of the current parameter and the temperature parameter that occurred N minutes before the first fault type is extracted and used as a training data set. Then, the initial weight matrix between the current parameter A and the temperature parameter B is trained and corrected using the training data set to obtain the final correlation weight matrix W=[[W_AA, W_AB], [W_BA, W_BB]].

[0092] It should be noted that the number of rows and columns in the parameter association weight matrix varies depending on the parameters. That is, when the parameters corresponding to the first fault type involve current, temperature, and pressure, the association weight matrix ultimately calculated and trained based on historical fault data is a 3*3 matrix. When calculating the coupling between parameters, a machine learning algorithm such as supervised learning or semi-supervised learning can be used to train the parameter association weight matrix, which is not limited in this disclosure.

[0093] Step b3: Calculate the dynamic threshold of each parameter based on the parameter association matrix and the threshold of each parameter.

[0094] Specifically, the parameter correlation matrix is ​​first normalized, and then the upper or lower limit of each parameter threshold is calculated based on the coupling influence between the parameters. Taking the correlation weight matrix W of the first fault type as an example, step b3 specifically includes the following steps:

[0095] Step b31, normalizing the association weight matrix to obtain a normalized matrix w=[[w_aa, w_ab],[w_ba, w_bb]];

[0096] Step b32: Calculate the dynamic lower limit of the current threshold according to the second formula, and calculate the dynamic upper limit of the current threshold according to the third formula. The second formula is:

[0097]

[0098] The third formula is:

[0099]

[0100] Step b33: Calculate the dynamic lower limit of the temperature threshold according to the fourth formula, and calculate the dynamic upper limit of the temperature threshold according to the fifth formula. The fourth formula is:

[0101]

[0102] The fifth formula is:

[0103]

[0104] in, and is the dynamic coefficient, The value range is 0.3~0.5, usually 0.3. The value range is 0.8~1.2, usually 0.8; What is calculated is the coupling effect of temperature parameter B on current parameter A. What is calculated is the coupling influence of current parameter A on temperature parameter B.

[0105] Step b34: determine the dynamic current threshold as [TA_min, TA_max] based on the dynamic lower limit and dynamic upper limit of the current threshold, and determine the dynamic temperature threshold as [TW_min, TW_max] based on the dynamic lower limit and dynamic upper limit of the temperature threshold.

[0106] It should be noted that the above embodiment uses temperature and current as an example to describe steps b1 through b3 in detail. Similarly, the principle of determining dynamic thresholds for two parameters in steps b1 through b3 can also be used to determine dynamic thresholds for two other parameters, or for more than two parameters. Furthermore, the parameter thresholds in the first rule layer and the parameter thresholds in the second rule layer need to be updated in real time based on the timeline and the real-time fault data generated by the device to ensure that subsequent real-time fault detection is highly effective.

[0107] In the above method, the static threshold is determined in combination with the aging degree or health status of the equipment at the first rule layer, thereby improving the real-time detection accuracy of faults; the dynamic threshold is determined according to the correlation strength of the associated parameters in the historical fault data at the second rule layer, thereby achieving early warning of equipment failures; by setting the first rule layer and the second rule layer to perform dual judgments on the static threshold and the dynamic threshold, early warning of equipment failures can be achieved while detecting the faulty equipment in real time, thereby improving the fault detection effect.

[0108] Step S14: sending early warning information of steel plant equipment failure based on the failure data.

[0109] Specifically, it can be understood that the fault data includes the fault detection results of the first rule layer and the fault detection results of the second rule layer. Since the first rule layer is used to perform static threshold judgment on standardized data and the second rule layer is used to perform dynamic threshold judgment on standardized data, that is, the fault detection results of the first rule layer are used to indicate whether the equipment has failed, and the fault detection results of the second rule layer are used to indicate whether the equipment is about to fail. Therefore, when the warning information occurs, different warning information needs to be sent for specific detection results to display the fault data to the staff more comprehensively and accurately, thereby improving the warning effect.

[0110] In some embodiments, the fault data includes at least one of primary-labeled fault data and secondary-labeled fault data, where the primary-labeled fault data is the fault detection result of the first rule layer, and the secondary-labeled fault data is the fault detection result of the second rule layer; the warning information includes first warning information and second warning information, where the first warning information is used to indicate that a certain type of fault has occurred in a certain device, and the second warning information is used to indicate that a certain type of fault is about to occur in a certain device. In this case, step S14, "sending warning information of steel plant equipment failure based on the fault data," at least includes:

[0111] Step S141: If the fault data is level 1 marked fault data, a first warning message is sent;

[0112] Step S142: If the fault data is fault data with a secondary mark, a second warning message is sent;

[0113] Step S143: If the fault data is fault data with a first-level mark and the fault data is also fault data with a second-level mark, a first warning message and a second warning message are sent simultaneously.

[0114] Specifically, it can also be understood that the fault data that can be determined in step S13 include various types of fault data determined by the first rule layer and various types of fault data determined by the second rule layer. That is, the fault data finally obtained includes fault data determined according to two methods. It is possible that a fault is not detected by the first rule layer, but it is detected through the second rule layer that the equipment is about to have a certain fault, which provides great convenience for equipment maintenance and greatly improves the fault detection results.

[0115] In some embodiments, when there are multiple fault types at the same time, the priority of sending the warning information can be set according to the fault type. For example, the fault type is divided into three types: high-level, intermediate and primary faults according to the degree of impact on the equipment; when sending warning information of steel plant equipment failure, the warning information corresponding to the high-level fault type is sent first, and is sent directly to the on-site operation screen to display it to the staff in the fastest and most intuitive way. It can also be sent to the electronic devices of the associated staff; for warnings of intermediate faults and primary faults, it can be sent only to the electronic devices of the associated staff, or different warning methods can be used for the equipment. This disclosure does not limit this.

[0116] In this method, by sending different warning messages based on the type of fault data being marked, warnings of existing or impending faults can be issued more accurately, improving warning effectiveness. By sending warning messages based on the priority of the fault type, high-priority "high-level faults" are prioritized and sent to the staff's equipment or other related devices, allowing staff to prioritize these faults and perform rapid fault resolution or equipment maintenance, thus enriching the fault warning method.

[0117] It should be noted that after sending the warning information of the steel plant equipment failure according to the fault data, the real-time fault detection method further includes:

[0118] Step S15 , generating index data according to the first fault detection result and the second fault detection result, wherein the index data is unlabeled multi-source data, and generating a device status report according to the index data and the fault data.

[0119] Specifically, the equipment status report includes information such as equipment health status, fault prediction results, maintenance records, etc. The equipment status report can be stored in the form of charts or dashboards, so that users can clearly and intuitively view the equipment's operating status and fault trends.

[0120] In the above embodiment, the data quality of the operation status data is improved by standardizing and aligning the timestamps of the real-time operation status data; the effect of equipment fault detection is effectively improved by dividing the standardized data into operation stages and setting different fault judgment models for different operation stages; the stability of real-time data processing is guaranteed by adaptively caching the data of each data queue on the cache device; real-time, accurate and efficient detection of various equipment faults is achieved by dual judgment of static thresholds and dynamic thresholds on standardized data of different operation stages, and early warning of faults is also achieved through dynamic threshold judgment, which can be used to warn the fault in advance before it occurs, effectively improving the fault detection effect.

[0121] Figure 2 This is a flow chart of a processing method for a fault judgment model provided by the present disclosure. The fault judgment model includes a first rule layer and a second rule layer. The first rule layer is used to perform static threshold judgment on standardized data, and the second rule layer is used to perform dynamic threshold judgment on standardized data. The processing method of the fault judgment model refers to a method for processing the standardized data of the four data queues, that is, see Figure 2 Step S13 "inputting the standardized data of the four data queues into the corresponding equipment fault judgment model for processing to obtain fault data" specifically includes:

[0122] Step S21 : for any data queue, obtain multi-source data corresponding to the device model from the standardized data aligned with the timestamp according to the device model.

[0123] Among them, the multi-source data corresponding to the device model can be multi-source data at the current moment, or within the current period of time, such as 30s, 1min, 5min; the multi-source data refers to the obtained data including at least multiple types of parameters such as temperature, current, vibration, pressure, etc. about the operating status of the same device.

[0124] Step S22: input the multi-source data into the first rule layer to perform static judgment of parameter thresholds, and output a first fault detection result.

[0125] The first fault data includes: unmarked multi-source data and / or first-level marked fault data. The first-level marked fault data indicates that the abnormal data has been marked with a fault and a first-level mark, the fault mark is the corresponding fault type, and the first-level mark is the corresponding fault determination method at the first rule layer.

[0126] Specifically, it can be understood that when the first rule layer is used to judge multi-source data, when the parameters in the multi-source data do not meet the parameter threshold set by the first rule layer, the fault judgment result "first fault data" is unmarked multi-source data; when the parameters in the multi-source data all meet the parameter threshold set by the first rule layer, the fault judgment result "first fault data" is first-level marked fault data.

[0127] Specifically, it can also be understood that step S22 specifically includes the following steps: according to the device model, reading the parameter combination corresponding to various fault types under the device model from the first rule layer, and determining the static threshold of each parameter in the parameter combination; performing fault judgment based on the multi-source data corresponding to the device model and the static threshold of each parameter, and generating a first-level fault mark.

[0128] In some embodiments, the multi-source data corresponding to the equipment model are parameters such as temperature, vibration, pressure, etc. of a certain model of rolling mill. At this time, it is necessary to read the parameters and thresholds that need to be monitored for each fault type under the equipment model from the first rule layer, and then judge between the data values ​​of multiple parameters and the parameter thresholds based on the data type to determine whether the rolling mill has a fault. In the specific judgment, multi-threaded concurrent technology can be used to simultaneously process the judgment between the data values ​​of multiple parameters and the parameter thresholds to improve the processing speed. Other technologies can also be used for processing. The present disclosure does not make any special restrictions on this.

[0129] Step S23: input the multi-source data into the second rule layer to perform dynamic judgment of parameter thresholds, and output a second fault detection result.

[0130] The second fault detection result includes: unlabeled multi-source data and / or secondary labeled fault data. The secondary labeled fault data indicates that the abnormal data has been fault-labeled and secondary-labeled, where the fault label is the corresponding fault type and the secondary label is the corresponding fault determination method, which is the second rule layer.

[0131] Specifically, it can be understood that when the second rule layer is used to judge multi-source data, when the parameters in the multi-source data do not meet the parameter threshold set by the second rule layer, the fault judgment result "second fault data" is unmarked multi-source data; when the parameters in the multi-source data all meet the parameter threshold set by the second rule layer, the fault judgment result "second fault data" is secondary marked fault data.

[0132] In some embodiments, step S23 specifically includes the following steps: according to the device model, reading the parameter combination corresponding to various fault types under the device model from the second rule layer, and determining the dynamic threshold of each parameter in the parameter combination; performing fault judgment based on the multi-source data corresponding to the device model and the dynamic threshold of each parameter, and generating a secondary fault mark.

[0133] Specifically, it can also be understood that the first rule layer and the second rule layer are two independent layers. The multi-source data can be input into the first rule layer and the second rule layer at the same time for static threshold judgment and dynamic threshold judgment, or the multi-source data can be input into the first rule layer and the second rule layer successively for static threshold judgment and dynamic threshold judgment. The present disclosure does not make any special limitations on this.

[0134] It should be noted that in the process of "performing fault judgment based on the multi-source data corresponding to the device model and the static thresholds of each parameter to generate a first-level fault mark" or "performing fault judgment based on the multi-source data corresponding to the device model and the dynamic thresholds of each parameter", the number of decision rules can be used for matching judgment, or other commonly used methods can be used for matching judgment, and the present disclosure does not limit this.

[0135] Step S24 : determining fault data corresponding to the multi-source data according to the first fault detection result and the second fault detection result.

[0136] Specifically, the fault data includes fault type and tag information. The tag information refers to whether a detected fault type is determined according to the first rule layer or the second rule layer. The tag information corresponding to the fault type determined according to the first rule layer is a first-level tag, and the tag information corresponding to the fault type determined according to the second rule layer is a second-level tag.

[0137] In some embodiments, a third rule layer can also be used to replace the first rule layer and the second rule layer, that is, a device fault judgment model including a third rule layer is constructed for each data queue. The third rule layer is a multi-threshold judgment rule in a multi-level key value form based on the device type, device model, fault type, and parameter combination. The multiple thresholds of each parameter in the parameter combination of the third rule layer include: static thresholds and dynamic thresholds.

[0138] The static thresholds are the thresholds for each parameter in the parameter combination of the first rule layer, and the dynamic thresholds are the thresholds for each parameter in the parameter combination of the second rule layer. At this point, step S13, "inputting the standardized data from the four data queues into the corresponding device fault judgment models for processing to obtain fault data," specifically includes: for any data queue, obtaining multi-source data corresponding to the device model from the timestamp-aligned standardized data based on the device model; inputting the multi-source data into the third rule layer for multi-threshold judgment of parameter thresholds, and outputting a third fault detection result, which includes unlabeled multi-source data and / or first-level labeled fault data and / or second-level labeled fault data.

[0139] In the above embodiment, by applying static threshold parameter judgment to standardized data at different operating stages, real-time, accurate, and efficient detection of various equipment fault types is achieved. Furthermore, dynamic threshold judgment also enables early warning of faults, effectively improving fault detection effectiveness by providing early warning before they occur.

[0140] Figure 3 This is a structural diagram of a real-time fault detection device provided by the present disclosure. Figure 3 The real-time fault detection device 300 includes: a data acquisition module 310, a data cache module 320, a processing module 330 and an early warning module 340.

[0141] The data acquisition module 310 is configured to acquire operating status data of at least one steel mill equipment, perform standardization processing on the operating status data to generate standardized data with aligned timestamps, and classify the standardized data into standardized data for four operating stages: raw material preparation, smelting processing, forming processing, and shutdown.

[0142] a data caching module 320 configured to construct four data queues on a cache device for receiving standardized data in the four operating stages, calculate data flow fluctuation coefficients of the four data queues based on the standardized data cache inflow rates of the four data queues and the remaining capacity of the cache device, and adjust cache policies of the four data queues based on the data flow fluctuation coefficients;

[0143] a processing module 330 configured to input the standardized data of the four data queues into corresponding equipment fault judgment models for processing to obtain fault data; wherein different data queues correspond to different equipment fault judgment models, and the equipment fault judgment models include a first rule layer and a second rule layer, wherein the first rule layer is configured to perform static threshold judgment on the standardized data, and the second rule layer is configured to perform dynamic threshold judgment on the standardized data;

[0144] The early warning module 340 is used to send early warning information of steel plant equipment failure based on the fault data.

[0145] In some embodiments, the real-time fault detection apparatus further includes a reporting module configured to generate indicator data based on the first fault detection result and the second fault detection result, and to generate an equipment status report based on the indicator data and the fault data. The indicator data is unlabeled multi-source data, and the equipment status report includes information such as equipment health status, fault prediction results, and maintenance records. The equipment status report can be stored in the form of a chart or dashboard, allowing users to clearly and intuitively view the equipment's operating status and fault trends.

[0146] For a detailed description of the above-mentioned real-time fault detection device, please refer to the description of the relevant detection method in the above-mentioned embodiment, and the repeated parts will not be repeated. The embodiments of the real-time fault detection method and the real-time fault detection device described above are merely illustrative, and the "modules" used as separate components can be a combination of software and / or hardware that implements a predetermined function, and may or may not be physically separated. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without any creative work.

[0147] Figure 4 This is a structural diagram of an electronic device provided in an embodiment of the present disclosure. The electronic device 400 includes: a processor 410 and a memory 420. The processor 410 and the memory 420 are connected via a bus 430. The memory 420 stores machine-readable instructions executable by the processor 410. The processor 410 reads the machine-readable instructions from the memory 420 and executes the machine-readable instructions to implement the real-time fault detection method provided in an embodiment of the present disclosure.

[0148] It should be understood that the electronic device may be a personal computer (PC), a tablet computer, a smart phone, or other electronic device with logic computing capabilities.

[0149] The embodiment of the present disclosure further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program is used to execute the real-time fault detection method provided by the embodiment of the present disclosure.

[0150] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than to limit them. Although the present disclosure has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present disclosure, and should be covered by the protection scope of the present invention.

Claims

1. A real-time fault detection method, characterized in that: The method comprises: Obtaining operating status data of at least one steel mill equipment, standardizing the operating status data to generate standardized data with aligned timestamps, and dividing the standardized data into standardized data of four operating stages: raw material preparation, smelting processing, forming processing, and shutdown; Establishing four data queues on the cache device for receiving standardized data of the four operating stages, calculating data flow fluctuation coefficients of the four data queues according to the standardized data cache inflow rates of the four data queues and the remaining capacity of the cache device, and adjusting the cache strategies of the four data queues according to the data flow fluctuation coefficients; Inputting the standardized data of the four data queues into corresponding equipment fault judgment models for processing to obtain fault data; wherein different data queues correspond to different equipment fault judgment models, and the equipment fault judgment models include a first rule layer and a second rule layer, the first rule layer is used to perform static threshold judgment on the standardized data, and the second rule layer is used to perform dynamic threshold judgment on the standardized data; Sending early warning information of steel plant equipment failure based on the fault data; The dynamic threshold and the static threshold are of the same type. When the static threshold is a unilateral threshold, the dynamic threshold is a unilateral threshold but the threshold value is different. When the static threshold is a bilateral threshold, the dynamic threshold is a bilateral threshold but the upper and lower limits of the bilateral threshold are different. The standardized data of the four data queues are respectively input into the corresponding equipment fault judgment model for processing to obtain fault data, specifically including: For any data queue, obtain multi-source data corresponding to the device model from the timestamp-aligned standardized data according to the device model; Inputting the multi-source data into the first rule layer for static judgment of parameter thresholds, and outputting a first fault detection result, the first fault detection result including: unlabeled multi-source data and / or first-level labeled fault data; Inputting the multi-source data into the second rule layer for dynamic determination of parameter thresholds, and outputting a second fault detection result, the second fault detection result including: unlabeled multi-source data and / or secondary labeled fault data; Determine fault data corresponding to the multi-source data according to the first fault detection result and the second fault detection result, wherein the fault data includes a fault type and tag information; Determining the threshold of each parameter in the parameter combination of the second rule layer includes: Obtaining the threshold of each parameter in the parameter combination determined by the first rule layer; Calculate the coupling between parameters in the parameter combination based on the historical fault data of the equipment and obtain the parameter association weight matrix; The dynamic threshold of each parameter is calculated according to the parameter association matrix and the threshold of each parameter.

2. The real-time fault detection method according to claim 1, wherein: The first rule layer includes static judgment rules in a multi-level key-value format constructed according to device type, device model, fault type, and parameter combination, wherein the determination of the threshold value of each parameter in the parameter combination of the first rule layer includes at least one of the following: Determine the first static threshold of the parameter based on the basic threshold set in the equipment manufacturing specifications and the age of the equipment; or Determine a second static threshold value of the parameter based on the historical fault data of the device and the health of the device; or, A third static threshold is determined according to the first static threshold and the second static threshold.

3. The real-time fault detection method according to claim 2, wherein: The second rule layer includes dynamic judgment rules in the form of multi-level key values ​​constructed according to device type, device model, fault type, and parameter combination.

4. The real-time fault detection method according to claim 1, wherein: Calculating the data flow fluctuation coefficients of the four data queues respectively according to the standardized data cache inflow rates of the four data queues and the remaining capacity of the cache device includes: A flow monitoring sensor at the data transmission interface of the cache device is used to monitor the cache inflow rate of the standardized data of the four data queues; Using a capacity monitor integrated within the cache device to monitor the remaining capacity of the cache device, the remaining capacity of the cache device includes the remaining capacity of each data queue; The data flow fluctuation coefficient of each data queue is calculated based on the cache inflow rate and remaining capacity of the standardized data of each data queue.

5. The real-time fault detection method according to claim 4, characterized in that: The data flow fluctuation coefficient of each data queue is calculated based on the cache inflow rate and remaining capacity of the standardized data of each data queue, including: The data flow fluctuation coefficient of each data queue is calculated according to the first formula, which is: Among them, the data inflow rate standard deviation is the standard deviation of the data inflow rate within the preset time period, the data inflow rate average is the average inflow rate of the data inflow rate within the preset time period, α is the capacity weight coefficient, "1-remaining capacity percentage" is the cache capacity adjustment factor, and the remaining capacity percentage is the ratio between the remaining capacity of each data queue and the allocated capacity.

6. The real-time fault detection method according to claim 4, characterized in that: The adjusting the cache strategies of the four data queues according to the data traffic fluctuation coefficient includes: According to the data flow fluctuation coefficient and fluctuation threshold of each data queue, the data cache inflow of each data queue is divided into stable state, violent fluctuation state, and extreme fluctuation state; When the data cache flow of any data queue is in an extremely fluctuating state, the standardized data to be cached in the data queue is temporarily stored in the cloud cache device using a high-speed data transmission link; When the data cache inflow of any data queue is in a state of violent fluctuation and there is at least one data queue in a state of stable data cache inflow among the four data queues, the allocation amount of the data queue in a state of violent fluctuation is increased while the allocation amount of the data queue in a state of stable data cache inflow is reduced; When the data cache inflow of any data queue is in a state of violent fluctuation and there is no data queue with a stable data cache inflow among the four data queues, the standardized data to be cached in the data queue is temporarily stored in the cloud cache device using a high-speed data transmission link; When the data cache inflow of any data queue is in a stable state, data that has not been accessed for a long time is cleared and storage fragments of the local cache device are defragmented.

7. A real-time fault detection device, characterized in that: The device comprises: A data acquisition module is configured to acquire operating status data of at least one steel mill equipment, perform standardization processing on the operating status data to generate standardized data with aligned timestamps, and divide the standardized data into standardized data of four operating stages: raw material preparation, smelting processing, forming processing, and shutdown; a data caching module, configured to construct four data queues on a cache device for receiving standardized data of the four operating stages, calculate data flow fluctuation coefficients of the four data queues according to standardized data cache inflow rates of the four data queues and remaining capacity of the cache device, and adjust cache strategies of the four data queues according to the data flow fluctuation coefficients; a processing module, configured to input the standardized data of the four data queues into corresponding equipment fault judgment models for processing to obtain fault data; wherein different data queues correspond to different equipment fault judgment models, and the equipment fault judgment models include a first rule layer and a second rule layer, wherein the first rule layer is configured to perform static threshold judgment on the standardized data, and the second rule layer is configured to perform dynamic threshold judgment on the standardized data; An early warning module, configured to send early warning information of steel plant equipment failure based on the fault data; The dynamic threshold and the static threshold are of the same type. When the static threshold is a unilateral threshold, the dynamic threshold is a unilateral threshold but the threshold value is different. When the static threshold is a bilateral threshold, the dynamic threshold is a bilateral threshold but the upper and lower limits of the bilateral threshold are different. The standardized data of the four data queues are respectively input into the corresponding equipment fault judgment model for processing to obtain fault data, specifically including: For any data queue, obtain multi-source data corresponding to the device model from the timestamp-aligned standardized data according to the device model; Inputting the multi-source data into the first rule layer for static judgment of parameter thresholds, and outputting a first fault detection result, the first fault detection result including: unlabeled multi-source data and / or first-level labeled fault data; Inputting the multi-source data into the second rule layer for dynamic determination of parameter thresholds, and outputting a second fault detection result, the second fault detection result including: unlabeled multi-source data and / or secondary labeled fault data; Determine fault data corresponding to the multi-source data according to the first fault detection result and the second fault detection result, wherein the fault data includes a fault type and tag information; Determining the threshold of each parameter in the parameter combination of the second rule layer includes: Obtaining the threshold of each parameter in the parameter combination determined by the first rule layer; Calculate the coupling between parameters in the parameter combination based on the historical fault data of the equipment and obtain the parameter association weight matrix; The dynamic threshold of each parameter is calculated according to the parameter association matrix and the threshold of each parameter.

8. An electronic device, characterized in that: The electronic device includes: a processor and a memory, wherein the memory stores machine-readable instructions executable by the processor, and the processor is used to read the machine-readable instructions from the memory and execute the machine-readable instructions to implement the real-time fault detection method described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and the computer program is used to execute the real-time fault detection method according to any one of claims 1 to 6.

Images