An internet of things multi-source data edge computing processing system and method

Abstract

The application discloses an Internet of Things multi-source data edge computing processing system and method in the technical field of Internet of Things data processing, and relates to the technical field of edge computing.The system comprises a centralized production scheduling unit that continuously monitors and analyzes production plans, identifies processing tasks to be started in a future preset time period, and extracts material batch identifiers corresponding to the processing tasks.Based on the material batch identifiers, simplified key background information matching the processing tasks is extracted from a global background information database.The simplified key background information is the minimum data set required for edge side anomaly detection.The application simultaneously solves the problem of misjudgment caused by the lack of global background information on the edge side and the problem of communication delay caused by real-time request information, significantly improves detection accuracy on the premise of ensuring millisecond-level response capability, and avoids redundant data transmission and storage overhead.

Description

Technical Field

[0001] This invention relates to an edge computing processing system and method for multi-source data in the Internet of Things (IoT), belonging to the field of IoT data processing technology. Background Technology

[0002] In modern industrial production, IoT technology is widely used to improve efficiency and equipment reliability. In industrial IoT scenarios, edge computing gateways need to perform millisecond-level anomaly detection (such as tool failure early warning) on ​​multi-source sensor data, including machine tool vibration and current, to ensure equipment safety and production continuity.

[0003] However, existing technologies have the following core contradictions: On the one hand, the edge side lacks global background information (such as the batch characteristics of the currently processed materials), which can easily misjudge normal fluctuations caused by differences in material characteristics as abnormalities, leading to unnecessary downtime; on the other hand, if the edge side requests background information from the central system in real time during detection, the communication delay will exceed the millisecond-level response requirement, resulting in the inability to promptly block sudden failures; if the central system distributes all redundant background data, it will cause a waste of network bandwidth and a burden on edge storage, and there is still a risk of information lag. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide an IoT multi-source data edge computing processing system and method. It solves the problem of misjudgment caused by the lack of global background information on the edge side, as well as the communication delay caused by real-time information requests. It significantly improves detection accuracy while ensuring millisecond-level response capability and avoids redundant data transmission and storage overhead.

[0005] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0006] In a first aspect, the present invention provides a method for edge computing processing of multi-source data in the Internet of Things, comprising:

[0007] The production plan is continuously monitored and analyzed by the centralized production scheduling unit to identify the processing tasks that will be started within a preset time period in the future and extract the material batch identifiers corresponding to the processing tasks.

[0008] Based on the material batch identifier, simplified key background information matching the processing task is extracted from the global background information database. The simplified key background information is the minimum dataset required for edge-side anomaly detection.

[0009] At a preset time point before the processing task starts, the simplified key background information is proactively pushed to the corresponding edge computing gateway through a low-latency communication channel.

[0010] The simplified key background information is received and temporarily stored through the edge computing gateway. When the start of the processing task is detected, the matching local judgment logic calibration parameters are loaded.

[0011] Based on the calibration parameters, the local anomaly detection threshold is dynamically adjusted to perform millisecond-level anomaly judgment on the real-time acquired multi-source sensor data, and a control command is triggered when an anomaly is determined.

[0012] Furthermore, the simplified key background information includes the average physical property conversion parameters at the material batch level; the extraction of simplified key background information matching the processing task from the global background information database specifically includes: querying the material property database corresponding to the material batch identifier, obtaining at least one attribute value among average hardness and toughness, converting the attribute value into a threshold correction coefficient for the edge-side anomaly detection algorithm, and removing redundant detailed quality inspection reports and full material data.

[0013] Furthermore, it also includes a pilot cutting micro-feature prediction step:

[0014] After the machining task is started but before the main cutting begins, the edge computing gateway sends a pilot cutting command to the machine tool control system, controlling the tool to perform a trial cut at a preset distance with low feed and shallow cutting depth;

[0015] Simultaneously acquire vibration and spindle current data during the pilot cutting process, and extract instantaneous fluctuation characteristics that reflect the local microscopic properties of the material;

[0016] A local correction factor is generated based on the instantaneous fluctuation characteristics, and combined with the batch-level calibration parameters, a dynamic anomaly detection threshold is obtained.

[0017] Furthermore, the local microscopic properties include at least one of localized hard spots and stress concentration regions within the material; the instantaneous fluctuation characteristics are extracted in the following manner:

[0018] Calculate the instantaneous amplitude envelope of the vibration signal and the instantaneous power spectral density of the current signal;

[0019] Identify energy abrupt changes, amplitude spikes, duration and rate of change in specific high-frequency bands, and match them with preset material inhomogeneity feature templates.

[0020] Furthermore, it also includes a flexible correction step for micro-feedback cutting parameters:

[0021] During the main cutting process, the transient spikes of vibration and current signals are monitored in real time. If early signs that conform to the local characteristic fluctuation mode of the material are detected, the temporary cutting parameter fine-tuning amount is calculated.

[0022] The fine-tuning command is sent to the machine tool control system via the industrial bus, and the feed rate and spindle speed are temporarily adjusted at the current cutting point, and the original parameters are automatically restored after passing through a local uneven area.

[0023] The anomaly detection threshold is dynamically updated based on real-time cutting parameters to distinguish between material fluctuations and actual tool failures.

[0024] Furthermore, the temporary cutting parameter fine-tuning amount is a minor correction within ±2% of the feed rate reduction and spindle speed adjustment; the dynamic update adopts the following formula:

[0025] V_threshold_dynamic = V_base + K_H·H_avg + K_F·F_actual + K_S·S_actual

[0026] Where V_threshold_dynamic is the dynamic vibration threshold, V_base is the base threshold, H_avg is the batch average hardness, F_actual is the actual feed rate of the current machine tool, S_actual is the spindle speed, and K_H, K_F, and K_S are correction coefficients predetermined based on material and process characteristics.

[0027] Furthermore, the low-latency communication channel adopts at least one of industrial Ethernet, 5G industrial private network, and MQTT protocol; the total latency of the edge computing gateway's anomaly detection and instruction triggering is controlled within 5 milliseconds.

[0028] Secondly, the present invention provides an IoT multi-source data edge computing processing system for implementing the IoT multi-source data edge computing processing method described above, comprising:

[0029] Central scheduling module: Used to continuously monitor and parse production plans through a centralized production scheduling unit, identify processing tasks that will be started within a preset time period in the future, and extract the material batch identifiers corresponding to the processing tasks;

[0030] Information extraction module: used to extract concise key background information matching the processing task from the global background information library based on the material batch identifier. The concise key background information is the minimum dataset required for edge-side anomaly detection.

[0031] Information push module: used to proactively push the simplified key background information to the corresponding edge computing gateway through a low-latency communication channel at a preset time point before the start of the processing task;

[0032] Information receiving module: used to receive and temporarily store the simplified key background information through the edge computing gateway, and load matching local judgment logic calibration parameters when the processing task is detected to be started;

[0033] Anomaly detection module: used to dynamically adjust the local anomaly detection threshold based on the calibration parameters, perform millisecond-level anomaly detection on real-time multi-source sensor data, and trigger control commands when an anomaly is detected.

[0034] Thirdly, the present invention provides an IoT multi-source data edge computing processing device, including a processor and a storage medium;

[0035] The storage medium is used to store instructions;

[0036] The processor is configured to operate according to the instructions to perform the steps of the method according to any of the foregoing.

[0037] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0038] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0039] I. This solution employs a "production scheduling-driven predictive background information broadcasting" mechanism. A centralized production scheduling unit pre-analyzes future processing tasks, extracts the minimum dataset (simplified key background information) required for edge-side anomaly detection, and proactively pushes it to the edge computing gateway via a low-latency channel before task initiation. The edge side dynamically adjusts its local anomaly detection threshold based on this information, achieving millisecond-level anomaly judgment and command triggering. This solution simultaneously addresses the misjudgment problem caused by the lack of global background information on the edge side, as well as the communication latency issue caused by real-time information requests. It significantly improves detection accuracy while ensuring millisecond-level response capabilities, avoiding redundant data transmission and storage overhead.

[0040] II. Based on the pilot cutting micro-feature prediction mechanism, this solution effectively overcomes the deficiency of insufficient granularity of batch-level average background information. By performing trial cutting with low feed and shallow depth of cut before formal machining, the edge computing gateway can simultaneously collect vibration and current data, extract instantaneous fluctuation features reflecting micro-characteristics such as local hard points and internal stress concentrations within the material, and generate local correction factors accordingly. This mechanism enables the anomaly detection threshold on the edge side to no longer be limited to the static average parameters issued by the central system, but can dynamically adapt to the actual micro-characteristics of the current workpiece, accurately distinguishing between normal fluctuations caused by material inhomogeneity and actual tool failures, fundamentally solving the problem of continuous misjudgment caused by micro-characteristics, and further improving the detection reliability under complex working conditions.

[0041] Third, by introducing a flexible correction mechanism for micro-feedback cutting parameters and a multi-dimensional dynamic threshold adjustment mechanism, this solution achieves high-precision detection while also considering machining continuity and process adaptability. When early signs of local material property fluctuations are detected, the edge computing gateway can instantly calculate and execute minute corrections such as a 1% to 5% reduction in feed rate and a ±2% adjustment in spindle speed, helping the tool smoothly pass through uneven areas and avoiding unnecessary downtime caused by sudden spikes in vibration. Simultaneously, the abnormal detection threshold is dynamically updated based on real-time cutting parameters and batch average hardness, ensuring accurate differentiation between material fluctuations and actual faults even as process parameters change. Furthermore, the application of a low-latency communication channel strictly controls the total latency of edge-side judgment and command triggering to within 5 milliseconds, meeting the extreme requirements of real-time response in precision machining scenarios. Attached Figure Description

[0042] The accompanying drawings, which form part of this specification, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0043] Figure 1 This is a flowchart illustrating an IoT multi-source data edge computing processing method provided in Embodiment 1 of the present invention. Detailed Implementation

[0044] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0045] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this invention is for describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0046] Example 1:

[0047] Please see Figure 1 This embodiment proposes an edge computing processing method for multi-source data in the Internet of Things (IoT), aiming to solve the problem of misjudgment caused by the lack of global background information when edge computing units analyze local data at high speed and in real time in industrial production scenarios, while avoiding large-scale data transmission and high communication latency. This solution achieves intelligent information collaboration between the central system and edge units through a "production scheduling-driven predictive background information broadcasting" mechanism. The specific method includes the following steps:

[0048] 1. Central Production Planning Perception and Analysis:

[0049] Centralized production scheduling and background information management units, such as a factory's Manufacturing Execution System (MES), continuously monitor and parse production plans and work order scheduling information across the entire shop floor. This MES system can be configured with a scheduled task, for example, running every minute, to scan its internal production work order database. This database typically contains a "production task table," which records upcoming processing tasks, including the task's unique identifier, the target machine number, the type of part to be processed, and the batch identifier of the materials to be used. By reading these records, the MES system identifies new tasks that will begin within a certain timeframe (e.g., within the next 30 minutes).

[0050] 2. Extraction of predictive background information:

[0051] Based on the future production task information parsed in step 1, the centralized production scheduling and background information management unit intelligently extracts concise key background information closely related to these upcoming tasks from its stored global background information. For example, if a task will use a specific batch of steel, the MES system will query the batch's "material property database" for key attributes such as hardness and toughness based on the batch's unique identifier. Then, it converts these attributes into calibration parameters required by the edge computing gateway's local decision-making logic, such as a value representing a material hardness correction, instead of sending out the entire material database or detailed quality inspection reports. This extraction process ensures that the amount of information transmitted is minimized, containing only key data that directly impacts edge decision-making.

[0052] 3. Proactive background information push:

[0053] Before a processing task is predicted to begin (e.g., a pre-set time before the scheduled start time, such as 5 minutes), the centralized production scheduling and background information management unit proactively pushes the concise key background information extracted in step 2 to the corresponding edge computing gateway responsible for the task through a pre-set, low-latency communication channel. This communication channel can use industrial Ethernet (e.g., Profinet or EtherCAT) or a highly reliable wireless network (e.g., 5G industrial private network) to ensure timely data transmission. The push method can be point-to-point, where the MES system directly sends information to the target edge gateway; or it can be a broadcast to a group of related edge gateways if multiple gateways will perform similar tasks. The push data packet structure can be a lightweight JSON or Protobuf format, containing a task identifier, material batch identifier, and corresponding calibration parameters.

[0054] 4. Temporary storage and loading of local background information at the edge:

[0055] After receiving these predictive background information pushes, the edge computing gateway stores them in a local background information buffer. This buffer can be a memory area or a temporary file on a high-speed solid-state drive. When the edge gateway detects that a connected machine tool has actually started executing a new task (e.g., by listening to a task start signal from the machine tool control system, which might be a Modbus register state change or an OPC UA variable update), it immediately searches for and loads the background information matching the current task from the local buffer. If no matching information is found in the buffer, or if the information has expired, the edge gateway can trigger an emergency request to request the latest background information from the central system.

[0056] 5. Dynamic adjustment of local decision logic:

[0057] The judgment logic adjustment module inside the edge computing gateway adjusts its local anomaly detection judgment logic or threshold in real time based on the current task background information loaded in step 4. For example, for tool vibration anomaly detection, if the received calibration parameters indicate that the hardness of the currently processed material is H_current, the upper limit threshold V_max of the normal amplitude of the vibration signal will be adjusted to V_base + k × H_current, where V_base is the base threshold and k is a correction coefficient predetermined based on the material characteristics. This adjustment process is achieved by modifying a configuration variable or parameter table in the real-time data processing program running within the gateway, without the need for redeployment or program recompilation. In this way, the edge gateway can adaptively adjust its judgment criteria for "normal" and "abnormal" according to the characteristics of the actual processed material, avoiding misjudging normal cutting vibration caused by increased material hardness as tool anomaly.

[0058] 6. Real-time anomaly detection and decision-making:

[0059] The adjusted local decision-making logic continues to analyze high-frequency sensor data (such as vibration and current) in real time. The edge gateway continuously collects data from connected vibration sensors, for example, at a frequency of thousands of times per second. This raw data, after preprocessing (such as filtering and feature extraction), is compared with the dynamically adjusted thresholds from step 5. Once an abnormal signal exceeding the adjusted threshold is detected, the edge gateway immediately sends a command to the machine tool control system, such as sending an emergency stop signal via an industrial bus (e.g., EtherCAT), to suspend machining. This decision-making and command transmission process is completed within milliseconds, enabling timely response to sudden failures and avoiding significant losses, such as severe tool damage or workpiece scrap.

[0060] The working principle of this solution is similar to that of an intelligent scheduler in a factory. It not only knows the arrangement of all production tasks but also prepares the necessary "reference materials" for each workstation in advance. In a precision machining workshop, machine tools process different parts using batches of different materials. The subtle differences in the properties of these materials (such as hardness) directly affect the vibration performance of the cutting tool. If the edge computing gateway does not know the specific characteristics of the material being processed, it may misjudge vibrations within the normal range as tool malfunctions, leading to unnecessary downtime.

[0061] Traditionally, either the edge gateway only queries the central system when it detects an "anomaly," but this back-and-forth communication delay is too slow for sudden failures like microcracks in cutting tools that require millisecond-level response; by the time the information returns, damage may have already occurred. Alternatively, the central system sends all possible material information to each edge gateway, but this creates a huge network burden and storage pressure on the edge gateways, and most of the information is redundant.

[0062] This approach differs. It leverages the explicitness and predictability of shop floor production plans. The centralized Manufacturing Execution System (MES) acts like an intelligent scheduler, continuously monitoring the production schedule and knowing which machine tool will process which parts and what materials. For example, the MES system detects that a machine tool will switch from processing a batch of aluminum alloy with hardness H_A to processing a batch of steel with hardness H_B in 5 minutes. The MES system immediately extracts the key characteristics of H_B steel from its material database and converts them into a lightweight calibration parameter package. Before the task switch, the MES system proactively pushes this calibration parameter package to the edge computing gateway corresponding to that machine tool.

[0063] When the machine tool actually begins machining H_B steel, the edge gateway immediately loads this calibration parameter from its local temporary storage. Its local anomaly detection program dynamically adjusts the threshold for judging vibration signals based on this parameter. For example, if H_B steel is harder than H_A aluminum alloy, the vibration amplitude will naturally increase during normal cutting. The edge gateway will correspondingly increase the threshold for judging vibration anomalies to avoid misjudging normal vibrations caused by increased material hardness as tool malfunctions. Simultaneously, if a microcrack in the tool does occur, generating abnormal vibrations exceeding the adjusted threshold, the edge gateway can still issue a stop command within milliseconds to prevent losses.

[0064] This approach cleverly resolves the contradiction between "real-time" and "global awareness." It doesn't acquire information only when needed at the edge; instead, the information is already in place before the edge requires it. It's like equipping the edge gateway with a "predictive" assistant, always providing the most accurate "cheat sheets" in advance at critical moments, enabling it to make more informed decisions while simultaneously making rapid judgments. This solution is particularly suitable for industrial scenarios with clearly defined production plans, frequent material batch changes, and extremely high requirements for real-time response and judgment accuracy, such as monitoring tool status in precision machining workshops. Scenarios with fixed production tasks and relatively stable material characteristics may not require such sophisticated predictive information distribution; simple fixed thresholds or periodic manual updates to model parameters may suffice.

[0065] In a precision machining workshop, a CNC machine tool is high-speed machining a batch of aero-engine blades. This machine tool is connected to an edge computing gateway, which runs a real-time tool condition monitoring program. This program analyzes data from vibration sensors on the spindle to determine if the tool has experienced abnormal wear or microcracks.

[0066] Without this solution: The machine tool first processes a batch of aluminum alloy blades with batch number "AL-202301". The edge gateway's tool status detection program is set with a fixed vibration amplitude threshold, for example, V_threshold_fixed = 100 units. When the vibration signal exceeds this threshold, the gateway will issue a tool abnormality alarm and pause the machine tool.

[0067] Next, the production plan requires the machine tool to switch to machining a batch of titanium alloy blades with batch number "TI-202302". Titanium alloys are much harder than aluminum alloys. When the machine tool starts machining titanium alloys, even if the tool is in perfect working order, the increased material hardness will naturally lead to a significant increase in the vibration amplitude during cutting, for example, reaching 120 units. At this point, the edge gateway's detection program will immediately issue a tool malfunction alarm and force the machine tool to stop because the vibration amplitude (120) exceeds a fixed threshold (100).

[0068] Upon arrival, the on-site engineers inspected the cutting tools and found them intact. Further investigation revealed that the false alarm was caused by a material change leading to an increase in normal vibration amplitude, while the edge gateway lacked awareness of the current material characteristics. This false alarm not only disrupted production and delayed the delivery of an urgent order but also wasted the engineers' valuable time.

[0069] After adopting this solution:

[0070] 1. Central Production Planning Awareness and Analysis: The factory's Manufacturing Execution System (MES) continuously monitors the production plan. By analyzing its internal "Production Work Order Sheet," the MES system discovers that machine tool M001 will complete the processing of batch "AL-202301" aluminum alloy at 10:00 AM and begin processing batch "TI-202302" titanium alloy at 10:05 AM.

[0071] 2. Predictive Background Information Extraction: Based on the batch number "TI-202302", the MES system retrieves key properties such as hardness and toughness of the titanium alloy from its "Material Property Database". It converts these properties into lightweight calibration parameters, such as a "hardness correction factor" F_titanium = 1.2. This factor indicates that the normal vibration amplitude will be 20% higher when machining titanium alloys compared to machining aluminum alloys.

[0072] 3. Proactive background information push: At 10:00 AM, 5 minutes before machine tool M001 is about to switch tasks, the MES system proactively pushes a simplified data packet containing the task identifier, material batch number, and hardness correction factor F_titanium = 1.2 to the edge computing gateway corresponding to machine tool M001 via industrial Ethernet.

[0073] 4. Edge Local Background Information Storage and Loading: After receiving this data packet, the edge gateway stores it in its local memory buffer. When the control system of machine tool M001 issues a "task start" signal at 10:05 (e.g., by updating a status variable via the OPC UA protocol), the edge gateway immediately loads the background information matching the current task from the buffer, i.e., F_titanium = 1.2.

[0074] 5. Dynamic Adjustment of Local Judgment Logic: The tool status detection program inside the edge gateway adjusts its judgment logic adjustment module in real time based on the loaded hardness correction factor to determine the vibration anomaly threshold. Assuming a base threshold V_base = 100 units (applicable to aluminum alloys), the new threshold V_threshold_titanium = V_base × F_titanium = 100 × 1.2 = 120 units. This adjustment is achieved by modifying a configuration parameter within the program.

[0075] 6. Real-time Anomaly Detection and Decision-Making: Machine tool M001 begins machining titanium alloy. Due to the high hardness of the material, the normal cutting vibration amplitude may reach 115 units. At this time, the edge gateway's detection program compares it with the adjusted threshold of 120 units. Because 115 is less than 120, the program determines that the tool condition is normal and does not issue an alarm. Production proceeds smoothly.

[0076] If a microcrack does occur in the tool at this point, causing the vibration amplitude to suddenly spike to 150 units, the edge gateway will immediately detect that 150 exceeds the threshold of 120 and issue a stop command to the machine tool control system within milliseconds. This avoids false alarms caused by changes in material properties while ensuring timely response to real, sudden failures, effectively protecting the equipment and workpiece, and significantly improving production efficiency and reliability.

[0077] This solution cleverly resolves the inherent contradiction between edge computing units' pursuit of high-speed real-time response and the central system's global awareness in industrial production. Traditional approaches often involve edge units passively requesting background information from the central system when needed, or the central system distributing all possible information. However, in scenarios like precision machining workshops where millisecond-level tool status monitoring is required, the former's communication latency is unacceptable, while the latter would create a huge network burden and waste edge unit resources, and there remains a risk of information lag in real-time assessment of sudden faults.

[0078] This solution changes the traditional information flow model by introducing a "production scheduling-driven predictive background information broadcasting" mechanism. Instead of simply pushing background information from the central system to the edge, or requiring the edge to request it every time it's needed, it endows the centralized Manufacturing Execution System (MES) with a "predictive" capability. The MES system no longer passively waits for requests from edge units, but instead, through deep insights into the production plan, anticipates which edge units will face changes in background information (e.g., machine tools are about to switch material batches).

[0079] This "proactive" information distribution strategy enables the central system to intelligently extract concise key background information closely related to future tasks (e.g., hardness correction parameters for specific material batches) and proactively push it to the corresponding edge computing gateways before the task actually begins. This "time-shifting" strategy allows edge units to immediately obtain and apply the latest background information when switching tasks, ensuring that information is available "just in time" when actually needed, achieving highly accurate judgments with near-zero real-time communication latency.

[0080] This intelligent scheduling of time resources and information flow is key to its non-obviousness. It transforms a seemingly irreconcilable contradiction into an efficient and elegant solution through a deep dive into the "predictability" of industrial production planning. It avoids the loss of real-time performance due to waiting for background information, and also avoids misjudgments caused by missing background information, thus significantly improving the wisdom and reliability of decision-making while ensuring millisecond-level response capabilities.

[0081] This solution significantly improves the accuracy of anomaly detection by edge computing units in industrial production scenarios, effectively avoiding misjudgments caused by a lack of global background information. By predictively acquiring key background information in advance, the local judgment logic of the edge unit can dynamically and adaptively adjust, thereby maintaining millisecond-level real-time response capabilities while greatly reducing unnecessary production interruptions, equipment inspections, and wasted human resources.

[0082] In modern industrial production, the Internet of Things (IoT) technology is widely used to improve efficiency and equipment reliability. However, in industrial IoT scenarios, massive amounts of data generated by devices from different sources and in various formats suffer from problems such as inconsistent formats, high transmission latency, and difficulties in real-time analysis. Traditional centralized data processing models, while having advantages in some aspects, often lead to information asymmetry and struggle to meet the real-time response capabilities required for equipment status prediction and process optimization in smart manufacturing. Especially in environments with extremely high real-time requirements, such as precision manufacturing, effectively processing this data and ensuring the accuracy and timeliness of decision-making has become a pressing challenge.

[0083] For example, in a large precision machining workshop, hundreds of CNC machine tools are deployed to produce metal parts. To achieve intelligent management of the production process and proactive maintenance of the equipment, various types of sensors are installed on key components of each machine tool, such as the spindle, feed motor, and tool turret. The data generated by these sensors differ in format, unit, sampling frequency, and communication protocol.

[0084] To address the network transmission latency and data processing queuing delays inherent in traditional centralized processing models when dealing with sudden failures, the factory introduced an edge computing solution. An edge computing gateway was deployed next to each processing unit in the workshop. This gateway directly connects to the sensors of several machine tools within that unit, handling high-frequency vibration, current, and other data locally in real time, enabling rapid anomaly identification and timely response.

[0085] This solution proposes a centralized Manufacturing Execution System (MES) that extracts concise background information related to the upcoming machining tasks in advance based on the production plan, and proactively distributes it to the corresponding edge computing gateways before the task begins. In modern precision manufacturing workshops, to ensure machining accuracy and equipment reliability, the edge computing gateways connected to the machine tools can perform real-time analysis of data collected by the spindle vibration sensors to assess the tool status and obtain advance knowledge of the characteristics of the current machining materials, thereby laying the foundation for adjusting their local tool status judgment logic.

[0086] However, in actual operation, we observed a more nuanced scenario. When the MES system recognizes that the machine tool is about to switch from its current task to a new machining task—for example, from processing a specific batch of aluminum alloy to processing a different batch of titanium alloy with higher hardness—it initiates a pre-processing information flow. Based on the batch identification information of the titanium alloy to be used, the MES system queries and extracts key background information related to that batch of material from its maintained material property database, such as its hardness correction parameters. This extracted information is encapsulated into a concise data structure. Subsequently, at a pre-set time before the machine tool actually begins executing the titanium alloy machining task, the MES system proactively distributes this concise background information to the edge computing gateway responsible for that machine tool via a pre-established communication channel. This ensures that the edge gateway, upon receiving the new task initiation signal, has prior knowledge of the characteristics of the material being processed, thus laying the foundation for adjusting its local tool status judgment logic.

[0087] Although the edge gateway has adjusted the threshold for detecting abnormal tool vibration based on the average hardness correction parameters distributed by the MES, new batches of titanium alloy materials, due to microscopic local inhomogeneities generated during their upstream smelting or forging processes, result in tiny areas within the same batch or even the same workpiece where the hardness or internal stress distribution deviates significantly from the average. These microscopic differences are far smaller than the batch-level average attributes typically recorded by the MES system and are difficult to fully identify before machining using conventional rapid quality inspection methods. When the tool is machining subsequent workpieces, if its cutting edge accidentally and precisely contacts a localized high-hardness area within the material, even if the tool itself is intact, this sudden change in material properties will cause a rapid increase in cutting force, leading to a spindle vibration signal amplitude that significantly exceeds the normal threshold adjusted by the edge gateway. At this point, the edge gateway's anomaly detection program will immediately respond, determining it as a tool malfunction and quickly issuing a stop command. Maintenance personnel arrive at the scene after receiving the alarm and inspect the tool, only to find no signs of damage. Such false alarms not only disrupted production of urgent orders and wasted human resources, but also, more fundamentally, revealed that while the peripheral system correctly applied the information provided by the central system, this information lacked sufficient granularity and precision to describe "unexpected" changes at the microscopic level of materials, failing to support completely accurate judgments at the peripheral end. These "normal anomalies" caused by the microscopic properties of materials may repeatedly occur before the entire batch of materials is exhausted, continuously disrupting production and limiting the effectiveness of advance information distribution.

[0088] In industrial production scenarios where a central system distributes batch-level background information in advance according to the production plan, and the edge computing unit dynamically adjusts the local judgment threshold, how can the edge computing unit quickly and accurately identify and distinguish between instantaneous vibration signal changes caused by such material characteristic fluctuations and vibration anomalies caused by actual tool failures when faced with unpredictable microscopic local characteristic inhomogeneities within the processed materials? This would effectively avoid continuous misjudgments caused by insufficient granularity of existing background information, while also avoiding additional communication delays or excessive computational burden on the edge.

[0089] This solution aims to enhance the accuracy of anomaly detection in industrial production scenarios, particularly after the central system provides batch-level average information. It addresses the challenge of effectively identifying and distinguishing between transient vibration anomalies caused by the inhomogeneity of microscopic local characteristics within the processed material and actual tool failures. This solution utilizes a "pilot cutting micro-feature prediction" mechanism to achieve proactive perception and precise calibration of the edge-end's logic for judging the microscopic characteristics of the current workpiece material. The specific method includes the following steps:

[0090] 1. Central system task information distribution and edge preloading:

[0091] The centralized Manufacturing Execution System (MES) continuously monitors the production plan and identifies upcoming processing tasks. Before a task begins, the MES system extracts batch-level average background information (e.g., material average hardness correction parameters) from its material property database based on task information (e.g., material batch number) and proactively distributes it to the edge computing gateway responsible for that task. Upon receiving the information, the edge gateway temporarily stores it locally and loads it when the task starts to initially adjust its local decision-making logic.

[0092] Implementation: The MES system can be configured with a scheduled service, for example, scanning the production work order database every minute to identify tasks that will begin within the next 30 minutes. For each identified task, the MES queries its material property database (e.g., an SQL database table) to obtain the average hardness value H_avg of the batch of materials. The MES then encapsulates H_avg into a lightweight JSON data packet and pushes it to a specific topic on the target edge gateway via the MQTT protocol. Upon receiving the JSON data, the edge gateway (e.g., an industrial gateway based on an ARM Cortex-A series processor) parses it and stores it in a hash table in memory, with the task identifier as the key and H_avg as the value. When the machine tool control system (e.g., a Siemens 840D CNC system) issues a task start signal via the OPC UA protocol, the edge gateway's local program looks up and loads the corresponding H_avg from the hash table.

[0093] 2. Generation and execution of pilot cutting commands:

[0094] Once the edge computing gateway loads the batch-level background information for the current task, before the machine tool begins its main cutting process, it issues a command to the machine tool control system to perform a "pilot cut" or "probe" action, based on the current task type and material characteristics. This pilot cut is a very short, low-feed-speed, shallow-depth-of-cut controlled action, the purpose of which is not for machining, but to "probe" the local microscopic properties of the current workpiece material with low risk.

[0095] Implementation: Upon receiving the task start signal, the control module of the edge gateway (e.g., a C++ program running on a real-time operating system (RTOS) generates a set of CNC machine tool G-code instructions based on the material type (e.g., titanium alloy) and preset pilot cutting strategy provided by the MES. For example, for titanium alloy, it might generate G01 X10 Y10 F100 S5000D0.1 (representing a very short-distance straight-line cut at a feed rate of 100 mm / min, a spindle speed of 5000 rpm, and a depth of cut of 0.1 mm). These G-code instructions are sent to the machine tool's CNC system via Industrial Ethernet (e.g., EtherCAT or Profinet). The machine tool's CNC system receives and executes these instructions to complete the pilot cutting action.

[0096] 3. Micro-feature data acquisition and analysis:

[0097] During the pilot cutting action, the edge computing gateway acquires data from the tool vibration sensor (e.g., piezoelectric accelerometer) and the spindle motor current sensor at ultra-high frequency in real time. After preprocessing (e.g., high-pass filtering and noise reduction), the edge gateway's micro-feature analysis module performs fine analysis on this raw data to extract micro-fluctuation patterns that reflect the local resistance distribution of the material or the location of potential hard spots.

[0098] Implementation: The high-speed data acquisition card built into the edge gateway (e.g., 20kHz sampling rate) synchronously acquires current signals from a vibration sensor (e.g., a PCB Piezotronics 352C33 accelerometer) and a spindle motor driver (e.g., a Siemens SINAMICS S120 driver). The acquired raw vibration signal V(t) and current signal I(t) are preprocessed by a digital signal processor (DSP). The micro-feature analysis module calculates the instantaneous amplitude envelope A(t) of the vibration signal and the instantaneous power spectral density P(f,t) of the current signal. By analyzing the local peak values, rates of change, and energy distribution of specific frequency components of A(t) and P(f,t) during the pilot cutting process, the characteristics of local hard spots or internal stress concentrations in the material are identified. For example, when the tool cuts into a local hard spot, A(t) and P(f,t) show a transient and significant increase in a specific high-frequency band.

[0099] 4. Fine calibration of local vibration threshold:

[0100] Based on the micro-fluctuation patterns obtained from the analysis, the judgment logic adjustment module of the edge computing gateway generates a more refined local vibration threshold correction parameter for the current workpiece. This correction parameter is combined with the batch-level average hardness correction parameter provided by the MES system to dynamically adjust the tool vibration anomaly judgment threshold during subsequent main cutting processes.

[0101] Implementation: The micro-feature analysis module calculates a local hardness correction factor F_local based on the maximum instantaneous vibration amplitude A_pilot_max detected during pilot cutting and the spindle current fluctuation characteristics. For example, F_local = 1 + C1(A_pilot_max / A_base_pilot) + C2 (I_peak / I_avg), where C1 and C2 are preset weighting coefficients, A_base_pilot is the baseline pilot cutting amplitude, I_peak is the peak current, and I_avg is the average current. Then, the edge gateway combines the batch-level threshold V_threshold_batch provided by the MES with F_local to obtain the final dynamic judgment threshold V_threshold_dynamic = V_threshold_batch·F_local. This V_threshold_dynamic is immediately updated in the configuration register of the real-time tool status detection program running within the edge gateway.

[0102] 5. Real-time anomaly detection and decision-making during the main cutting process:

[0103] During the main cutting process, the edge computing gateway continuously collects tool vibration data at millisecond-level frequencies and compares it with the finely calibrated dynamic judgment threshold V_threshold_dynamic from step 4. Once the vibration signal exceeds V_threshold_dynamic, the edge gateway immediately determines it as a real tool failure and issues an emergency stop command to the machine tool control system, thereby promptly avoiding significant losses.

[0104] Implementation: The real-time data processing unit of the edge gateway (e.g., an FPGA or a high-performance MCU) continuously receives vibration sensor data. Whenever a new vibration data point arrives, it calculates its instantaneous amplitude and compares it with the current V_threshold_dynamic. If the instantaneous amplitude V_current > V_threshold_dynamic, an emergency stop (E-STOP) signal is immediately sent to the machine tool CNC system via the EtherCAT bus. This process, from data acquisition to command transmission, has a total latency controlled within 5 milliseconds, ensuring rapid response to sudden failures.

[0105] The working principle of this solution is similar to equipping a machine tool's cutting tool with "pathfinding eyes," allowing it to "understand" the material's characteristics before actual machining. In a precision machining workshop, even if the central system has already distributed the average characteristics (such as average hardness) of the batch material to the edge computing gateway according to the production plan, those localized hard spots or inhomogeneities within the material that are invisible to the naked eye may still suddenly appear during machining, causing a sudden increase in vibration signals. This can lead the edge gateway to mistakenly believe that there is a problem with the cutting tool and stop the machine. Such misjudgment not only wastes time but may also damage expensive workpieces.

[0106] Traditionally, edge gateways either only query the central system when an "anomaly" is detected, but this back-and-forth communication latency is too slow for sudden faults like microcracks in cutting tools that require millisecond-level response; by the time the information returns, damage may have already occurred. Alternatively, the central system sends all possible material information to each edge gateway, but this creates a huge network burden and storage pressure on the edge gateways, and most of the information is redundant. Even if the central system pre-distributes batch-level average information, it cannot solve the problem of instantaneous misjudgment caused by microscopic material inhomogeneities.

[0107] This solution differs. It utilizes the precise motion control capabilities of CNC machine tools. When the edge gateway receives information from the central system that a new batch of material is about to be processed, it does not immediately instruct the machine tool to perform high-speed main cutting. Instead, it first directs the machine tool to perform a very short, very gentle "pilot cutting" motion. This motion is like the tool gently "probing" the material surface, rather than actually cutting. During this "probing" process, the edge gateway carefully "listens" at extremely high frequencies to the vibration of the tool and the current changes of the spindle motor. By analyzing these extremely subtle fluctuations, the edge gateway can, with low risk, "sense" in advance whether there are local hard spots or uneven areas inside the current workpiece material.

[0108] Once the "pathfinding" is complete, the edge gateway will immediately adjust its local tool anomaly judgment criteria more precisely based on this microscopically sensed information. For example, if the "pathfinding" reveals that a certain area of ​​the material has a higher hardness than average, the edge gateway will correspondingly increase the vibration anomaly judgment threshold for that area. This way, when the machine tool begins its main cutting, even if the tool cuts into a localized hard spot in the material, causing a sudden increase in the vibration signal, the edge gateway's judgment threshold has already "adapted" to this change and will not misjudge this normal vibration caused by material properties as a tool anomaly. Simultaneously, if tool damage does occur, producing abnormal vibrations exceeding the adjusted threshold, the edge gateway can still issue a stop command within milliseconds to avoid losses.

[0109] This method cleverly addresses the challenges posed by the microscopic inhomogeneities within materials during precision machining. Instead of reactively responding after a problem occurs, it proactively gathers information and prepares for potential issues before they arise. Compared to relying solely on batch-level average information from a central system, this approach adds a proactive "microscopic probing" step at the edge, enabling the edge system to maintain millisecond-level response times while gaining a deeper, more real-time understanding of the local characteristics of the material being processed, thus allowing for more accurate judgments. This approach is particularly suitable for precision machining environments with extremely high requirements for machining accuracy, tool life, and production efficiency, where the processed material may exhibit microscopic fluctuations. However, for scenarios with highly uniform material properties or where machining efficiency requirements are so stringent that any additional "pathfinding" steps are intolerable, such a sophisticated predictive mechanism may not be necessary; simple batch-level information distribution or fixed thresholds may suffice.

[0110] In modern industrial production, the Internet of Things (IoT) technology is widely used to improve efficiency and equipment reliability. However, in industrial IoT scenarios, massive amounts of data generated by devices from different sources and in various formats suffer from problems such as inconsistent formats, high transmission latency, and difficulties in real-time analysis. Traditional centralized data processing models, while having advantages in some aspects, often lead to information asymmetry and struggle to meet the real-time response capabilities required for equipment status prediction and process optimization in smart manufacturing. Especially in environments with extremely high real-time requirements, such as precision manufacturing, effectively processing this data and ensuring the accuracy and timeliness of decision-making has become a pressing challenge.

[0111] To address the network transmission latency and data processing queuing delays inherent in traditional centralized processing models when dealing with sudden failures, the factory introduced an edge computing solution. An edge computing gateway was deployed next to each processing unit in the workshop. This gateway directly connects to the sensors of several machine tools within that unit, handling high-frequency vibration, current, and other data locally in real time, enabling rapid anomaly identification and timely response.

[0112] This solution proposes a centralized production scheduling and background information management unit. By continuously monitoring and parsing the production plan, it identifies the processing tasks that are about to begin and intelligently extracts concise key background information closely related to these tasks from the global background information. This information is used to provide timely and accurate context for the edge computing unit, effectively avoiding problems caused by information lag or redundancy.

[0113] In a precision machining workshop equipped with multiple CNC machine tools, the workshop's central Manufacturing Execution System (MES) plays a crucial scheduling role to maintain efficient and stable production. This system is programmed to periodically scan its internal production work order database to understand the current task status of all machine tools and the machining tasks to be performed in the near future. For example, by analyzing the production task list, the MES system identifies a specific CNC machine tool that, after completing its current task, will switch to a new machining task—producing aerospace components using titanium alloy material with batch number "TI-202302." Once the MES system anticipates this task switch, it immediately initiates a background information extraction process. Based on the batch identifier "TI-202302," it queries and extracts the specific physical properties of that batch of titanium alloy, such as its hardness value and toughness grade, from its maintained material property database. The MES system does not send this raw, detailed material data in its entirety. Instead, it intelligently processes and transforms these extracted attributes, condensing them into a lightweight calibration parameter that directly guides the decision-making logic of the edge computing gateway. For example, it converts the hardness value of titanium alloy into a "cutting vibration threshold correction value," which indicates how the normal cutting vibration amplitude should be adjusted when machining this batch of titanium alloy. This process ensures that the amount of information transmitted is minimized, containing only core data useful for real-time decision-making at the edge, laying the foundation for the accuracy of subsequent edge gateway judgments.

[0114] However, in actual production, even though the central system had proactively distributed batch-level material background information according to this plan, new challenges still emerged. First, the titanium alloy material with batch number "TI-202302" experienced a slight decrease in the uniformity of its grain structure and alloy element distribution at the microscopic level due to recent adjustments in the smelting process by its supplier. While still meeting macroscopic batch quality standards, the physical properties of localized areas showed increased volatility. Second, to meet an urgent order for aerospace components with stringent delivery deadlines, the machine tool was required to operate with cutting parameters close to its design limits. For example, both spindle speed and feed rate were set to the upper limits of conventional machining, significantly increasing the sensitivity of the cutting process to local material properties. Under these dual conditions, when the tool accidentally came into contact with locally high-hardness areas within the material due to microstructural inhomogeneities during high-speed cutting, even if these areas were only tens of micrometers in size, the instantaneous cutting resistance would increase dramatically. This sudden localized resistance can instantly cause a surge in the amplitude of the spindle vibration signal, with its peak value far exceeding the normal threshold adjusted by the edge computing gateway based on the batch average hardness information provided by the MES. At this point, the edge gateway's real-time anomaly detection program, despite having applied batch-level background information distributed by the central system for threshold calibration, cannot distinguish between this instantaneous vibration caused by localized fluctuations in material microstructure and the abnormal vibration truly caused by tool wear or microcracks. It immediately identifies it as a tool malfunction and quickly issues a stop command to the machine tool control system. Maintenance personnel then inspect the tool and find it intact, but production has already been interrupted, delaying urgent orders and causing unnecessary resource waste. This persistent false alarm reveals a deeper problem: the central system only provides batch-level average background information, which is insufficient in granularity to support completely accurate real-time judgments at the edge when facing extreme processing conditions and material micro-inhomogeneities.

[0115] In precision machining scenarios where a centralized Manufacturing Execution System (MES) proactively distributes batch-level material background information based on production plans and edge computing units dynamically adjust local judgment thresholds, how can edge computing units quickly and accurately identify and distinguish between instantaneous vibration signal changes caused by fluctuations in the local micro-characteristics of the material and vibration anomalies caused by actual tool failures, given the microstructural inhomogeneities within the processed material due to supplier process adjustments and extreme cutting parameters used to meet urgent orders? This would effectively avoid continuous misjudgments caused by insufficient granularity of existing batch-level background information, without increasing communication latency or computational burden at the edge.

[0116] This solution aims to enhance the accuracy of anomaly detection in precision machining scenarios, particularly after the central system provides batch-level average information. It addresses the challenge of effectively identifying and distinguishing between transient vibration anomalies caused by microscopic local inhomogeneities within the machined material and actual tool failures. This solution utilizes a "micro-feedback cutting parameter flexible correction" mechanism to achieve proactive perception and precise calibration of the edge's logic for judging the microscopic characteristics of the current workpiece material.

[0117] The specific method includes the following steps:

[0118] 1. Central system task information distribution and edge preloading:

[0119] Description: A centralized Manufacturing Execution System (MES) continuously monitors the production schedule and identifies upcoming processing tasks. Before a task begins, the MES system extracts batch-level average background information (e.g., material average hardness correction parameters) from its material property database based on task information (e.g., material batch number) and proactively distributes it to the edge computing gateway responsible for that task. Upon receiving this information, the edge gateway temporarily stores it locally and loads it when the task starts to initially adjust its local decision-making logic.

[0120] Implementation: The MES system is configured with a scheduled service that scans the production work order database every minute to identify tasks that will begin within the next 30 minutes. For each identified task, the MES queries its material property database (e.g., an SQL database table) to obtain the average hardness value H_avg of the material for that batch. The MES then encapsulates H_avg into a lightweight JSON data packet and pushes it to a specific topic on the target edge gateway via the MQTT protocol. Upon receiving the JSON data, the edge gateway (e.g., an industrial gateway based on an ARM Cortex-A series processor) parses it and stores it in a hash table in memory, with the task identifier as the key and H_avg as the value. When the machine tool control system (e.g., a Siemens 840D CNC system) issues a task start signal via the OPC UA protocol, the edge gateway's local program looks up and loads the corresponding H_avg from the hash table. This step provides an initial, batch-level reference baseline for subsequent micro-feedback adjustments.

[0121] 2. Real-time perception of micro-features of the cutting path:

[0122] Description: After the machine tool begins its main cutting process, the edge computing gateway collects data from the tool vibration sensor and the spindle motor current sensor at an ultra-high frequency in real time. Instead of waiting for the vibration signal to reach a high threshold, it focuses on analyzing the instantaneous change patterns of these data at each new contact between the tool and the material or at the beginning of the cutting path to identify early signs of local microscopic properties of the material, such as localized hard spots or areas of concentrated internal stress.

[0123] Implementation: The high-speed data acquisition card built into the edge gateway (e.g., sampling rate 20kHz) synchronously acquires current signals from a vibration sensor (e.g., a PCB Piezotronics 352C33 accelerometer) and a spindle motor driver (e.g., a Siemens SINAMICS S120 driver). The edge gateway's real-time processing unit (e.g., an FPGA or high-performance DSP) continuously calculates the instantaneous amplitude envelope A(t) of the vibration signal and the instantaneous rate of change dI / dt of the current signal. When A(t) or dI / dt exhibits a specific amplitude (e.g., exceeding 20%) exceeding its local average (calculated based on the first 100 milliseconds of data) and a duration of extremely short duration (e.g., less than 10 milliseconds) within a very short time (e.g., less than 5 milliseconds), this is judged as an early indication of fluctuations in the local microscopic properties of the material. This judgment logic avoids complex pattern recognition of the entire cutting process and instead focuses on the characteristics of the transient response.

[0124] 3. Instantaneous cutting parameter fine-tuning decision:

[0125] Description: Based on the early indications of local microscopic material properties detected in step 2, the decision module of the edge computing gateway immediately calculates and generates a minimal-amplitude cutting parameter correction command for the current cutting point. This correction aims to flexibly adjust the interaction between the tool and the material to smoothly traverse local inhomogeneities and avoid excessive spikes in vibration signals.

[0126] Implementation: If step 2 detects early signs of localized fluctuations in the material's microstructure, the edge gateway's decision logic calculates a temporary fine-tuning amount based on the current cutting parameters (e.g., feed rate F_current, spindle speed S_current) and the material's average hardness H_avg provided by the MES. For example, it might decide to temporarily reduce the feed rate F_current by a small percentage (e.g., 5%), or fine-tune the spindle speed S_current by a small amount (e.g., 2%). The specific feed rate adjustment F_adjusted and spindle speed adjustment S_adjusted can be determined using a preset lookup table or a simple linear function, for example:

[0127] F_adjusted = F_current (1 - K_f)

[0128] S_adjusted = S_current (1 + K_s)

[0129] Here, F_adjusted and S_adjusted are the feed rate correction and spindle speed correction, respectively; F_current and S_current are the feed rate and spindle speed, respectively; and K_f and K_s are small correction coefficients pre-calibrated based on material type and cutting parameters, used for flexible adjustments when early signs are detected. This decision-making process is completed within milliseconds to ensure timeliness.

[0130] 4. Micro-feedback execution of the machine tool control system:

[0131] Description: The edge computing gateway sends the instantaneous cutting parameter fine-tuning instructions generated in step 3 to the machine tool's CNC system in real time via a preset low-latency industrial communication protocol. After receiving the instructions, the CNC system immediately applies these fine-tuning parameters to the current cutting point and automatically restores the original cutting parameters after the tool passes through a local non-uniform area.

[0132] Implementation: The edge gateway sends a command packet containing temporary feed rate and spindle speed correction values ​​to the machine tool CNC system (e.g., FANUC 0i-F Plus CNC system) via Industrial Ethernet (e.g., EtherCAT or Profinet). This command packet can be a local modification command in G-code (e.g., G01 F_adjusted S_adjusted) or an update of real-time parameter variables within the CNC system via the OPC UA protocol. Upon receiving these commands, the CNC system immediately applies these parameters to the current machining path and automatically switches back to the original cutting parameters after a preset short time window (e.g., 50 milliseconds) or a specific distance (e.g., 0.5 mm). This approach ensures the locality and immediacy of fine-tuning without significantly affecting the overall machining process.

[0133] 5. Dynamic thresholds and real-time anomaly detection:

[0134] Description: During the main cutting process, the edge computing gateway continuously collects tool vibration data. Its anomaly detection logic dynamically adjusts the judgment threshold for tool vibration anomalies based on batch-level background information provided by the MES and the actual cutting parameters of the current machine tool application (including any instantaneous fine-tuning). Once the vibration signal exceeds this dynamically calibrated threshold and does not conform to the characteristics of local material fluctuations, the edge gateway immediately determines it as a real tool failure and issues an emergency stop command to the machine tool control system.

[0135] Implementation: The edge gateway's real-time data processing unit continuously receives vibration sensor data. It calculates a dynamic vibration threshold V_threshold_dynamic based on the batch average hardness H_avg provided by the MES and the current machine tool's actual feed rate F_actual and spindle speed S_actual. For example:

[0136] V_threshold_dynamic = V_base + K_H·H_avg + K_F·F_actual + K_S·S_actual

[0137] Where V_threshold_dynamic is the dynamic vibration threshold, V_base is the base threshold, H_avg is the batch average hardness, F_actual is the current actual feed rate of the machine tool, S_actual is the spindle speed, and K_H, K_F, and K_S are correction coefficients predetermined based on material and process characteristics. Whenever a new vibration data point arrives, its instantaneous amplitude V_current is calculated and compared with V_threshold_dynamic. If V_current > V_threshold_dynamic, and the vibration characteristic does not conform to the material local micro-characteristic fluctuation pattern defined in step 2 (e.g., excessive duration, continuously increasing amplitude), an emergency stop (E-STOP) signal is immediately sent to the machine tool CNC system via the EtherCAT bus. This process, from data acquisition to command transmission, has a total delay controlled within 5 milliseconds, ensuring rapid response to real sudden faults.

[0138] The working principle of this solution is like equipping a machine tool with an experienced "fine-tuner." It not only understands the overall "temperament" of the material but also senses local "emotional fluctuations" in real time during machining and immediately provides gentle reassurance. In a precision machining workshop, even if the central system has distributed the average characteristics (such as average hardness) of batch materials to the edge computing gateway in advance according to the production plan, those invisible local hard spots or inhomogeneities within the material can still suddenly appear during machining. This is especially true when using near-limit cutting parameters for urgent orders. Such local inhomogeneities can cause a sudden and sharp increase in cutting force, leading to a surge in the amplitude of the spindle vibration signal. If only a fixed threshold adjusted based on batch-level information is used, the edge gateway may misjudge this instantaneous vibration caused by local material characteristics as a tool malfunction and shut down the machine. This misjudgment not only wastes time but may also damage expensive workpieces.

[0139] Traditionally, edge gateways either only query the central system when they detect an "anomaly," but this back-and-forth communication latency is too slow for sudden events requiring millisecond-level responses; by the time the information returns, damage may have already occurred. Alternatively, the central system sends all possible material information to each edge gateway, but this creates a huge network burden and storage pressure on the edge gateways, and most of the information is redundant. Even if the central system pre-distributes batch-level average information, it cannot solve the problem of instantaneous misjudgments caused by microscopic inhomogeneities in the material.

[0140] This solution differs. It leverages the precise motion control capabilities of CNC machine tools and the ultra-high-speed sensing capabilities of edge computing. Once the edge gateway receives the average characteristics of the new batch of material to be processed from the central system, it enters a state of "high alert" as soon as the machine tool begins its main cutting operation. Instead of waiting for the vibration signal to reach a stop threshold, it meticulously "listens" to the tool's vibration and the spindle motor's current changes at extremely high frequencies at the very beginning of each new contact between the tool and the material or at the start of the cutting path. By analyzing these extremely subtle instantaneous fluctuations, the edge gateway can detect early signs of localized hard spots or uneven areas within the current workpiece material before problems escalate.

[0141] Once these early signs are detected, the edge gateway's "fine-tuner" acts immediately. It quickly calculates a tiny adjustment to the cutting parameters, such as temporarily and slightly reducing the feed rate or fine-tuning the spindle speed. These correction commands are sent to the machine tool's CNC system within milliseconds via high-speed industrial communication channels. Upon receiving the commands, the CNC system immediately applies these fine-tuning parameters to the current cutting point, allowing the tool to pass through localized hard spots more gently. This effectively suppresses vibration spikes caused by localized material inhomogeneities, preventing them from reaching the shutdown threshold. After the tool passes through the localized area, the machine tool automatically returns to its original cutting parameters.

[0142] Meanwhile, the anomaly detection threshold of the edge gateway is dynamically adjusted based on batch-level information provided by the central system and the machine tool's current actual cutting parameters (including any fine-tuning). This ensures that even during fine-tuning, if tool damage occurs, resulting in abnormal vibrations exceeding the adjusted threshold, the edge gateway can still issue a stop command within milliseconds, preventing significant losses.

[0143] This method ingeniously addresses the challenges posed by the microscopic inhomogeneities within materials under extreme processing conditions. Instead of reacting passively after a problem occurs, it proactively gathers information and prepares before problems arise, even intervening to prevent escalation. Compared to relying solely on batch-level average information from a central system, this solution adds a proactive "microscopic sensing and flexible intervention" step at the edge. This allows the edge system to maintain millisecond-level response speeds while gaining a deeper, more real-time understanding of the local characteristics of the material being processed. It can then proactively adjust its systems to make more accurate judgments and significantly reduce unnecessary production interruptions. This approach is particularly suitable for precision manufacturing environments with extremely high requirements for processing quality and equipment reliability, where the processed materials may exhibit microscopic fluctuations, and where processing must be performed near-limit parameters. However, for scenarios with highly uniform material properties or where processing efficiency requirements are so stringent that no proactive intervention is tolerated, such a sophisticated micro-feedback mechanism may not be necessary; simple batch-level information distribution or fixed thresholds may suffice.

[0144] Example 2:

[0145] An IoT multi-source data edge computing processing system, which can implement the IoT multi-source data edge computing processing method described in Embodiment 1, includes:

[0146] Central scheduling module: Used to continuously monitor and parse production plans through a centralized production scheduling unit, identify processing tasks that will be started within a preset time period in the future, and extract the material batch identifiers corresponding to the processing tasks;

[0147] Information extraction module: used to extract concise key background information matching the processing task from the global background information library based on the material batch identifier. The concise key background information is the minimum dataset required for edge-side anomaly detection.

[0148] Information push module: used to proactively push the simplified key background information to the corresponding edge computing gateway through a low-latency communication channel at a preset time point before the start of the processing task;

[0149] Information receiving module: used to receive and temporarily store the simplified key background information through the edge computing gateway, and load matching local judgment logic calibration parameters when the processing task is detected to be started;

[0150] Anomaly detection module: used to dynamically adjust the local anomaly detection threshold based on the calibration parameters, perform millisecond-level anomaly detection on real-time multi-source sensor data, and trigger control commands when an anomaly is detected.

[0151] Example 3:

[0152] This invention also provides an IoT multi-source data edge computing processing device, which can implement the IoT multi-source data edge computing processing method described in Embodiment 1, including a processor and a storage medium;

[0153] The storage medium is used to store instructions;

[0154] The processor is configured to operate according to the instructions to perform the steps of the following method:

[0155] The production plan is continuously monitored and analyzed by the centralized production scheduling unit to identify the processing tasks that will be started within a preset time period in the future and extract the material batch identifiers corresponding to the processing tasks.

[0156] Based on the material batch identifier, simplified key background information matching the processing task is extracted from the global background information database. The simplified key background information is the minimum dataset required for edge-side anomaly detection.

[0157] At a preset time point before the processing task starts, the simplified key background information is proactively pushed to the corresponding edge computing gateway through a low-latency communication channel.

[0158] The simplified key background information is received and temporarily stored through the edge computing gateway. When the start of the processing task is detected, the matching local judgment logic calibration parameters are loaded.

[0159] Based on the calibration parameters, the local anomaly detection threshold is dynamically adjusted to perform millisecond-level anomaly judgment on the real-time acquired multi-source sensor data, and a control command is triggered when an anomaly is determined.

[0160] Example 4:

[0161] This invention also provides a computer-readable storage medium that can implement the IoT multi-source data edge computing processing method described in Embodiment 1. The medium stores a computer program that, when executed by a processor, performs the steps of the following method:

[0162] The production plan is continuously monitored and analyzed by the centralized production scheduling unit to identify the processing tasks that will be started within a preset time period in the future and extract the material batch identifiers corresponding to the processing tasks.

[0163] Based on the material batch identifier, simplified key background information matching the processing task is extracted from the global background information database. The simplified key background information is the minimum dataset required for edge-side anomaly detection.

[0164] At a preset time point before the processing task starts, the simplified key background information is proactively pushed to the corresponding edge computing gateway through a low-latency communication channel.

[0165] The simplified key background information is received and temporarily stored through the edge computing gateway. When the start of the processing task is detected, the matching local judgment logic calibration parameters are loaded.

[0166] Based on the calibration parameters, the local anomaly detection threshold is dynamically adjusted to perform millisecond-level anomaly judgment on the real-time acquired multi-source sensor data, and a control command is triggered when an anomaly is determined.

[0167] As is known from common technical knowledge, this invention can be implemented through other embodiments that do not depart from its spirit or essential characteristics. Therefore, the disclosed embodiments described above are merely illustrative and not exhaustive. All modifications within the scope of this invention or its equivalents are included in this invention.

[0168] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0169] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0170] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0171] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0172] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for edge computing processing of multi-source data in the Internet of Things, characterized in that, include: The production plan is continuously monitored and analyzed by the centralized production scheduling unit to identify the processing tasks that will be started within a preset time period in the future and extract the material batch identifiers corresponding to the processing tasks. Based on the material batch identifier, simplified key background information matching the processing task is extracted from the global background information database. The simplified key background information is the minimum dataset required for edge-side anomaly detection. At a preset time point before the processing task starts, the simplified key background information is proactively pushed to the corresponding edge computing gateway through a low-latency communication channel. The simplified key background information is received and temporarily stored through the edge computing gateway. When the start of the processing task is detected, the matching local judgment logic calibration parameters are loaded. Based on the calibration parameters, the local anomaly detection threshold is dynamically adjusted to perform millisecond-level anomaly judgment on the real-time acquired multi-source sensor data, and a control command is triggered when an anomaly is determined.

2. The IoT multi-source data edge computing processing method according to claim 1, characterized in that, The simplified key background information includes the average physical property conversion parameters at the material batch level; the extraction of simplified key background information matching the processing task from the global background information database specifically includes: querying the material property database corresponding to the material batch identifier, obtaining at least one attribute value among average hardness and toughness, converting the attribute value into a threshold correction coefficient for the edge-side anomaly detection algorithm, and removing redundant detailed quality inspection reports and full material data.

3. The IoT multi-source data edge computing processing method according to claim 1, characterized in that, It also includes a pilot cutting micro-feature prediction step: After the machining task is started but before the main cutting begins, the edge computing gateway sends a pilot cutting command to the machine tool control system, controlling the tool to perform a trial cut at a preset distance with low feed and shallow cutting depth; Simultaneously acquire vibration and spindle current data during the pilot cutting process, and extract instantaneous fluctuation characteristics that reflect the local microscopic properties of the material; A local correction factor is generated based on the instantaneous fluctuation characteristics, and combined with the batch-level calibration parameters, a dynamic anomaly detection threshold is obtained.

4. The IoT multi-source data edge computing processing method according to claim 3, characterized in that, The local microscopic properties include at least one of localized hard spots and regions of internal stress concentration within the material; the instantaneous fluctuation characteristics are extracted in the following manner: Calculate the instantaneous amplitude envelope of the vibration signal and the instantaneous power spectral density of the current signal; Identify energy abrupt changes, amplitude spikes, duration and rate of change in specific high-frequency bands, and match them with preset material inhomogeneity feature templates.

5. The IoT multi-source data edge computing processing method according to claim 1, characterized in that, It also includes a micro-feedback cutting parameter flexible correction step: During the main cutting process, the transient spikes of vibration and current signals are monitored in real time. If early signs that conform to the local characteristic fluctuation mode of the material are detected, the temporary cutting parameter fine-tuning amount is calculated. The fine-tuning command is sent to the machine tool control system via the industrial bus, and the feed rate and spindle speed are temporarily adjusted at the current cutting point, and the original parameters are automatically restored after passing through a local uneven area. The anomaly detection threshold is dynamically updated based on real-time cutting parameters to distinguish between material fluctuations and actual tool failures.

6. The IoT multi-source data edge computing processing method according to claim 5, characterized in that, The temporary cutting parameter fine-tuning amount is a minor correction within ±2% for feed rate reduction and spindle speed adjustment; the dynamic update adopts the following formula: V_threshold_dynamic = V_base + K_H·H_avg + K_F·F_actual + K_S·S_actual Where V_threshold_dynamic is the dynamic vibration threshold, V_base is the base threshold, H_avg is the batch average hardness, F_actual is the actual feed rate of the current machine tool, S_actual is the spindle speed, and K_H, K_F, and K_S are correction coefficients predetermined based on material and process characteristics.

7. The IoT multi-source data edge computing processing method according to claim 1, characterized in that, The low-latency communication channel adopts at least one of the following: industrial Ethernet, 5G industrial private network, and MQTT protocol; the total latency of the edge computing gateway's anomaly detection and instruction triggering is controlled within 5 milliseconds.

8. An IoT multi-source data edge computing processing system, characterized in that, The method for implementing IoT multi-source data edge computing processing as described in any one of claims 1-7 includes: Central scheduling module: Used to continuously monitor and parse production plans through a centralized production scheduling unit, identify processing tasks that will be started within a preset time period in the future, and extract the material batch identifiers corresponding to the processing tasks; Information extraction module: used to extract concise key background information matching the processing task from the global background information library based on the material batch identifier. The concise key background information is the minimum dataset required for edge-side anomaly detection. Information push module: used to proactively push the simplified key background information to the corresponding edge computing gateway through a low-latency communication channel at a preset time point before the start of the processing task; Information receiving module: used to receive and temporarily store the simplified key background information through the edge computing gateway, and load matching local judgment logic calibration parameters when the processing task is detected to be started; Anomaly detection module: used to dynamically adjust the local anomaly detection threshold based on the calibration parameters, perform millisecond-level anomaly detection on real-time multi-source sensor data, and trigger control commands when an anomaly is detected.

9. An IoT multi-source data edge computing processing device, characterized in that, Including processor and storage media; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1 to 7.

Images