A SECS / GEM communication method and system supporting log parsing and dynamic computation tasks.

Abstract

This invention relates to a SECS / GEM communication method and system supporting log parsing and dynamic computation tasks. The method includes the following steps: high-frequency concurrent acquisition of heterogeneous data and standardized feature extraction; time alignment and virtual tag space mapping of multi-source data; multi-level routing distribution based on hash mapping tables and unified unpacking mechanisms; business parsing and hierarchical dynamic computation based on state machine verification; polymorphic reporting of analysis results and millisecond-level security closed-loop control under abnormal states. This invention unifies the acquisition of heterogeneous PLC and log data through a hardware abstraction layer, decouples physical addresses and business logic using a virtual tag space, and dynamically executes analysis tasks issued by the host using an edge computing engine. This solves the problems of outdated interfaces, coupled logic, difficult log utilization, and high cloud latency, improving system compatibility and real-time response capabilities.

Description

Technical Field

[0001] This invention relates to the field of cross-disciplinary technology of automated communication and intelligent operation and maintenance of semiconductor manufacturing equipment, and in particular to a SECS / GEM communication method and system that supports log parsing and dynamic calculation tasks. Background Technology

[0002] With the continuous evolution of semiconductor manufacturing processes, automated communication between Manufacturing Execution Systems (MES) and underlying production equipment has become a key component of intelligent production lines. Among them, the SECS / GEM protocol, which follows the SEMI standard, has become the core communication standard for enabling status monitoring, data acquisition, and remote control between semiconductor manufacturing equipment and the host system.

[0003] However, in the current semiconductor packaging and testing field, especially in representative packaging and testing industry clusters, a large number of production equipment are still heterogeneous machines provided by different eras and manufacturers. These machines differ significantly in communication interfaces, control logic, and data acquisition methods, posing numerous technical challenges to the automated integration of the equipment, mainly in the following aspects:

[0004] (a) Outdated communication interfaces of old equipment

[0005] In existing packaging and testing production lines, many early devices are only equipped with serial communication interfaces such as RS232, lacking native Ethernet interfaces and SECS / GEM communication capabilities. Traditional solutions often use Windows-based industrial PCs (IPCs) as communication relay platforms to convert serial data to Ethernet data. However, such solutions not only have high hardware and software licensing costs, but also suffer from poor stability and complex maintenance of Windows systems, making them susceptible to virus attacks or system updates, thereby reducing the reliability of equipment operation.

[0006] (ii) Strong coupling between PLC register address and business logic

[0007] In existing equipment communication architectures, equipment control logic typically relies directly on PLC register addresses for data reading and writing. However, different brands of PLCs exhibit significant differences in register structures and communication protocols. Developers must repeatedly write communication logic when interfacing with different devices, leading to high system development and maintenance costs. Furthermore, in the common "one box controls multiple machines" scenario in packaging and testing production lines, the lack of a unified data abstraction layer when multiple devices share the same communication platform can easily result in severe data coupling, communication resource conflicts, and difficulties in system expansion.

[0008] (iii) Equipment log data cannot be effectively utilized

[0009] In many semiconductor devices, PLCs typically store only limited discrete state information, while richer process data (such as pressure changes, mechanical motion trajectories, temperature profiles, etc.) are often stored as text logs in the log files of the device's local industrial control computer (IPC). This log data is mostly in unstructured text format, which traditional SECS / GEM communication systems struggle to parse and utilize directly. This results in a large amount of valuable process characteristic data remaining uncollected and unanalyzed by the higher-level system, limiting the capabilities of equipment fault detection (FDC) and predictive maintenance (PdM).

[0010] (iv) Cloud computing suffers from high latency and is difficult to support dynamic computing tasks.

[0011] With the development of intelligent manufacturing, production lines are placing higher demands on real-time equipment monitoring and data analysis capabilities. Traditional solutions often upload raw data collected by equipment to cloud servers for centralized analysis. However, when there are many devices or a high sampling frequency, the transmission of large amounts of raw data over the network will generate significant communication delays and bandwidth pressure, making it difficult to perform timely shutdown protection in the event of momentary anomalies. In addition, existing SECS / GEM communication interfaces mostly adopt static configuration methods, which are difficult to support analysis tasks or algorithm models dynamically issued by the host, and cannot meet the needs of online data analysis and dynamic calculation tasks in flexible manufacturing environments. Summary of the Invention

[0012] In view of the shortcomings of the prior art, the purpose of this invention is to provide a SECS / GEM communication method and system that supports log parsing and dynamic computing tasks, so as to solve one or more problems in the prior art.

[0013] To achieve the above objectives, the technical solution of the present invention is as follows:

[0014] A SECS / GEM communication method supporting log parsing and dynamic computation tasks, the method comprising the following steps:

[0015] S1. High-frequency concurrent acquisition of heterogeneous data and standardized feature extraction;

[0016] S2. Temporal alignment and virtual label space mapping of multi-source data;

[0017] S3. Multi-level route distribution based on hash mapping table and unified unpacking mechanism;

[0018] S4. Business parsing and hierarchical dynamic calculation based on state machine verification;

[0019] S5. Multi-state reporting of analysis results and millisecond-level safety closed-loop control under abnormal conditions.

[0020] Furthermore, data acquisition in S1 is based on the association configuration set of the underlying data source and the virtual tag system predefined in the system parsing device capability description file (ECD), including the PLC configuration set. With log configuration set It includes the following steps:

[0021] S11, PLC physical status data polling:

[0022] Based on PLC configuration set And instantiate the corresponding underlying driver using the dynamic link library;

[0023] Within a preset sampling period, the target physical address set Perform polling reads;

[0024] The read binary dataset is scaled by a factor. After standardizing the engineering units, a structured numerical physical state dataset is obtained. ;

[0025] S12. Event-driven log feature capture:

[0026] The system listens for kernel file append events, and when it captures the newly added log text stream... At that time, the resident memory regular expression capture group library is called. Perform concurrent matching;

[0027] Text fragments are extracted using regular expression capture groups and then converted using a type conversion function. The process transforms unstructured text into a structured set of log feature key-value pairs. .

[0028] Furthermore, the temporal alignment and virtual label space mapping of multi-source data in S2 includes the following steps:

[0029] S21, when or The system obtains the current absolute timestamp the moment the data arrives at the hardware abstraction layer entry point. ;

[0030] S22. The system constructs a global static mapping dictionary based on the ECD file. :

[0031] The PLC address is defined as follows:

[0032]

[0033] For log rules, the definition is as follows:

[0034]

[0035] In the formula: It serves as a globally unique virtual tag identifier. For SECS / GEM entity attributes, The target data type;

[0036] S23, System construction of multi-source mapping function ,when or Data in Upon arrival, the system queries Extract metadata and dynamic values With timestamp Combine them to generate standardized data tuples, as defined in the following formula:

[0037]

[0038] S24. Push the generated standardized data tuples into the virtual tag space. In the time-series queue.

[0039] Furthermore, the multi-level route distribution in S3 based on hash mapping tables and a unified unpacking mechanism includes the following steps:

[0040] S31. After the communication layer completes the TCP / IP session establishment and handshake with the host based on the HSMS protocol, it continuously listens for SECS-II messages.

[0041] S32. After receiving the message, parse the header type field. If the judgment is The underlying link layer responds directly; if Extract the business stream number With Function Number ;

[0042] S33, The system construction time complexity is... Global route hash table Calculate message feature key values And use a hash function to calculate the index. ;

[0043] S34, the system passes The addressing processing entity is routed to the standard GEM processor if a standard instruction is hit, and to the extended task manager if a custom extended analysis protocol family is hit.

[0044] Furthermore, in S4, based on state machine verification, business parsing and hierarchical dynamic calculation, for the results distributed in S3, the system concurrently executes differentiated dual-track logic, including the following steps:

[0045] S41 and SECS-II messages are uniformly recursively parsed by calling the underlying GEM protocol parser. The parser is based on a recursive descent algorithm to peel off the nested List and Item data structures in the message body layer by layer and extract the pure business parameters at the underlying level.

[0046] S42. Standard control flow execution: After receiving the clean instruction parameters, the standard GEM processor calls the E30 state machine validity function. The system performs verification and performs logical interception based on the three control levels defined in the SECS / GEM specification, including OFF-LINE, LOCAL, and REMOTE states.

[0047] S43. Edge analysis stream execution: After receiving the parsed pure parameters of the analysis task, the extended task manager sends the pure computing power task to the edge computing engine for hierarchical computing, including lightweight simple computing, predefined statistical process control, and dynamic parameterized script inference.

[0048] Furthermore, in S5, the analysis results are reported in multiple states and millisecond-level safety closed-loop control is implemented under abnormal conditions. The system generates closed-loop feedback based on the execution results of S4, including the following steps:

[0049] S51, Standard Control Flow Feedback: According to the SECS / GEM specification, if the physical control execution is successful, a standard response message is encapsulated and reported; if the underlying execution fails or the E30 state machine verification function fails... If the return is invalid, the system will encapsulate the corresponding exception response message or interrupt communication message and report it to the host;

[0050] S52, Edge computing flow feedback, including:

[0051] Regular reporting occurs when the edge engine's inference results are reported. Within the preset safety threshold range During this time, the engine encapsulates the results into an extended event report message and reports it, and the host replies with an acknowledgment message;

[0052] Abnormal closed-loop interception, when the calculation result The device control layer will short-circuit the regular communication reporting link; during the call After verifying that the current device mode allows intervention, the engine directly generates the optimal blocking or fine-tuning command. Through the hardware abstraction layer The atom is written into the predefined protection register of the underlying PLC.

[0053] A communication system, wherein the communication system is applied to the aforementioned SECS / GEM communication method supporting log parsing and dynamic computing tasks, the system includes a device edge control unit deployed between the host system and the underlying device; the device edge control unit includes:

[0054] The hardware abstraction layer is configured to perform high-frequency concurrent acquisition of heterogeneous data and standardized feature extraction, including a PLC communication module and a log parsing module;

[0055] The standardized information model layer is configured to perform time alignment and mapping of multi-source data. It builds a unified data abstraction model for devices and decouples the upper-layer business logic from the lower-layer physical address by establishing a virtual label space. It includes a mapping management module and a dynamic configuration module.

[0056] The device control layer is configured to perform device control logic management and edge computing task processing, including a standard GEM processor and an edge computing engine;

[0057] The communication layer is configured to enable SECS / GEM protocol communication between the device edge control unit and the host system, including a session management module, a protocol parsing module, and a command distribution module.

[0058] Furthermore, the standard GEM processor includes:

[0059] The state management module is configured to manage the device's operating and communication states based on a finite state machine.

[0060] The compliance verification module is configured to verify the device status before executing device control commands.

[0061] An electronic device includes: a processor, a memory, and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the aforementioned SECS / GEM communication method supporting log parsing and dynamic calculation tasks.

[0062] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned SECS / GEM communication method supporting log parsing and dynamic computation tasks.

[0063] Compared with the prior art, the beneficial technical effects of the present invention are as follows:

[0064] (i) This invention introduces a log parser in the hardware abstraction layer and establishes a unified virtual tag space in the standardized information model layer to realize the fusion collection of PLC data and equipment log data, thereby improving the integrity of equipment process data.

[0065] (ii) By establishing a virtual tag space, this invention decouples physical addresses from business logic, enabling device variables to be accessed and managed in a unified manner, thereby improving system scalability.

[0066] (iii) By extending the analysis task message mechanism, this invention enables the host to dynamically issue computing tasks and perform data analysis at the device edge, thereby improving the device anomaly detection and process optimization capabilities.

[0067] (iv) This invention reduces reliance on cloud computing resources by performing data analysis and control decisions at the device edge, thereby reducing system communication latency and improving real-time control capabilities. Attached Figure Description

[0068] Figure 1 This illustration shows a flowchart of a SECS / GEM communication method and system that supports log parsing and dynamic calculation tasks according to an embodiment of the present invention.

[0069] Figure 2 This paper illustrates a SECS / GEM communication method and system for log data parsing and virtual tag mapping that supports log parsing and dynamic calculation tasks, according to an embodiment of the present invention.

[0070] Figure 3 This illustration shows a schematic diagram of the device edge analysis task execution flow of a SECS / GEM communication method and system that supports log parsing and dynamic calculation tasks according to an embodiment of the present invention.

[0071] Figure 4 This illustration shows a schematic diagram of the virtual tag space mapping relationship of a SECS / GEM communication method and system that supports log parsing and dynamic calculation tasks according to an embodiment of the present invention.

[0072] Figure 5 This paper illustrates an embodiment of the SECS / GEM communication method and system that supports log parsing and dynamic calculation tasks, and provides an edge analysis engine full-trajectory anomaly detection model architecture diagram.

[0073] Figure 6 This illustration shows a schematic diagram of the overall system architecture of a SECS / GEM communication method and system that supports log parsing and dynamic calculation tasks according to an embodiment of the present invention. Detailed Implementation

[0074] To make the objectives, technical solutions, and advantages of this invention clearer, the following detailed description, in conjunction with the accompanying drawings and specific embodiments, provides a further detailed explanation of the SECS / GEM communication method and system supporting log parsing and dynamic computation tasks proposed by this invention. The advantages and features of this invention will become clearer from the following description. It should be noted that the accompanying drawings are in a very simplified form and use non-precise proportions, used only to facilitate and clearly illustrate the purpose of the embodiments of this invention. Please refer to the accompanying drawings to make the objectives, features, and advantages of this invention more apparent and understandable. It should be understood that the structures, proportions, sizes, etc., depicted in the accompanying drawings are only used to complement the content disclosed in the specification, for those skilled in the art to understand and read, and are not intended to limit the implementation conditions of this invention. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportional relationships, or adjustments to the size, without affecting the effects and objectives achieved by this invention, should still fall within the scope of the technical content disclosed in this invention.

[0075] Please refer to the following: Figures 1 to 5 A SECS / GEM communication method that supports log parsing and dynamic computation tasks, the method comprising the following steps:

[0076] This embodiment uses a bonding machine PLC with a certain brand of MC communication protocol as an example to illustrate the above communication method.

[0077] S1. High-frequency concurrent acquisition of heterogeneous data and standardized feature extraction.

[0078] In system runtime, the Hardware Abstraction Layer (HAL) within the system device edge control unit actively parses the pre-configured Device Capability Description File (ECD). The ECD predefines the associated configuration set between the underlying data source and the virtual tag system, including the PLC configuration set. With log configuration set The system initiates a dual-channel data acquisition mechanism in parallel based on this, specifically including the following steps:

[0079] S11, PLC physical status data polling:

[0080] Based on PLC configuration set And instantiate the corresponding underlying driver using the dynamic link library. PLC configuration set middle, The target physical absolute address is the PLC data identifier. It is a primitive register data type. For polling frequency, This is a linear scaling factor.

[0081] In this embodiment, to achieve high-fidelity feature extraction across the entire trajectory, the configuration subset not only includes static parameters but also incorporates high-frequency dynamic features, such as the blade heating temperature, the ball-burning current (EFO), ultrasonic power, and Z-axis displacement. The specific configuration subset is as follows:

[0082]

[0083] The Hardware Abstraction Layer instantiates the MC protocol driver plugin using the operating system's dynamic link library (dlopen).

[0084] Furthermore, within a preset sampling period, the target physical address set is... The process involves polling and reading data, where m is the total number of target physical addresses to be collected, and is a positive integer. In this example, within a 20-millisecond sampling period, data is collected from the target address set. Initiate polling. Further, scale the read binary dataset. After standardizing the engineering units, a structured numerical physical state dataset is obtained. ,in For the current moment The physical value. In this embodiment, the sampling time is a specific instant during system operation. For example, after reading the underlying binary data, it is scaled by a scaling factor. The transformed, structured, multidimensional physical state dataset is as follows:

[0085]

[0086] S12. Event-driven log feature capture:

[0087] Based on configuration set ,in A unique identifier for the matching rule. For the log path, It is a regular expression. To capture group indexes. The system listens for kernel file append events, and when a new log text stream is captured... At that time, the resident memory regular expression capture group library is called. Perform concurrent matching. For the unstructured process logs generated by the bonding machine's industrial control computer (IPC), the system extracts the configuration set. For example, the configuration for bonding pressure is as follows: The log parsing module continuously listens to the log directory by calling the underlying operating system's Inotify interface. When a file append event is captured and a new text stream is added, the module will respond accordingly. When the expression is "[INFO] BondForce: 45.2g, Time: 12ms", the regular expression capture group is called to extract the string "45.2".

[0088] and through type conversion function Process the data as floating-point numbers, transforming unstructured text into a structured set of log feature key-value pairs. ,in The extracted and transformed feature values ​​are, where, The total number of captured log feature key-value pairs, and must be a positive integer. For the first The feature values ​​are extracted and transformed by the matching rules. In this embodiment, a specific moment is captured when the above-mentioned log incremental event is recorded. For example, the final output is a set of structured log feature key-value pairs at that moment. .

[0089] The purpose of this step is to introduce a log parser into the hardware abstraction layer through the unified distribution of underlying capability configurations. Its beneficial effect is that, through dual-channel concurrent acquisition, it achieves deep integration of PLC discrete state data and continuous process data from equipment logs, thereby improving the integrity and high fidelity of equipment process characteristic data from the source.

[0090] S2. Temporal alignment and virtual label space mapping of multi-source data, including the following steps:

[0091] S21, when or The moment the data arrives at the hardware abstraction layer entry point, the system immediately calls a high-precision formula to obtain the current absolute timestamp. (For example, 1710324000123 milliseconds).

[0092] S22. The system constructs a global static mapping dictionary based on the ECD file. Let the system's virtual tag space be... In memory, it manifests as a time-series data queue supporting high-concurrency read and write operations. The system constructs a global static mapping dictionary in memory based on the ECD file. This dictionary defines a static mapping between physical identifiers and SECS / GEM entity attributes:

[0093] The PLC address is defined as follows:

[0094]

[0095] For log rules, the definition is as follows:

[0096]

[0097] In the formula: It serves as a globally unique virtual tag identifier. For SECS / GEM entity attributes such as VID / CEID, The target data type.

[0098] S23, System construction of multi-source mapping function ,when or Data in Upon arrival, the system queries Extract metadata and dynamic values With timestamp Combine them to generate standardized data tuples, as defined in the following formula:

[0099]

[0100] By using a global static mapping dictionary Retrieve the corresponding metadata in the middle:

[0101] Hit ,

[0102] Hit ,

[0103] Hit ,

[0104] Hit .

[0105] S24. Push the generated standardized data tuples into the virtual tag space. In the time-series queue.

[0106] The system integrates the aforementioned metadata and dynamic values ​​with a unified... Combining them, the standard tuples are generated as follows:

[0107]

[0108]

[0109]

[0110]

[0111] The above standard tuples are pushed into the virtual tag space. In the process of continuous operation, the PLC physical status data in S11 is continuously polled at a preset sampling period (e.g., 20ms), while the log features in S12 are asynchronously triggered and captured based on the underlying file append event. These two continuously generated data streams undergo time alignment and mapping processing in S2 and are then pushed into... In the time-series queue, a continuous multi-dimensional time-series data stream containing the complete historical operating trajectory of the device is accumulated in memory. This step completely shields the physical differences of the underlying data source (whether...). still This establishes a decoupling relationship between "data source identifier - virtual label - GEM entity", enabling multi-source data to be accessed simultaneously in the same time dimension. Absolute alignment and structured integration.

[0112] This step aims to eliminate timing errors from heterogeneous data sources by using a high-precision clock for unified marking; and to achieve mapping by establishing a virtual label space. Its beneficial effect is to completely decouple the underlying physical address and log rule identifiers from the upper-layer business logic, enabling device variables to be accessed and managed in a unified manner, greatly improving the system's scalability and compatibility when interfacing with heterogeneous devices.

[0113] S3. Multi-level route distribution based on hash mapping tables and a unified unpacking mechanism. Specifically, it includes the following steps:

[0114] S31. After the communication layer completes the TCP / IP session establishment and handshake with the host based on the HSMS protocol, it continuously listens for SECS-II messages.

[0115] S32. After receiving the SECS-II message, the protocol parsing module first reads and parses the header type field. In this embodiment, the message header is 10 bytes. If... (For example, the heartbeat packet Linktest) is directly responded to by the underlying link layer, without consuming computing power from the business layer. If Then proceed to the business routing and extract the business stream number. With Function Number .

[0116] S33, The system construction time complexity is... Global route hash table Calculate message feature key values And use a hash function to calculate the index. ;

[0117] S34, the system passes When addressing a processing entity, if a standard instruction is hit, the process is routed to the standard GEM processor; if a custom extended analysis protocol family is hit, the process is routed to the extended task manager. In this case, the message body remains in its unpacked, raw byte stream state.

[0118] For example, the host issues a remote control command to modify parameters. Feature key value The command distribution module within the communication layer utilizes... In the global routing hash table In the middle addressing, unpacked packets are routed to the standard GEM processor. If the host issues extended instructions defining the analysis script... Then it will route to the extended Task Manager. The host requests the device to execute a specified remote control command and carries the temperature parameter 190.0. These are instructions sent from the host to the device, used to define the analysis script task.

[0119] This step utilizes hash lookup logic to achieve extremely low-latency packet routing. Its benefit is ensuring the isolation of the standard control flow (Standard Stream) and the edge computing analysis flow (Stream 22) on the physical processing channel, so that when high-order data analysis tasks are introduced, the system will not block or affect the efficient response of the standard SECS / GEM protocol.

[0120] S4. Business parsing and hierarchical dynamic calculation based on state machine verification. For the results distributed in S3, the system concurrently executes differentiated dual-track logic, including the following steps:

[0121] S41 and SECS-II messages are uniformly parsed recursively. Whether it is the GEM processor or the extended task manager, the underlying GEM protocol parser is first called. The parser is based on the recursive descent algorithm to peel off the nested List and Item data structures in the message body layer by layer and extract the pure business parameters at the bottom layer (such as the VID number of the target register or the Python script string SCRIPTTEXT).

[0122] S42, Standard Control Flow Execution (Processing) After receiving the clean instruction parameters (the GEM processor extracts the target cleaver heating temperature parameter 190.0), the standard GEM processor calls the E30 state machine validity function. Perform verification, among which This indicates which control level (off-line, local, or remote) the device is currently in within the E30 state machine. The specific operation that the host attempts to perform on the device via the SECS-II message; It is a mapping matrix or rule engine that solidifies the lookup table of allowed and prohibited operations in the SEMI E30 standard; This is the function's output value. The system performs logical interception based on three control levels defined in the SECS / GEM specification, including OFF-LINE, LOCAL, and REMOTE states, as detailed below:

[0123] In OFF-LINE mode: As the lowest level of control, the system strictly restricts host access and operation. Except for replying to messages used to establish communication or request to change the device status from OFF-LINE to ON-LINE, the compliance verification module intercepts any other host commands received and replies directly with a communication interruption message.

[0124] In LOCAL (Local Online) status: As an intermediate level of control status, the host can access all information of the device, but the system will intercept the operation commands issued by the host to the device. At this time, all operations of the device can only be performed directly by the field operator.

[0125] When in REMOTE (remote online) state: This is the highest level of control, granting the host maximum control. The system does not restrict any functions of the host, allowing full access and physical control actions. In this embodiment, compliance is verified only when in REMOTE state, and then the temperature parameter 190.0 is converted to the underlying format and written to PLCD100.

[0126] S43, Edge Analysis Stream Execution (Processing Stream 22 Tasks): After receiving the pure parameters of the analysis task, the extended task manager sends the pure computing power task to the edge computing engine. The engine, based on the task definition instructions issued by the host (including...),... and ,in These are instructions sent from the host to the device to define predefined analysis tasks; These are instructions sent from the host to the device, used to define the analysis script task, and task binding instructions. ,in These are instructions sent from the host to the device to bind analysis tasks to device events, automatically executing tasks when an event is triggered. This establishes a dynamic mapping between analysis tasks and device-specific collected events (CEIDs); when the corresponding event is triggered, hierarchical computations are performed, including lightweight simple computations, predefined statistical process control, and dynamically parameterized script inference.

[0127] Lightweight and simple computation (based on) ):

[0128] For high-frequency sensor data, stateless basic statistical operations (such as min, max, average, stdev) are performed directly in memory to achieve threshold determination with extremely low latency.

[0129] Predefined Statistical Process Control (SPC) (based on) ):

[0130] The SPC algorithm library embedded in the engine is invoked, and based on the event collection (CEID) triggering mechanism, it performs tasks such as Xbar-R control chart anomaly detection or Cpk / Ppk process capability index calculation to conduct standardized monitoring of process stability.

[0131] Dynamic parameterized script inference (based on) ):

[0132] The engine loads the Python script sent from the host and the model weights passed in, and associates them through the input interface. A specific group of virtual tags is used to pull time-aligned feature data matrices from an internal sandbox. These tasks perform multivariate correlation analysis or AI inference and prediction. All three types of computing tasks are completely decoupled from the physical control flow of the device.

[0133] This embodiment uses dynamic parameterized inference (i.e., full-trajectory anomaly detection based on a 1D-CNN model) as an example for detailed explanation. When the internal event of the end of a single bonding action is triggered, the edge computing engine traces back to the historical time sequence according to the input interface and from... The time-series queue is used to extract complete batches of the bonding cycle (i.e., containing consecutive cycles). A slice of standard tuples for VT_TEMP (temperature), VT_POWER (ultrasonic power), VT_Z_POS (Z-axis trajectory), and VT_FORCE (bonding pressure) strictly aligned at each sampling time point.

[0134] In this forward reasoning process, the periodic parameters used to characterize the mechanical and electrical states of the equipment (such as VT_TEMP temperature, VT_POWER ultrasonic power, and VT_Z_POS Z-axis trajectory) originate from high-frequency polling of the underlying PLC. Meanwhile, the parameters characterizing the transient feedback of key processes (such as VT_FORCE bonding pressure) originate from real-time regularized capture of the industrial computer's local unstructured logs.

[0135] For the acquisition path of multi-source input data during the inference process of this model, please refer to [link / reference]. Figure 5 .

[0136] These are periodic parameters used to characterize the mechanical and electrical fundamental state of the equipment, including: VT_TEMP (cutting blade temperature), VT_POWER (ultrasonic power), and VT_Z_POS (Z-axis trajectory), which are acquired in real time by the PLC's high-frequency polling channel. Meanwhile, extremely short-time parameters characterizing the transient feedback of key processes include: VT_FORCE (actual bonding pressure), which is captured by the log parsing extraction channel through an event-driven mechanism in the unstructured logs of the industrial control computer using regular expressions.

[0137] The engine integrates these rich dynamic process features in memory and reconstructs them into a form of size [size missing]. ( This is a two-dimensional time-series input tensor (4 being the number of sampling points within a single bonding cycle and 4 being the feature dimension). The engine then feeds this high-dimensional time-series tensor into a lightweight one-dimensional convolutional neural network (1D-CNN) deployed on the NPU for forward inference. It outputs the real-time probability of solder debonding anomalies in the current process. .

[0138] In this invention, the 1D-CNN model sequentially passes through a one-dimensional convolutional layer, a max-pooling layer, and a depthwise convolutional layer, enabling it to fully traverse and process the complete trajectory from the aforementioned four parameters. Finally, after the feature data is synthesized and mapped by a fully connected layer, the output layer outputs the probability of solder debonding anomaly in the current process in real time. .

[0139] The purpose of this step is to achieve deep decoupling and dual-track parallel processing between equipment control logic and edge analytics computing power. Its beneficial effect is that by extending the analytics task messaging mechanism (Stream 22), the host computer can break through static configuration, dynamically distribute computing tasks (such as models or scripts), and independently execute data analysis at the equipment edge, significantly improving the intelligent capabilities of equipment anomaly detection and process optimization.

[0140] S5. Multi-state reporting of analysis results and millisecond-level safety closed-loop control under abnormal conditions. The system generates closed-loop feedback based on the execution result of S4, including the following steps:

[0141] S51, Standard Control Flow Feedback: Based on the SECS / GEM specification, such as for... If the physical control executes the instruction successfully, a standard reply message is encapsulated and reported. If the underlying execution fails or the E30 state machine verification function fails... If the return is invalid, the system will encapsulate the corresponding exception response message or interrupt communication message and report it to the host.

[0142] S52, Edge computing flow feedback, including:

[0143] Regular reporting occurs when the edge engine's inference results are reported. Within the preset safety threshold range During the period, among which, and These are the lower and upper limits of the preset safety range, respectively. The engine encapsulates the results into an extended event report message and reports it, and the host replies with an acknowledgment message.

[0144] Abnormal closed-loop interception, when the calculation result The device control layer will short-circuit the regular communication reporting link. (This occurs during the call.) After verifying that the current device mode allows intervention, the system short-circuits the SECS roaming link, and the engine directly generates the optimal blocking or fine-tuning command. Through the hardware abstraction layer The atom is written into the predefined protection register of the underlying bonding machine PLC, thereby completing local dynamic intervention within milliseconds and avoiding process accidents caused by reliance on the plant-level network. Subsequently, the engine encapsulates the high-dimensional anomaly characteristics and analysis results into an extended report (such as...). ,in (This refers to the device sending task results to the host asynchronously and reporting them to the host.)

[0145] This step completes the closed-loop feedback from computation to physical control. Its beneficial effect is that by performing data analysis and abnormal control decisions directly at the device edge, it avoids the network latency caused by uploading data to the cloud and then issuing commands when transient anomalies occur, reducing reliance on cloud computing resources and thus achieving millisecond-level local security protection, greatly improving real-time control capabilities.

[0146] Please refer to the following: Figure 6 A communication system is provided, which is applied to a SECS / GEM communication method that supports log parsing and dynamic computing tasks. The system includes an Edge Control Unit (ECU), deployed between a factory host system (such as MES / EAP) and underlying semiconductor devices, thereby enabling heterogeneous data acquisition, standardized data modeling, and edge AI computing functions.

[0147] Furthermore, the device edge control unit is preferably implemented using various high-performance industrial-grade embedded platforms. This embodiment uses an embedded core board based on the RK3588 SoC. The RK3588 is a 64-bit ARM architecture multi-core processor, whose integrated high-performance NPU unit is dedicated to accelerating the edge feature inference task described herein. The device edge control unit hardware includes at least one gigabit Ethernet interface, multiple RS232 / RS485 serial communication interfaces, and local memory. The device edge control unit software architecture adopts a layered design, from bottom to top including: Hardware Abstraction Layer (HAL), Standardized Information Model Layer, Device Control Layer, and Communication Layer, specifically:

[0148] The hardware abstraction layer, configured to perform high-frequency concurrent acquisition of heterogeneous data and standardized feature extraction, includes a PLC communication module and a log parsing module, wherein:

[0149] The PLC communication module is used to establish communication connections with heterogeneous PLC devices through different protocols, and to read device register data and write control instructions.

[0150] The log parsing module is configured to perform real-time monitoring of the local log files in the host device environment, and trigger the log parsing algorithm when an incremental log event is captured to extract process feature data from unstructured text data.

[0151] In one embodiment, the hardware abstraction layer immediately appends a high-precision timestamp to all collected data the instant it captures the underlying data, and achieves time series alignment between PLC data and log data through a time synchronization mechanism. Its core function lies in achieving deep fusion and acquisition of underlying discrete PLC status data and continuous equipment log process data by introducing a log parser. The beneficial effects are a significant improvement in the integrity and high fidelity of equipment process characteristic data, and ensuring absolute consistency of multi-source heterogeneous data in subsequent analysis.

[0152] The standardized information model layer is configured to perform time alignment and mapping of multi-source data. It constructs a unified data abstraction model for the device and decouples upper-layer business logic from the underlying physical address by establishing a virtual tag space. This layer includes a mapping management module and a dynamic configuration module. This layer builds and maintains the system's virtual tag space in memory. With global static mapping dictionary .

[0153] The mapping management module is used to establish a three-layer mapping relationship of "GEM variable - virtual tag - physical address or log feature".

[0154] The dynamic configuration module is used to parse the device capability description file (ECD) and generate device data mapping rules based on the description file.

[0155] In one embodiment, the standardized information model layer dynamically generates underlying device data parsing rules based on the Equipment Capability Description File (ECD). These parsing rules include PLC data polling parsing rules and log feature parsing rules. The rules are then distributed to the PLC communication module and the log parsing module in the hardware abstraction layer to achieve automatic collection and parsing of multi-source heterogeneous device data.

[0156] In one embodiment, the underlying physical status data collected by the PLC communication module and the unstructured log data extracted by the log parsing module are both uniformly mapped to corresponding virtual tag values ​​and jointly participate in the data organization of the GEM variable space. The role of this layer is to shield the communication differences of physical devices from above. By establishing a unified virtual tag space, the complex underlying physical addresses and business logic are completely decoupled, enabling device variables to be accessed and managed in a unified manner. This not only greatly improves the scalability of the system when connecting to heterogeneous machines, but also realizes the deep integration of multi-source device data in the same information model.

[0157] The device control layer, configured to perform device control logic management and edge computing task processing, includes a standard GEM processor and an edge computing engine, wherein the standard GEM processor includes:

[0158] State Management Module: Used to manage the operating and communication states of devices based on finite state machines.

[0159] Compliance verification module: Configured to verify the device status before executing device control commands. It performs strict interception verification of the device status according to the SEMI E30 specification before executing device control commands. When in OFF-LINE, it only allows communication establishment. Or change the device status from OFF-LINE to ON-LINE. The message, and the rest of the message. Interruption response. When in LOCAL mode, only information access requests are allowed, while control operations are denied. When in REMOTE mode, no host functionality is restricted; all requests are allowed. Used to establish a communication request, typically after device startup or a sudden interruption of communication. It is sent periodically until communication is established. Used by the host to request the device to enter the ON-LINE state. It is a standard rejection action used by the device control layer to proactively block and respond to unauthorized business commands when performing strong security and compliance checks in the OFF-LINE state.

[0160] The edge computing engine is used to perform data analysis tasks at the device edge. It supports receiving tasks issued by the host through extended analysis task messages (such as custom Stream22 messages) and dynamically loading algorithm models locally to execute edge intelligent analysis, anomaly detection, or predictive maintenance algorithms.

[0161] The role of this layer is to achieve deep decoupling and dual-track parallel processing of device control logic and edge analytics computing power. Its beneficial effect is that by processing Standard Stream and Stream 22 separately, the device can directly perform high-intensity data analysis and control decisions at the edge, effectively reducing the dependence of large amounts of raw data on cloud computing resources and network bandwidth. This significantly reduces system communication latency and improves millisecond-level local real-time control capabilities under transient anomalies.

[0162] The communication layer is configured to enable SECS / GEM protocol communication between the device edge control unit and the host system. The communication layer supports both the standard SECS / GEM communication protocol and extended analysis task messages to enable dynamic configuration of device edge computing tasks by the host. It includes a session management module, a protocol parsing module, and a command distribution module.

[0163] The session management module is used to establish a long TCP / IP connection with the host system based on the HSMS protocol and to maintain the communication session state.

[0164] The protocol parsing module is used to parse SECS-II messages and generate corresponding GEM instruction objects.

[0165] The command distribution module is used to distribute commands to the device control layer for execution based on the instruction type.

[0166] This layer's role is to build high-concurrency, highly compatible internal and external interaction channels. Its beneficial effects include enabling the host not only to perform low-level link state maintenance (such as heartbeat packet command interaction) and issue routine device status monitoring and service control commands (such as Standard Stream service flow commands) based on the SECS / GEM protocol, but also to break through the static limitations of traditional protocols, dynamically issuing high-order computing tasks (such as Stream 22 extended commands containing model weights or inference scripts) and executing them in real time at the device edge. This completely breaks the limitations of static configuration in traditional communication gateways, greatly improving the host's ability to detect process anomalies and dynamically schedule computing power.

[0167] An electronic device includes: a processor, a memory, a computer program, and a computer-readable storage medium. The computer program is stored on the computer-readable storage medium. The computer program is stored in the memory and configured to be executed by the processor to implement the aforementioned SECS / GEM communication method supporting log parsing and dynamic computing tasks. When executed by the processor, the computer program performs the following steps: receiving SECS-II messages from a host system; recursively parsing the messages using a protocol parsing module to extract pure service parameters; distributing the parsed instructions to a standard GEM processor or an extended task manager according to the instruction type; executing device control instructions by the standard GEM processor, or assigning data analysis tasks to an edge computing engine; encapsulating the execution results into SECS-II messages and feeding them back to the host system, or triggering local millisecond-level security closed-loop control in abnormal conditions. By implementing the above methods, electronic devices can achieve efficient acquisition, standardized modeling, and edge intelligent analysis of data from heterogeneous devices. This effectively solves technical problems such as outdated communication interfaces of old equipment, strong coupling between PLC register addresses and business logic, ineffective utilization of equipment log data, and high latency in cloud computing that makes it difficult to support dynamic computing tasks. This significantly improves the intelligence level and process control capabilities of semiconductor manufacturing equipment.

[0168] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0169] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A SECS / GEM communication method supporting log parsing and dynamic computation tasks, characterized in that: The method includes the following steps: S1. High-frequency concurrent acquisition of heterogeneous data and standardized feature extraction; S2. Temporal alignment and virtual label space mapping of multi-source data; S3. Multi-level route distribution based on hash mapping table and unified unpacking mechanism; S4. Business parsing and hierarchical dynamic calculation based on state machine verification; S5. Multi-state reporting of analysis results and millisecond-level safety closed-loop control under abnormal conditions.

2. The SECS / GEM communication method supporting log parsing and dynamic computation tasks as described in claim 1, characterized in that: Data acquisition in S1 is based on the association configuration set of the underlying data source and the virtual tag system predefined in the system parsing device capability description file (ECD), including the PLC configuration set. With log configuration set It includes the following steps: S11, PLC physical status data polling: Based on PLC configuration set And instantiate the corresponding underlying driver using the dynamic link library; Within a preset sampling period, the target physical address set Perform polling reads; The read binary dataset is scaled by a factor. After standardizing the engineering units, a structured numerical physical state dataset is obtained. ; S12. Event-driven log feature capture: The system listens for kernel file append events, and when it captures the newly added log text stream... At that time, the resident memory regular expression capture group library is called. Perform concurrent matching; Text fragments are extracted using regular expression capture groups and then converted using a type conversion function. The process transforms unstructured text into a structured set of log feature key-value pairs. .

3. The SECS / GEM communication method supporting log parsing and dynamic computation tasks as described in claim 2, characterized in that: The temporal alignment and virtual label space mapping of multi-source data in S2 includes the following steps: S21, when or The system obtains the current absolute timestamp the moment the data arrives at the hardware abstraction layer entry point. ; S22. The system constructs a global static mapping dictionary based on the ECD file. : The PLC address is defined as follows: The definition of a log rule is as follows: In the formula: It serves as a globally unique virtual tag identifier. For SECS / GEM entity attributes, The target data type; S23, System construction of multi-source mapping function ,when or Data in Upon arrival, the system queries Extract metadata and dynamic values With timestamp Combine them to generate standardized data tuples, as defined in the following formula: S24. Push the generated standardized data tuples into the virtual tag space. In the time-series queue.

4. The SECS / GEM communication method supporting log parsing and dynamic computation tasks as described in claim 3, characterized in that: The multi-level route distribution in S3 based on hash mapping tables and a unified unpacking mechanism includes the following steps: S31. After the communication layer completes the TCP / IP session establishment and handshake with the host based on the HSMS protocol, it continuously listens for SECS-II messages. S32. After receiving the message, parse the header type field. If the judgment is The underlying link layer responds directly; if Extract the business stream number With Function Number ; S33, The system construction time complexity is... Global route hash table Calculate message feature key values And use a hash function to calculate the index. ; S34, the system passes The addressing processing entity is routed to the standard GEM processor if a standard instruction is hit, and to the extended task manager if a custom extended analysis protocol family is hit.

5. The SECS / GEM communication method supporting log parsing and dynamic computation tasks as described in claim 4, characterized in that: In S4, business parsing and hierarchical dynamic calculation based on state machine verification, for the results distributed in S3, the system concurrently executes differentiated dual-track logic, including the following steps: S41 and SECS-II messages are uniformly recursively parsed by calling the underlying GEM protocol parser. The parser is based on a recursive descent algorithm to peel off the nested List and Item data structures in the message body layer by layer and extract the pure business parameters at the underlying level. S42. Standard control flow execution: After receiving the clean instruction parameters, the standard GEM processor calls the E30 state machine validity function. Verification is performed. The system uses logical interception based on the three control levels defined in the SECS / GEM specification, including OFF-LINE, LOCAL, and REMOTE states; S43. Edge analysis stream execution: After receiving the parsed pure parameters of the analysis task, the extended task manager sends the pure computing power task to the edge computing engine for hierarchical computing, including lightweight simple computing, predefined statistical process control, and dynamic parameterized script inference.

6. The SECS / GEM communication method supporting log parsing and dynamic computation tasks as described in claim 5, characterized in that: In S5, the analysis results are reported in multiple states and millisecond-level safety closed-loop control is implemented under abnormal conditions. The system generates closed-loop feedback based on the execution results of S4, including the following steps: S51, Standard Control Flow Feedback: According to the SECS / GEM specification, if the physical control execution is successful, a standard response message is encapsulated and reported; if the underlying execution fails or the E30 state machine verification function fails... If the return is invalid, the system will encapsulate the corresponding exception response message or interrupt communication message and report it to the host; S52, Edge Computing Flow Feedback, including: Regular reporting occurs when the edge engine's inference results are reported. Within the preset safety threshold range During this time, the engine encapsulates the results into an extended event report message and reports it, and the host replies with an acknowledgment message; Abnormal closed-loop interception, when the calculation result The device control layer will short-circuit the regular communication reporting link; during the call After verifying that the current device mode allows intervention, the engine directly generates the optimal blocking or fine-tuning command. Through the hardware abstraction layer The atom is written into the predefined protection register of the underlying PLC.

7. A communication system, wherein the communication system applies the SECS / GEM communication method supporting log parsing and dynamic computation tasks as described in any one of claims 1 to 6, characterized in that, The system includes a device edge control unit deployed between the host system and the underlying devices; the device edge control unit includes: The hardware abstraction layer is configured to perform high-frequency concurrent acquisition of heterogeneous data and standardized feature extraction, including a PLC communication module and a log parsing module; The standardized information model layer is configured to perform time alignment and mapping of multi-source data. It builds a unified data abstraction model for devices and decouples the upper-layer business logic from the lower-layer physical address by establishing a virtual label space. It includes a mapping management module and a dynamic configuration module. The device control layer is configured to perform device control logic management and edge computing task processing, including a standard GEM processor and an edge computing engine; The communication layer is configured to enable SECS / GEM protocol communication between the device edge control unit and the host system, including a session management module, a protocol parsing module, and a command distribution module.

8. A communication system as described in claim 7, characterized in that: The standard GEM processor includes: The state management module is configured to manage the device's operating and communication states based on a finite state machine. The compliance verification module is configured to verify the device status before executing device control commands.

9. An electronic device, characterized in that, include: A processor, a memory, and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement a SECS / GEM communication method supporting log parsing and dynamic computation tasks as described in any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When executed by the processor, the program implements the SECS / GEM communication method as described in any one of claims 1 to 6, which supports log parsing and dynamic computation tasks.

Images