A method and system for evaluating the adaptation of skill talent training and manufacturing transformation

Abstract

This invention discloses a method and system for assessing the adaptation of skilled personnel training to manufacturing transformation, relating to the field of data management technology. The method includes the following steps: S1, acquiring event stream data and performing timeline correction and normalization; S2, constructing a multi-device state interlock diagram to generate standard event sequences and operational event sequences; S3, evaluating the deviation of the operational event sequences and generating deviation type labels; S4, determining the current event segment and generating adaptation deviation labels. This invention effectively improves the accuracy of the adaptation assessment between skilled personnel training process data and manufacturing transformation collaborative equipment data, solving the problem that evaluation based solely on single-device logs and task completion results cannot identify repeated triggering and interlock dependency violations in multi-device collaborative tasks.

Description

Technical Field

[0001] This invention relates to the field of data management technology, specifically to a method and system for assessing the adaptation of skilled personnel to manufacturing transformation. Background Technology

[0002] As intelligent manufacturing production lines evolve from single-equipment operation to collaborative operation of robots, CNC machine tools, PLCs, testing terminals, and MES work orders, operational event data during skills training needs to be correlated with the collaborative status of equipment after manufacturing transformation. Existing technologies often rely on single-equipment logs, training scores, or task completion markers for evaluation. However, existing systems lack modeling of the dependencies between allow signals, wait signals, completion signals, and abnormal signals, and cannot perform structured analysis of the sources of deviation between operational event sequences and standard event sequences. This results in a lack of data support and traceability for adaptation evaluation results.

[0003] For example, the invention patent with publication number CN111738586B discloses a talent assessment method and apparatus, including: obtaining information about the talent to be assessed; the information about the talent to be assessed includes at least one of basic information, learning experience, work experience, skill information or self-evaluation; obtaining assessment results based on the information about the talent to be assessed and a pre-trained talent assessment model; the assessment results include quantified work performance; quantified work performance is used to indicate the predicted value of the quantified work performance of the talent corresponding to the information of the talent to be assessed.

[0004] For example, the invention patent with announcement number CN118228947B discloses an intelligent analysis method and system for enterprise digital transformation, including: analyzing relevant information of the target enterprise to determine its type level and type coefficient, then monitoring the basic status information of the target enterprise over a period of time to evaluate the transformation process, thereby obtaining a transformation process evaluation value, combining the type coefficient and the transformation process evaluation value to calculate and analyze the actual transformation value, and introducing reference case analysis to obtain the target transformation value. If the actual transformation value of the target enterprise is equal to or greater than the target transformation value, the transformation status of the target enterprise over that period of time is judged as normal; otherwise, it is judged as abnormal. Based on the judgment of abnormality, the transformation strategy of the target enterprise is adjusted and processed, thereby helping enterprises identify and solve key problems in digital transformation and ensuring the smooth progress of enterprise digital transformation.

[0005] In existing technologies, while existing systems may perform well in single-device training, they may encounter problems such as waiting timeouts, erroneous preemption, duplicate command transmissions, and failure to release device resources in multi-device collaborative tasks. Furthermore, systems often evaluate capabilities based on single-device logs, failing to identify whether they truly understand the state interlocking relationships of collaborative units.

[0006] Therefore, in response to the above problems, there is an urgent need for a method and system for assessing the adaptation of skilled personnel training to manufacturing transformation. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a method and system for assessing the adaptation of skilled personnel training to manufacturing transformation. This solves the problem that relying solely on single-device logs and task completion results for evaluation fails to identify repeated triggering and interlock dependency violations in multi-device collaborative tasks.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a method for assessing the adaptation of skilled personnel training to manufacturing transformation, comprising: S1, acquiring event stream data and performing time axis correction, noise reduction, and normalization on the event stream data to generate a unified event stream; S2, traversing the unified event stream, constructing a multi-device state interlocking diagram, calculating a blocking risk index for each node, and generating a standard event sequence and an operational event sequence; S3, based on the operational event sequence, assessing the deviation of the operational event sequence, filtering out deviation event segments and matching event segments, and generating deviation type labels; S4, based on the deviation event segments and matching event segments, calculating an adaptation deviation judgment value, judging the current event segment, and generating an adaptation deviation label.

[0009] Further, the process of acquiring event stream data and performing time-axis correction, denoising, and normalization to generate a unified event stream is as follows: The process involves collecting data on state transitions, action commands, signal triggers, waiting times, abnormal events, and parameter adjustments occurring in collaborative tasks involving robots, CNC machine tools, PLCs, inspection terminals, and MES work orders. This generates event stream data, which includes device number, state number, action number, trigger signal number, feedback signal number, operating entity number, start time, end time, duration, abnormal marker, and parameter values. The delay of a single event is obtained by subtracting the device's local time from the gateway's received time. The median of the delays of the M most recent events from the same device is then used as the current correction delay to uniformly map the events to the global system time. A sliding window first-order derivative constraint method is used to calculate the rate of state change at each time point for denoising. When the rate of state change exceeds a threshold defined by the device specifications, the abnormal peak value is replaced with the previous sampled value, generating a smooth event sequence. Finally, the event stream data is normalized using linear scaling to generate a unified event stream.

[0010] Furthermore, the specific process of traversing the unified event stream and constructing a multi-device state interlocking graph is as follows: Traverse the unified event stream; for each event, determine the corresponding node based on the device number and state number; map the triggering order and dependencies between nodes in the event sequence to edges of the graph; simultaneously, calculate the waiting time, signal consistency, and number of exceptions for each node, and populate the node attributes; obtain the waiting time by subtracting the end time of the previous state from the start time of the node state; calculate the proportion of matching times within a window to the total number of times as signal consistency; obtain the number of exceptions by counting the abnormal events triggered by the node in its state; output... A state-interlocked graph consists of a set of nodes and a set of edges, with each node and edge carrying attribute information. A node dependency matrix is ​​constructed based on the state-interlocked graph: the rows and columns of the matrix correspond to all nodes, and each element represents the dependency type from node i to node j. Dependency types include allowed signal dependency, waiting signal dependency, and abnormal signal dependency. A basic weight value is assigned to each dependency. If there are multiple dependencies between node i and node j, the maximum value of all dependency weights is taken as the initial topological connection strength. The initial topological connection strength is multiplied by the node importance coefficient to obtain the final topological connection strength.

[0011] Furthermore, the specific process for calculating the congestion risk index for each node is as follows: Divide the product of the waiting times of each pair of nodes by the sum of the difference in node signal consistency and 1 to obtain the node coupling congestion baseline value; subtract the topology connection strength from the constant term 1 to obtain the topology difference value; square the topology difference value to obtain the topology difference squared term; multiply the topology difference squared term by the topology sensitivity adjustment coefficient to obtain the topology influence term; perform a negative exponential operation on the topology influence term to obtain the exponential decay coefficient; multiply the node coupling congestion baseline value by the exponential decay coefficient to obtain the nonlinear congestion contribution of the node pair; take the maximum value of the nonlinear congestion contribution calculated for all node pairs to obtain the congestion risk index; combine the waiting time, signal consistency, and number of anomalies by column to form the feature vector of each node; arrange the feature vectors of all nodes sequentially according to their node numbers to generate a node feature matrix; the rows of the matrix represent different nodes, and the columns represent the node attribute dimensions.

[0012] Further, the specific process for generating standard event sequences and operation event sequences is as follows: The blocking risk index is jointly stored with the node feature matrix to form a node-level blocking risk feature set; nodes are sorted according to the blocking risk index to generate a node priority list; based on the device list for the adaptation assessment task, the states and actions that each device needs to execute are listed and arranged in the order required by the task to form an initial event list; combined with the dependencies in the state interlocking graph, the event sequence is topologically sorted, and the expected trigger time and completion time for each event are calculated simultaneously and represented by a unified system time; all events are sequentially generated into a structured array, with each row containing the device number, state number, action number, permitted signal number, and time information to form a standard event sequence; the node priority list is compared with the unified event stream and the standard event sequence, and the corresponding operation event record is located in the unified event stream based on the node number; by retrieving events in the operation log of the operation subject whose device number and state number match the priority list, operation events involving blocking risk nodes are extracted to form an operation event sequence.

[0013] Furthermore, the specific process for evaluating the deviation of the operation event sequence based on the operation event sequence is as follows: Locate the corresponding event in the operation log and standard event sequence according to the event number and status number, then calculate the difference in trigger time between the two, normalize the time difference to the range of zero to one, and obtain the time deviation value; simultaneously compare whether the action numbers are consistent; if they are inconsistent, assign a value of one, otherwise assign zero; combine the time deviation value and the action deviation value, and calculate the matching distance through a weighted average; traverse each event number and status number in the standard event sequence, and check whether a corresponding event exists in the operation event sequence; if a corresponding event does not exist in the operation event sequence or the event order is incorrect, record the event. For unmatched key events, multiply the matching distance of each event in the operation event sequence by the matching distance scaling factor, invert the negative sign, perform exponential operation, add 1, and take the reciprocal to obtain the weighted matching value. Then, exponentiate the weighted matching value of each event with the corresponding event weight coefficient to obtain the event weighting term. Assign the value 1 to each unmatched key event in the operation event sequence, perform negative exponential operation, add 1, and take the reciprocal to obtain the unmatched influence factor. Accumulate all unmatched influence factors, multiply by the unmatched event penalty coefficient, and obtain the total unmatched event value. Add the product of the event weighting terms to the total unmatched event value, take the negative value, and add a constant term 1 to obtain the operation event sequence deviation.

[0014] Furthermore, the specific process of filtering deviation event segments and matching event segments, and generating deviation type labels is as follows: For the current event, when the deviation of the operation event sequence is greater than the current deviation threshold, instead of directly marking it as an anomaly, the process is centered on the current event, backtracking to the most recent event with a deviation less than or equal to the current deviation threshold and a matching status number, and then scanning backward to the first event whose deviation recovers to below the current deviation threshold, thus capturing the deviation event segment; when the deviation of the operation event sequence is less than or equal to the current deviation threshold, the current event is written into the stable event buffer. If M consecutive events are all less than or equal to the current deviation threshold, the event segment is recorded as a matching event segment, and deviation type labels are generated through component proportion comparison and rule mapping methods.

[0015] Furthermore, the specific process for calculating the adaptation deviation judgment value based on the deviation event segment and the matching event segment is as follows: Read the deviation degree of each operation event sequence within the same deviation event segment from the deviation event segment table, and calculate the maximum and average values ​​respectively to obtain the peak deviation degree and the average deviation degree of the deviation event segment; read the deviation degree of each operation event sequence within adjacent matching event segments from the matching event segment table, and calculate the average value to obtain the average deviation degree of the matching event segment; count the number of events within the deviation event segment to obtain the number of events in the deviation event segment; determine the number of events within the deviation event segment that do not satisfy the dependency relationship according to the state interlock diagram, and divide it by the total number of events in the deviation event segment to obtain the interlock dependency violation ratio; count the number of unmatched events according to the standard event sequence, and divide it by the number of events in the standard event sequence to obtain the ratio of unmatched key events; count the number of events with inconsistent action numbers within the deviation event segment. The deviation percentage is calculated by dividing the number of events in the deviation event segment by the total number of events in the deviation event segment. The percentage of repeated event numbers within the deviation event segment is calculated and divided by the total number of events in the deviation event segment to obtain the percentage of repeated operations. The peak deviation is added to the average deviation of the deviation event segment and then added to the zero-prevention item to obtain the deviation intensity item. The average deviation of the matching event segment is added to the zero-prevention item to obtain the matching benchmark item. The deviation intensity item is divided by the matching benchmark item to obtain the deviation amplification ratio. The number of events in the deviation event segment is incremented by one and then the natural logarithm is taken to obtain the event amplification item. The percentage of interlock dependency violations is incremented by one to obtain the interlock amplification item. The percentage of unmatched key events, the percentage of action number deviation, and the percentage of repeated operations are added together and then square-rooted to obtain the deviation aggregation item. The deviation amplification ratio, the event amplification item, the interlock amplification item, and the deviation aggregation item are multiplied sequentially to obtain the adaptation deviation judgment value.

[0016] Furthermore, the specific process of determining the current event segment and generating adaptation deviation labels is as follows: The adaptation deviation judgment value is compared with the adaptation deviation threshold in real time. When the adaptation deviation judgment value is less than or equal to the adaptation deviation threshold, the current event segment is written into the adaptation pass event segment table. When the adaptation deviation judgment value is greater than the adaptation deviation threshold, the current event segment is written into the adaptation deviation event segment table, and component judgment is performed. Adaptation deviation labels are generated according to the component judgment results. Adaptation deviation labels include interlock violation labels, critical event missing labels, action mismatch labels, repeated triggering labels, and trigger delay labels. When an adaptation deviation event segment meets the generation conditions of more than two deviation labels at the same time, the labels are arranged in the order of interlock violation label, critical event missing label, action mismatch label, trigger delay label, and repeated triggering label, and a complex adaptation deviation label is generated.

[0017] Furthermore, a second aspect of the present invention provides a skills training and manufacturing transformation adaptation assessment system, applied to a skills training and manufacturing transformation adaptation assessment method, comprising: a data acquisition and preprocessing module, used to acquire event stream data and perform time axis correction, noise reduction, and normalization processing on the event stream data to generate a unified event stream; a state modeling and blockage identification module, used to traverse the unified event stream, construct a multi-device state interlocking diagram, calculate a blockage risk index for each node, and generate a standard event sequence and an operation event sequence; a collaborative task analysis module, used to evaluate the deviation of the operation event sequence based on the operation event sequence, filter out deviation event segments and matching event segments, and generate deviation type labels; and an adaptation assessment module, used to calculate an adaptation deviation judgment value based on the deviation event segments and matching event segments, and judge the current event segment to generate an adaptation deviation label.

[0018] The present invention has the following beneficial effects: (1) This invention constructs a multi-device state interlocking graph, representing the device state as nodes, and representing allowed signal dependencies, waiting signal dependencies, and abnormal signal dependencies as edges. It also records the waiting time, signal consistency, and number of abnormalities for each node, thus transforming the device interlocking relationship in collaborative tasks from a single log record into computable graph structure data. This solves the problem that existing systems cannot express the state dependency relationship between robots, CNC machine tools, PLCs, detection terminals, and MES work orders.

[0019] (2) This invention distinguishes between allowed signal dependencies, waiting signal dependencies and abnormal signal dependencies between nodes by using a node dependency matrix and a topology connection strength calculation method. It also calculates the node importance coefficient by combining the node out-degree, in-degree, waiting time and abnormal number of terms, so that different dependencies in the state interlock graph can be quantified and processed, avoiding processing all device state relationships with the same weight.

[0020] (3) The present invention intercepts the deviation event segment by means of deviation threshold comparison, forward backtracking and backward scanning, and generates deviation type labels based on the proportion of trigger time deviation, the proportion of action number deviation, the proportion of unmatched key events, the proportion of repeated operations and the proportion of interlock dependency violation, so that a single deviation event can be expanded into a continuous event segment for parsing, avoiding isolated marking of only single point anomalies.

[0021] (4) This invention reads the deviation event segment table and the matching event segment table, and further calculates the adaptation deviation judgment value to generate adaptation deviation label and matching adaptation deviation label, so that the operation event data in the process of cultivating skilled talents can form a corresponding relationship with the collaborative equipment requirements in the manufacturing transformation scenario, thereby outputting a structured adaptation classification data table.

[0022] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0023] Figure 1 This is a flowchart of a method for assessing the adaptation of skilled personnel to manufacturing transformation, as described in this invention. Figure 2 This is an architecture diagram of a skills training and manufacturing transformation adaptation assessment system according to the present invention; Figure 3 This is a matrix diagram showing the interlocking relationship between the device states of the present invention. Figure 4 This is a three-dimensional scatter plot showing the correlation between node blocking risk and the present invention. Figure 5 This is a stacking analysis diagram of the off-event segment components of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figures 1-5This invention provides a technical solution: a method for assessing the adaptation of skilled personnel training to manufacturing transformation, comprising: S1, acquiring event stream data and performing time axis correction, denoising, and normalization on the event stream data to generate a unified event stream; S2, traversing the unified event stream, constructing a multi-device state interlocking diagram, calculating a blocking risk index for each node, and generating a standard event sequence and an operation event sequence; S3, based on the operation event sequence, assessing the deviation of the operation event sequence, filtering out deviation event segments and matching event segments, and generating deviation type labels; S4, based on the deviation event segments and matching event segments, calculating an adaptation deviation judgment value, judging the current event segment, and generating an adaptation deviation label.

[0026] Specifically, the process of acquiring event stream data and performing timeline correction, noise reduction, and normalization on the event stream data to generate a unified event stream is as follows: The system collects data on state transitions, action commands, signal triggers, waiting times, abnormal events, and parameter adjustments occurring during collaborative tasks involving robots, CNC machine tools, PLCs, inspection terminals, and MES work orders via an edge gateway at fixed intervals or through event triggering. This data generates event stream data, which includes device number, status number, action number, trigger signal number, feedback signal number, operator number, start time, end time, duration, anomaly flag, and parameter values. The operator number identifies the evaluated object participating in the skills training process and can be associated with a specific operator or agent in training for skills analysis and performance evaluation. The delay of a single event is obtained by subtracting the device's local time from the gateway's received time. The median of the delays of the M most recent events from the same device is taken as the current correction delay. This median is used to compensate for the deviation between the device's local clock and the global system clock, uniformly mapping events to the global system time. M represents the number of most recent events from the same device used to calculate the median correction delay. The typical value range is 5 to 20, with the specific value determined based on network stability and clock drift characteristics. Using the sliding window first-order derivative constraint method, the difference is calculated in the neighborhood before and after each sampling point to approximate the first-order derivative. The state change rate at each time point is calculated for noise reduction. When the state change rate exceeds the threshold defined by the device specifications, the abnormal peak value is replaced with the previous sample value to generate a smooth event sequence. The event stream data is normalized by linear scaling, mapping the minimum value to 0 and the maximum value to 1. The ratio of the signal consistency index is mapped to 0 to 1. A unified event stream is generated.

[0027] In this implementation scheme, an edge gateway is used to uniformly collect state transitions, action commands, signal triggers, waiting times, abnormal events, and parameter adjustments in multi-device collaborative tasks. An operation entity number is introduced, enabling the event stream data to be traceable simultaneously in terms of device origin, state process, operation origin, and time process. This allows heterogeneous events generated by robots, CNC machine tools, PLCs, inspection terminals, and MES work orders to form a unified event stream with a consistent structure, providing a stable and computable data foundation for constructing state interlock diagrams, calculating congestion risk indices, and assessing operational event sequence deviations.

[0028] Specifically, the process of traversing the unified event flow and constructing a multi-device state interlocking graph is as follows: Traverse the unified event stream. For each event, determine the corresponding node based on the device number and state number. The node represents the operational stage of the device in a specific state. Map the triggering order and dependencies between nodes in the event sequence to edges of a graph. Edges represent the control or signal transmission relationship from the source node to the target node. Simultaneously, count the waiting time, signal consistency, and number of anomalies for each node and populate the node attributes. The waiting time is obtained by subtracting the end time of the previous state from the start time of the node state. The waiting time reflects the idle interval of the node before obtaining execution permission. Count the number of matches within a window as the proportion of the total number of matches as signal consistency. Count the number of anomalies triggered by the node in the state. Output a state interlocking graph, which consists of a set of nodes and a set of edges. Each node and edge has attribute information.

[0029] A node dependency matrix is ​​constructed based on the state interlocking graph: the rows and columns of the matrix correspond to all nodes, and each element represents the dependency type from node i to node j. Dependency types include allowed signal dependency, waiting signal dependency, and abnormal signal dependency. Allowed signal dependency means that node j can only execute after node i issues an allowed signal; waiting signal dependency means that node j must wait for node i's state to complete before it can start; abnormal signal dependency means that node j will be blocked due to an abnormal event of node i. A base weight value is assigned to each dependency: the base weight for allowed signal dependency is 1.0, the base weight for waiting signal dependency is 0.8, and the base weight for abnormal signal dependency is 0.6. If there are multiple dependencies between node i and node j, the maximum value of all dependency weights is taken as the initial topology connection strength, used to adjust the connection weights so that the dependencies of key nodes account for a higher proportion of the topology connection strength. The initial topology connection strength is multiplied by the node importance coefficient to obtain the topology connection strength. The waiting time and number of anomalies for each node are normalized and linearly mapped to the interval between 0 and 1 to avoid the influence of dimensions; the arithmetic mean of these two values ​​is calculated; at the same time, the sum of the out-degree and in-degree of each node is calculated according to the state interlocking graph and recorded as the total degree of the node; the node with the largest total degree in the entire graph is found; the node importance coefficient is equal to the total degree of the node divided by the maximum total degree, and then multiplied by the normalized average of the waiting time and the number of anomalies.

[0030] like Figure 3 The device state interlocking relationship matrix shown illustrates the topological connection strength between each preceding state node and the current state node in a multi-device collaborative task. The horizontal axis represents subsequent state nodes, and the vertical axis represents preceding state nodes. The values ​​in the matrix represent the dependency strength between two state nodes, ranging from zero to one. The larger the value, the stronger the constraint relationship between the preceding state and the subsequent state.

[0031] As shown in the diagram, the connection strength between MES binding and detection reception is 0.92, and the connection strength between detection reception and MES binding is 0.90, indicating a strong data dependency between the work order binding state and the detection reception state. The connection strength between CNC door opening and PLC enable is 0.88, and the connection strength between PLC enable and CNC door opening is 0.86, indicating a clear interlocking relationship between the equipment door opening state and the control enable signal. The connection strength between CNC machining and RBT unloading is 0.84, and the connection strength between RBT unloading and CNC machining is 0.82, indicating a close action connection between the machining completion state and the robot unloading state. This diagram can be used to illustrate how the system establishes a state interlocking diagram based on a unified event flow and converts the dependencies of enable signals, waiting signals, and abnormal signals into a computable topological connection strength matrix.

[0032] In this implementation scheme, by traversing a unified event flow, device numbers, status numbers, trigger sequences, and dependencies are transformed into a state interlocking graph. This transforms the state relationships between robots, CNC machine tools, PLCs, inspection terminals, and MES work orders from scattered event records into structured data composed of nodes and edges. This provides structured input for calculating the congestion risk index, enabling quantifiable, sortable, and traceable data representation of the pre-existing states, states, and signal dependencies in multi-device collaborative tasks.

[0033] Specifically, the process of calculating the congestion risk index for each node is as follows: The node coupling congestion baseline is obtained by dividing the product of the waiting times of each pair of nodes by the sum of the difference in node signal consistency and 1. This baseline reflects the degree of congestion caused by the accumulation of waiting time and signal deviation between the two nodes. Subtracting the topology connection strength from the constant term 1 yields the topology difference, which represents the relaxation of the dependency relationship between nodes; a larger value indicates a weaker connection. Squaring the topology difference yields the topology difference squared term. Multiplying this term by the topology sensitivity adjustment coefficient yields the topology influence term. Applying a negative exponentiation to the topology influence term yields the exponential decay coefficient. Multiplying the node coupling congestion baseline by the exponential decay coefficient yields the nonlinear congestion contribution of the node pair. The maximum value of the nonlinear congestion contribution calculated for all node pairs is taken to obtain the congestion risk index. Taking the maximum value instead of summing or averaging is chosen because the overall congestion risk in a collaborative task is limited by the most vulnerable pair of node couplings. The maximum value directly reflects the most critical congestion bottleneck in the entire task, preventing the average of multiple low-risk node pairs from masking a single high-risk node pair.

[0034] The specific formula for calculating the congestion risk index is as follows: ; In the formula, Congestion Risk Index Quantify the risk level of the most severely blocked node in the entire collaborative task. Waiting time of the current state node of device i Measuring the contribution of node stagnation to overall congestion risk The device number to which the current state node belongs is obtained through the device identifier field in the event stream and is used to determine the node's position and association in the state interlock diagram; The device number of the other node that depends on node i is determined by the event flow and the dependency edges of the state interlock graph, and is used to calculate the impact of inter-node coupling on blocking risk; The signal consistency of the current state node of device i is used to reflect whether the actual execution of the node conforms to the expected signal state; The topology-sensitive adjustment coefficient, set through parameter configuration, ranges from 0.1 to 10 and is used to control the nonlinear adjustment strength of the congestion risk index by the topology coupling between nodes. The topological connection strength between node i and node j is used to reflect the impact of the coupling strength between nodes on the overall blocking risk.

[0035] The waiting time, signal consistency, and number of anomalies are combined column-wise to form the feature vector for each node. All node feature vectors are then arranged sequentially according to their node numbers to generate a node feature matrix. This matrix is ​​used for node priority generation or standard sequence generation, serving as the source of structured data for federated storage. Rows in the matrix represent different nodes, and columns represent the node attribute dimensions.

[0036] Table 1 shows the network node operational status risk assessment data. Multi-dimensional quantitative analysis was conducted on each network node: Node 1 (N01): waiting time 0.42, signal consistency 0.91, number of anomalies 2, topology connection strength 0.82, congestion risk index 0.73; Node 2 (N02): waiting time 0.57, signal consistency 0.86, number of anomalies 3, topology connection strength 0.79, congestion risk index 0.88; Node 3 (N03): waiting time 0.31, signal consistency 0.94, number of anomalies 1, topology connection strength 0.68, congestion risk index 0.55; Node 4 (N04)... The waiting time is 0.63, signal consistency is 0.81, number of anomalies is 4, topology connection strength is 0.84, and congestion risk index is 0.97; Node 5, ID N05, has a waiting time of 0.48, signal consistency is 0.88, number of anomalies is 2, topology connection strength is 0.76, and congestion risk index is 0.71; Node 6, ID N06, has a waiting time of 0.54, signal consistency is 0.83, number of anomalies is 3, topology connection strength is 0.81, and congestion risk index is 0.86; Node 7, ID N07, has a waiting time of 0.37, signal consistency is 0.90, number of anomalies is 2, topology connection strength is 0.72, and congestion risk index is 0.62. The combined data shows that the longer the node wait time, the lower the signal consistency, and the more anomalies occur, the higher the corresponding congestion risk index. Among them, node N04 has the highest congestion risk index and the most prominent potential for operational anomalies. Node N03 performs well in all operational indicators and has the lowest congestion risk. The risk levels of the other nodes are between the two. This data can provide a quantitative reference for network node operation and maintenance management and risk warning.

[0037] Table 1 Network Node Operation Status Risk Assessment Data Table

[0038] like Figure 4 The 3D scatter plot illustrating the node congestion risk shows the correspondence between waiting time, signal consistency, and topology connectivity strength of different nodes in the state interlocking graph. The horizontal axis represents the normalized value of waiting time, the vertical axis represents signal consistency, and the color of the points represents the topology connectivity strength. The closer the color is to the top of the color bar, the stronger the dependency relationship between the node and other nodes. The plot shows that node N08 has the highest normalized value of waiting time, the lowest signal consistency, and the highest topology connectivity strength, indicating that this node simultaneously exhibits characteristics of long waiting time, low signal matching degree, and strong node dependency. Nodes N04, N10, and N02 are also located in the region of high waiting time and low signal consistency, and have high topology connectivity strength, making them key nodes to identify in the congestion risk calculation. In contrast, node N03 has low waiting time, high signal consistency, and low topology connectivity strength, resulting in a relatively small contribution to congestion. This demonstrates how to jointly represent node waiting time, signal consistency, and topology connectivity strength before calculating the congestion risk index, and provides data basis for node priority ranking and extraction of operational events involving nodes with congestion risk.

[0039] In this implementation scheme, a congestion risk index is constructed by jointly calculating the waiting time, signal consistency, and topology connection strength of each node. This transforms the congestion level in multi-device collaborative tasks from scattered individual indicators into a unified quantitative result. Through the joint construction of a node feature matrix, the congestion risk index, and a node priority list, high-risk nodes can be sorted, located, and associated, providing a structured data foundation for extracting operational events involving congestion risk nodes, generating standard event sequences, and analyzing operational event sequences.

[0040] Specifically, the process of generating the standard event sequence and the operation event sequence is as follows: The congestion risk index and node feature matrix are jointly stored to form a node-level congestion risk feature set. This feature set uses the node ID as the primary key, encapsulating the multi-dimensional attributes of each node with global risk indicators. The feature set includes each node's waiting time, signal consistency, number of anomalies, and corresponding congestion risk index. Nodes are sorted according to their congestion risk index to generate a node priority list; nodes with higher congestion risk indices have higher priority, indicating they require more attention or optimization in collaborative tasks. Based on the device list for the adaptation evaluation task, the states and actions each device needs to execute are listed and arranged in the order required by the task, forming an initial event list. Combining the dependencies in the state interlocking graph, the event sequence is topologically sorted to ensure that the sequence satisfies all allowable signals, waiting signals, and anomaly signal dependency constraints.

[0041] Each event simultaneously calculates its expected trigger time and completion time, represented by a unified system time. The expected trigger time is calculated based on the preset process cycle time, the time window constraints of the MES work order, and the statistical median of the equipment's historical response time. The completion time is based on the expected trigger time plus the standard action duration defined by the equipment specifications or the average duration learned from historical events. All events are sequentially generated into a structured array, with each row containing the equipment number, status number, action number, permitted signal number, and time information, forming a standard event sequence. The node priority list is compared with the unified event stream and the standard event sequence. Based on the node number, the corresponding operation event record is located in the unified event stream. By retrieving events in the operation log of the operation subject whose equipment number and status number match the priority list (the matching key is a combination of the equipment number, status number, and action number, and the time window is within the allowable deviation range before and after the trigger time of the corresponding event in the standard event sequence), operation events involving nodes with blocking risk are extracted to form an operation event sequence.

[0042] In this implementation scheme, by jointly storing the congestion risk index and the node feature matrix according to node number, the waiting time, signal consistency, number of anomalies, and congestion risk index of each node form a unified set of node-level congestion risk features, giving the standard event sequence clear time constraints. By using device number, status number, action number, and time window as matching conditions, operation events involving nodes with congestion risk are extracted from the operation log of the operation subject, forming an operation event sequence that corresponds to the standard event sequence, providing structured input for the calculation of the deviation of the operation event sequence.

[0043] Specifically, the process of evaluating the deviation of the operation event sequence based on the operation event sequence is as follows: Based on the event number and status number, the corresponding event is located in the operation log and standard event sequence of the operation entity. The difference between their trigger times is then calculated and normalized to a range of zero to one. The denominator of the normalization is the maximum allowable time deviation defined by the process or equipment. The maximum allowable time deviation can be taken from the equipment cycle time, the allowable delay window specified by the MES work order, or a multiple of the standard deviation of historical event time differences. Simultaneously, the action numbers are compared for consistency. If they are inconsistent, a value of one is assigned; otherwise, zero is assigned. The time deviation value and action deviation value are combined and a weighted average is used to calculate the matching distance. Each event number and status number in the standard event sequence is traversed to check if a corresponding event exists in the operation event sequence. If no corresponding event exists in the operation event sequence or the event order is incorrect, the event is recorded as an unmatched critical event. All unmatched critical events are arranged by sequence number to form a set, with each element of the set taking a value of zero or one, which is used for cumulative calculation in the deviation formula.

[0044] Multiply the matching distance of each event in the operation event sequence by the matching distance scaling factor, invert the negative sign, perform exponential operation, add 1, and take the reciprocal to obtain the weighted matching value. Then, exponentiate the weighted matching value of each event with the corresponding event weight coefficient to obtain the event weighting term. Assign a value of 1 to each unmatched key event in the operation event sequence, perform a negative exponential operation, add 1, and take the reciprocal to obtain the unmatched impact factor. Accumulate all unmatched impact factors, multiply by the unmatched event penalty coefficient, and obtain the total unmatched event value. Add the product of the event weighting terms to the total unmatched event value, take the negative value, and add a constant term of 1 to obtain the operation event sequence deviation.

[0045] The specific formula for calculating the deviation of the operation event sequence is as follows: ; In the formula, This indicates the deviation of the operational event sequence, used to quantify the degree of difference between the operational event sequence and the standard sequence. This represents the index of the event in the current sequence, determined by traversing the event sequence, and is used to locate the position of each event in the sequence; This represents the length of the standard event sequence, which is obtained through statistics of the standard event sequences for job tasks and is used to determine the total number of events for deviation calculation. This represents the matching distance scaling factor. It is calculated by statistically analyzing the matching results of the operational data and the standard event sequence, calculating the mean and variance of the matching distance distribution, and obtaining it through a linear mapping method. The value ranges from 0.1 to 10 and is used to adjust the sensitivity of the event matching distance to the degree of deviation. This represents the k-th event in the sequence of operation events, extracted through a unified event stream and the operation log of the operation subject, and used to participate in the matching distance calculation; This represents the k-th event in the standard event sequence, obtained through the job task standard sequence, and used for comparison with operational events; The event matching distance function is obtained by calculating the difference between the event trigger time and the standard event trigger time, as well as the difference in action number, and is used to quantify the degree of matching deviation for each event. The event weight coefficient is obtained by multiplying the topological connection strength of the node to which the event belongs by the normalized value of the node blocking risk index. The value ranges from zero to one and is used to weight and highlight the impact of key events on the overall deviation. This represents the penalty coefficient for unmatched events. It is obtained by statistically analyzing the number of unmatched key events in the operational event sequence and the standard event sequence, and by mapping the average impact of unmatched events using a normalization method. The value ranges from zero to one, and it is used to amplify the contribution of unmatched key events to the deviation. An index representing unmatched key events is generated by comparing the operational event sequence with a standard sequence, and is used to calculate the impact of each unmatched event. This represents the total number of unmatched key events, obtained by comparing the sequence of operational events with a standard sequence, and is used to determine the size of the set of unmatched events. This represents the u-th unmatched key event. Missing or out-of-order events in the operation log of the main operation subject are extracted by comparison and used to participate in the sparse penalty calculation.

[0046] In this implementation scheme, the sequence of operational events is matched with the standard sequence of events using event numbers, status numbers, and action numbers. The matching distance is calculated by combining the trigger time difference, ensuring that each operational event generates calculable deviation data. Missing events and out-of-order events can be organized into a set of unmatched key events, so that the deviation of the operational event sequence simultaneously includes event matching deviation, key node weights, and key event missing information.

[0047] Specifically, the process of filtering off-target event segments and matching event segments, and generating off-target type labels is as follows: For the current event, when the deviation of the operation event sequence is greater than the current deviation threshold, instead of simply marking it as an anomaly, the process starts by backtracking backward from the current event to the previous event, backtracking to the most recent event whose deviation is less than or equal to the current deviation threshold and whose status number matches, and then scanning backward to the end point of the deviation event segment as the end boundary, scanning backward to the first event whose deviation recovers to below the current deviation threshold, thus capturing the deviation event segment. The capture operation includes all events between the backtracking start event and the scanning end event.

[0048] When the deviation of an operation event sequence is less than or equal to the current deviation threshold, the current event is written to the stable event buffer. If M consecutive events are all less than or equal to the current deviation threshold, the event segment is recorded as a matching event segment. A deviation type label is generated using a component proportion comparison and rule mapping method: First, the trigger time deviation proportion, action number deviation proportion, unmatched key event proportion, repeated operation proportion, and interlock dependency violation proportion within the deviation event segment are read. These five proportion values ​​are then sorted from largest to smallest to obtain the deviation component sorting result. Then, the deviation component with the largest value is extracted as the primary deviation component, and a primary deviation type label is generated according to a preset component type mapping table. The specific rules for the component type mapping table are as follows: If the trigger time deviation ratio is greater than or equal to 0.4 and is the largest component, a trigger delay label is generated; if the action number deviation ratio is greater than or equal to 0.35 and is the largest component, an action mismatch label is generated; if the proportion of unmatched key events is greater than or equal to 0.3 and is the largest component, a key event missing label is generated; if the proportion of repeated operations is greater than or equal to 0.3 and is the largest component, a repeated trigger label is generated; if the proportion of interlock dependency violation is greater than or equal to 0.25 and is the largest component, an interlock violation label is generated. When the largest component does not meet the above thresholds, the corresponding label is still generated according to the largest proportion, but a slight marking is added.

[0049] If the trigger time deviation accounts for the largest proportion, a trigger delay label is generated. If the action number deviation accounts for the largest proportion, an action mismatch label is generated. If the unmatched critical event accounts for the largest proportion, a critical event missing label is generated. If the duplicate operation accounts for the largest proportion, a duplicate trigger label is generated. If the interlock dependency violation accounts for the largest proportion, an interlock violation label is generated.

[0050] When two or more deviation components have the same value, or when the difference between the largest deviation component and the second largest deviation component is less than the preset component discrimination threshold, the component discrimination threshold is set to 0.1. Based on experimental measurements of the resolution between components in typical deviation samples, the corresponding deviation components are simultaneously written into the label field to generate a composite deviation type label. The composite deviation type labels are arranged in the following order: interlock dependency violation, unmatched key event, action number deviation, trigger time deviation, and repeated operation. The deviation event segment number, main deviation type label, composite deviation type label, largest deviation component, second largest deviation component, deviation peak value, and number of events in the deviation event segment are written into the deviation event segment table.

[0051] like Figure 5The deviation event segment component stacking analysis diagram shown illustrates the composition of various deviation components within different deviation event segments. The horizontal axis represents the deviation event segment number, and the vertical axis represents the cumulative value of the deviation components. Different colors represent the proportions of trigger time deviation, action number deviation, unmatched critical events, repeated operations, and interlock dependency violations, respectively. The diagram shows that event segment S04 has the highest cumulative value of deviation components, with high proportions of trigger time deviation, action number deviation, unmatched critical events, and interlock dependency violations, indicating that this event segment is not caused by a single deviation but is formed by the superposition of multiple types of deviations. Event segment S06 also has a high cumulative deviation value, mainly contributed by trigger time deviation and interlock dependency violations. Event segment S05 has a lower cumulative deviation value, indicating that its deviation degree is relatively small. This diagram illustrates how the system decomposes the trigger time deviation, action number deviation, unmatched critical events, repeated operations, and interlock dependency violations within a deviation event segment after intercepting it, and provides data basis for generating main deviation type labels and composite deviation type labels.

[0052] In this implementation plan, by writing the deviation event segment number, main deviation type label, composite deviation type label, maximum deviation component, second largest deviation component, deviation peak value, and number of events in the deviation event segment table, the segmentation results, deviation composition, and deviation degree of the deviation event segment are formed into a unified structured record, providing a directly readable data foundation for the calculation of adaptation deviation judgment value and the generation of adaptation classification results.

[0053] Specifically, the process of calculating the adaptation deviation judgment value based on the deviation event segment and the matching event segment is as follows: The deviation of each operation event sequence within the same deviation event segment is read from the deviation event segment table, and the maximum and average values ​​are calculated to obtain the peak deviation and the average deviation of the deviation event segment. The peak value is obtained using a maximum value function, and the average value is calculated using an arithmetic mean. The deviation of each operation event sequence within adjacent matching event segments is read from the matching event segment table, and the average value is calculated to obtain the average deviation of the matching event segment. The number of events within the deviation event segment is counted to obtain the number of events in the deviation event segment. The number of events is obtained by counting each event entry in the deviation event segment line by line. The number of events that do not satisfy the dependency relationship within the deviation event segment is determined according to the state interlock diagram, and divided by the total number of events in the deviation event segment to obtain the interlock dependency violation ratio. The number of unmatched events is counted according to the standard event sequence, and divided by the number of events in the standard event sequence to obtain the ratio of unmatched critical events. The number of unmatched events is obtained by performing a difference operation between the standard event sequence and the operation event sequence according to the event number and state number. The number of events with inconsistent action numbers within the deviation event segment is counted and divided by the total number of events in the deviation event segment to obtain the action number deviation ratio. Inconsistent action numbers refer to the actual action number of the same device in the same state being different from the action number specified in the standard event sequence. The number of event numbers that appear repeatedly within the deviation event segment is counted and divided by the total number of events in the deviation event segment to obtain the percentage of repeated operations.

[0054] Adding the peak deviation to the average deviation of the deviation event segment and then adding the zero-prevention term yields the deviation intensity term; adding the average deviation of the matching event segment to the zero-prevention term yields the matching benchmark term; dividing the deviation intensity term by the matching benchmark term yields the deviation amplification ratio; adding one to the number of events in the deviation event segment and taking the natural logarithm yields the event amplification term; the engineering significance of taking the natural logarithm is to suppress the scale effect when the number of events is too large, so that the contribution of the increase in the number of events to the adaptation deviation judgment value decreases logarithmically, avoiding excessive amplification of the judgment result due to the quantity advantage of long event segments; adding one to the interlock dependency violation ratio yields the interlock amplification term; adding one to the proportion of unmatched key events, the proportion of action number deviation, and the proportion of repeated operations, and then performing a square root operation yields the deviation aggregation term; the square root operation is also used to mitigate the nonlinear expansion when multiple deviation components are superimposed, keeping the deviation aggregation term within a reasonable range; multiplying the deviation amplification ratio, event amplification term, interlock amplification term, and deviation aggregation term in sequence yields the adaptation deviation judgment value.

[0055] The specific formula for calculating the adaptation deviation judgment value is as follows: ; In the formula, This represents the adaptation deviation judgment value, used to indicate the overall degree of deviation of the current deviation event segment relative to the matching event segment; This indicates the peak deviation, used to represent the maximum deviation level of a single event within a deviation event segment. This indicates the average deviation of the deviation event segment, used to represent the overall average deviation level of that deviation event segment. This indicates the zero-prevention item, which is preset through the system parameter table and has a value range of 0.001 to 0.1, used to avoid the denominator being zero. This represents the average deviation of the matched event segment, which is used as the matching baseline value for the current deviation event segment; This indicates the number of events in the off-event segment, which is used to indicate the duration of the off-event segment. Indicates the percentage of interlock dependency violations, used to represent the degree of interlock relationship violation within the deviation event segment; This indicates the percentage of unmatched key events, used to indicate the degree of missing key events; This indicates the percentage of deviation in action numbers, used to represent the degree of mismatch in operational actions; This indicates the percentage of repeated operations, representing the degree to which the same event is triggered repeatedly.

[0056] In this implementation plan, the adaptation deviation judgment value is obtained by multiplying the deviation amplification ratio, event amplification term, interlock amplification term and deviation aggregation term in sequence. This allows the peak deviation, average deviation, duration and deviation source of the deviation event segment to be uniformly converted into comparable numerical results, providing a structured calculation basis for adaptation deviation threshold comparison, adaptation deviation label generation and re-inspection level classification.

[0057] Specifically, the process of determining the current event segment and generating adaptation deviation labels is as follows: The adaptation deviation judgment value is compared with the adaptation deviation threshold in real time. The adaptation deviation threshold is obtained by reading the adaptation deviation judgment values ​​of historical matching event segments and taking the 95th percentile. When the adaptation deviation judgment value is less than or equal to the adaptation deviation threshold, the current event segment is written to the adaptation passed event segment table, recording the event segment number, start event number, end event number, adaptation deviation judgment value, and matching flag. When the adaptation deviation judgment value is greater than the adaptation deviation threshold, the current event segment is written to the adaptation deviation event segment table, and component judgment is performed. If the proportion of interlock dependency violation is greater than the interlock violation threshold, it is recorded as an interlock violation deviation. If the proportion of unmatched critical events is greater than the critical event missing threshold, it is recorded as a critical event missing deviation. If the proportion of action number deviation is greater than the action mismatch threshold, it is recorded as an action mismatch deviation. If the proportion of repeated operations is greater than the repeated operation threshold, it is recorded as a repeated trigger deviation. If the proportion of trigger time deviation is greater than the trigger time deviation threshold, it is recorded as a trigger delay deviation. Adaptation deviation labels are generated based on the component judgment results. These labels include interlock violation labels, critical event missing labels, action mismatch labels, repeated triggering labels, and trigger delay labels. When an adaptation deviation event segment simultaneously meets the generation conditions for more than two deviation labels, the labels are arranged in the following order: interlock violation label, critical event missing label, action mismatch label, trigger delay label, and repeated triggering label, and then composite adaptation deviation labels are generated. These composite adaptation deviation labels are bound to the event segment number, adaptation deviation judgment value, peak deviation value, and number of events.

[0058] If the adaptation deviation judgment value is greater than the adaptation deviation threshold but less than or equal to 1.2 times the adaptation deviation threshold, the re-inspection level field is recorded as Level 1. If the adaptation deviation judgment value is greater than 1.2 times the adaptation deviation threshold but less than or equal to 1.5 times the adaptation deviation threshold, the re-inspection level field is recorded as Level 2. If the adaptation deviation judgment value is greater than 1.5 times the adaptation deviation threshold, the re-inspection level field is recorded as Level 3.

[0059] After all operation event sequences in the current batch have been processed, they are summarized according to the operation subject number, node number, deviation label, and re-inspection level fields to generate an adaptation classification data table. The adaptation classification data table includes the operation subject number, node number, event segment number, adaptation deviation judgment value, main deviation label, composite deviation label, re-inspection level field, start event number, end event number, and number of events.

[0060] In this implementation plan, first-level, second-level, and third-level re-inspection level fields are generated based on the multiple relationship between the adaptation deviation judgment value and the adaptation deviation threshold, enabling graded recording of the deviation degree of event segments. An adaptation classification data table is generated by summarizing according to the operation subject number, node number, deviation label, and re-inspection level fields, resulting in a unified data output for the adaptation deviation results of all operation event sequences in the current batch.

[0061] Specifically, the second aspect of this invention provides a skills training and manufacturing transformation adaptation assessment system, applied to a skills training and manufacturing transformation adaptation assessment method, comprising: a data acquisition and preprocessing module, used to acquire event stream data and perform time axis correction, denoising, and normalization on the event stream data to generate a unified event stream; the time axis correction adopts the gateway delay median compensation method, the denoising adopts the sliding window first derivative constraint method, and the normalization adopts linear scaling. A state modeling and blocking identification module, used to traverse the unified event stream, construct a multi-device state interlocking graph, calculate the blocking risk index for each node, and generate a standard event sequence and an operation event sequence; the state interlocking graph contains nodes and edges and their attribute information, and the blocking risk index is calculated based on the maximum value of the node coupling blocking contribution. A collaborative task analysis module, used to evaluate the deviation of the operation event sequence based on the operation event sequence, filter the deviated event segments and matching event segments, and generate deviation type labels; the deviation evaluation adopts a weighted matching distance and non-matching penalty mechanism, and the deviation type labels include main deviation type and composite deviation type. The adaptation evaluation module calculates the adaptation deviation judgment value based on the deviation event segment and the matching event segment, judges the current event segment, and generates an adaptation deviation label. The adaptation deviation judgment value is obtained by multiplying the deviation amplification ratio, the event amplification term, the interlock amplification term, and the deviation aggregation term. The adaptation deviation label is output based on the comparison result between the judgment value and a preset threshold.

[0062] In this implementation scheme, through the data transmission relationship between modules, multi-source event data generated by robots, CNC machine tools, PLCs, inspection terminals and MES work orders can be converted into computable adaptation evaluation results, and the adaptation deviation judgment value, adaptation deviation label and re-inspection level fields are written into the adaptation classification data table.

[0063] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0064] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A skill talent training and manufacturing transformation adaptation evaluation method, characterized in that, Includes the following steps: S1: Acquire event stream data and perform timeline correction, noise reduction, and normalization on the event stream data to generate a unified event stream; S2, traverse the unified event stream, construct a multi-device state interlocking graph, calculate the blocking risk index for each node, and generate standard event sequences and operation event sequences; S3, based on the operation event sequence, evaluates the deviation of the operation event sequence, filters the deviation event segments and matching event segments, and generates deviation type labels; S4. Based on the deviation event segment and the matching event segment, calculate the adaptation deviation judgment value, judge the current event segment, and generate the adaptation deviation label.

2. The skill personnel training and manufacturing transformation adaptation evaluation method according to claim 1, characterized in that: The specific process of acquiring event stream data and performing time-axis correction, noise reduction, and normalization on the event stream data to generate a unified event stream is as follows: The system collects data on state transitions, action commands, signal triggers, waiting times, abnormal events, and parameter adjustments that occur during collaborative tasks involving robots, CNC machine tools, PLCs, inspection terminals, and MES work orders. This data generates event stream data, which includes device number, status number, action number, trigger signal number, feedback signal number, operating entity number, start time, end time, duration, abnormal flag, and parameter values. The delay of a single event is obtained by subtracting the device's local time from the gateway's received time. The median of the delays of the M most recent events from the same device is then used as the current correction delay to uniformly map events to the global system time. The sliding window first derivative constraint method is used to calculate the rate of state change at each time point for noise reduction. When the rate of state change exceeds the threshold defined by the device specifications, the abnormal peak value is replaced with the previous sample value to generate a smooth event sequence. The event stream data is normalized by linear scaling to generate a unified event stream.

3. The skill personnel training and manufacturing transformation adaptation evaluation method according to claim 1, characterized in that: The specific process of traversing the unified event stream and constructing the multi-device state interlocking graph is as follows: Traverse the unified event stream, determining the corresponding node for each event based on the device number and status number. Map the triggering order and dependencies between nodes in the event sequence to edges in the graph. Simultaneously, count the waiting time, signal consistency, and number of anomalies for each node, and populate the node attributes. The waiting time is obtained by subtracting the end time of the previous state from the start time of the node state. The proportion of matching times within a window to the total number of times is used as the signal consistency. The number of anomalies is obtained by counting the abnormal events triggered by the node in the state. Output a state interlocking graph, which consists of a set of nodes and a set of edges, with each node and edge carrying attribute information. Construct a node dependency matrix based on the state interlocking graph: the rows and columns of the matrix correspond to all nodes, and each element represents the dependency type from node i to node j. The dependency types include allowed signal dependency, waiting signal dependency, and abnormal signal dependency. Assign a basic weight value to each type of dependency. If there are multiple dependencies between node i and node j, the maximum value of all dependency weights is taken as the initial topological connection strength. The initial topological connection strength is then multiplied by the node importance coefficient to obtain the topological connection strength.

4. The skill talent training and manufacturing transformation adaptation evaluation method according to claim 1, characterized in that: The specific process for calculating the blocking risk index for each node is as follows: Divide the product of the waiting times of each pair of nodes by the sum of the difference in node signal consistency and 1 to obtain the node coupling blocking base value; subtract the topology connection strength from the constant term 1 to obtain the topology difference; square the topology difference to obtain the topology difference square term. Multiply the squared topology difference term by the topology sensitivity adjustment coefficient to obtain the topology influence term; perform a negative exponential operation on the topology influence term to obtain the exponential decay coefficient; multiply the node coupling blocking base value by the exponential decay coefficient to obtain the nonlinear blocking contribution of the node pair; take the maximum value of the nonlinear blocking contribution calculated for all node pairs to obtain the blocking risk index. The waiting time, signal consistency, and number of anomalies are combined by column to form the feature vector of each node. The feature vectors of all nodes are arranged in order of node number to generate a node feature matrix. The rows of the matrix represent different nodes, and the columns represent the node attribute dimensions.

5. The skill personnel training and manufacturing transformation adaptation evaluation method according to claim 1, characterized in that: The specific process for generating the standard event sequence and the operation event sequence is as follows: The blocking risk index and the node feature matrix are jointly stored to form a node-level blocking risk feature set; the nodes are sorted according to the blocking risk index to generate a node priority list; based on the device list of the adaptation assessment task, the status and action number that each device needs to perform are listed and arranged in order according to the task requirements to form an initial event list; Based on the dependencies in the state interlocking graph, the event sequence is topologically sorted. The expected trigger time and completion time of each event are calculated simultaneously and represented by a unified system time. All events are generated into a structured array in sequence, with each row containing device number, status number, action number, permission signal number, and time information, forming a standard event sequence. The node priority list is compared with the unified event stream and the standard event sequence. The corresponding operation event record is located in the unified event stream according to the node number. By retrieving events in the operation log of the operation subject whose device number and status number match the priority list, operation events involving nodes with blocking risk are extracted to form an operation event sequence.

6. The method for assessing the adaptation of skilled personnel training to manufacturing transformation according to claim 1, characterized in that: The specific process for evaluating the deviation of the operation event sequence based on the operation event sequence is as follows: Based on the event number and status number, locate the corresponding event in the operation log and standard event sequence of the operation subject, then calculate the difference between their trigger times, normalize the time difference to the range of zero to one, and obtain the time deviation value; at the same time, compare whether the action numbers are consistent. If they are inconsistent, assign a value of one, otherwise assign a value of zero; combine the time deviation value and the action deviation value, and calculate the matching distance by weighted average; traverse each event number and status number in the standard event sequence to check whether there is a corresponding event in the operation event sequence; If the corresponding event does not exist in the sequence of operation events or the event order is incorrect, the event will be recorded as an unmatched critical event. Multiply the matching distance of each event in the operation event sequence by the matching distance scaling factor, invert the negative sign, perform exponential operation, add 1, and take the reciprocal to obtain the weighted matching value. Then, exponentiate the weighted matching value of each event with the corresponding event weight coefficient to obtain the event weighting term. Assign a value of 1 to each unmatched key event in the operation event sequence, perform a negative exponential operation, add 1, and take the reciprocal to obtain the unmatched impact factor. Accumulate all unmatched impact factors, multiply by the unmatched event penalty coefficient, and obtain the total unmatched event value. Add the product of the event weighting terms to the total unmatched event value, take the negative value, and add a constant term of 1 to obtain the operation event sequence deviation.

7. The method for assessing the adaptation of skilled personnel training to manufacturing transformation according to claim 1, characterized in that: The specific process of filtering deviation event segments and matching event segments, and generating deviation type labels is as follows: For the current event, when the deviation of the operation event sequence is greater than the current deviation threshold, instead of simply marking it as an anomaly, the system traces back to the most recent event with a deviation less than or equal to the current deviation threshold and a matching status number, and then scans backward to the first event whose deviation recovers to below the current deviation threshold, thus capturing the deviation event segment. When the deviation of the operation event sequence is less than or equal to the current deviation threshold, the current event is written into the stable event buffer. If M consecutive events are all less than or equal to the current deviation threshold, the event segment is recorded as a matching event segment, and a deviation type label is generated through component proportion comparison and rule mapping.

8. The method for assessing the adaptation of skilled personnel training to manufacturing transformation according to claim 1, characterized in that: The specific process for calculating the adaptation deviation judgment value based on the deviation event segment and the matching event segment is as follows: The deviation of each operation event sequence within the same deviation event segment is read from the deviation event segment table, and the maximum and average values ​​are calculated to obtain the peak deviation and the average deviation of the deviation event segment. The deviation of each operation event sequence within adjacent matching event segments is read from the matching event segment table, and the average value is calculated to obtain the average deviation of the matching event segment. The number of events within the deviation event segment is counted to obtain the number of events in the deviation event segment. The number of events that do not satisfy the dependency relationship within the deviation event segment is determined according to the state interlock diagram, and divided by the total number of events in the deviation event segment to obtain the interlock dependency violation ratio. The number of unmatched events is counted according to the standard event sequence, and divided by the number of events in the standard event sequence to obtain the ratio of unmatched key events. The number of events with inconsistent action numbers within the deviation event segment is counted and divided by the total number of events in the deviation event segment to obtain the action number deviation ratio. The number of repeated event numbers within the deviation event segment is counted and divided by the total number of events in the deviation event segment to obtain the repeated operation ratio. The deviation intensity term is obtained by adding the peak deviation to the average deviation of the deviation event segment and then adding it to the zero-prevention term; the matching benchmark term is obtained by adding the average deviation of the matching event segment to the zero-prevention term. Divide the deviation intensity term by the matching reference term to obtain the deviation magnification ratio; Add one to the number of events in the off-event segment and take the natural logarithm to obtain the event amplification term; Add one to the interlock dependency violation ratio to get the interlock amplification item; add one to the ratio of unmatched key events, the ratio of action number deviation, and the ratio of repeated operations, and then perform a square root operation to get the deviation aggregation item; multiply the deviation amplification ratio, the event amplification item, the interlock amplification item, and the deviation aggregation item in sequence to get the adaptation deviation judgment value.

9. The method for assessing the adaptation of skilled personnel training to manufacturing transformation according to claim 1, characterized in that: The specific process of determining the current event segment and generating adaptation deviation labels is as follows: The adaptation deviation judgment value is compared with the adaptation deviation threshold in real time. When the adaptation deviation judgment value is less than or equal to the adaptation deviation threshold, the current event segment is written into the adaptation pass event segment table. When the adaptation deviation judgment value is greater than the adaptation deviation threshold, the current event segment is written into the adaptation deviation event segment table, and component judgment is performed. Adaptation deviation labels are generated according to the component judgment results. Adaptation deviation labels include interlock violation labels, critical event missing labels, action mismatch labels, repeated triggering labels, and trigger delay labels. When an adaptation deviation event segment meets the generation conditions of more than two deviation labels at the same time, the labels are arranged in the order of interlock violation label, critical event missing label, action mismatch label, trigger delay label, and repeated triggering label, and a complex adaptation deviation label is generated.

10. A skills training and manufacturing transformation adaptation assessment system, employing the skills training and manufacturing transformation adaptation assessment method as described in any one of claims 1-9, characterized in that, include: The data acquisition and preprocessing module is used to acquire event stream data and perform time axis correction, noise reduction and normalization on the event stream data to generate a unified event stream. The state modeling and blockage identification module is used to traverse the unified event stream, construct a multi-device state interlocking graph, calculate the blockage risk index for each node, and generate standard event sequences and operation event sequences. The collaborative task analysis module is used to evaluate the deviation of operation event sequences based on the operation event sequence, filter deviation event segments and matching event segments, and generate deviation type labels; The adaptation evaluation module is used to calculate the adaptation deviation judgment value based on the deviation event segment and the matching event segment, and to judge the current event segment and generate an adaptation deviation label.

Images