Multi-station on-line detection system and detection method for impact rammer

Abstract

The application discloses an impact ram multi-station online detection system and a detection method, and belongs to the technical field of impact ram production detection; the system comprises a single-station basic detection unit, a multi-station traceability linkage core module, an abnormal conduction path analysis module and a self-adaptive threshold optimization module which are interacted through an industrial Ethernet, and further comprises a key process multi-dimensional re-inspection unit, a dynamic interference compensation module and the like; the method is based on the system and realizes detection through station data acquisition and preprocessing, dynamic interference compensation, comprehensive quality evaluation, quantitative traceability, hierarchical processing, threshold optimization and data storage. The core lies in: accurately positioning a quality problem root station through a quantitative model, tracing an abnormal conduction path, dynamically optimizing a detection threshold in combination with traceability data, forming a 'detection-traceability-optimization' closed loop, solving the problems of low accuracy, poor efficiency and disconnection between traceability and detection of traditional manual traceability, adapting to a flexible production scene, reducing resource waste and improving impact ram production quality and efficiency.

Description

Technical Field

[0001] This invention belongs to the field of impact rammer production testing technology, and particularly relates to an online multi-station testing system and testing method for impact rammers. Background Technology

[0002] Currently, impact rammer production often involves multi-station online assembly operations using manual labor or machines. Due to negligence in the previous process or quality issues with parts, defective workpieces can easily flow into the next process, resulting in wasted production resources and decreased production efficiency.

[0003] To address this issue, while some production lines have installed basic inspection devices such as CCD vision inspection at individual workstations, technical deficiencies in the quality problem tracing process have become a key bottleneck restricting production efficiency and quality control. Current technologies, after identifying quality problems, often rely on manual review of previous workstation records or simple inference of the root cause through process sequence correlation. However, the quality of impact ram assembly is affected by the coordination of multiple workstations and parameters. Manual tracing not only suffers from low positioning accuracy (often with root cause location errors reaching 2-3 workstations), but also easily leads to misjudging the root cause workstation and overlooking related processes, resulting in the recurrence of similar quality problems. Furthermore, the inspection data from each workstation are independent. When a defective workpiece is found during final inspection, it is impossible to trace the transmission process of the quality anomaly from the root cause workstation to the current workstation, making it difficult to accurately identify key nodes in the anomaly transmission and resulting in a lack of targeted corrective measures. Furthermore, manual traceability requires retrieving historical records from multiple workstations and comparing operational parameters, with an average traceability time of 5-10 minutes. During this time, the production line needs to be paused or run at low speed, resulting in production stoppages and losses. This is especially true in flexible production lines with multiple models operating simultaneously, where the process parameters and workstation adaptation requirements differ for different models of impact rams, further increasing the complexity and reducing the efficiency of manual traceability. More importantly, existing traceability technologies only address the handling of individual non-conforming workpieces. For frequently occurring quality problems originating from the same source, the traceability results cannot be used to optimize detection thresholds or adjust detection priorities, resulting in a disconnect between traceability and detection, making it difficult to fundamentally improve quality control. Summary of the Invention

[0004] To address the problems in the prior art, the present invention proposes the following technical solution: The impact ram multi-station online inspection system and method include a single-station basic inspection unit that achieves real-time data interaction via industrial Ethernet, a multi-station traceability and linkage core module, an anomaly propagation path analysis module, and an adaptive threshold optimization module. The single-station basic inspection unit collects assembly deviations and equipment operating parameters at each station, and transmits them after preprocessing by binding them with a unique workpiece identifier. The multi-station traceability and linkage core module integrates a quantitative traceability model, accurately locating the root cause station and associated affecting stations of quality problems by calculating the root cause attribution coefficient. The anomaly propagation path analysis module traces the propagation process of anomalies from the root cause station to the current station and identifies key nodes based on the time series of inspection data and the correlation relationship of parameters. The adaptive threshold optimization module dynamically updates the quality judgment threshold in combination with traceability data, forming a closed loop of "inspection-traceability-optimization".

[0005] As a preferred embodiment of the above technical solution, a multi-dimensional re-inspection unit for key processes is also included. This unit is set after the key process of impact ram assembly and integrates a CCD vision inspection module, a force sensor, a vibration sensor and a data fusion module to collect multi-dimensional data such as assembly space deviation, impact contact force and instantaneous vibration acceleration to generate comprehensive quality assessment data.

[0006] As a preferred embodiment of the above technical solution, a dynamic interference compensation module is also included. This module has a built-in interference compensation algorithm that calls synchronous detection data from the preceding workstation and calculates the compensation using the formula: Correct the raw data of the current workstation sensors; among which This is the interference compensation factor, with a value range of 0-0.8; The reference distance is fixed at 2m; , This refers to the raw force and vibration acceleration data collected by the sensors at the current workstation. , This is the corrected and valid data.

[0007] As a preferred embodiment of the above technical solution, the single-station basic detection unit includes a CCD vision sensor and a data preprocessing module. The CCD vision sensor acquires x / y / z axis deviation data (Δx, Δy, Δz), and the equipment sensor acquires assembly pressure (F). 工 ), working speed (v) 工 ), operational vibration (a 工 The data preprocessing module uses mean filtering with a window size of 5 for noise reduction. The unique identifier of the workpiece includes the model, online time, and transfer station sequence information.

[0008] As a preferred embodiment of the above technical solution, in the quantitative tracing model of the multi-station tracing linkage core module, the formula for calculating the root cause attribution coefficient R is as follows: in: This is the workstation weighting coefficient. =0.8 is the distance attenuation coefficient. For workstation spacing, For process correlation; The time decay factor, =0.5 is the time decay coefficient. , The current and the number are respectively Inspection time per workstation; This represents the overall quality assessment value for the j-th workstation. The adaptive threshold is set to n, which is the number of preceding associated workstations (default 3).

[0009] As a preferred embodiment of the above technical solution, the abnormal propagation path analysis module uses the following formula: Tracing the transmission of abnormalities; among which For conduction strength, , This refers to the deviation value of the core parameters of the workstation. The conduction attenuation coefficient, =0.3 is the attenuation coefficient. For workstation spacing, S represents the standard deviation threshold for the parameter; S≥0.3 indicates strong conduction, 0.1≤S<0.3 indicates weak conduction, and S<0.1 indicates no conduction.

[0010] As a preferred embodiment of the above technical solution, the threshold iteration formula of the adaptive threshold optimization module is: in For learning rate, when production fluctuates greatly =0.03-0.05, when production is stable =0.01-0.02; For sample size; For source tracing weight coefficients, when there are high-frequency root cause anomalies =1.5, when the fixed conduction path is abnormal =1.2, during occasional anomalies =0.8.

[0011] As a preferred embodiment of the above technical solution, it also includes a data storage module and a flexible production parameter database; the data storage module stores data on the entire process of detection, traceability, processing and optimization; the flexible production parameter database stores assembly standard parameters, process correlation and abnormal conduction attenuation coefficient data of different types of impact rammers, providing support for each module.

[0012] A multi-station online detection method for impact tamping, based on the above-mentioned system implementation, includes: Step 1: Basic detection and data synchronization at each workstation. Sensors at each workstation collect parameter data, which is then preprocessed and bound to the workstation number, timestamp, and unique workpiece identifier code before transmission and storage. Step 2: The dynamic interference compensation module calls the synchronous detection data of the adjacent workstations in the previous step, and corrects the original force and vibration data of the current workstation through the correction formula; Step 3: The multi-dimensional re-inspection unit for key processes receives valid data, calculates the comprehensive quality assessment value Q through a multi-physics field coupled detection model, and determines whether to trigger the traceability process; Step 4: The multi-workstation traceability linkage core module calculates the root cause attribution coefficient through the quantitative traceability model, locates the root cause workstation by combining the judgment rules, and the anomaly propagation path analysis module traces the propagation process and generates a path diagram. Step 5: Execute the coordinated processing strategy according to the source tracing results and Q-values; Step 6: The adaptive threshold optimization module iteratively updates the detection threshold based on the source data; Step 7: The data storage module stores all relevant data throughout the process and supports multi-dimensional query and traceability.

[0013] As a preferred embodiment of the above technical solution, the hierarchical processing strategy in step 5 is as follows: Level 1 anomalies (0.8≤ < Send compensation instructions and deviation warnings; Level 2 anomalies ( ≤ <1.2) Suspend the corresponding workstation and formulate a rectification plan; Level 3 abnormality ( ≥1.2) Lock the workpiece and strengthen the detection of key nodes.

[0014] The beneficial effects of this invention are as follows: 1. Accurate and efficient source tracing significantly improves quality control: By calculating the root cause attribution coefficient through a quantitative source tracing model and combining it with anomaly transmission path analysis, the root cause workstation and related affected workstations of quality problems are accurately located, significantly reducing the root cause location error and avoiding misjudgment and omissions in manual source tracing; at the same time, the source tracing time is shortened from the traditional 5-10 minutes to real-time response, reducing production line downtime losses and adapting to flexible production line multi-model mixed production scenarios.

[0015] 2. Forming a closed loop of "detection-traceability-optimization" to fundamentally reduce the recurrence of quality problems: Dynamically updating the detection threshold based on the traceability results, enhancing the detection sensitivity of parameters associated with high-frequency problems, enabling the detection system to have autonomous learning and dynamic adaptation capabilities, solving the problem of the disconnect between traceability and detection in existing technologies, and continuously improving the stability of production quality.

[0016] 3. The rectification measures are highly targeted and reduce the waste of production resources: Through a hierarchical and coordinated handling strategy, differentiated rectification plans are implemented according to the level of abnormality and the results of source tracing. At the same time, the key nodes of abnormality transmission are accurately identified, providing a clear direction for rectification, reducing ineffective rectification actions, and avoiding the waste of resources caused by unqualified workpieces flowing into subsequent processes. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments.

[0018] Multi-station online inspection system for impact tamping The system includes a single-station basic inspection unit, a multi-dimensional re-inspection unit for key processes, a dynamic interference compensation module, an adaptive threshold optimization module, a multi-station traceability and linkage core module, an anomaly propagation path analysis module, a data storage module, and a flexible production parameter database. All modules interact in real time via industrial Ethernet. Basic inspection unit: Located at each station of the impact rammer assembly, it includes a CCD vision sensor and a data preprocessing module to collect x / y / z axis deviation data (Δx, Δy, Δz) of the assembled parts at each station, and simultaneously collect the assembly pressure (F). 工 ), working speed (v) 工 ), operational vibration (a 工 The data preprocessing module collects equipment operating parameters such as the "parameter data + workstation number + timestamp" and binds them with the workpiece's unique identifier code for transmission after noise reduction through mean filtering (window size 5), providing basic data support for traceability.

[0019] Multi-dimensional re-inspection unit for key processes: Located after the key process of impact ram assembly, it integrates a CCD vision inspection module, force sensor, vibration sensor and data fusion module to collect multi-dimensional data such as assembly space deviation, impact contact force, instantaneous vibration acceleration, etc. After fusion processing, it generates comprehensive quality assessment data to provide a basis for anomaly judgment and source tracing.

[0020] Dynamic interference compensation module: It has a built-in interference compensation algorithm, calls synchronous detection data from the previous workstation, and corrects the original data of the current workstation sensor through the compensation formula, eliminating the impact of cross-workstation dynamic interference and ensuring the accuracy of the detection data.

[0021] Adaptive threshold optimization module: Based on the Bayesian iterative algorithm, the quality judgment threshold is dynamically updated by combining the source data, so as to realize the autonomous learning and dynamic adaptation of the detection system.

[0022] Multi-workstation traceability linkage core module: As the core of the system, it integrates a quantitative traceability model, a root cause workstation determination algorithm, and a traceability result output unit. It receives multi-dimensional detection data and multi-workstation correlation parameters, calculates the root cause attribution coefficient and the influence weight of related workstations through quantitative formulas, and accurately locates the root cause workstation and related influencing workstations of quality problems.

[0023] Anomaly propagation path analysis module: Based on the time series and parameter correlation of multi-station detection data, it traces the propagation process of quality anomalies from the root station to the current station, identifies key nodes of anomaly propagation, deviation amplification coefficient and coupling interference parameters, and generates anomaly propagation path diagram.

[0024] Data storage module: Stores real-time information such as detection data, correction data, comprehensive quality assessment values, threshold iteration records, traceability data, and linkage processing logs from each workstation, supporting multi-dimensional query and traceability.

[0025] Flexible production parameter database: Stores assembly standard parameters, workstation spacing parameters, initial values ​​of weighting coefficients, process correlation, abnormal transmission attenuation coefficients, and other data for different models of impact rammers, providing data support for each module.

[0026] Multi-station online inspection method for impact tamping The method is implemented based on the above system, with a core focus on strengthening the traceability steps, specifically including the following steps: Step 1: Basic testing and data synchronization at a single workstation CCD vision sensors at each workstation collect x / y / z axis deviation data (Δx, Δy, Δz) of the assembled parts, while equipment sensors collect assembly pressure (F). 工 ), working speed (v) 工 ), operational vibration (a 工 The data preprocessing module uses mean filtering (window size 5) to reduce noise, and then binds the "parameter data + workstation number + timestamp" to the multi-workstation traceability linkage core module and data storage module. Simultaneously, the system automatically records the unique identifier code of each workpiece (including model information, online time, and workflow sequence), achieving a one-to-one correspondence between workpieces and multi-workstation data.

[0027] Step 2: Dynamic Interference Compensation. The dynamic interference compensation module calls the synchronous detection data from the preceding adjacent workstation (force sensor data F from the preceding workstation). 前序 Vibration sensor data a 前序 ), combined with the distance between the current workstation and the previous workstation The original force and vibration data for the current workstation are corrected using the following formula: in, This is the interference compensation factor, with a value range of 0-0.8; The reference distance is fixed at 2m; , This refers to the raw force and vibration acceleration data collected by the sensors at the current workstation. , This is the corrected and valid data.

[0028] Step 3: Multi-dimensional Coupling Evaluation The multi-dimensional re-inspection unit for key processes receives valid data after dynamic interference compensation and calculates the comprehensive quality assessment value Q using a multi-physics field coupled detection model, as shown in the following formula: in, This represents the actual assembly space deviation value. The maximum allowable assembly deviation threshold for this type of impact ram is retrieved from the flexible production parameter database; , , The weight coefficients for each dimension parameter satisfy the following conditions: + + =1, the initial value is determined by the analytic hierarchy process (AHP). =0.5, =0.3, =0.2), and then dynamically adjusted using an adaptive algorithm; This is the standard assembly contact force reference value for this process. The standard vibration acceleration reference values ​​for this process are all retrieved from the flexible production parameter database. This was obtained through statistical analysis of over 1000 sets of qualified product data. Similarly). When ≥ When the (adaptive threshold) is reached, the source tracing process is triggered; < At that time, only the data is stored, and the tracing process is not initiated. Step 4: Multi-station quantitative traceability The multi-station traceability and linkage core module receives abnormal workpieces. Values, detection data for each workstation, and workstation-related parameters (spacing) Process correlation This is achieved through the following quantitative model for accurate source tracing, including: 4.1 Calculation of Root Cause Attribution Coefficient Formula symbol definition: This is the root cause attribution coefficient, with a value range of 0-n (n is the number of preceding associated workstations, the default is n=3, and it can be adjusted adaptively according to the length of the production line). The closer the value is to a certain integer The higher the confidence level, the more likely the workstation is to be the root cause workstation. =0 indicates the current workstation. =1-n represents the preceding 1-n workstations; For the first The overall quality assessment value of the workpiece at each workstation ( =0 indicates a currently abnormal workpiece. value, ≥1 indicates the detection of the same workpiece at the preceding workstation. value), This is the adaptive judgment threshold for the current process.

[0029] The workstation weighting coefficient is determined by both the "distance attenuation factor" and the "process correlation". For the first Distance between each workstation and the current workstation (unit: m). =0.8 is the distance attenuation coefficient (calibrated through 1000 sets of traceability cases); For the first The degree of process correlation between each workstation and the current workstation (e.g., the process correlation between the preceding core component assembly workstation and the final inspection workstation). =0.95, auxiliary workstation =0.3-0.6, retrieved from the flexible production parameter database). The time decay factor, This is the current inspection time at the workstation. For the first Inspection time per workstation =0.5 is the time decay coefficient, which is used to correct the impact of the time difference between different workstations on traceability (the closer the detection time, the higher the weight). Root cause identification rules: ∈[ -0.2, +0.2]( =0,1,...,n): Determine the first... Each workstation is the sole source workstation; ∈( +0.2, +1-0.2): Determine the first The workstation and the first +1 workstation is the joint root cause workstation (the anomaly is caused by the combined effect of two workstations), and is classified as follows: The ratio allocation affects the weight; When all corresponding When the value is ≤0, it is determined to be a sudden anomaly at the current workstation (such as a momentary equipment failure or a batch defect in the parts), and the traceability result is marked as "independent anomaly at the current workstation".

[0030] 4.2 Tracing the Abnormal Transmission Path The abnormal transmission intensity formula is used to trace the transmission process of deviation from the root station to the current station: Formula symbol definition: For the first From the workstation to the first The abnormal conduction intensity of +1 workstation is determined as follows: when S≥0.3, it is judged as "strong conduction" (the deviation is significantly amplified in this path); when 0.1≤S<0.3, it is "weak conduction"; and when S<0.1, it is "no conduction". For the first Deviation values ​​of core parameters for each workstation (such as assembly space deviation) Force parameter deviation ), For the first +1 corresponding parameter deviation value for each workstation; The conduction attenuation coefficient, For the first The workstation and the first +1 workstation spacing, =0.3 is the attenuation coefficient (calibrated based on deviation transmission test of different workstation spacing). The standard deviation threshold for this parameter (retrieved from the flexible production parameter database). Path tracing output: The system automatically generates an anomaly propagation path diagram, marking the root cause workstation, key propagation workstations (path nodes with S≥0.3), deviation amplification factor, and the contribution of parameter deviation of each workstation, intuitively presenting the complete process from the generation of the anomaly to its manifestation.

[0031] Step 5: Hierarchical linkage processing The process linkage control module executes targeted processing strategies based on the traceability results (root cause station, joint root cause station, transmission path), replacing the traditional unified processing method: Level 1 anomaly (0.8≤ < If the source is traced back to a slight deviation in a single workstation, calculate the compensation parameters. ( (To maximize the transmission intensity), a compensation command is sent to the next workstation, and a deviation warning is pushed to the source workstation at the same time; Level 2 abnormality ( ≤ <1.2): If it is a single root cause workstation, suspend the operation at that workstation and send out a source tracing report (including the root cause attribution coefficient). (Abnormal parameters, transmission paths), guide staff to make targeted rectifications; if it is a joint root cause workstation, suspend related workstations, analyze the collaborative impact mechanism, and formulate a comprehensive rectification plan; Level 3 abnormality ( ≥1.2): Lock the non-conforming workpiece and trigger the scrapping process. At the same time, based on the source tracing results, analyze the key nodes of abnormal transmission and strengthen the detection parameters of the workstation at that node (such as increasing the detection frequency and reducing the detection threshold). Step 6: Source tracing result-driven detection optimization The adaptive threshold optimization module incorporates source data into the threshold iteration model, achieving a closed loop of "source tracing-optimization": in, For the first The new detection threshold after the next iteration The threshold for pre-detection. The learning rate is controlled by dynamically adjusting the threshold value: this is useful when production fluctuates greatly (e.g., during line changes or batch replacements of parts). =0.03-0.05; Production is stable (no level 2 or higher abnormalities for 1 consecutive hour) =0.01-0.02, the system automatically adjusts based on production stability. The number of data samples for threshold iteration. Let be the source tracing weight coefficient, if the th If the traceability result for a workpiece is "high-frequency root cause station anomaly" (the traceability frequency for this station is ≥3 times in the past hour), then... 1.5; If it is "abnormal fixed conduction path", then =1.2; if it is an "occasional anomaly", then =0.8. This formula tilts the detection threshold towards parameters associated with high-frequency problems, enhancing the detection sensitivity for root cause anomalies. Step 7: End-to-end data storage and traceability The data storage module, serving as the system's "data hub," is based on an industrial-grade distributed database architecture. It stores traceability-related data from the entire process of impact ram assembly and testing in real time and completely, constructing a full-link data archive encompassing "data acquisition, traceability analysis, linked processing, and optimization iteration." This thoroughly resolves the pain points of existing technologies, such as "fragmented traceability data, inefficient querying, and inability to support production optimization in reverse."

[0032] The stored data not only includes core traceability results, but also links basic data from each preceding stage with optimization data from subsequent stages, forming a complete data chain: Basic raw inspection data: raw x / y / z axis deviation values ​​(Δx, Δy, Δz) collected by CCD vision sensors at each workstation, and raw force values ​​from force sensors. Vibration sensor raw acceleration And the noise-reduced data after data preprocessing, and the corrected data after dynamic interference compensation ( , Each data entry is bound to a workstation number, a data collection timestamp, and a unique workpiece identifier code. Multi-dimensional evaluation data: Comprehensive quality assessment value of key processes Adaptive decision threshold Historical iteration records; Core source attribution data: Root cause attribution coefficient The complete calculation process, root cause location determination results, abnormal conduction intensity S, conduction attenuation coefficient k, deviation amplification factor, and structured data of the abnormal conduction path diagram (node ​​location, conduction sequence, and key conduction node markers). Data linkage processing includes: hierarchical linkage processing instructions, staff rectification operation logs (rectification time, rectification measures, and review results), and non-conforming workpiece disposal records (scrap label, unique identifier, and disposal time). Optimize iterative data: trace the weight coefficients The assignment criteria include (abnormality type, high-frequency abnormality frequency statistics, and determination of the fixedness of the transmission path) and threshold optimization logs.

[0033] The system supports query needs based on different scenarios such as production management, quality analysis, and problem investigation, providing multiple core query dimensions, and the query results can be exported as structured reports (Excel / CSV format).

[0034] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it.

Claims

1. A multi-station online inspection system for impact rammers, characterized in that, It includes a single-station basic inspection unit that enables real-time data interaction via industrial Ethernet, a multi-station traceability and linkage core module, an anomaly propagation path analysis module, and an adaptive threshold optimization module. The single-station basic inspection unit collects assembly deviations and equipment operating parameters from each station, and transmits them after preprocessing by binding them with a unique workpiece identifier. The multi-station traceability and linkage core module integrates a quantitative traceability model, which accurately locates the root cause station and associated affected stations of quality problems by calculating the root cause attribution coefficient. The anomaly propagation path analysis module traces the propagation process of anomalies from the root cause station to the current station and identifies key nodes based on the time series of inspection data and the correlation between parameters. The adaptive threshold optimization module dynamically updates the quality judgment threshold by combining the traceability data, forming a closed loop of "detection-traceability-optimization"; In the quantitative tracing model of the multi-station tracing linkage core module, the formula for calculating the root cause attribution coefficient R is as follows: in: This is the workstation weighting coefficient. =0.8 is the distance attenuation coefficient. For workstation spacing, For process correlation; The time decay factor, =0.5 is the time decay coefficient. , The current and the number are respectively Inspection time per workstation; This represents the overall quality assessment value for the j-th workstation. The adaptive threshold is set to n, which is the number of preceding associated workstations (default 3). The abnormal propagation path analysis module uses the following formula: Tracing the transmission of abnormalities; among which For conduction strength, , This refers to the deviation value of the core parameters of the workstation. The conduction attenuation coefficient, =0.3 is the attenuation coefficient. For workstation spacing, S represents the standard deviation threshold for the parameter; S≥0.3 indicates strong conduction, 0.1≤S<0.3 indicates weak conduction, and S<0.1 indicates no conduction.

2. The multi-station online detection system for impact rammers according to claim 1, characterized in that, It also includes a multi-dimensional re-inspection unit for key processes. This unit is set after the key process of impact tamping assembly. It integrates a CCD vision inspection module, force sensor, vibration sensor and data fusion module to collect multi-dimensional data such as assembly space deviation, impact contact force and instantaneous vibration acceleration, and generate comprehensive quality assessment data.

3. The multi-station online detection system for impact rammers according to claim 1, characterized in that, The single-station basic detection unit includes a CCD vision sensor and a data preprocessing module. The CCD vision sensor acquires x / y / z axis deviation data Δx, Δy, Δz, and the equipment sensor acquires assembly pressure F. 工 Working speed v 工 , Operational vibration a 工 The data preprocessing module uses mean filtering with a window size of 5 for noise reduction. The unique identifier of the workpiece includes the model, online time, and transfer station sequence information.

4. The multi-station online detection system for impact rammers according to claim 1, characterized in that, The threshold iteration formula of the adaptive threshold optimization module is: in For learning rate, when production fluctuates greatly =0.03-0.05, when production is stable =0.01-0.02; For sample size; This is the overall quality assessment value; This is the time decay factor; For source tracing weight coefficients, when there are high-frequency root cause anomalies =1.5, when the fixed conduction path is abnormal =1.2, during occasional anomalies =0.

8.

5. The multi-station online inspection system for impact rammers according to claim 1, characterized in that, It also includes a data storage module and a flexible production parameter database; the data storage module stores data from the entire process of detection, traceability, processing, and optimization. The flexible production parameter database stores assembly standard parameters, process correlations, and abnormal conduction attenuation coefficient data for different models of impact rammers, providing support for each module.

6. A multi-station online detection method for impact tamping, characterized in that, The system implementation based on any one of claims 1-5 includes: Step 1: Basic detection and data synchronization at each workstation. Sensors at each workstation collect parameter data, which is then preprocessed and bound to the workstation number, timestamp, and unique workpiece identifier code before transmission and storage. Step 2: The dynamic interference compensation module calls the synchronous detection data of the adjacent workstations in the previous step, and corrects the original force and vibration data of the current workstation through the correction formula; Step 3: The multi-dimensional re-inspection unit for key processes receives valid data, calculates the comprehensive quality assessment value Q through a multi-physics field coupled detection model, and determines whether to trigger the traceability process; Step 4: The multi-workstation traceability linkage core module calculates the root cause attribution coefficient through the quantitative traceability model, locates the root cause workstation by combining the judgment rules, and the anomaly propagation path analysis module traces the propagation process and generates a path diagram. Step 5: Execute the coordinated processing strategy according to the source tracing results and Q-values; Step 6: The adaptive threshold optimization module iteratively updates the detection threshold based on the source data; Step 7: The data storage module stores all relevant data throughout the process and supports multi-dimensional query and traceability.

7. The multi-station online detection method for impact rammers according to claim 6, characterized in that, The tiered processing strategy in step 5 is as follows: Level 1 anomalies 0.8 ≤ < Send compensation instructions and deviation warnings; Level 2 anomaly ≤ <1.2 Suspend the corresponding workstation and formulate a rectification plan; Level 3 abnormality ≥1.2 Lock the workpiece and strengthen the detection of key nodes.

Images