Method for monitoring state of cold header for bolt machining

Abstract

The invention provides a method for monitoring the state of a cold header for bolt machining, and belongs to the technical field of cold header monitoring, and the method comprises the steps: obtaining the specification information of a bolt product machined by the cold header, and extracting an initial forming parameter set and a dynamic adjustment parameter set based on the geometric dimension and material characteristics of the bolt; identifying a process trigger signal in the machining process based on the parameter set, and extracting parameter change nodes which comprise a parameter fine tuning node triggered by equipment state change during machining of the same bolt specification and a parameter overall updating node triggered during switching machining of different bolt specifications; distinguishing and classifying the same and different parameter change nodes, and calculating a process offset; based on the process offset, the process generality characteristic parameters and the process exclusive characteristic parameters are separated, a state early warning mechanism is triggered, the process generality characteristic parameters are suitable for machining of different bolt specifications, and the process exclusive characteristic parameters are only suitable for machining of specific bolt specifications. The method is suitable for machining bolts of different specifications, and the monitoring adaptability and accuracy are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of cold header monitoring, more specifically, it relates to a cold header state monitoring method for bolt processing. BACKGROUND

[0002] In the field of modern mechanical manufacturing, as a basic connecting piece, the quality of the bolt directly affects the safety and reliability of the whole machine equipment. Cold heading process has become the core process means of bolt processing due to its high material utilization rate, fast production efficiency, excellent product mechanical properties and other characteristics, and the cold header is the key equipment to realize this process. With the rapid development of high-end equipment manufacturing industries such as automobiles, high-speed rails, aerospace, etc., the market has put forward more stringent requirements on the precision, strength and consistency of the bolt, which also puts higher challenges on the stable running state of the cold header and the precise control of the processing process. The processing process of the cold header involves the synergistic action of multiple dies, and the continuous heading and forging of the punch and the concave die are required to realize the key processes such as head forming and rod diameter reduction of the bolt. Its running state is influenced by a variety of factors, including material properties (such as material hardness, ductility), equipment mechanical properties (such as die wear, slide gap), process parameter settings (such as heading force, feed speed, die temperature), etc. In actual production, even a small parameter fluctuation or equipment state change may cause the bolt to have quality problems such as head cracking, size out-of-tolerance, thread defects, etc., not only causing waste of raw materials and increase in production cost, but also possibly causing serious safety hazards due to the inflow of unqualified products into the market. The state monitoring of the cold header mainly relies on traditional manual inspection and offline detection methods. The operator needs to periodically make subjective judgments on the equipment running sound and vibration, or indirectly assess the equipment state by sampling and detecting the geometric dimensions of the bolt finished product. This method has obvious limitations: on the one hand, manual monitoring relies on the experience of the operator and is subjective, making it difficult to capture early subtle abnormalities of the equipment; on the other hand, offline detection has time lag, and cannot reflect the state change in the processing process in real time. When quality problems are found, a batch of unqualified products has usually been produced. With the development of industrial automation technology, some cold header machines are equipped with simple sensor monitoring systems that can collect single parameters such as upsetting force and temperature in real time. However, the existing monitoring methods still have significant shortcomings: they do not differentiate according to the diversity of bolt product specifications. Different specifications of bolts (such as different diameters, lengths, and head shapes) correspond to different cold heading process parameters. The universal parameter monitoring threshold is difficult to adapt to multi-specification production scenarios, which may lead to false positives or false negatives. The dynamic characteristics of parameter change nodes are ignored. In actual production, when processing the same specification of bolts, factors such as equipment mold wear and material batch changes may cause parameter fine-tuning nodes. When switching to process different specifications of bolts, parameter overall updating nodes may occur. The existing system cannot effectively identify and distinguish these nodes, resulting in inaccurate process deviation calculation. The existing methods lack scientific classification of process characteristic parameters. In cold heading processing, some parameters (such as equipment idle running temperature and basic lubrication pressure) are general stable parameters, while upsetting force peak and feed speed are strongly related to specific bolt specifications. The existing methods do not separate the two types of parameters, resulting in insufficient pertinence and accuracy of state early warning. With the deepening of the concept of intelligent manufacturing, cold heading production is developing towards multi-variety, small-batch, and flexibility. The equipment needs to frequently switch between processing specifications, and the traditional static threshold monitoring mode cannot meet the dynamic production demand. Developing a cold header machine state monitoring method that can combine bolt product specification information, accurately identify process parameter change nodes, and separate common and exclusive characteristic parameters is of great significance for improving the quality stability of bolt processing, reducing production costs, and promoting the intelligent upgrading of cold heading processing. SUMMARY

[0003] The purpose of the present application is to provide a cold header machine state monitoring method for bolt processing, which can be applied to different specifications of bolts and improve monitoring adaptability and accuracy.

[0004] The embodiments of the present application provide a cold header machine state monitoring method for bolt processing, which comprises: obtaining bolt product specification information of a cold header machine; based on the bolt product specification information, extracting an initial forming parameter set and a dynamic adjustment parameter set of the cold header machine; the bolt product specification information includes bolt geometric dimensions and material properties; Based on the initial forming parameter set and the dynamic adjustment parameter set, identifying a process trigger signal of the cold header machine during processing; based on the process trigger signal, extracting a parameter change node of the cold header machine; the parameter change node includes a parameter fine-tuning node caused by equipment state changes when processing the same specification of bolts and a parameter overall updating node triggered when processing different specifications of bolts; Based on the parameter change node, distinguishing and classifying the same parameter change node and the different parameter change node, and calculating the process deviation; Based on the process offset, separate process common characteristic parameters and process exclusive characteristic parameters; based on the process common characteristic parameters and process exclusive characteristic parameters, trigger the state early warning mechanism of the cold header; the process common characteristic parameters refer to general stable parameters applicable to the processing of different bolt specifications, and the process exclusive characteristic parameters refer to key control parameters applicable only to the processing of specific bolt specifications.

[0005] In an embodiment of the present application, the specification information of the bolt product processed by the cold header is acquired, and based on the specification information of the bolt product, an initial forming parameter set and a dynamic adjustment parameter set of the cold header are extracted, including: setting the reference parameter values and their effective operation time lengths of the cold header in the feeding, pre-forming and final forging stages according to the bolt geometric dimensions and material properties; According to the switching processing nodes of different bolt specifications, the correction parameter values and their action time periods of the cold header in the feeding, pre-forming and final forging stages are recorded; According to the reference parameter values and their effective operation time lengths, the verified correction parameter values are calculated by applying parameter drift verification rules; The reference parameter values and the verified correction parameter values are merged to form a complete process parameter record of the cold header.

[0006] In an embodiment of the present application, based on the initial forming parameter set and the dynamic adjustment parameter set, the process trigger signal of the cold header in the processing process is identified, including: The initial forming parameter set is analyzed to generate a first type of process trigger signal; The dynamic adjustment parameter set is analyzed to generate a second type of process trigger signal; The first type of process trigger signal and the second type of process trigger signal are fused to analyze the signal trigger rule in the feeding, pre-forming and final forging stages when the same bolt specification processing and different bolt specification switching processing are performed; According to the signal trigger rule, the capture time of the cold header process trigger signal is dynamically adjusted In a fourth aspect of the embodiments of the present application, based on the process trigger signal, the parameter change node of the cold header is extracted, including: The trend similarity of the process trigger signal and historical parameter fine-tuning node reference data is calculated, and when the trend similarity exceeds a set threshold, it is determined as a device state fine-tuning node and classified as a parameter fine-tuning node of the same bolt specification; The trend similarity of the process trigger signal and historical parameter update node reference data is calculated, and when the trend similarity is lower than a set threshold, it is determined as a specification switching update node and classified as a parameter overall update node of different bolt specifications.

[0007] In an embodiment of the present application, based on the parameter change node, the same parameter change node and the different parameter change node are distinguished and classified, and the process offset is calculated, including: The parameter fine-tuning node caused by the change of the equipment state during the same bolt specification processing is marked as the same parameter change node. The parameter overall updating node triggered during the different bolt specification switching processing is marked as the different parameter change node. Based on the same parameter change node, the parameter cumulative change amount in the same specification processing period is counted as the first type of process offset. Based on the different parameter change node, the parameter mutation difference in the specification switching process is calculated as the second type of process offset. The first type of process offset and the second type of process offset are fused to generate a comprehensive process offset.

[0008] In an embodiment of the present application, based on the process offset, the process common characteristic parameter and the process exclusive characteristic parameter are separated, including: Taking feeding, preforming and final forging as the analysis dimensions, the stable characteristics repeatedly appearing in the comprehensive process offset are identified, the parameter set that can be reused across bolt specifications is extracted, and is marked as the process common characteristic parameter. The unique characteristics changing with the bolt specification in the comprehensive process offset are identified, the parameter set bound to a specific specification is extracted, and is marked as the process exclusive characteristic parameter.

[0009] Based on the process common characteristic parameter and the process exclusive characteristic parameter, a state early warning mechanism of the cold header is triggered, including: The process common characteristic parameter is used as a steady-state monitoring benchmark and applied to the processing process monitoring of all bolt specifications. The process exclusive characteristic parameter is combined to perform real-time parameter calibration on the processing process of a specific bolt specification, and the steady-state monitoring benchmark is linked to realize abnormal state determination.

[0010] In an embodiment of the present application, according to the signal trigger rule, the capture timing of the cold header process trigger signal is dynamically adjusted, including: If the strength of the signal trigger rule is higher than the preset strength threshold, the process trigger signal of the cold header is captured in advance; If the strength of the signal trigger rule is lower than the preset strength threshold, the process trigger signal of the cold header is captured with delay.

[0011] The trend similarity is obtained by calculating the covariance relationship between the process trigger signal timing data and the historical node reference data.

[0012] In an embodiment of the present application, the method further includes: establishing a device degradation evaluation model according to the comprehensive process deviation; generating a maintenance decision instruction based on the device degradation evaluation model and a real-time captured process trigger signal; performing a preventive maintenance operation of the cold header according to the maintenance decision instruction.

[0013] The cold header state monitoring method for bolt processing provided by the embodiment has the beneficial effects that: The cold header state monitoring method starts from bolt product specification information, realizes accurate monitoring of the processing process through systematic parameter extraction, signal recognition, node differentiation and feature separation. Based on the bolt geometric size and material property, the initial forming parameter set and the dynamic adjustment parameter set are extracted, so that the monitoring process is closely combined with the specific processing demand, and the problem of disconnection between parameter setting and actual processing in traditional monitoring is avoided. This product specification-oriented parameter extraction method ensures that the parameters for monitoring can truly reflect the core control elements of the processing process, laying a foundation for subsequent state evaluation.

[0014] In the process of identifying the process trigger signal and the parameter change node, the method can accurately capture the parameter fine-tuning node during processing of the same bolt specification and the parameter overall updating node during switching of different specifications. This detailed differentiation of parameter change nodes changes the general treatment of parameter changes in traditional monitoring, so that each parameter adjustment in the processing process can be clearly identified. By distinguishing between the same parameter change node and the different parameter change node and calculating the process deviation, the degree and trend of parameter change can be quantified, so that the change of the device state is presented in a measurable way, which is convenient for more intuitive judgment of whether the processing process is in a stable state.

[0015] Separating process common characteristic parameters and process exclusive characteristic parameters makes the monitoring more targeted. Process common characteristic parameters are suitable for processing of different bolt specifications, and their stability is directly related to the basic state of the overall operation of the cold header. Monitoring of such parameters can timely find common problems affecting the processing quality of multiple specifications of bolts. Process exclusive characteristic parameters are only suitable for processing of specific specifications of bolts, and monitoring of such parameters can accurately locate the abnormalities in the processing of specific specification products. This way of classified monitoring avoids the problem of early warning ambiguity caused by parameter mixing in traditional monitoring, so that the early warning signal can accurately point to the problem.

[0016] Based on the above characteristic parameter trigger state early warning mechanism, parameter abnormalities can be found in time during the machining process. When the common characteristic parameters deviate, the potential problems of the overall operation state of the equipment can be prompted, facilitating early maintenance and adjustment; when the exclusive characteristic parameters are abnormal, the machining problems of the specific specification bolt can be quickly located, facilitating targeted optimization of parameters. This precise early warning mechanism can intervene in time before the problem expands, reduce product quality problems caused by parameter abnormalities, and reduce the probability of waste product generation. At the same time, the method is suitable for the machining monitoring of different specification bolts, and does not need to build a monitoring model for each specification, thereby enhancing the adaptability of the monitoring system, flexibly coping with diversified production demands, reducing the cumbersome process of frequently adjusting the monitoring settings due to product specification switching, and helping to maintain the continuity and stability of the production process. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creating laborious work.

[0018] Figure 1 The timing diagram of the bolt machining cold header state monitoring method provided by an embodiment of the present application; Figure 2 The flowchart of initial and dynamic parameter set extraction provided by an embodiment of the present application; Figure 3 The flowchart of process offset calculation provided by an embodiment of the present application; Figure 4 The flowchart of cold header state early warning mechanism triggering provided by an embodiment of the present application. DETAILED DESCRIPTION

[0019] In the following description, specific details such as specific system structures, techniques, etc. are presented in order to thoroughly understand the embodiments of the present application. However, it should be clear to those skilled in the art that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits and methods are omitted to avoid unnecessary details that hinder the description of the present application.

[0020] In order to make the purpose, technical solutions and advantages of the present application clearer, the following will be described by specific embodiments in conjunction with the drawings.

[0021] Please refer to Figure 1 The present application provides a bolt machining cold header state monitoring method, which comprises: By real-time acquisition and analysis of process parameters in the cold heading machine processing, the precise monitoring and early warning of the equipment state are realized. First, the bolt product specification information is acquired, including bolt geometric dimensions and material properties, based on which the initial forming parameter set and the dynamic adjustment parameter set of the cold heading machine are extracted. The initial forming parameter set includes the baseline parameter values and their effective running time of the feeding, preforming and final forging stages, and the dynamic adjustment parameter set records the correction parameter values and their action time period when different bolt specifications are switched for processing. Through the identification of process trigger signals, parameter change nodes are extracted, including parameter fine-tuning nodes when the same bolt specification is processed and parameter overall updating nodes when different bolt specifications are switched. Based on the parameter change nodes, the process offset is calculated, and the process common characteristic parameters and the process exclusive characteristic parameters are separated, finally triggering the state early warning mechanism of the cold heading machine. The process common characteristic parameters are suitable for the general stable parameters for processing different bolt specifications, and the process exclusive characteristic parameters are only suitable for the key control parameters for processing a specific bolt specification.

[0022] Example 1: refer to Figure 2 The bolt product specification information includes bolt geometric dimensions and material properties. After obtaining the specification information, the baseline parameter values of the feeding, preforming and final forging stages of the cold heading machine are set according to the material deformation characteristics and size tolerance range. The baseline wire feeding speed and clamping force threshold are set in the feeding stage; the initial stroke pressure of the punch and the positioning accuracy of the concave die are set in the preforming stage; the number of blows and the standard value of the die closing gap are set in the final forging stage. The effective running time of each baseline parameter is determined according to the single-piece processing cycle corresponding to the bolt specification multiplied by the batch quantity, for example, a certain specification takes 0.8 seconds for single-piece processing, and the standard batch quantity is 3000 pieces, so the effective running time is 40 minutes.

[0023] When the cold heading machine switches to process different bolt specifications, the correction parameter values are recorded at the specification change time. The correction parameters cover the compensation feeding rate in the feeding stage, the punch pressure offset in the preforming stage, and the die temperature calibration value in the final forging stage. The action time period of these correction values is accurate to the millisecond timestamp, for example, the preforming pressure needs to be increased by 5% from 2.3 seconds to 45 seconds after the specification is switched. When the parameter drift verification rule is executed, the same specification processing records in the cold heading machine historical database are called, and the current correction parameters are compared with the historical same working condition parameter distribution interval. If the current feeding speed correction value exceeds the historical same specification speed fluctuation range ±10%, it is determined that the correction value is abnormal and the historical average value is replaced; if it is within the allowed range, the current correction value is retained. The baseline parameter values and the verified correction parameter values are aligned and merged according to the time axis: the specification switching point is taken as the boundary, the baseline parameter record before switching is retained, the correction parameter record after switching is inserted after passing the verification, and the process parameter set covering the complete processing cycle is formed.

[0024] The initial set of forming parameters is processed by the signal analysis engine to generate the first type of process trigger signal. This signal is mapped to three-stage parameter steady-state operating characteristics: the feeding signal corresponds to the material delivery uniformity identifier, the preforming signal reflects the punch pressure fluctuation frequency value, and the final forging signal outputs the forging vibration energy level index. The dynamic adjustment parameter set is analyzed to generate the second type of process trigger signal, which is strongly related to the parameter correction behavior: the feeding signal mutates into a feed rate step pulse, the preforming signal presents a pressure adjustment ramp waveform, and the final forging signal is converted into a die gap change differential value. The two types of signals are matched and fused through time stamp matching to form a composite waveform sequence on the signal graph.

[0025] In the same bolt specification continuous processing period, the feeding signal presents a low-frequency fluctuation below 0.5 Hz, the preforming signal maintains a stable oscillation with a pressure amplitude of ±3%, and the final forging signal shows a vibration energy level decay pulse every 50 processing periods. When switching between different specifications, the feeding signal produces a rate step with an amplitude exceeding the reference value by 30% within 100 milliseconds after switching, the preforming signal forms a pressure ramp rising edge lasting 2 seconds, and the final forging signal generates a die gap mutation spike. This law reflects the transient response of the signal at the specification switching node and the steady-state maintenance characteristics during continuous processing.

[0026] According to the above law, the dynamic adjustment signal capture strategy is as follows: when the preforming pressure amplitude exceeds ±5% for three consecutive cycles, it is determined that the signal strength exceeds the preset threshold, and the early capture mechanism is started. The early capture uses a signal pre-trigger mode, and a high-speed sampling unit is deployed 50 milliseconds before the start of the next processing cycle. When the final forging vibration energy level index is continuously below the historical average by 20% for 10 cycles, it is determined that the signal strength is insufficient, and the delayed capture mechanism is activated. The delayed capture is executed by a downsampling filter, which switches the original 1 millisecond sampling interval to a 10 millisecond interval, and only records effective signals with an amplitude exceeding the noise threshold of 0.1V. This process achieves strategy switching by real-time monitoring of the frequency energy distribution of the signal graph.

[0027] During signal processing, the rate step pulse in the feeding stage needs to be identified. A sliding time window algorithm is used to set an analysis window of 10 milliseconds in the 0-200 millisecond interval after specification switching, and when the signal slope change rate in the window exceeds 5V / ms, it is marked as the step start point. The preforming pressure ramp waveform is fitted by linear regression, and the continuous rising / falling edge with a duration exceeding 500 milliseconds and a fitting goodness greater than 0.95 is extracted as the effective adjustment signal. The processing of the final forging die gap differential signal needs to be associated with the displacement sensor data, converting the electrical signal amplitude into micron-level displacement, and triggering the record when the displacement mutation exceeds 10 microns.

[0028] The data storage adopts a hierarchical structure: the original process parameters are compressed and stored at a minute level, the trigger signal characteristic values are stored in the cache area at a millisecond level, and the signal waveform fragments are stored in the historical database after analysis. During the continuous 8-hour operation cycle, the system maintains a signal capture delay of less than 500 microseconds, and the data integrity rate reaches 99.2% coverage of the process parameter points. The historical database stores typical signal patterns by bolt specifications, and when a new specification is processed, it automatically associates similar historical signal characteristics through a nearest neighbor matching algorithm.

[0029] The process parameter record update mechanism includes bidirectional verification: after updating the reference parameters, the signal reanalysis process is triggered, and the dynamic correction parameters are written in synchronization with the historical data compatibility verification. When the material characteristics change, the system automatically traces back to the same batch processing record within 24 hours, and compensates and calibrates the offset of the affected dynamic correction parameters. This closed-loop process ensures the spatiotemporal consistency of the process parameter set in the continuous operation state of the device.

[0030] Example 2: refer to Figure 3 The trend similarity calculation of the process trigger signal uses the following formula:

[0031] Here the symbols are defined as: represents the trend similarity calculation result, the numerical interval is [-1, 1]; is the total amount of time series data, covering the entire processing period; is the measured value of the current process trigger signal at time point ; is the average value of the current process trigger signal in the time period ; is the reference value of the historical reference data at time point ; is the average value of the historical reference data in the time period .

[0032] The historical parameter fine-tuning node reference data construction needs to call the same specification processing records of the previous three months of the cold header. Select stable processing period data as the basic template, and exclude abnormal fragments during start and stop. The reference data preprocessing includes three steps: the feeding speed sequence is filtered with a sliding average filter with a window width of 0.5 seconds; the preforming pressure data is removed by percentile screening to remove the first 5% extreme values; the final forging vibration signal performs frequency domain noise reduction to remove high-frequency components above 100 Hz. The processed data is segmented according to the processing stage, and each segment is labeled with the device state type as a classification label.

[0033] The current process trigger signal and the trend similarity of the fine-tuning node reference data require timestamp forced alignment. The alignment method uses dynamic time warping algorithm to compensate for the time difference of different processing rhythms. After alignment, the data is calculated by 200 milliseconds equal length slice by segment, and the slice length is the basic period of single cold heading action. When the similarity of more than 70% of the slice segment is greater than 0.85 , it is determined that the device state fine-tuning node is classified into the same bolt specification parameter fine-tuning node set. If the specification switching scene, the similarity of the continuous three slice segments is less than 0.2, it is determined that the specification switching update node is classified into the different bolt specification parameter overall update node set. The node classification result is stored as a four-dimensional coordinate: timestamp, processing stage, node type, and similarity evidence value.

[0034] The identification of the same parameter change node needs to be associated with the device operation log. During the continuous processing of the same specification bolt in the cold header, the temperature sensor records the device spindle temperature rise curve, and the bearing wear index output by the vibration monitoring system. When the temperature rise rate is greater than 0.8℃ / min or the wear index single day increment is greater than 5, and the process parameters are adjusted at this time, the time point is marked as "parameter fine-tuning node caused by device state change". The node data packet includes the parameter difference value before and after adjustment, device state index, and duration.

[0035] The capture of different parameter change nodes depends on the specification switching instruction. When the host computer issues a product specification change code, the system starts the parameter snapshot function 50 milliseconds before the instruction execution, and records the final state parameter value before switching. Within 300 milliseconds after switching is completed, the initial parameter value of the new specification is collected. The parameter mutation difference value calculation uses the absolute value difference method, and the wire feeding amount percentage change :

[0036] Wherein: is the final state feeding speed of the old specification, is the initial feeding speed of the new specification. The preforming stage calculates the actual deviation of the pressure set value, and the final forging stage compares the adjustment amplitude of the die gap. All difference values are associated with the material property change index, such as the additional material hardness correction coefficient when switching from carbon steel to stainless steel.

[0037] The first type of process offset statistics is performed within the same specification processing period. The offset is divided into a basic value and a derivative: the basic value records the actual change in the parameter, such as the cumulative offset of the feeding speed +0.2 mm / s; the derivative calculates the change trend, and the hourly offset rate is obtained by linear regression. The statistical process sets three filters: filter out transient disturbances with a duration of less than 5 seconds; shield parameter jumps caused by manual intervention; ignore calibration data within 24 hours after equipment maintenance. The cumulative change is stored as an incremental matrix, with rows corresponding to processing stages and columns representing time dimensions.

[0038] The second type of process offset calculation sets dynamic weights. When the specification is switched, the feeding parameter weight accounts for 40%, the preforming weight accounts for 35%, and the final forging weight accounts for 25%. The weight distribution is based on the history of cold header failures: feeding abnormalities account for 38% of downtime, preforming defect rate accounts for 42%, and final forging die damage accounts for 20%. The parameter mutation difference is weighted to generate a standard offset. When switching from carbon steel M8 bolts to stainless steel M10 bolts, if the feeding speed difference is -12%, the preforming pressure difference is +8%, and the final forging gap difference is +15%, the comprehensive offset is calculated as: This value is stored in the specification switching offset feature library as a reference for subsequent similar switching.

[0039] The comprehensive process offset generation requires time sequence superposition processing. The time axis is divided by the specification switching point, with the left side continuously accumulating the first type of offset and the right side recording the second type of offset from zero. The superposition area is the specification switching transition period, and the data in this period is marked with double axes: the main time axis records the offset of the new specification, and the auxiliary time axis retains the cumulative amount of the old specification. The transition period length is set according to the material type, with 15 seconds for carbon steel processing and 30 seconds for alloy steel processing. On the comprehensive graph, the parameter stable area shows low slope linear growth, the switching point presents stepwise transition, and the transition area contains the crossfade dynamics of the two types of offsets.

[0040] The data verification mechanism includes hardware redundancy design. Each type of offset is calculated and sent to three independent controllers: the main control PLC executes device control decisions; the edge computing node stores raw data; and the cloud server performs logical verification. When the difference between the three results exceeds 5%, the re-computation process is triggered. The re-computation adopts a degradation strategy: close the signal acquisition channel with a real-time requirement lower than 50 milliseconds, release the computing power to ensure the accuracy of the offset calculation. At the same time, the device operating condition scanning is started to exclude abnormal offsets caused by sensor failures.

[0041] The offset storage structure adopts a tree index. The root node is the device number, the first branch distinguishes the same / different parameter change node type, the second branch is divided into three stages of feeding / preforming / terminal forging, and the leaf node stores the specific offset value and its occurrence timestamp. In an 8-hour working period, the system generates an average of 1500 offset data points, with an original accuracy of 0.1% measurement error range.

[0042] The learning mechanism of the specification switching offset involves model feedback. After each specification switch, the actual offset data is compared with the predicted model output value, and if the difference exceeds the safety threshold, the model is updated. The training set contains the records of the previous 20 similar switches, and the feature parameters are extracted using principal component analysis to reduce the dimension to three core dimensions: material hardness difference, size change rate, and equipment running time. The training period is set to the idle period of the equipment after the switch is completed, avoiding the occupation of computing resources during the production peak period.

[0043] The visualization output of the process offset includes a two-dimensional graph and a three-color warning zone. The horizontal axis is the processing timeline, and the vertical axis is the normalized offset value. The green safe zone is ±5% of the baseline value, the yellow warning zone is ±5-15%, and the red warning zone is more than 15%. The same specification offset is displayed as a solid line, and the different specification switching points are marked as a dashed line with a triangle marker. The interface refreshes the data every 30 seconds, and the historical data backtracking supports a 12-hour deep trend playback.

[0044] Example 3: refer to Figure 4 The extraction algorithm of the process common feature parameters is as follows:

[0045] The formula symbol definitions are as follows: represents the process common feature parameter matrix, with a dimension of 3×M (3 analysis dimensions × feature quantity); is the total number of bolt specification categories, covering all processing history records; represents the number of comprehensive process offset data points for specification k; is the i-th offset value of specification k; is the duration of the i-th offset; is the total processing duration of specification k; is the weight coefficient matrix of specification k, calculated according to the specification usage frequency.

[0046] In the feeding dimension analysis, the stable feature that repeatedly appears is the wire feeding speed offset. Extracting continuous 100 batches of processing data, the feeding speed offset of carbon steel M6 to M24 specifications is calculated: when the environmental temperature is in the range of 18-25℃, the speed offset is maintained within the range of ±0.15mm / s of the baseline value, and this parameter is marked as a process common feature parameter. Similarly, in the preforming dimension, the pressure offset of different bolt lengths shows a same-direction change trend within the mold life cycle, and the pressure threshold offset converges in the interval of ±0.8MPa, forming a common parameter set. The vibration energy level offset in the final forging dimension shows similar exponential decay curves in aluminum alloy and carbon steel processing, and this model parameter is included in the common feature library. The common feature parameters are stored in the public area independent of the specification, and the historical records are managed through version numbers.

[0047] The binding rules for process-specific feature parameters are based on the core factor of bolt geometry. Feeding-specific parameters are associated with bolt shank diameter: when the shank diameter is ≤8mm, the specific parameter is the "elastic compensation coefficient"; when the shank diameter is >8mm, it is bound to the "rigid damping factor". Pre-forming-specific parameters are determined in three levels: bolt length <20mm uses the short-piece forming curve, 20-50mm uses the medium-long-piece pressure gradient, and >50mm activates the long-piece support mode parameter. The final forging die temperature compensation coefficient is a specific parameter, and its assignment is only effective when the bolt material is stainless steel and the head diameter is ≥12mm. Specific feature parameters are stored in a specification-related encrypted partition; access requires verification of the product specification code.

[0048] The steady-state monitoring baseline is constructed using a two-layer architecture. The base layer defines the allowable fluctuation range for common characteristics: the feeding speed baseline range is set to the absolute value range of the current specification baseline value ±0.3 mm / s; the preforming pressure baseline is set with a relative threshold fluctuation band of ±5%; the final forging vibration level baseline uses dynamic envelope control, with an upper limit of the historical average + 2 standard deviations. The reinforcement layer establishes association rules: when the ambient humidity >70%, the feeding baseline range is automatically compressed to ±0.1 mm / s; when the equipment has been running for more than 8 hours, the preforming pressure baseline is relaxed to ±8%. The baseline data is automatically checked and updated every 24 hours, and change records generate a version history.

[0049] Real-time parameter calibration executes a three-step closed-loop control. Step 1: Acquire actual operating values: feeding speed sampling interval is 50ms, preforming pressure is measured at 5 equally spaced points per stroke, and the peak value of a single forging is recorded at the final forging vibration level. Step 2: Calculate the specific characteristic deviation rate:

[0050] in: These are measured values. These are standard values ​​for parameters specific to the current specification. This is the compensation base for the current equipment status. Step 3 triggers the calibration command: when the feeding speed... Adjust the servo motor torque after 3 seconds; when preforming... During two consecutive strokes, the hydraulic system pressure valve is adjusted; during final forging... Furthermore, if the temperature exceeds the limit, inject coolant and reduce the speed by 20%.

[0051] The abnormal state determination adopts complex logic. The primary determination monitors the steady-state reference: any parameter exceeding the limit for 5 consecutive sampling points triggers a yellow warning. The secondary determination implements correlation analysis: when the preforming pressure and the feeding speed exceed the reference value of 120% at the same time, the material hardness is determined to be abnormal. The tertiary determination checks the calibration effect of the exclusive parameter: if the deviation rate of the exclusive parameter after calibration is still higher than the threshold of 150%, a red stop alarm is triggered. The determination results are classified into four levels of response: level 1 records the log; level 2 sounds and lights alarm; level 3 runs at a reduced speed; level 4 emergency brakes.

[0052] The feature parameter storage architecture is designed as a three-dimensional tensor. The X-axis stores the data according to the feeding / preforming / forging dimensions; the Y-axis distinguishes between common and exclusive feature types; and the Z-axis records the time version sequence. Each feature parameter contains 12 attribute fields: parameter ID, applicable specification range, data source batch, last calibration time, invalidation identifier, historical maximum value, historical minimum value, associated sensor list, responsible person code, modification lock status, reference citation number, and data fingerprint. Data compression uses differential encoding technology, with a storage space compression rate of 35% of the original data.

[0053] The real-time calibration system runs in an independent safety domain. Hardware deployment has triple redundancy: the main controller executes calibration decisions; the backup controller synchronously calculates deviation rates; and the safety monitor continuously checks the consistency of the results. Data flow implements signature verification: a 256-bit digital signature is attached when updating feature parameters, and the certificate chain is verified during calibration. The system clock maintains a microsecond-level error, and the timestamp is used as the key sequencing basis for calibration instruction execution.

[0054] The cold header running state visualization interface includes three-color rendering: common parameters are displayed on a blue background, exclusive parameters are marked with a yellow border, and parameters in real-time calibration flash red for warning. Operators can directly modify exclusive parameter values through force feedback devices, and the modification process is forced to record operation videos and synchronize to the audit system. After each parameter change, the system automatically initiates a 48-hour data trace to detect changes in associated equipment indicators.

[0055] The feature parameter recycling mechanism includes an automatic forgetting algorithm. When a certain exclusive parameter is not called for six consecutive batches, the system marks it as a recycling state. Before recycling, data value evaluation is performed: the contribution weight of the parameter in historical abnormal diagnosis is calculated, and when the weight is less than 0.05, it is transferred to the archive storage. Common parameters perform importance sorting every quarter, and the last 10% of parameters are transferred to the observation library to make room for new features. Recycling is performed during equipment maintenance period to avoid affecting normal production.

[0056] Example 4: The signal trigger rule strength evaluation system uses multi-dimensional quantitative analysis to achieve dynamic signal capture adjustment by monitoring the parameter change characteristics in the cold heading machine processing. The system establishes a signal strength evaluation index system, including amplitude change rate, frequency stability, and waveform repeatability. The amplitude change rate calculates the deviation of the current signal peak value from the historical reference value, the frequency stability analyzes the retention ability of the signal periodicity characteristics, and the waveform repeatability detects the similarity characteristics of adjacent processing period signals. The three indexes are fused by weighting to form a comprehensive strength score, which is used as the decision basis for signal capture timing adjustment.

[0057] The signal capture timing dynamic adjustment mechanism includes two parallel working modes. The early capture mode is activated when the strength score exceeds the preset threshold, and the system automatically performs three operations: increasing the power supply voltage of the high-speed sampling module to 12V to improve the sampling rate; expanding the signal buffer capacity to 8MB; starting the dedicated digital signal processing core for real-time feature extraction. The delay capture mode runs when the strength score is below the threshold, adopting a degradation processing strategy: closing two redundant sampling channels; reducing the analog signal filter cutoff frequency to 500Hz; using a 1 / 4 compression rate sampling algorithm to reduce data volume. The mode switching response time is controlled within 5 milliseconds to ensure signal acquisition continuity. See Table 1.

[0058] Table 1: Typical signal strength score and capture strategy correspondence.

[0059]

[0060] The hardware architecture of the signal capture system adopts modular design. The main control unit is equipped with a dual-core processor, running a real-time operating system and a general Linux system respectively. The real-time system is responsible for signal strength evaluation and capture strategy execution, and the general system handles data storage and network communication. The signal acquisition front end includes eight channels of 24-bit ADC, supporting dynamic adjustment of sampling rate from 10kHz to 200kHz. Each channel is equipped with an independent programmable gain amplifier with a gain range of 1-1000 times adjustable. The system power supply adopts an intelligent power distribution scheme, dynamically allocating power consumption to each module according to the capture mode, and prioritizing power supply to the signal processing unit in the early capture mode.

[0061] The signal feature database construction adopts a hierarchical storage structure. The original signal waveform is saved as a 16-bit signed integer array, with a timestamp and device status code. The feature parameters are stored as structured records, including 32 feature fields such as peak position, zero-crossing rate, and harmonic components. The database index is fragmented by signal type and processing stage, supporting millisecond-level feature retrieval. After each signal capture is completed, the system automatically performs data quality detection, removes abnormal segments with a signal-to-noise ratio below 20dB, and adds integrity check codes to the valid data.

[0062] The adaptive learning mechanism of signal capture timing is based on historical data analysis. The system maintains a reference library containing 2000 typical signal features, each feature associated with the optimal capture strategy. When a new signal arrives, a nearest neighbor search algorithm is used to match similar patterns in the reference library, and the current capture parameters are fine-tuned based on the matching results. The learning process continuously optimizes the content of the reference library, eliminating obsolete patterns with a usage frequency of less than 1% every 24 hours, and adding typical signal features from the current production environment. The reference library is updated simultaneously with the adjustment of the intensity score threshold, maintaining the sensitivity of strategy adjustment.

[0063] The operation interface provides visual monitoring of signal capture status. The main display screen real-time draws signal waveform and intensity score curve, and the auxiliary panel displays detailed configuration parameters of the current capture strategy. The operator can manually adjust the capture sensitivity, with a range of 0-100%, and the system automatically converts it to internal intensity threshold. All manual interventions are recorded as audit logs, including operation time, personnel ID, modification content and system response results. The interface supports signal playback function, which can compare the signal quality under different capture strategies.

[0064] The abnormal processing flow contains a three-level response mechanism. The primary response is for transient signal loss, automatically switching to the backup sampling channel and reissuing the capture instruction. The intermediate response handles persistent signal abnormalities, triggering device speed reduction and starting the diagnostic program. The advanced response is executed in the case of severe signal distortion, immediately stopping the current processing cycle and locking the device. A fault analysis report is generated after each abnormal processing, recording signal abnormality characteristics, processing measures and recovery time, for optimizing subsequent capture strategies.

[0065] The system calibration maintenance cycle is synchronized with the device maintenance plan. The monthly routine calibration includes: sampling clock accuracy verification to ensure that the time reference error is less than 1ppm; ADC linearity test to verify the conversion accuracy within the full range; channel isolation detection to ensure the independence of multi-channel acquisition. Calibration data is entered into the device health file for predictive maintenance decision-making. When the calibration parameters deviate from the standard value by more than 5%, a special repair work order is triggered, requiring the device manufacturer to perform hardware-level repair.

[0066] The network communication of the signal capture system adopts a dual-channel redundant design. The main channel transmits real-time acquisition data through industrial Ethernet, and the backup channel transmits compressed feature parameters using RS485 bus. The network load balancing algorithm dynamically allocates data transmission paths, automatically switching to the backup channel when the Ethernet is congested. All communication data is time-synchronized, ensuring that the time consistency of distributed acquisition nodes is within 50 microseconds of error. The local cache mechanism is started when communication is interrupted, supporting up to 30 minutes of offline data saving.

[0067] Environmental adaptability design considers the actual working conditions of the workshop. In terms of electromagnetic compatibility, the shell of the acquisition equipment reaches IP54 protection level, and the internal circuit adopts multi-layer shielding structure. The vibration protection adopts three-point suspension installation method to isolate mechanical vibration with a frequency greater than 25 Hz. The temperature compensation module monitors the internal temperature of the equipment in real time and automatically calibrates the ADC reference voltage. The system ensures that the acquisition accuracy is not worse than 0.1% FS within the environmental temperature range of -10°C to 55°C.

[0068] Example 5: The equipment degradation evaluation model is established on the basis of time series analysis of comprehensive process deviation. The model input layer receives the process deviation data stream of the three dimensions of feeding, preforming, and final forging, with a data sampling interval of 10 seconds. The data preprocessing module first performs outlier cleaning to remove sudden changes exceeding 300% of the historical maximum value; then performs sliding window standardization processing, with a window width of 480 sample points corresponding to 80 minutes of working condition. The feature extraction engine captures key degradation indicators from the standardized data: the cumulative integral value of speed deviation in the feeding dimension; the covariance coefficient of pressure fluctuation frequency and peak value in the preforming dimension; and the duration ratio of vibration energy level exceeding the threshold in the final forging dimension. These features form a 128-dimensional feature vector input into the core evaluation algorithm.

[0069] The evaluation algorithm adopts a deep recurrent neural network architecture. The network contains three bidirectional LSTM hidden layers with 64, 32, and 16 neurons respectively, and processes a continuous 4.8-hour process deviation sequence at each time step. The output layer generates a device health status score, with a score interval of 0-100, and a score above 85 indicating normal operation, a score between 60 and 85 indicating an observation period, and a score below 60 triggering a maintenance alarm. The score calculation combines the degradation contribution weights of the three dimensions: feeding abnormality weight coefficient 0.35, preforming defect weight 0.4, and final forging fault weight 0.25. When a certain M12 stainless steel bolt is continuously processed, the model detects a 12% increase in the preforming pressure deviation week-on-week, an 8% decrease in the feeding speed integral value, and finally outputs a score of 72, showing a clear deterioration trend compared to the initial value of 98.

[0070] The maintenance decision instruction generation module contains a four-level response mechanism. The first level response is for scores between 70 and 85, sending device parameter calibration instructions to the controller, such as increasing the lubrication frequency by 20%; the second level response is for scores between 60 and 70, generating a preventive maintenance work order containing a mold inspection checklist and a component replacement plan; the third level response handles scores between 40 and 60, forcing a scheduled maintenance and locking the production plan; the fourth level response is for scores below 40, triggering an automatic shutdown program and pushing an emergency repair request. The decision instruction is accompanied by detailed operation guidelines: when the second level response is triggered, the system automatically retrieves the mold maintenance procedures corresponding to the current bolt specification, accurately marking the 3rd punch guide column gap standard value of 0.05±0.003mm.

[0071] The preventive maintenance operation execution system implements instruction issuance through a device networking platform. The control instruction transmission adopts a dual-channel redundancy protocol, the main channel transmits maintenance parameters through the OPCUA protocol, and the standby channel sends simple instructions through ModbusTCP. The execution terminal includes a multifunctional maintenance mechanical arm and a manual operation interface: the mechanical arm receives a standardized instruction set to perform repetitive operations such as mold cleaning and bearing oiling; the manual interface displays three-dimensional disassembly animations and torque calibration parameters to guide technicians to handle complex maintenance. When the system issues the "preforming station slider rail maintenance" instruction, the mechanical arm automatically performs the rail dust removal and special grease coating process, and the manual calibration synchronously calibrates the slider gap to an accuracy of 0.1 mm.

[0072] The real-time process trigger signal processing unit adopts a dynamic priority scheduling mechanism. The signal receiving end sets up three levels of cache areas: level 0 cache stores signals in the score critical value ±2 minute interval, with millisecond-level response; level 1 cache stores regular monitoring signals, with a delay tolerance of 200 milliseconds; level 2 cache stores historical data synchronization signals, allowing for second-level delay. The processing flow performs intelligent filtering of signals: temperature fluctuation signals unrelated to the current degradation mode are automatically downgraded, and vibration signals related to key bearing wear are prioritized for analysis. During the processing of SCM435 material in a certain case, when the final forging vibration signal abnormally increases, the signal breaks through the cache level and directly activates the model for real-time evaluation.

[0073] The model online updating mechanism is automatically activated in the device idle window. The trigger conditions include: 20 batches of new materials are continuously processed, the cumulative running time of the device exceeds 500 hours, or the first piece of processing is completed after maintenance. The updating process first compares the consistency of the new data and the model prediction value, and calculates the root mean square error of the parameters exceeding 10% to trigger network structure adjustment. The transfer learning module retains the basic feature layer of the degradation mode and adjusts the weights of the fully connected layer to adapt to the new working condition. The update verification uses the cross-validation method: the first three days of data are used for retraining, and the last three days of data are used for testing, and the accuracy rate fluctuation needs to be controlled within ±3% before deployment.

[0074] The maintenance execution tracking system generates a three-dimensional verification report. The sensor network collects the first processing data after maintenance: the laser displacement meter detects the mold installation precision to 0.01 mm level; the acoustic emission sensor records the bearing vibration spectrum and the matching degree with the standard template; the infrared thermal imager monitors the temperature distribution state of each station. The verification data are generated by the algorithm to score the execution quality: above 98 points is automatically released from maintenance lock; 90-98 points triggers local rework; below 90 points restarts the maintenance process. Each maintenance generates a digital archive package, including pre-maintenance diagnosis parameters, execution process images, and acceptance test data, with an archive period covering the entire life cycle of the device. In a typical maintenance case, the difference rate between the first processing data after hydraulic system maintenance and the predicted curve is only 0.7%, and the system automatically closes the maintenance work order to resume full-load production.

[0075] The maintenance decision optimization module establishes a knowledge base iteration mechanism. Within 72 hours after each maintenance action, the system collects the operating stability indicators of the component: statistical abnormal downtime frequency, parameter exceeding rate, mold wear rate, and other parameters. These data are analyzed in association with the original decision basis, and when the actual failure rate is 25% higher than the predicted value, the decision rule revision is triggered. The new rule is preferentially used in subsequent similar maintenance, and after three verification cycles, it is written into the formal decision tree. The entire system runs maintenance instructions in a network isolation environment, the communication interface uses 256-bit encryption authentication, and the operation log is written in real time to the blockchain storage platform.

[0076] The above merely provides a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present application, and these modifications or replacements shall be encompassed within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A method for monitoring the condition of a cold heading machine used for bolt processing, characterized in that, Includes the following steps: Obtain the bolt product specification information processed by the cold heading machine, and based on the bolt product specification information, extract the initial forming parameter set and dynamic adjustment parameter set of the cold heading machine; the bolt product specification information includes the bolt geometric dimensions and material properties; Based on the initial forming parameter set and the dynamic adjustment parameter set, the process trigger signals of the cold heading machine during the processing are identified; based on the process trigger signals, the parameter change nodes of the cold heading machine are extracted; the parameter change nodes include parameter fine adjustment points caused by changes in equipment status when processing the same bolt specifications and overall parameter update nodes triggered when switching processing of different bolt specifications. Based on the parameter change nodes, distinguish between nodes with the same parameter change nodes and nodes with different parameter change nodes and classify them to calculate the process offset. Based on the process offset, common process feature parameters and process-specific feature parameters are separated. Based on the common process characteristic parameters and process-specific characteristic parameters, a status early warning mechanism for the cold heading machine is triggered; the common process characteristic parameters refer to general stable parameters applicable to the processing of different bolt specifications, and the process-specific characteristic parameters refer to key control parameters applicable only to the processing of specific bolt specifications.

2. The method for monitoring the condition of a cold heading machine for bolt processing according to claim 1, characterized in that, The step of obtaining the bolt product specification information processed by the cold heading machine, and based on the bolt product specification information, extracting the initial forming parameter set and dynamic adjustment parameter set of the cold heading machine, includes: Based on the bolt geometry and material properties, the reference parameter values ​​and effective running time of the cold heading machine in the feeding, preforming, and final forging stages are set. Based on the switching processing nodes for different bolt specifications, record the correction parameter values ​​and their duration of action of the cold heading machine in the feeding, preforming, and final forging stages; Based on the baseline parameter values ​​and their effective runtime, the corrected parameter values ​​after verification are calculated using the parameter drift verification rules. The baseline parameter values ​​and the verified corrected parameter values ​​are combined to form a complete record of process parameters for the cold heading machine.

3. The method for monitoring the condition of a cold heading machine for bolt processing according to claim 1, characterized in that, The process triggering signal identification during the cold heading process based on the initial forming parameter set and the dynamic adjustment parameter set includes: The initial molding parameter set is analyzed to generate the first type of process trigger signal; The second type of process trigger signal is generated by parsing the set of dynamically adjusted parameters; By integrating the first type of process triggering signal and the second type of process triggering signal, the signal triggering rules in the three stages of feeding, preforming and final forging are analyzed when processing the same bolt specification and switching between different bolt specifications. Based on the signal triggering pattern, the timing of capturing the process trigger signal of the cold heading machine is dynamically adjusted.

4. The method for monitoring the condition of a cold heading machine for bolt processing according to claim 3, characterized in that, The step of extracting parameter change nodes of the cold heading machine based on the process trigger signal includes: Calculate the trend similarity between the process trigger signal and the historical parameter fine-tuning point reference data. When the trend similarity exceeds the set threshold, it is determined to be a device status fine-tuning point and classified as a parameter fine-tuning point with the same bolt specification. Calculate the trend similarity between the process trigger signal and the historical parameter update node reference data. When the trend similarity is lower than the set threshold, it is determined to be a specification switching update node and classified as a parameter overall update node for different bolt specifications.

5. The method for monitoring the condition of a cold heading machine for bolt processing according to claim 1, characterized in that, Based on the parameter change nodes, the process offset is calculated by distinguishing and classifying nodes with the same parameter change and nodes with different parameter change, including: Mark the parameter fine-tuning points caused by changes in equipment status when processing bolts of the same specification, and define them as the same parameter change nodes; Mark the parameter update node triggered when switching between different bolt specifications, and define it as a different parameter change node; Based on the same parameter change node, the cumulative change of parameters within the same specification processing cycle is counted as the first type of process offset. Based on different parameter change nodes, the parameter abrupt change difference during the specification switching process is calculated as the second type of process offset; The first type of process offset and the second type of process offset are combined to generate a comprehensive process offset.

6. The method for monitoring the condition of a cold heading machine for bolt processing according to claim 5, characterized in that, Based on the process offset, common process characteristic parameters and process-specific characteristic parameters are separated, including: Using feeding, preforming, and final forging as the analysis dimensions, the stable features that repeatedly appear in the comprehensive process offset are identified, and the parameter set that can be reused across bolt specifications is extracted and marked as common process feature parameters. Identify the unique features of the integrated process offset that vary with bolt specifications, extract the parameter set that is only bound to specific specifications, and mark it as process-specific feature parameters.

7. The method for monitoring the condition of a cold heading machine for bolt processing according to claim 6, characterized in that, The status early warning mechanism for the cold heading machine, based on the common process characteristic parameters and process-specific characteristic parameters, includes: The common process characteristic parameters are used as steady-state monitoring benchmarks and applied to the monitoring of the processing of all bolt specifications. By combining the process-specific characteristic parameters, real-time parameter calibration is performed on the processing of specific bolt specifications, and the abnormal state is determined in conjunction with the steady-state monitoring benchmark.

8. The method for monitoring the condition of a cold heading machine for bolt processing according to claim 3, characterized in that, The step of dynamically adjusting the acquisition timing of the cold heading machine process trigger signal according to the signal triggering rule includes: If the intensity of the signal triggering pattern is higher than the preset intensity threshold, the process triggering signal of the cold heading machine will be captured in advance. If the intensity of the signal triggering pattern is lower than the preset intensity threshold, the process triggering signal of the cold heading machine will be delayed.

9. The method for monitoring the condition of a cold heading machine for bolt processing according to claim 4, characterized in that, The trend similarity is obtained by calculating the covariance relationship between the time series data of the process trigger signal and the historical node reference data.

10. The method for monitoring the condition of a cold heading machine for bolt processing according to claim 1, characterized in that, The method further includes: A device degradation assessment model is established based on the comprehensive process offset. Based on the equipment degradation assessment model and the real-time captured process trigger signals, maintenance decision instructions are generated; Preventive maintenance operations are performed on the cold heading machine according to the maintenance decision instructions.

Images