Data acquisition and processing method and system for digital delivery of intelligent factory

Abstract

The invention relates to the technical field of data processing, in particular to a data collecting and processing method and system for intelligent factory digital delivery. The method comprises the following steps: connecting each execution terminal in the intelligent factory, and collecting working data points; timing sequences of all the working data points are calibrated in a unified mode, and timing sequence abnormal data points are recognized and marked; counting a first process stage in which time sequence abnormal data points occur at high frequency, executing sliding window detection, and determining and eliminating interference noise points to obtain first cleaned data points; identifying and eliminating transition state abnormal points in the first cleaned data points to obtain second cleaned data points; counting a second process stage in which time sequence abnormal data points occur at high frequency based on the second cleaning data points, and determining a sensor signal drift point of the second process stage; and determining the actual delivery time according to the interference noise point, the transition state abnormal point and the sensor signal drift point. According to the invention, high integrity of delivery data and high accuracy of delivery time determination can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a data acquisition and processing method and system for digital delivery of an intelligent factory. BACKGROUND

[0002] At present, in the construction and delivery process of an intelligent factory, a full life cycle digital delivery mode covering design modeling, equipment integration, process debugging, trial operation evaluation and other links has been gradually formed. In order to ensure the real-time visibility, controllability and traceability of the running state, line rhythm, environmental parameters and process data of various factory assets (including equipment, workstations, production lines, energy systems, etc.) in the delivery process, it is urgent to collect, process and map all data of multi-source heterogeneous terminal equipment (such as PLC, industrial camera, sensor, robot controller, etc.) in the field.

[0003] However, abnormal data such as equipment communication interruption, sensor drift, interference noise and start-up transient state often occur in the field. If these data are directly entered into the database without identification, it is easy to cause a large deviation between the expected delivery time and the actual delivery time, thereby affecting the customer's vehicle experience. SUMMARY

[0004] Therefore, it is necessary to provide a data acquisition and processing method and system for digital delivery of an intelligent factory to solve at least one of the above technical problems.

[0005] To achieve the above-mentioned purpose, a data acquisition and processing method for digital delivery of an intelligent factory comprises the following steps: Step S1: connecting each execution terminal in the intelligent factory, and collecting work data points generated by each execution terminal during work; Step S2: uniformly timing the time sequence of each work data point according to the collection time stamp, identifying and marking time sequence abnormal data points; Step S3: counting a first process stage of high-frequency time sequence abnormal data points; performing sliding window detection on each work data point in the first process stage, determining and removing interference noise points, and obtaining first cleaned data points; Step S4: identifying and removing transient state abnormal points in the first cleaned data points according to the start-up time sequence relationship of each execution terminal in the first process stage, and obtaining second cleaned data points; Step S5: counting a second process stage of high-frequency time sequence abnormal data points based on the second cleaned data points, determining sensor signal drift points in the second process stage, and determining the actual delivery time according to the interference noise points, the transient state abnormal points and the sensor signal drift points.

[0006] The application establishes a terminal data access basis on site, sequentially carries out time correction, abnormal positioning, process segmentation identification and multi-level abnormal repair on the collected work data points, forms continuous time sequence data after cleaning, and then determines the actual delivery time combined with the abnormal influence period, and dynamically adjusts and synchronously pushes the delivery plan. It can effectively solve the technical problem that multi-source data abnormality affects the delivery result in the intelligent factory delivery process, realize high integrity of delivery data, high consistency of time reference and high accuracy of delivery time determination, thereby significantly improving the reliability, controllability and customer experience of the delivery process.

[0007] Optionally, the application also provides a data acquisition processing system for intelligent factory digital delivery for executing the data acquisition processing method for intelligent factory digital delivery as described above, which comprises: A data acquisition module is configured to connect each execution terminal in the intelligent factory and acquire work data points generated by each execution terminal during work; A time correction module is configured to uniformly correct the time sequence of each work data point according to the acquisition timestamp, identify and mark the time sequence abnormal data points; A denoising module is configured to count the first process stage in which high-frequency time sequence abnormal data points appear; perform sliding window detection on each work data point in the first process stage, determine and eliminate interference noise points, and obtain first cleaned data points; A transition state removal module is configured to identify and eliminate transition state abnormal points in the first cleaned data points according to the start time sequence relationship of each execution terminal in the first process stage, and obtain second cleaned data points; A delivery time correction module is configured to count the second process stage in which high-frequency time sequence abnormal data points appear based on the second cleaned data points, determine sensor signal drift points in the second process stage, and determine the actual delivery time according to the interference noise points, the transition state abnormal points and the sensor signal drift points.

[0008] The data acquisition processing system for intelligent factory digital delivery of the application can implement any one of the data acquisition processing methods for intelligent factory digital delivery of the application, and is configured to jointly operate and transmit signals between the modules to complete the data acquisition processing method for intelligent factory digital delivery. The modules in the system cooperate with each other, thereby realizing high integrity of delivery data, high consistency of time reference and high accuracy of delivery time determination, thereby significantly improving the reliability, controllability and customer experience of the delivery process. BRIEF DESCRIPTION OF DRAWINGS

[0009] Other features, objects and advantages of the application will become more apparent from the following detailed description of non-limiting embodiments made with reference to the accompanying drawings: Fig. 1 A flowchart of a data acquisition and processing method for intelligent factory digital delivery in an embodiment; Fig. 2 A schematic diagram for determining a sensor signal drift point in an embodiment; Fig. 3 A module diagram of a data acquisition and processing system for intelligent factory digital delivery in an embodiment; The implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0010] The technical method of the present application will be described clearly and completely below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0011] In addition, the accompanying drawings are only schematic illustrations of the present application, and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus repeated descriptions thereof will be omitted. Some block diagrams shown in the drawings are functional entities, which do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0012] It should be understood that although the terms "first", "second" and the like can be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, without departing from the scope of the exemplary embodiments, a first element can be referred to as a second element, and similarly a second element can be referred to as a first element. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0013] To achieve the above-mentioned object, please refer to Figs. 1-3 The present application provides a data acquisition and processing method for intelligent factory digital delivery, which comprises the following steps: Step S1: connecting each execution terminal in the intelligent factory, and collecting work data points generated by each execution terminal during work; In an embodiment, the welding robot controller on the main assembly line, the visual inspection industrial camera on the detection branch line, the air pressure sensor in the energy system, and the PLC controller inside the station execution unit are sequentially connected in the electronic assembly workshop, the connection is performed through the factory network built by the mixed architecture of the Ethernet bus and the field bus, and after completing the identity authentication and IP address allocation, the welding current and voltage signals, the image detection frame sequence, the air pressure change value, and the PLC control state code are respectively collected at a sampling frequency of 5 ms to 500 ms, the real-time working data point set is formed through the field host upload, and is used as the original data input for delivery evaluation.

[0014] Step S2: uniformly time the time sequence of each working data point according to the collection timestamp, identify and mark the time sequence abnormal data point; In the embodiment, after receiving the original working data points of each execution terminal, a time correction instruction is issued by using the NTP time synchronization protocol to uniformly correct the local sampling timestamps of the welding robot controller, the visual inspection industrial camera, the air pressure sensor, and the PLC controller to millisecond level precision; then the time sequence after the time correction is compared, if a jump or a time sequence that does not conform to the time monotone increasing rule is detected, the data point is identified as a time sequence abnormal data point and an abnormal mark is added, thereby completing the construction of a uniform time sequence reference.

[0015] Step S3: count the first process stage of high-frequency time sequence abnormal data points; perform sliding window detection on each working data point in the first process stage to determine and remove the interference noise points, and obtain the first cleaned data points; In an embodiment, according to the distribution of the time sequence abnormal data points in the process segment number, the station segment with the highest abnormal occurrence frequency is counted as the first process stage (for example, welding stations 1-3); then, the sliding time window detection is performed on the working data points in the first process stage with a 5s window and a 1s sliding step, the adjacent window sampling mean difference and the instantaneous fluctuation amplitude are calculated, and when the amplitude exceeds the set threshold of 15%, it is determined as an interference noise point; further, the interference noise points are removed from the working data points in this stage, and the mean value of the two data points before and after the interference noise points is extracted to replace the interference noise points, thereby generating the first cleaned data point set with continuous structure for subsequent identification.

[0016] Step S4: identify and remove the transition state abnormal points in the first cleaned data points according to the start time sequence relationship of each execution terminal in the first process stage, and obtain the second cleaned data points; In this embodiment, according to the preset starting sequence of the equipment (for example, PLC→robot controller→vision camera→air pressure sensor), the timestamps of each data point are compared with each other, if any robot controller data point timestamp is ahead of the corresponding process PLC starting time by more than 120 ms, or lags behind the industrial camera trigger time by more than 150 ms, it is determined that it is a transition state abnormal point; the transition state abnormal point is removed from the first cleaned data point, and the work data set of different windows in the historical time period is extracted from the corresponding execution terminal of the data point, the time window with a mean value fluctuation range lower than 8% is identified as a steady state window, the value closest to the end of the steady state window is selected as the extension replacement value, and the value is filled to the transition state abnormal point position to obtain continuous and reliable second cleaned data points.

[0017] Step S5: based on the second cleaned data points, a second process phase of high-frequency occurrence timing abnormal data points is counted, sensor signal drift points of the second process phase are determined, and actual delivery times are determined according to the interference noise points, transition state abnormal points and sensor signal drift points.

[0018] In some embodiments, the second cleaned data points are used to statistically analyze the abnormal point proportion of each sub-process segment in the first process phase, and when the proportion of the sub-process segment is 10% to 15%, the sub-process segment is determined as the second process phase (such as detection branch line station 4-5); subsequently, the sensor sampling value sequences of the industrial camera, air pressure sensor and PLC controller in this phase are extracted, a time window of 3s is used, the mean value of the continuous sampling points in the window is calculated, and a mean value change curve is constructed, the change rate is obtained by first-order difference processing of the mean value change curve, if the direction remains consistent and the absolute values of the change rates of two consecutive windows both exceed 0.02, it is determined that the corresponding sensor point is a sensor signal drift point, and the third cleaned data points are generated by replacing the mean values of the nearest two normal sampling points before and after with the mean values. Finally, the sampling time period caused by the three types of abnormalities is labeled, the sampling abnormality occupation time is calculated, the second ideal delivery time is obtained by deducting the corresponding abnormal influence time period from the initial sampling starting time, if the difference between the first ideal delivery time calculated using the third cleaned data points is not more than 10 min, the earlier time is selected as the actual delivery time, otherwise the later time is selected as the actual delivery time and automatically uploaded to the delivery management platform to trigger manual verification, so as to improve the delivery accuracy.

[0019] It is noted that the actual delivery time refers to the time when the manufacturing process in the smart factory is completed, and does not refer to the time when the product is delivered to the user.

[0020] Optionally, the method further comprises: obtaining a delivery plan of the smart factory, and reconstructing the corresponding delivery plan according to the actual delivery time to obtain a reconstructed delivery plan and a corresponding reconstructed delivery time; The reconstructed delivery plan is uploaded to the smart factory delivery management platform to adjust the delivery plan, and the reconstructed delivery time is transmitted to the application program in the user's mobile terminal to correct the delivery time displayed by the application program.

[0021] In some embodiments, after determining the actual delivery time, the factory delivery schedule table is retrieved from the smart factory delivery management platform, wherein the factory delivery schedule table includes planned delivery nodes, equipment delivery batches, process delivery completion times, etc. The actual delivery time is compared with the original planned delivery time of the corresponding process in the delivery schedule table, and if the difference exceeds the set time tolerance (e.g. ±30 min), the delivery sequence and delivery node time are automatically adjusted according to the difference result, and the reconstructed delivery plan and corresponding reconstructed delivery time are generated based on the sequence correction.

[0022] Especially important is that the reconstructed delivery plan is specifically: According to the difference calculation result of the actual delivery time and the original planned delivery time of the corresponding process in the delivery schedule table, the difference calculation result is sequentially added to the planned time of the subsequent process node, and the delivery sequence of each delivery number is adjusted according to the added planned time to generate the reconstructed delivery plan.

[0023] In one embodiment, if the difference calculation result of the actual delivery time compared to the original planned delivery time of the process in the delivery schedule table exceeds the preset tolerance threshold (e.g. ±30 min), the difference calculation result is used as the delay correction amount, which is sequentially added to the planned time of each delivery preparation node after the process; for example, the original planned manufacturing process completion time is 10:00, the process A (moving the product to the designated delivery buffer) planned completion time is 11:00, and the process B (sending delivery notification to the user) planned completion time is 12:00. If the actual delivery time of the manufacturing process determined by the above data collection and processing method is 10:30, the delivery plan time corresponding to process A is automatically adjusted to 11:30, the delivery plan time corresponding to process B is adjusted to 12:30, and the reconstructed delivery plan and new reconstructed interaction time are output, thereby realizing automatic rearrangement and time synchronization update of the delivery plan. It is worth noting that the actual delivery time determined by the present scheme refers to the time of the internal manufacturing process completion node of the smart factory, not the time node of the completion of delivery by the smart factory to the customer, and the subsequent delivery plan pushing and interaction time correction are all based on this internal manufacturing completion time as the basis.

[0024] In another embodiment, the reconstructed delivery plan is uploaded to the smart factory delivery management system via the delivery management platform interface, replacing the original plan in real time in the delivery board to adjust the arrangement of each process delivery execution node, logistics transfer node and customer receiving node; at the same time, the reconstructed delivery time is pushed to the service application program in the user mobile terminal through the API interface, covering the original displayed delivery time, so as to ensure that the end user can obtain the delivery time change information in real time, and improve the perceptibility and delivery transparency of the delivery process.

[0025] Optionally, step S3 comprises: According to the process segment number corresponding to each timing abnormal data point, the frequency of the timing abnormal data point in each process segment is counted, and the process segment with the highest frequency is taken as the first process stage; The sliding time window detection is performed on the working data points in the first process stage, and the mean difference and fluctuation amplitude of adjacent time windows are calculated to determine the interference noise points; The interference noise points are removed from the working data points in the first process stage, and the mean value of the working data points closest to the interference noise points in time is substituted for the interference noise points, to obtain the first cleaned data points.

[0026] In some embodiments, each timing abnormal data point is classified according to the process segment number; then, the frequency of the timing abnormal data point with time stamp abnormal jump (such as time stamp rollback, jump, etc.) in each process segment is counted, and when the frequency of the timing abnormal data point in the welding process segment is the highest, the welding process segment is determined as the first process stage. For the working data points in the welding process segment, the sliding time window detection is performed with a window length of 2s and a sliding step of 0.5s, the mean value of the sampling values in each window is calculated, and the difference between the mean value of the current window and the mean value of the previous window is calculated, and the fluctuation amplitude of the sampling values in the window is calculated. When the mean difference exceeds the set threshold (such as 10%) and the fluctuation amplitude is higher than the set fluctuation threshold (such as 15%), the data point in the maximum fluctuation interval in the window is marked as an interference noise point. Further, the interference noise point is removed from the current data column in the welding process segment, and the sampling values of the adjacent two sampling time points before and after the noise point are extracted and the mean value is calculated (for example, the current of the previous sampling point is 110A, and the current of the next sampling point is 112A, then the mean value is 111A), and the mean value is substituted for the removed noise point, to form the first cleaned data point sequence after noise repair, providing a continuous and effective data basis for subsequent anomaly identification.

[0027] Optionally, step S4 comprises: According to the preset starting sequence of each execution terminal in the first process stage, the time stamp of each data point in the first cleaned data is arranged in sequence. If the time stamp of any data point is ahead of or behind the time stamp of the data point corresponding to the previous process in the starting sequence by more than a preset allowable offset range, it is determined to be a transition state abnormal point; The transition state abnormal point is removed from the first cleaned data, and the end data value of the nearest steady state window corresponding to the execution terminal of the transition state abnormal point is extended to replace the transition state abnormal point, to obtain the second cleaned data.

[0028] In some embodiments, if the robot welding process segment is taken as the first process stage, the execution terminals in the process segment include PLC controllers, welding robot controllers, industrial cameras and temperature sensors corresponding to different starting sequences, respectively. The starting of the PLC controller should be earlier than that of the welding robot controller, the starting of the robot controller should be earlier than that of the industrial camera, and the starting of the industrial camera should be earlier than that of the temperature sensor. The sampling time stamps of the four types of execution terminals in the first cleaned data are read and compared in sequence according to the execution terminal sequence. If the sampling time stamp 10:00:20 of a certain welding robot controller is earlier than the sampling time stamp 10:00:25 of the corresponding PLC controller by more than the allowable offset range (set to ±3s), or the sampling time stamp 10:02:01 of the temperature sensor is later than the sampling time stamp 10:01:55 of the industrial camera by more than the allowable offset range (set to ±5s), the abnormal data point is determined to be a transition state abnormal point. Further, the transition state abnormal point is removed from the first cleaned data sequence, and the historical sampling records of the execution terminal to which the transition state abnormal point belongs are used to extract a plurality of candidate time windows with a window of 10s and a step length of 2s. The fluctuation range of the mean value of the sampling values in each time window is calculated, and the window with a fluctuation less than a set threshold (such as 5%) is selected as a steady state window. The sampling value at the end of the steady state window closest to the time of the transition state abnormal point (for example, the steady state window end value is 108A) is selected as the extension reference value, and the second cleaned data point sequence is formed after filling the extension reference value at the original time stamp position, so as to ensure the time sequence continuity and steady state value consistency of the repaired time sequence.

[0029] In another embodiment, if the first process stage is an assembly online process segment, the executing terminals include a conveying line PLC, a robot controller, an RFID reader and a torque sensor, wherein the start of the conveying line PLC is the starting mark of the process segment, the robot controller and the RFID reader are started subsequently, and the torque sensor is started before the workpiece enters the tightening station. After obtaining the sampling time stamps of each executing terminal in the first cleaning data points, the start sequence of PLC→robot controller→RFID reader→torque sensor is compared. If the sampling time stamp of a certain robot controller at 09:15:02 is earlier than the start signal time recorded by the PLC at 09:15:05 and the offset exceeds the preset threshold of 2s, it is determined that the sampling value of the robot controller is a transient state abnormal point. Similarly, if the sampling time stamp of the RFID reader at 09:18:12 is later than the normal start time of the robot controller at 09:18:00 and the offset exceeds the set time deviation of 5s, it is determined that the corresponding sampling point of the RFID is a transient state abnormal point. After removing the identified transient state abnormal points from the first cleaning data points, a set of sampling windows at different times is extracted based on the executing terminal to which the abnormal point belongs in the historical wave peak stable stage. When the mean square deviation of the sampling values in the window is less than the set stable state threshold (for example, 1.5), it is determined that the sampling window is a stable state window. The sampling value at the end of the stable state window closest to the transient state abnormal point in time (such as torque of 50N·m) is selected as a replacement value to fill in the original abnormal point sampling position, and the second cleaning data point sequence is obtained after the correction.

[0030] Optionally, the end data value of the executing terminal corresponding to the transient state abnormal point in the nearest stable state window is extended to replace the transient state abnormal point, including: extracting a set of data values of the executing terminal corresponding to the transient state abnormal point under different time windows from the first cleaning data points to obtain a transient state candidate window data set; According to the mean value fluctuation range of each time window in the transient state candidate window data set, a time window smaller than a preset mean value fluctuation threshold is selected as a stable state window, and the end data value of the stable state window closest to the transient state abnormal point in time is taken as an extension reference value; The extension reference value is filled into the position of the transient state abnormal point in the first cleaning data points in time sequence to obtain the second cleaning data points.

[0031] In some embodiments, when it is determined that the data point corresponding to the sampling time 09:16:35 of the robot controller is a transition state abnormal point, first, based on the historical sampling data of the controller within the previous 10 minutes, time windows are divided from the first cleaned data point in a manner of window length = 5s, sliding step = 1s, and sampling value sets in each time window are extracted in turn, so as to build a transition state candidate window data set containing multiple time window sampling sequences. Subsequently, for each time window, the mean fluctuation range of the window sampling value is calculated. For example, if the mean value fluctuation range of the sampling value in a window is 2%, which is lower than the preset mean value fluctuation threshold of 5%, it is marked as a steady state window; if the fluctuation range exceeds the threshold, it is determined as a non-steady state window. Further, the nearest steady state window to the transition state abnormal point is selected from the several steady state windows determined (for example, the latest window end time is 09:16:30), and the end sampling value (for example, the robot pose execution angle θ is 35°) is obtained as the extrapolation reference value. Finally, the extrapolation reference value is filled in the original sampling time stamp position of 09:16:35 to cover the transition state abnormal point, so that the data sequence maintains continuity and numerical stability on the time axis, and a new second cleaned data point sequence is generated, providing a reliable data basis for subsequent drift identification.

[0032] In another embodiment, if the transition state abnormal point appears in the sampling sequence of the temperature sensor (for example, the sampling time 09:45:12 is determined to be abnormal), first, the historical sampling sequence of the temperature sensor within the previous 20 minutes is intercepted in the first cleaned data point, and multiple candidate time windows are constructed in a manner of window length 5s, sliding step 1s, each window containing multiple temperature sampling values; the mean value fluctuation range of the data in the candidate window is calculated in turn, and when it is detected that the temperature sampling mean value fluctuation in a window is less than 3℃ (less than the preset fluctuation threshold of 5℃), it is determined as a steady state window, and the rest is determined as a non-steady state window, then the window closest in time to the abnormal point is searched in all steady state windows (for example, the end time is 09:45:09), and the temperature sampling value at the end of the steady state window (for example, 75.2℃) is extracted as the extrapolation reference value, and is replaced and filled in the original abnormal time 09:45:12, so as to generate a continuous repaired second cleaned data point sequence, so as to realize the time sequence continuity and steady state extrapolation of the temperature signal.

[0033] Fig. 2 A schematic diagram for determining the sensor signal drift point in an embodiment; for example, Fig. 2As shown, the vertical axis is the sampling value rate of change, the horizontal axis is the sampling time, and the columnar bar represents the value condition of the mean rate of change in each sampling time window. The two pink dotted lines in the figure represent the preset upper and lower threshold values of the rate of change, the column body located in the threshold value range (between the upper threshold value and the lower threshold value) is identified by green, representing that the sampling window is in the normal fluctuation range; the column body exceeding the threshold value range is identified by red, representing that the mean rate of change of the corresponding sampling window deviates to a large extent. According to the change rate direction and threshold value condition judgment, the 08:01-08:04 and 08:24-08:29 segments in the figure continuously appear regions with consistent change rate direction and higher than the threshold value, which are identified as sensor signal drift intervals; and segments such as 08:13-08:18, although the rate of change deviates from the threshold value, the direction alternately changes, and is still considered as a normal sampling fluctuation state. Through the above trend identification, the sensor signal drift point can be accurately located and abnormal compensation can be realized, providing a reliable basis for subsequent delivery time judgment.

[0034] Optionally, the determining of the sensor signal drift point in the second process stage in step S5 comprises: Based on the second cleaning data points, the proportion of time sequence abnormal data points in each sub-process segment in the first process stage is counted, and the sub-process segment with a proportion of 10% to 15% is taken as the second process stage; The sensor sampling value sequence of each execution terminal in the second process stage is extracted, the mean value of the continuous sampling points in the time window of each sensor sampling value sequence is calculated, and a mean value change curve is constructed; According to the mean value change curve, the sensor signal drift point and the sensor signal sampling point are determined, and the mean value of the sensor signal sampling point closest to the sensor signal drift point in time sequence is replaced by the sensor signal drift point in the second cleaning data point, to obtain third cleaning data points.

[0035] In some embodiments, the timing anomaly data point marks of each sub-process segment in the first process stage are read, and the proportion of the number of anomaly data points in each sub-process segment to the total number of data points in the segment is calculated in sequence to form an anomaly proportion data table of each sub-process segment. Then, according to the anomaly proportion data, all sub-process segments with a proportion between 10% and 15% are screened out as the second process stage. According to the time stamp and sub-process segment number index, the corresponding sensor sampling values are completely extracted to form a multi-dimensional time series data set. Then, a fixed length time window (such as 5s) is set, and each sensor sampling sequence is divided into a sliding window starting from the starting time point in chronological order. The sampling points in the window are all corrected by time synchronization to ensure time continuity. The arithmetic mean of the sampling values in each window is calculated to obtain the mean sampling point of the time period. The mean values of each window are arranged in chronological order to form a mean value change curve. For the mean value change curve obtained in the second embodiment, the difference between adjacent mean values is calculated to form a first-order difference sequence, which represents the mean value change rate. Checking the change rate sequence by time window will get two cases: 1) If the change rate direction is consistent (all positive or all negative) in continuous multiple time windows, and the absolute value of the change rate is higher than the set threshold (for example, 0.05 units / minute), it is determined that the data corresponding to these windows is a sensor signal drift point. 2) If the sign of the change rate alternates between time windows, or the absolute value of the change rate is lower than the threshold, it is determined to be a normal sampling point. For the time windows determined to be drift points, the mean value of the nearest normal sampling point (i.e. the original sampling value corresponding to the non-drift point) near the time of the window is found in the second cleaned data points, and the mean value of the normal sampling point is obtained. The mean value of the normal sampling point is used to replace the sampling value of the drift point time window. This replacement is based on time stamp alignment to ensure the continuity of the replaced data. After all drift point replacements are completed, the third cleaned data points are generated as the basis data for the next step of time judgment.

[0036] In some embodiments, if the first process stage corresponding to the second cleaned data point is the precision injection molding process segment, the six sub-process segments (raw material heating segment, glue melting segment, injection segment, pressure maintaining segment, cooling segment, and demolding segment) are respectively counted to obtain the proportion of the number of time series abnormal data points to the total number of sampling points. When the abnormal proportions of the injection segment and the pressure maintaining segment are 11% and 14% respectively, the injection segment and the pressure maintaining segment are determined as the second process stage. Further, for the sensor sampling value sequence of the temperature sensor, the pressure sensor, and the displacement sensor in the second process stage, a time window of 5 seconds and a sliding step of 1 second are used to calculate the continuous multi-point sampling mean value in the window to construct the mean value change curve of each sensor. When the mean value change curve of the temperature sensor shows a monotonous rising trend in the continuous 4 windows and the absolute value of the change rate is greater than 0.8℃ / s, it is determined that the temperature sensor signal drift point is determined. At the same time, when the mean value change curve of the pressure sensor continuously decreases and the absolute value of the change rate exceeds 2.0bar / s, it is determined that the pressure signal drift point is determined. Subsequently, the temperature drift point corresponding timestamp 09:42:05 is located from the second cleaned data point, the sampling value at the end of the front window is 221.6℃, and the sampling value at the beginning of the rear window is 222.1℃. The mean value 221.85℃ of the two is taken to replace the drift point. Similarly, the pressure drift point is repaired and replaced according to the mean value of the adjacent normal sampling points (such as the mean value 196.5 bar of 196 bar and 197 bar), thereby generating a third cleaned data point set. The data set maintains the overall stable and continuous fluctuation trend, which is convenient for accurate determination of the subsequent delivery time.

[0037] Optionally, determining the sensor signal drift point comprises: performing first-order difference processing on the mean value change curve to obtain the mean value change rate in the time window; if the change rate direction of the mean value change rate in the time window remains unchanged, and the absolute value of the mean value change rate exceeds the preset change rate threshold, it is determined that the sensor sampling value in the corresponding time window is a sensor signal drift point; if the change rate direction of the mean value change rate in the time window changes alternately, or the absolute value of the mean value change rate does not exceed the preset change rate threshold, it is determined that the sensor sampling value in the corresponding time window is a sensor signal sampling point.

[0038] In some embodiments, for each adjacent time window mean in the mean variation curve, the difference value is calculated, i.e. the current time window mean is subtracted from the previous time window mean, forming a mean variation rate sequence. The sequence reflects the change amplitude and direction of each time window mean compared to the previous window. By traversing the continuous time window segments in the mean variation rate sequence, the sign (positive or negative) and amplitude of the variation rate are detected to confirm the signal trend: for example, in a plurality of consecutive time windows, the sign of the variation rate is consistent (all positive or all negative), indicating that the signal presents a one-way stable rising or falling trend. A predetermined variation rate threshold (for example 0.05 units / minute) can be set, and only when the absolute value of the variation rate exceeds the variation rate threshold, the variation is considered to be significant. If a plurality of consecutive time windows meet the conditions of keeping the same variation rate direction and the amplitude exceeding the threshold, the sampling data in these time windows are determined as sensor signal drift points, which indicates that the sensing signal has an abnormal continuous drift. If the sign of the variation rate alternates (i.e. positive and negative alternately) or the absolute value of the variation rate does not exceed the threshold in the analyzed time window, the data in the time window is determined as normal sensor sampling points, reflecting normal signal fluctuations.

[0039] Optionally, determining the actual delivery time according to the interference noise points, the transition state abnormal points and the sensor signal drift points in step S5 comprises: determining the delivery time according to the third cleaned data points to obtain a first ideal delivery time; counting the abnormal sampling time distribution corresponding to the interference noise points, the transition state abnormal points and the sensor signal drift points, and combining a preset process duration reference to calculate the data influence period of each abnormal type to obtain an abnormal data influence correction time; based on the abnormal data influence correction time, eliminating the corresponding abnormal influence period from the collection time range of the first process stage, and correcting the collection time sequence to obtain fourth cleaned data points, and determining the delivery time according to the fourth cleaned data points to obtain a second ideal delivery time; comparing the time difference between the first ideal delivery time and the second ideal delivery time, if the time difference is less than or equal to a preset time difference threshold, selecting the earlier one as the actual delivery time; if the time difference is greater than the preset time difference threshold, selecting the later one as the actual delivery time, and triggering a manual review process.

[0040] Especially important is that determining the delivery time comprises: extracting the effective sampling segment cumulative duration of each data point in the third cleaned data points, and adding it to the start time of the corresponding process in the start sequence to calculate the first ideal delivery time; The effective sampling duration of each data point in the fourth cleaned data point is extracted and added to the start time of the corresponding process in the start sequence to calculate the second ideal delivery time.

[0041] In some embodiments, the third cleaned data points are sorted in order of sampling time, and continuous effective sampling segments are divided, each effective sampling segment being defined as being in the same effective sampling segment if the interval between two adjacent sampling time stamps is less than or equal to a set threshold (such as 1 s); if the interval is greater than the threshold, it is considered that the time sequence has a discontinuous jump and needs to be divided into a new sampling segment. Then, the duration of all effective sampling segments, i.e., the difference between the start time and the end time of each sampling segment, is calculated and summed to obtain the cumulative duration of the effective sampling segments. The duration is added to the start time of the corresponding process (such as the start time of the conveyor PLC 09:00:00) in the start sequence to obtain the first ideal delivery time under the third cleaned data point. For example, if the cumulative duration of the effective sampling segments is 1 hour and 14 minutes (4440 seconds), and the manufacturing process only includes the conveyor PLC, the first ideal delivery time is calculated as 09:00:00+01:14:00=10:14:00.

[0042] In another embodiment, after identifying the abnormal data influence period and removing it, a fourth cleaned data point can be constructed through time advance processing, the time stamp of which no longer includes the stall time caused by the abnormal influence period, thereby obtaining a shortened collection time sequence. According to the segmentation principle defined in the above effective sampling segment, the sampling time sequence of the fourth cleaned data point is traversed, and the effective sampling segments are identified in the time continuous interval; the cumulative value of the duration of all effective sampling segments is calculated as the cumulative duration of the effective sampling segments under the fourth cleaned data point. Subsequently, the duration is added to the start time of the corresponding process (such as 09:00:00) in the start sequence to calculate the second ideal delivery time. For example, after removing the abnormal influence, the cumulative effective sampling segment duration becomes 1 hour and 10 minutes (4200 seconds), and the manufacturing process only includes the conveyor PLC, then the second ideal delivery time = 09:00:00+01:10:00=10:10:00.

[0043] It is worth noting that the cumulative duration of the effective sampling segments refers to the total length of the data segment with continuous sampling in the time sequence, which reflects the actual working time of the process equipment without abnormal stop and data omission. Its calculation method is to calculate the total duration of all such segments by subtracting the start time from the end time of a segment with continuously increasing time stamps (and the interval between adjacent samples does not exceed a set upper limit), and the total duration obtained by adding all such segments does not include the time interval caused by abnormal sampling, communication delay, interruption, etc.

[0044] In this embodiment, the first ideal delivery time is defined according to the third cleaned data points. Specifically, the first ideal delivery time can be calculated by reading the sampling time stamps marked as key process completion or nodes in the data, and following the pre-defined process rules (for example, the sampling time of the last key execution terminal completion is the delivery time). Based on the three types of abnormal data points (interference noise points, transition state abnormal points and sensor signal drift points) marked in the previous processing stage, the sampling time stamps of each abnormal point are traversed from the second process stage. Combined with the process duration reference (such as the standard duration of each process and the allowed floating range), the abnormal sampling time is aggregated and analyzed to determine the time interval in which the abnormal data may affect the entire process progress. For example, if the abnormal sampling points appear continuously and the interval is shorter than the upper limit of the process duration floating, these time points are combined into an impact time period. The abnormal impact time periods corresponding to all abnormal data are counted to form a complete set of abnormal data impact correction time periods.

[0045] In another embodiment, according to the set of abnormal impact correction time periods, the time intervals corresponding to all abnormal impact time periods are removed from the collection time sequence within the collection time range of the first process stage. Specifically, the sampling data corresponding to the start and end time intervals of the abnormal time period are all deleted, and then the sampling data time stamps after the removal interval are shifted forward to fill the time interval caused by the deleted abnormal time period. After completing the time sequence correction, the fourth cleaned data points are formed, which represent the cleaned data set after removing the abnormal impact time. Based on the fourth cleaned data points, the delivery time is re-determined according to the same process rules as the first ideal delivery time, and the second ideal delivery time is obtained.

[0046] In other embodiments, the first ideal delivery time and the second ideal delivery time are subtracted to calculate the time difference (absolute value). If the time difference is less than or equal to a pre-set time difference threshold (for example, 5 minutes), the earlier one is selected as the final actual delivery time from the two times. If the time difference is greater than the pre-set threshold, the later one is selected as the actual delivery time from the two times, and an artificial review process is automatically triggered. The review process is intervened by the quality management personnel through the system notification, and the abnormal data and the delivery time are checked and confirmed on site to ensure the accuracy of the delivery time determination and the safety of the process.

[0047] Optionally, the corrected collection time sequence includes: The start and end time intervals of the abnormal impact period are deleted in the time sequence of the first process stage, and the subsequent data points are sequentially shifted forward to fill the start and end time intervals of the deleted abnormal impact period, to obtain the fourth cleaned data points.

[0048] In some embodiments, all sampling time stamps in the first process phase time sequence are traversed, and for each abnormal influence time period, all sampling points falling within the interval are identified and marked according to the starting time and ending time of the abnormal influence time period. According to the marking, all sampling points falling within the abnormal influence time period are deleted one by one, and after the deletion, the length of the time sequence is shortened accordingly. The deletion operation keeps the order of the data points unchanged, preventing subsequent processing from being chaotic. The time lengths of all abnormal influence time periods are summed up to calculate the total time interval occupied by these time periods. For example, if there are three abnormal time periods [10:05:00-10:10:00], [10:20:00-10:22:00], and [10:30:00-10:35:00], the total elimination time is 5+2+5=12 minutes. Starting from the next data point after the ending time of the first abnormal influence time period, the time stamps of each subsequent data point are adjusted forward by the cumulative value of all deleted abnormal time periods (e.g., 12 minutes). The adjusted time stamps should be strictly increasing, and the time interval between adjacent sampling points should remain unchanged, ensuring that the time sequence is continuous and has no gaps. After the adjustment is completed, the entire fourth cleaned data point time sequence is subjected to integrity verification to verify the orderliness of the time stamps, the uniformity of the intervals, and the absence of missing sampling values. After the verification passes, the fourth cleaned data point generation is complete.

[0049] In another embodiment, assuming that the process phase is an assembly online process segment, the time sequence sampling data covers multiple execution terminals such as conveyor line PLC, robot controller, RFID reader and torque sensor. First, according to the start and end time points in the abnormal data influence correction time set, the sampling points in the corresponding time period are located and deleted from the first process phase time sequence. Assuming that the abnormal influence time period includes 10:05:00 to 10:10:00 and 10:20:00 to 10:22:00, all sampling points in these two time periods are deleted. Then, to ensure the continuity of the time sequence, for all sampling points after 10:10:00 to 10:20:00 and after 10:22:00, their time stamps are adjusted forward by 5 minutes (corresponding to the length of the first abnormal time period) and 2 minutes (corresponding to the length of the second abnormal time period) respectively, achieving the compression of the time axis. During the adjustment, the original time interval between adjacent sampling points is kept unchanged, and only the subsequent sampling points are uniformly moved forward by the corresponding time length. In the adjusted time sequence, the originally broken time period is effectively filled, avoiding the occurrence of time jumps or data missing. After the time stamp adjustment is completed, the system performs integrity detection on the entire time sequence to confirm that the time stamps are strictly increasing and the sampling values are continuous and correct, and finally forms the fourth cleaned data point sequence for subsequent accurate calculation of the delivery time.

[0050] Optionally, the application also provides a data acquisition processing system for intelligent factory digital delivery, which is used for executing the data acquisition processing method for intelligent factory digital delivery as described above, and comprises: a data acquisition module, configured to connect each execution terminal in the intelligent factory and acquire work data points generated by each execution terminal during work; a time correction module, configured to uniformly correct the time sequence of each work data point according to the acquisition time stamp, identify and mark time sequence abnormal data points; a denoising module, configured to count a first process stage in which high-frequency time sequence abnormal data points appear; perform sliding window detection on each work data point in the first process stage, determine and eliminate interference noise points, and obtain first cleaned data points; a transition state removing module, configured to identify and eliminate transition state abnormal points in the first cleaned data points according to the start time sequence relationship of each execution terminal in the first process stage, and obtain second cleaned data points; a delivery time correction module, configured to count a second process stage in which high-frequency time sequence abnormal data points appear based on the second cleaned data points, determine sensor signal drift points in the second process stage, and determine the actual delivery time according to the interference noise points, the transition state abnormal points and the sensor signal drift points.

[0051] Therefore, from any viewpoint, the embodiments should be considered as exemplary and non-limiting, the scope of the application being defined by the appended claims and not by the above description, and it is intended to encompass all variations falling within the meaning and the scope of the equivalent elements of the claims.

[0052] The above description is merely one specific implementation of the application, which enables a person skilled in the art to understand or implement the application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the application. Therefore, the application will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data acquisition and processing method for digital delivery in smart factories, characterized in that, Includes the following steps: Step S1: Connect each execution terminal in the smart factory and collect the work data points generated by each execution terminal during operation; Step S2: According to the timestamp of the collected data points, synchronize the timing of each working data point, identify and mark the data points with abnormal timing; Step S3: The first process stage involves statistically analyzing data points with high-frequency timing anomalies; a sliding window detection is performed on each working data point in the first process stage to identify and remove interference noise points, thus obtaining the first cleaned data points; Step S4: Identify and remove transitional abnormal points in the first cleanup data point based on the startup timing relationship of each execution terminal in the first process stage, and obtain the second cleanup data point; Step S5: Based on the second cleanup data points, statistically analyze the high-frequency timing anomaly data points in the second process stage, determine the sensor signal drift point in the second process stage; determine the actual delivery time based on the interference noise points, transition anomaly points, and sensor signal drift points.

2. The data acquisition and processing method for digital delivery in smart factories according to claim 1, characterized in that, The method also includes: Obtain the delivery plan of the smart factory, and reconstruct the corresponding delivery plan based on the actual delivery time to obtain the reconstructed delivery plan and the corresponding reconstructed delivery time; The restructured delivery plan is uploaded to the smart factory delivery management platform to adjust the delivery plan, and the restructured delivery time is simultaneously transmitted to the application on the user's mobile device to correct the delivery time displayed in the application.

3. The data acquisition and processing method for digital delivery in smart factories according to claim 2, characterized in that, Step S3 includes: Based on the process segment number corresponding to each time series abnormal data point, the frequency of time series abnormal data points in each process segment is counted, and the process segment with the highest frequency is taken as the first process stage. A sliding time window detection is performed on the work data points of the first process stage, and the mean difference and fluctuation amplitude of adjacent time windows are calculated to determine the interference noise points; The interference noise points are removed from the working data points in the first process stage, and the mean of the working data points closest to the interference noise point in time sequence is used to replace the interference noise point, thus obtaining the first cleaned data point.

4. The data acquisition and processing method for digital delivery in smart factories according to claim 1, characterized in that, Step S4 includes: According to the preset start order of each execution terminal in the first process stage, the timestamp arrangement relationship in the first cleanup data point is compared. If the timestamp of any data point is earlier or later than the data point corresponding to the start of the previous process in the start order by more than the preset allowable offset range, it is determined to be a transitional abnormal point. The transitional anomaly is removed from the first cleanup data point, and the data value at the end of the nearest steady-state window of the execution terminal corresponding to the transitional anomaly is extended to replace the transitional anomaly, thus obtaining the second cleanup data point.

5. The data acquisition and processing method for digital delivery in smart factories according to claim 4, characterized in that, The transition state anomaly point is replaced by the data value at the end of the nearest steady-state window of the execution terminal. This includes: Extract the set of data values ​​of the execution terminal corresponding to the transitional anomaly point under different time windows from the first cleaned data point to obtain the transitional candidate window dataset; Based on the mean fluctuation range of each time window in the transition state candidate window dataset, time windows with a mean fluctuation threshold less than the preset mean fluctuation threshold are selected as steady-state windows, and the last data value of the steady-state window closest to the transition state anomaly point is used as the extension benchmark value. The extended baseline value is filled into the position of the transitional anomaly point in the first cleanup data point according to the time sequence to obtain the second cleanup data point.

6. The data acquisition and processing method for digital delivery in smart factories according to claim 1, characterized in that, Step S5, determining the sensor signal drift point in the second process stage, includes: Based on the second cleanup data points, the percentage of time-series abnormal data points in each sub-process segment of the first process stage is statistically analyzed, and the sub-process segments with a percentage between 10% and 15% are designated as the second process stage. Extract the sensor sampling value sequence of each execution terminal in the second process stage, calculate the mean of each sensor sampling value sequence at continuous sampling points within the time window, and construct the mean change curve; The sensor signal drift point and sensor signal sampling point are determined based on the mean change curve. The mean value of the sensor signal sampling point that is closest to the sensor signal drift point in the time sequence before and after the second cleanup data point is used to replace the sensor signal drift point to obtain the third cleanup data point.

7. The data acquisition and processing method for digital delivery in smart factories according to claim 6, characterized in that, Determining sensor signal drift points includes: Perform first-order differencing on the mean change curve to obtain the rate of change of the mean within the time window; If the direction of the rate of change of the mean remains unchanged within the time window, and the absolute value of the rate of change of the mean exceeds the preset rate of change threshold, then the sensor sampling value within the corresponding time window is determined to be a sensor signal drift point. If the direction of the rate of change of the mean change within the time window is alternating between positive and negative, or if the absolute value of the rate of change of the mean does not exceed the preset rate of change threshold, then the sensor sampling value within the corresponding time window is determined to be a sensor signal sampling point.

8. The data acquisition and processing method for digital delivery in smart factories according to claim 1, characterized in that, Step S5, which determines the actual delivery time based on interference noise points, transient anomaly points, and sensor signal drift points, includes: The delivery time is determined based on the third cleanup data point, thus obtaining the first ideal delivery time. The abnormal sampling time distribution corresponding to the interference noise points, transitional abnormal points and sensor signal drift points is statistically analyzed, and the data impact period of each abnormal type is calculated in combination with the preset process duration benchmark to obtain the abnormal data impact correction time. Based on the abnormal data impact correction time, the corresponding abnormal impact period is removed from the collection time range of the first process stage, and the collection time series is corrected to obtain the fourth cleaned data point. The delivery time is determined based on the fourth cleaned data point to obtain the second ideal delivery time. Compare the time difference between the first ideal delivery time and the second ideal delivery time. If the time difference is less than or equal to the preset time difference threshold, the earlier time is selected as the actual delivery time. If the time difference is greater than the preset time difference threshold, the later time is selected as the actual delivery time, and a manual review process is triggered.

9. The data acquisition and processing method for digital delivery in smart factories according to claim 8, characterized in that, The correction of the collected time series includes: The start and end time intervals of the abnormal impact period are deleted from the time series of the first process stage, and subsequent data points are shifted forward to fill the deleted start and end time intervals of the abnormal impact period, resulting in the fourth cleanup data point.

10. A data acquisition and processing system for digital delivery in smart factories, characterized in that, For performing the data acquisition and processing method for digital delivery of a smart factory as described in claim 1, the data acquisition and processing system for digital delivery of a smart factory includes: The data acquisition module is used to connect to each execution terminal in the smart factory and collect the working data points generated by each execution terminal during operation. The time synchronization module is used to synchronize the time sequence of each working data point according to the collection timestamp, and to identify and mark data points with abnormal time sequence. The noise reduction module is used in the first process stage to count data points with high frequency of timing anomalies; it performs sliding window detection on each working data point in the first process stage to identify and remove interference noise points, thus obtaining the first cleaned data points. The transition state removal module is used to identify and remove transition state abnormal points in the first cleanup data points based on the startup timing relationship of each execution terminal in the first process stage, so as to obtain the second cleanup data points; The delivery time correction module is used to determine the sensor signal drift point in the second process stage based on the statistical analysis of high-frequency timing anomaly data points from the second cleanup data points; and to determine the actual delivery time based on the interference noise points, transition anomaly points, and sensor signal drift points.

Images