An adas support detection equipment production optimization method based on big data

Abstract

The application relates to the technical field of production management, and discloses an ADAS support detection equipment production optimization method based on big data, which comprises the following steps: establishing a production data chain with a device serial number as a primary key; extracting positioning surface coordinates and optical center coordinates to generate eccentricity parameters; dividing time windows such as locking, loading, calibration and final inspection according to assembly and adjustment nodes; calculating a load index according to the eccentricity parameters and the time windows and generating a target beat; combining the load index and the time boundary of the loading time window to allocate an assembly station and a calibration time; mapping the load index to identify bottleneck procedures and update process parameters; and finally outputting production control instructions containing the target beat and the station number. The application quantifies interference factors such as eccentric layout and waiting cycle into production scheduling data, and optimizes assembly rhythm and station resource allocation.

Description

Technical Field

[0001] This invention relates to the field of production management technology, and more specifically, to a method for optimizing the production of ADAS bracket testing equipment based on big data. Background Technology

[0002] Advanced driver assistance system (ADAS) bracket testing equipment mainly consists of a reference frame, bracket positioning and clamping module, testing head, target plate, and cable drag chain. The assembly process of this type of equipment involves the cascaded integration of mechanical positioning chain, optical measurement chain, and data calibration chain. In actual component layout, to avoid the sensor field of view and mounting holes of the bracket under test, the positioning surface of the clamping module is marginalized, while the optical measurement center is often located in the central area of ​​the working surface, resulting in an eccentric load-bearing configuration where their spatial centers of gravity do not coincide. After loading peripheral components such as the cable drag chain and protective cover, the additional load will transmit small torsional stresses to the clamping plate along the eccentric lever arm. Conventional manufacturing execution systems only record assembly yield and calibration residuals, easily attributing reference drift caused by stress release to human operational fluctuations or single material batch tolerances. Furthermore, to adapt to multiple vehicle models, the equipment often uses replaceable target drawings and quick-change fixtures; the assembly gaps of these detachable components are easily masked by the calibration compensation algorithms of the software system. If the work order scheduling alters the waiting time between component loading and optical calibration, the compensation amount from a single calibration cannot reflect the true degree of degradation of the reference component. Existing manufacturing management systems lack an algorithmic link that integrates eccentric stress, waiting time, and assembly reference into their management, and therefore cannot output quantified workstation scheduling instructions and cycle control parameters. Summary of the Invention

[0003] This invention provides a production optimization method for ADAS bracket testing equipment based on big data, which solves the technical problems mentioned in the background art.

[0004] This invention provides a method for optimizing the production of ADAS bracket testing equipment based on big data, including:

[0005] Using the equipment serial number as the primary key, establish a production data chain that includes structural nodes, assembly adjustment points, and quality nodes, and associates them with the work orders to be executed; Extract the surface coordinate data of the positioning surface and the center coordinate data of the optical center from the digital model of the equipment to generate an eccentric feature to characterize the spatial offset feature; The assembly and adjustment time window is divided according to the timestamp data of the assembly and adjustment point. The assembly and adjustment time window includes a locking time window, a load time window, a calibration time window and a final inspection time window, each with a start time boundary and an end time boundary. The load index is calculated based on the correlation data between the eccentricity feature and the assembly and adjustment time window, and a target cycle time is generated based on the load index to eliminate the assembly and adjustment disturbance caused by eccentricity. Based on the load index, a scheduling instruction is generated to allocate assembly and adjustment positions on the production site for each of the work orders to be executed, and the calibration time is determined in combination with the end time boundary of the load time window. The load index is mapped to the corresponding assembly and adjustment time window to identify bottleneck processes with speed-limiting characteristics and update the process parameters of the bottleneck processes. The output includes the target cycle time, the calibration time, and the assembly station number.

[0006] The beneficial effects of this invention are as follows: by integrating production data chains, assembly and adjustment time windows, and eccentricity parameters, and generating target cycle times through load index calculation, the assembly and adjustment disturbances caused by spatial center of gravity offset are converted into quantified scheduling instructions and bottleneck process parameters. This invention avoids misjudging assembly and adjustment drift as ordinary process fluctuations or material batch differences, achieves priority allocation of stable workstation positions for sensitive equipment and timed control of calibration timing, suppresses drift values ​​of finished products during repeated positioning and changeover acceptance, and improves the first-pass yield of finished products and the calibration stability during long-term operation. Attached Figure Description

[0007] Figure 1 This is a flowchart of a production optimization method for ADAS bracket testing equipment based on big data, according to the present invention. Detailed Implementation

[0008] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0009] For ease of understanding, ADAS in this application refers to Advanced Driver Assistance Systems. Specifically, it is an automotive electronic system that relies on onboard sensors (cameras, millimeter-wave radar, lidar, ultrasonic sensors, etc.) to perceive the vehicle's surrounding environment in real time. Combined with algorithms, it performs environmental recognition, risk assessment, and decision output, providing drivers with warning prompts, operational assistance, and even limited active intervention to improve driving safety and reduce driver workload. Common ADAS features include Adaptive Cruise Control (ACC), Lane Keeping Assist (LKA), Automatic Emergency Braking (AEB), and Blind Spot Detection (BSD).

[0010] like Figure 1 As shown, a method for optimizing the production of ADAS bracket testing equipment based on big data includes: Using the equipment serial number as the primary key, establish a production data chain that includes structural nodes, assembly adjustment points, and quality nodes, and associates them with the work orders to be executed; Extract the surface coordinate data of the positioning surface and the center coordinate data of the optical center from the digital model of the equipment to generate an eccentric feature to characterize the spatial offset feature; The assembly and adjustment time window is divided according to the timestamp data of the assembly and adjustment point. The assembly and adjustment time window includes a locking time window, a load time window, a calibration time window and a final inspection time window, each with a start time boundary and an end time boundary. The load index is calculated based on the correlation data between the eccentricity feature and the assembly and adjustment time window, and a target cycle time is generated based on the load index to eliminate the assembly and adjustment disturbance caused by eccentricity. Based on the load index, a scheduling instruction is generated to allocate assembly and adjustment positions on the production site for each of the work orders to be executed, and the calibration time is determined in combination with the end time boundary of the load time window. The load index is mapped to the corresponding assembly and adjustment time window to identify bottleneck processes with speed-limiting characteristics and update the process parameters of the bottleneck processes. The output includes the target cycle time, the calibration time, and the assembly station number.

[0011] S201 collects component codes from the bill of materials (BOM) to generate structural nodes. The BOM is synchronized from the PLM system hourly and must include component code, component name, version number, and supplier code fields. Component codes use a 16-digit encoding rule: the first 8 digits are the component type code, the middle 4 digits are the version code, and the last 4 digits are the supplier code. The assembly order is extracted from the hierarchical relationship of the BOM; the lower the hierarchy, the earlier the assembly order. Structural nodes are stored in a structural node table in the database, which includes equipment serial number, component code, version number, and assembly order fields. When a component version changes, a new version structural node is automatically generated and associated with the corresponding equipment serial number.

[0012] S202 collects timestamp data and process results to generate assembly adjustment points during the execution of each assembly and adjustment process, which includes the main locking process, component assembly process, optical calibration process, and final inspection and verification process. Each assembly and adjustment process must be executed in the order of main locking process, component assembly process, optical calibration process, and final inspection and verification process. The timestamp for the main locking process is automatically triggered when the torque wrench reaches the set torque. The timestamps for other processes are triggered by the operator clicking in the manufacturing execution system. The timestamp data includes the process start timestamp and end timestamp, with millisecond precision. The process result is coded as 1 (qualified), 2 (unqualified), and 3 (rework). After a non-qualified process is reworked, an independent assembly adjustment point is generated, marked with a rework identifier, and associated with the original process node. The operator logs into the system using their employee ID, and the equipment number is automatically read via RFID and bound to the assembly adjustment point. The assembly adjustment point is stored in the assembly adjustment point table in the database, which includes fields for equipment serial number, process type, start timestamp, end timestamp, process result, operator employee ID, equipment number, and rework identifier.

[0013] The number of rework processes must be recorded. If the same process is reworked more than 3 times, a quality warning will be triggered. If the RFID fails to read the device number, the operator is allowed to manually enter the device number, which must be confirmed twice. If the process result is unqualified, the execution of subsequent processes will be automatically suspended until the rework is completed and passes inspection.

[0014] The S203 acquires optical calibration residuals and final inspection verification residuals as residual data to jointly generate quality nodes. The optical calibration residual is measured using a standard ceramic target with a detection head alignment accuracy of 0.001 mm, while the final inspection verification residual is measured using a coordinate measuring machine. Both are Euclidean norms of the residuals in the X, Y, and Z directions, and the residual data are retained to three decimal places. When the residual exceeds the limit, the quality node is marked as unqualified, triggering a rework process. The quality node is stored in a quality node table in the database, which includes fields for equipment serial number, optical calibration residual, final inspection verification residual, and measurement timestamp.

[0015] S204 binds structural nodes, assembly adjustment points, and quality nodes to equipment serial numbers and stores them in the database to form a production data chain. The equipment serial number uses a 12-digit alphanumeric code: the first four digits represent the production year, the middle two digits represent the production month, and the last six digits represent the serial number. The production data chain is created synchronously when the equipment serial number is generated. A transaction mechanism ensures the atomicity of node data writes. The database uses read-committed isolation level, performs a full backup every morning, and incremental backups for real-time data. The main table of the production data chain uses the equipment serial number as the primary key, and foreign keys link to the structural node table, assembly adjustment point table, and quality node table. It also links to pending work orders through the equipment serial number field, which contains information such as production batch, planned output, and delivery time.

[0016] S301 acquires the equipment's 3D model and reads the coordinates of the main positioning points and main locking points to construct the surface coordinate data of the corresponding positioning surface. The equipment's 3D model uses the STEPAP242 format. The main positioning points are named LOC_PNT_01, LOC_PNT_02, and LOC_PNT_03, with a total of three points that are not collinear. The main locking points are named LOCK_PNT_01 to LOCK_PNT_n, with a number of 4-6 determined based on the size of the positioning surface. If multiple points with the same name exist in the model, the point with the smallest deviation from the design value is selected. The point coordinates are read using the BRep_Tool::Pnt() method of the OpenCASCADE library, and the plane equation of the positioning surface is constructed based on the coordinates of the three main positioning points. The plane equation parameters are rounded to six decimal places. The surface coordinate data of the positioning surface includes the plane equation parameters, the coordinates of the main positioning point, and the coordinates of the main locking point.

[0017] After the plane equation parameters are calculated, the results are verified by substituting the coordinates of the three main positioning points. The deviation of the substituted results from zero must be less than 1e-6. The number of main locking points is determined according to the area of ​​the positioning surface. Four points are set when the area is less than 0.1 square meters, five points are set when the area is between 0.1 and 0.3 square meters, and six points are set when the area is greater than 0.3 square meters.

[0018] The S302 reads the coordinates of the functional measurement center of the detection head to construct the center coordinate data of the corresponding optical center. The functional measurement center is named OPT_CENTER_01, initially using the theoretical design value from the 3D model, and corrected by actual measurement using a laser interferometer after the detection head is installed. When the detection head is replaced, the optical center coordinates are remeasured and updated to the equipment parameter library. For multi-detection head equipment, a weighted average optical center is calculated based on the measurement weight of each detection head. The weight is determined according to the measurement range of the detection head; the larger the measurement range, the higher the weight.

[0019] The measurement weight of a detection head is calculated as a proportion of its measurement range. The weight of a single detection head is the measurement range of all detection heads divided by the sum of their measurement ranges. S303 establishes a unified equipment coordinate system, mapping the coordinate data corresponding to the positioning surface and optical center to the equipment coordinate system. A 2mm diameter steel stamp is affixed to the lower left corner of the reference frame, serving as the origin of the equipment coordinate system. The X-axis points to the right along the frame length, the Y-axis points forward along the frame width, and the Z-axis points upward along the frame height. The coordinate transformation matrix is ​​obtained by multiplying the translation and rotation matrices. The translation vector is the coordinate of the origin of the reference coordinate system in the equipment coordinate system, and the rotation matrix is ​​calculated from the angles between the three axes of the reference coordinate system and the three axes of the equipment coordinate system. After the coordinate transformation matrix is ​​calculated, the deviation is verified by measuring the coordinates of the three reference points. A deviation less than 0.01mm is considered acceptable, and the allowable range for coordinate system deviation is ±0.01mm. The coordinate transformation matrix converts the local coordinates of the positioning surface and optical center into global coordinates in the equipment coordinate system.

[0020] The positional tolerance of the steel stamp mark on the reference frame shall not exceed 0.1 mm; the coordinate transformation matrix shall be stored in the equipment parameter library in the form of a two-dimensional array; when the coordinate system deviation exceeds the limit, the coordinates of the reference point shall be remeasured and a new coordinate transformation matrix shall be calculated until the deviation meets the requirements.

[0021] S304 calculates the coordinates of the projection point of the optical center onto the positioning surface. Based on the coordinates of the projection point and the geometric center coordinates of the positioning surface, a spatial distance and direction vector model is constructed to generate the eccentricity feature. If the optical center is located within the positioning surface, the projection point is the optical center itself; otherwise, a perpendicular line is drawn from the optical center to the positioning surface, and the foot of the perpendicular is the projection point. The formula for calculating the projection point coordinates is: ; ; in, The coordinates of the projection point, For the optical center coordinates, These are the parameters of the plane equation for the positioning surface. The formula for calculating spatial distance is: ; in, It is the spatial distance between the optical center projection point of the j-th device and the geometric center of the positioning surface. It is the optical center of the jth device. On the positioning surface The coordinates of the projection point on the surface, It is the positioning surface of the j-th device. The geometric center coordinates, It is an operator for calculating the Euclidean norm of a vector. The direction vector is the vector pointing from the geometric center of the positioning plane to the projection point of the optical center, and is represented by a three-dimensional array. The data is stored for adjusting the force direction during subsequent process parameter optimization. The eccentricity feature is composed of spatial distance and direction vector. If the spatial distance is greater than 50% of the maximum span of the positioning surface, it is considered an outlier and triggers manual review.

[0022] S401 generates a locking time window based on the completion time of the main locking process as the end time boundary. The start time of the production batch is the start time of the main locking process of the first machine in the batch, which serves as the start time boundary of the locking time window, and the completion time of the main locking process is the end time boundary. The standard torque for the main locking process is 25 N·m, with a holding time of 5 seconds. For multiple bolts, a diagonal tightening method is used, tightening in three stages to the set torque. The maximum allowable duration of the locking time window is twice the standard duration; an alarm is triggered if the timeout is exceeded. If the main locking process is reworked, the rework completion time is used as the new end time boundary of the locking time window, and a new locking time window is generated.

[0023] The main locking process is divided into three tightening torque ratios: 50% for the first tightening, 75% for the second tightening, and 100% for the third tightening.

[0024] S402 generates a load time window using the completion time of the component assembly process as the end time boundary. The end time boundary of the locked time window serves as the start time boundary of the load time window, and the completion time of the component assembly process serves as the end time boundary. Load components include cable chains, protective covers, target boards, cables, and control cabinets. The assembly sequence is cable chains, protective covers, target boards, cables, and control cabinets. When a component is temporarily disassembled, the load time window pauses timing and resumes timing after reassembly. After a new load component is assembled, the completion time of the last newly added component is used as the end time boundary of a new load time window, and a new load time window is generated. The minimum allowable duration of the load time window is 10 minutes.

[0025] S403 generates a calibration time window based on the start and end time boundaries of the optical calibration process. The process start time is the moment the operator clicks the process start button in the manufacturing execution system, and the end time is the moment the process finishes clicking the process finish button. The optical calibration process includes focal length calibration, distortion calibration, and field of view calibration; a calibration residual of less than 0.02 mm is considered acceptable. If the process is reworked, the rework start time is used as the new start time boundary, and the rework completion time is used as the new end time boundary to regenerate the calibration time window.

[0026] Focal length calibration is achieved by adjusting the distance between the detection head and the standard target until the image sharpness reaches its maximum value; distortion calibration is achieved by acquiring a standard checkerboard image, calculating the distortion coefficient, and compensating for it; field of view calibration is achieved by adjusting the angle of the detection head so that the standard target is completely centered in the field of view.

[0027] In S404, a final inspection time window is generated based on the start time boundary and the end time boundary of the final inspection verification process. The final inspection verification process includes repeated positioning accuracy test and temperature drift test, and a repeated positioning accuracy less than 0.01 mm is qualified. After the process rework is completed, the final inspection time window is regenerated with the rework start time as the new start time boundary of the time window and the rework completion time as the new end time boundary. When the time window times out, the time window is automatically extended and the abnormality is marked.

[0028] The repeated positioning accuracy test is carried out 5 times continuously, and the standard deviation of the 5 measurement results is taken as the final value; the temperature drift test is carried out in the temperature range of 20°C to 40°C, and the residual is measured every 5°C, with a test duration of no less than 2 hours.

[0029] In S405, the timestamp data and residual data in the production data chain are mapped to the corresponding assembly and adjustment time window of the time period respectively according to the equipment serial number. Each assembly node and quality node in the production data chain is traversed, and the assembly and adjustment time window it belongs to is determined according to the timestamp of the node. When the timestamp is equal to the end time boundary of the previous time window, it is classified into the previous time window. Cross-time-window nodes are classified according to their main execution time period. When data mapping fails, the original data is retained and manual processing is triggered. The mapped data set is verified by checking the timestamp continuity to form a data set organized by time window, which is stored in a time window mapping table. The table includes fields of equipment serial number, time window type, time window start time, time window end time, and associated node ID.

[0030] In S501, the preset maximum span of the positioning surface, length tolerance, residual upper limit, standard tact and benchmark quantile, as well as the preset compensation coefficient and calculation tolerance are obtained. The preset parameters can only be modified by the production supervisor, and the modification needs to go through an approval process. If the parameter exceeds the preset range, it is an abnormal value and will be automatically restored to the default value. The maximum span of the positioning surface is the larger value of the maximum distances of the positioning surface in the X-axis and Y-axis directions in the equipment coordinate system. The maximum span of the special-shaped positioning surface is the maximum distance between all vertices. Length tolerance is the allowable dimensional deviation during the processing of the positioning surface, with a value of 0.02 mm, which is obtained from the machining process document and verified by a coordinate measuring machine. The residual upper limit is the maximum allowable value of the optical calibration residual and the final inspection verification residual. The standard tact is the standard time interval from the end of the load time window to the start of the calibration time window specified in the production process. The initial value is 30 minutes, and adjustment needs to be approved by the process engineer and updated to the process document. The benchmark quantile is the 95% quantile of the historical load index data, which is calculated based on all equipment load index data of the past 6 months and updated once a month. The compensation coefficient is a preset parameter used to adjust the compensation range of the target tact, with a value ranging from 0.1 to 0.5 and an initial value of 0.3, which is determined based on regression analysis of historical production data. The calculation tolerance is a very small positive number used to avoid zero denominator, which is fixed at 1e-6 and cannot be adjusted.

[0031] The least squares method was used to determine the compensation coefficients through regression analysis, based on production data from the past three months. S502 constructs a computational model incorporating spatial distance, maximum span of the positioning surface, and length tolerance to solve for eccentricity, which is used to quantify the degree of structural offset. The formula for calculating eccentricity is: ; in, It is the eccentricity of the j-th device. It is the spatial distance between the optical center projection point of the j-th device and the geometric center of the positioning surface. It is the maximum span of the positioning surface of the j-th device. This refers to the length tolerance. The eccentricity ranges from 0 to 1; values ​​greater than 0.8 are considered outliers and trigger a structural review.

[0032] The S503 extracts the drift rate by normalizing the final verification residual, optical calibration residual, and upper limit of the residual, which characterizes the change in calibration status. The final verification residual is measured within 24 hours after optical calibration, and the absolute value of the residual is used for calculation. The formula for calculating the drift rate is: ; in, It is the drift rate of the j-th device. It is the final inspection and verification residual of the j-th device. It is the optical calibration residual of the j-th device. This is the upper limit of the residual for the j-th device. The drift rate ranges from 0 to 1; a value greater than 0.5 is considered an outlier and triggers calibration rework.

[0033] S504 uses a time deviation comparison algorithm based on the start time boundary of the calibration time window, the end time boundary of the load time window, and the standard beat rate to obtain the beat deviation used to characterize the waiting period disturbance. If the start time of the calibration time window is earlier than the end time of the load time window, the beat deviation is set to 10.0, and it is marked as a severe anomaly. The formula for calculating the beat deviation is: ; in, It is the cycle time deviation of the j-th device. This is the start time boundary of the calibration time window for the j-th device. It is the end time boundary of the load time window of the j-th device. This is the standard cycle time of the j-th device. The cycle time deviation ranges from 0 to positive infinity; a deviation greater than 2.0 is considered an anomaly and triggers production scheduling adjustments.

[0034] The S505 system integrates drift rate, beat deviation, and eccentricity to construct a load assessment model, outputting a load index to characterize the combined impact of structure and beat rate. The load index formula is derived based on structural mechanics and statistical analysis. Eccentricity reflects inherent structural risk, while drift rate and beat deviation reflect process risk. The formula for calculating the load index is: ; in, It is the load index of the j-th equipment. It is the eccentricity of the j-th device. It is the drift rate of the j-th device. It is the cycle time deviation of the j-th device. The load index is classified as follows: less than 0.2 is low risk, 0.2-0.5 is medium risk, and greater than 0.5 is high risk.

[0035] S506 performs a limited comparison between the load index and the baseline quantile, and introduces a standard cycle time to execute a dynamic compensation algorithm to generate a target cycle time. The target cycle time is at least 0.8 times the standard cycle time and at most 3 times the standard cycle time, and is updated once at the end of the load time window. If the target cycle time causes the calibration time to exceed the daily production plan, the target cycle time is adjusted to the maximum value allowed by the plan, and the work order is marked as high-risk. The formula for calculating the target cycle time is: ; in, It is the target cycle time of the j-th device. This is the standard cycle time of the j-th device. It is the compensation coefficient. It is the minimum value operator. It is the load index of the j-th equipment. It is the baseline quantile. It calculates the tolerance. The target cycle time is used to eliminate assembly and adjustment disturbances caused by eccentricity. When the load index exceeds the reference quantile, the target cycle time will be extended accordingly to ensure that the load is fully released before calibration.

[0036] When a production batch starts, S601 arranges all pending work orders in descending order of load index. Pending work orders are equipment orders that need to be produced in the current batch according to the production plan. Each work order corresponds to one piece of equipment, and work orders for different equipment models are uniformly sorted according to their load index. When load indices are the same, they are sorted from earliest to latest delivery time, with urgent work orders marked as having the highest priority and placed before all ordinary work orders. When a work order is canceled or changed, the sorting result is regenerated in real time, and the sorting result is valid for one production batch.

[0037] The S602 acquires historical operating data of the workstations, extracts the workstation stability, and arranges each assembly / adjustment workstation in descending order of stability value. Historical operating data includes information such as the number of devices produced at the workstation in the past 30 days, first-pass yield, and average calibration residual, updated daily at midnight. The first-pass yield is the number of devices that passed final inspection on the first attempt at the workstation in the past 30 days divided by the total production quantity. The average calibration residual is the arithmetic mean of the optical calibration residuals of all devices at the workstation in the past 30 days. Workstation stability is calculated as the product of the first-pass yield and the reciprocal of the average calibration residual. The grading standard is: greater than 0.9 for Grade A, 0.8-0.9 for Grade B, and less than 0.8 for Grade C. The initial stability of a new workstation is the average of the stability values ​​of all workstations. After major maintenance at a workstation, the stability is reset to the initial value and recalculated.

[0038] Based on the sorted order, S603 binds each work order to its corresponding assembly / adjustment station. The compatibility between stations and equipment models is pre-entered into the system, and only compatible work orders are bound. The first sorted work order is bound to the first sorted station, the second to the second, and so on, prioritizing high-load work orders for high-stability stations. Work orders are queued according to the sorting result. When a station is idle, the work order at the front of the queue is assigned. When the number of work orders waiting to be produced at a station exceeds twice the average, the last work order is automatically transferred to an idle station. If a station fails, work orders already assigned to that station but not yet started production are moved to the head of the queue and reassigned to other idle stations. After binding, the work order status changes to "assigned," and after production is completed, it changes to "completed."

[0039] S604 calculates the calibration time by superimposing the end time boundary of the load time window with the target cycle execution time. The formula for calculating the calibration time is: ; in, This is the calibration time for the j-th device. It is the end time boundary of the load time window of the j-th device. This is the target cycle time for the j-th device. The calibration time accuracy is in minutes. For every minute the load time window ends, the calibration time is delayed by 1 minute. A reminder is sent to the operator at the corresponding workstation 10 minutes in advance via the manufacturing execution system. If the calibration process starts early, the calibration time remains unchanged; if it starts late, the subsequent processes are recalculated based on the actual start time. In case of calibration time conflicts, the higher-risk work order is executed first.

[0040] The S701 maps drift rate and beat deviation to the corresponding assembly and adjustment time windows based on the equipment serial number. The time granularity of drift rate and beat deviation is 1 minute, and linear interpolation is used to fill in missing data. It iterates through the drift rate and beat deviation data of all equipment, and associates the data with its corresponding assembly and adjustment time window based on the equipment serial number and timestamp, forming a multidimensional time-series dataset organized by time window and equipment serial number, and storing it according to equipment serial number and time window type.

[0041] S702 uses the time period during which the drift rate value corresponding to the load index reaches its extreme value as the associated time window. The statistical granularity of the drift rate extreme value is a single time window for a single device, which is the maximum drift rate of all devices in the current production batch. The time window to which the drift rate of the device corresponding to this maximum value belongs is the associated time window. When there are multiple extreme points, the time window corresponding to the earliest extreme point is selected, and the length of the associated time window does not exceed 2 hours.

[0042] S703 When multiple assembly and adjustment operations are included in the same associated time window, the assembly and adjustment operation with the timestamp closest to the calibration time is designated as the target operation. The operation timestamp is taken as the operation end time. For operations that cross associated time windows, their main execution time period is used to determine whether they belong to the associated time window. When multiple operation timestamps are equidistant from the calibration time, the operation with the earlier timestamp is selected as the target operation, and the number of target operations shall not exceed 3.

[0043] S704 calculates the number of devices and load index for each target process, identifying the process with the highest cumulative load index as the bottleneck process. The cumulative load index is weighted by the number of devices, with high-risk devices having a weight of 2, medium-risk devices 1, and low-risk devices 0.5. For small-batch production, data from three historical batches are used for joint determination, with the statistical timeframe being the current production batch. When multiple processes have the same total load index, the process with the most devices is selected as the bottleneck process, and the bottleneck process is updated every production batch.

[0044] For pending work orders, S705 updates and locks the tightening sequence, workstation number, and calibration waiting time of the bottleneck process as process parameters. Bolts are numbered clockwise from the top left corner. The diagonal tightening sequence for four bolts is 1-3-2-4, and for six bolts it is 1-4-2-5-3-6, tightened in three stages to the set torque. The calibration waiting time is equal to the target cycle time, with an adjustment step of 5 minutes. Parameter locking is achieved through the manufacturing execution system's access control; only administrators can modify the locked process parameters. The locking is automatically released after the current batch of production is completed; re-approval is required for any continuation.

[0045] After completing the assembly and adjustment process corresponding to the load time window, S801 generates production control instructions using the equipment serial number as an index. Component assembly process inspection includes connection firmness check, cable continuity check, and visual inspection. All load components have been assembled and passed process inspection; instructions cannot be generated if inspection fails.

[0046] S802 encapsulates the target cycle time, calibration time, and assembly / adjustment station number into a production control instruction. The production control instruction uses JSON format, with a unique identifier consisting of the equipment serial number and a timestamp, valid for 24 hours and automatically expired. Each field is defined as follows: equipment serial number is a 12-digit string; target cycle time is a floating-point number in minutes; calibration time is a string in the format yyyy-MM-ddHH:mm:ss; and assembly / adjustment station number is an integer. Instruction fields are validated using non-empty checks, format checks, and range checks.

[0047] The S803 sends production control commands to the Manufacturing Execution System (MES). Commands are sent using the MQTT protocol, with the MQTT server address, port, username, and password pre-configured. Upon receiving the command, the MES updates the status of the corresponding work order and displays the target cycle time, calibration time, and workstation number on the production dashboard. After executing the command, it provides feedback on success or failure.

[0048] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. A method for optimizing the production of ADAS bracket testing equipment based on big data, characterized in that, include: Using the equipment serial number as the primary key, a production data chain is established that includes structural nodes, assembly adjustment points, and quality nodes, and is associated with the work orders to be executed. During the execution of each assembly adjustment process, which includes the main locking process, component assembly process, optical calibration process, and final inspection and verification process, timestamp data and process results are collected to generate the assembly adjustment points. The optical calibration residual and the final inspection and verification residual are collected as residual data to jointly generate the quality nodes. Extract the surface coordinate data of the positioning surface and the center coordinate data of the optical center from the digital model of the device to generate an eccentric feature to characterize the spatial offset feature. The eccentric feature is composed of spatial distance and direction vector. The assembly and adjustment time window is divided according to the timestamp data of the assembly and adjustment point. The assembly and adjustment time window includes a locking time window, a load time window, a calibration time window and a final inspection time window, each with a start time boundary and an end time boundary. The load index is calculated based on the correlation data between the eccentricity characteristics and the assembly / adjustment time window, and a target cycle time for eliminating assembly / adjustment disturbances caused by eccentricity is generated based on the load index, including: Obtain the preset maximum span, length tolerance, residual upper limit, standard cycle time, and reference quantile of the positioning surface; A computational model is constructed that includes the spatial distance, the maximum span of the positioning surface, and the length tolerance to solve for the eccentricity, which is used to quantify the degree of structural offset. The drift rate is extracted by normalizing the final verification residual, the optical calibration residual, and the upper limit of the residual, and is used to characterize the change in calibration state. Based on the comparison algorithm of the start time boundary of the calibration time window, the end time boundary of the load time window, and the standard beat construction time deviation, a beat deviation is obtained to characterize the waiting period disturbance. A load assessment model is constructed by integrating the drift rate, the beat deviation, and the eccentricity, and the load index is output to characterize the combined effect of structure and beat. The load index is compared with the benchmark quantile with a limited amplitude, and the standard beat is introduced to execute a dynamic compensation algorithm to generate the target beat; Based on the load index, scheduling instructions are generated to allocate assembly and adjustment positions on the production floor for each of the work orders to be executed, and the calibration time is determined in conjunction with the end time boundary of the load time window, including: When a production batch is started, the work orders to be executed are arranged in descending order of the load index. Obtain historical operating data of the assembly stations, extract the stability of the assembly stations, and arrange each assembly station in descending order of the stability values. Based on the sorted order, the work orders to be executed are bound one by one to the corresponding assembly and adjustment station; The end time boundary of the load time window is superimposed with the target cycle execution time to generate the calibration time corresponding to the work order to be executed; Mapping the load index to the corresponding assembly and adjustment time window to identify bottleneck processes with speed-limiting characteristics, and updating the process parameters of the bottleneck processes, including: The drift rate and the beat deviation are mapped to the assembly and adjustment time window corresponding to the time period according to the equipment serial number; The time period during which the drift rate value corresponding to the load index reaches its extreme value is used as the associated time window; When the same associated time window contains multiple assembly and adjustment processes, the assembly and adjustment process with the timestamp closest to the calibration time is designated as the target process; The number of equipment corresponding to each target process and the load index are counted, and the target process with the highest cumulative load index value is determined as the bottleneck process. For the work order to be executed, update and lock the tightening sequence, station number and calibration waiting time of the bottleneck process as the process parameters; The output includes the target cycle time, the calibration time, and the assembly station number.

2. The method for optimizing the production of ADAS bracket testing equipment based on big data according to claim 1, characterized in that, The process of establishing a production data chain, using the equipment serial number as the primary key, which includes structural nodes, assembly adjustment points, and quality nodes, and linking them to work orders to be executed, includes: Collect component codes from the bill of materials to generate the structural nodes; The structural nodes, the assembly adjustment points, and the quality nodes are bound according to the equipment serial number and stored in the database to form the production data chain.

3. The method for optimizing the production of ADAS bracket testing equipment based on big data according to claim 2, characterized in that, The extraction of surface coordinate data of the positioning surface and center coordinate data of the optical center in the digital model of the extraction device generates an eccentric feature to characterize the spatial offset feature, including: Obtain the 3D model of the device, and read the coordinates of the main positioning point and the main locking point to construct the surface coordinate data corresponding to the positioning surface; Read the coordinates of the functional measurement center of the detection head to construct the center coordinate data corresponding to the optical center; Construct a unified device coordinate system and map the coordinate data corresponding to the positioning surface and the optical center to the device coordinate system; Solve for the coordinates of the projection point of the optical center on the positioning surface, construct a spatial distance and direction vector model based on the coordinates of the projection point and the geometric center coordinates of the positioning surface, and generate the eccentric feature.

4. The method for optimizing the production of ADAS bracket testing equipment based on big data according to claim 3, characterized in that, The step of dividing the assembly and adjustment time window based on the timestamp data of the assembly and adjustment point includes: The locking time window is generated using the completion time of the main locking process as the end time boundary; The load time window is generated using the completion time of the component assembly process as the end time boundary. The calibration time window is generated from the start time boundary to the end time boundary of the optical calibration process; The final inspection time window is generated from the start time boundary to the end time boundary of the final inspection verification process. The timestamp data and residual data in the production data chain are mapped to the assembly and adjustment time windows corresponding to the time periods according to the equipment serial number.

5. The method for optimizing the production of ADAS bracket testing equipment based on big data according to claim 1, characterized in that, The output includes production control instructions containing the target cycle time, the calibration time, and the assembly / adjustment station number, including: After completing the assembly and adjustment process corresponding to the load time window, the production control instruction is generated using the equipment serial number as an index. The target cycle time, the calibration time, and the assembly station number are encapsulated into the production control instruction; The production control command is sent to the manufacturing execution system.

Images