Online quality detecting and tracing method and system for intelligent brake sensing line

Abstract

The invention discloses an online quality detection tracing method for an intelligent brake sensing line, and the method comprises the steps: S1, obtaining and processing sensor data in a production process, carrying out the detection of the sensor data, executing a next step S2 when the detection is passed, and triggering an early warning mechanism if the detection is not passed; s2, recording and storing information data in the production process of the brake sensing line; and S3, the real-time analysis result of the brake sensing line is compared with a preset quality standard, when the quality standard is not passed, production parameters are adjusted or the production process is improved, and otherwise, the production line of the brake sensing line continues to operate. According to the invention, the sensor data, the production parameters, the quality detection data and the like in the brake sensing line production process are monitored and analyzed in real time, so that abnormal conditions in the production process can be found in time, and corresponding measures are taken for adjustment, so that the product quality is improved.

Description

Technical Field

[0001] The invention relates to the technical field of automobile parts, and in particular to an online quality detection and tracing method and system for an intelligent brake sensor line. Background Art

[0002] As an important part of the automobile braking system, the quality of brake sensor wires directly affects the safety of vehicle driving. During the production process of traditional brake sensor wires, quality inspection mainly relies on offline manual inspection or simple online inspection devices. However, with the development of the automobile industry and the continuous improvement of consumers' requirements for driving safety, traditional quality inspection methods have gradually shown many shortcomings.

[0003] 1. Lack of real-time monitoring: Traditional quality inspection methods are usually intermittent and cannot achieve continuous monitoring of every link in the production process. This means that some minor quality problems may not be discovered in a short period of time, and will not appear until they accumulate to a certain extent, resulting in production delays and increased costs.

[0004] 2. Data tracing difficulties: In traditional production lines, due to the lack of a unified data management platform, various types of data generated during the production process are often scattered in different systems and cannot be effectively integrated and traced. When quality problems occur, it is difficult to quickly locate the root cause of the problem, which affects the efficiency of problem solving and the speed of quality improvement.

[0005] 3. Too much manual intervention: Existing testing methods are highly dependent on manual operations, which are not only inefficient but also prone to inconsistent results due to human factors. In addition, manual testing is difficult to cover all details and cannot achieve comprehensive quality control.

[0006] 4. Low system integration: Existing testing equipment usually operates in isolation, lacking effective communication and coordination mechanisms between each other. This not only increases the difficulty of coordination between equipment, but also limits the comprehensive use of data, which is not conducive to forming a complete quality management system.

[0007] The Chinese invention patent with the patent number CN102419347A discloses a brake pad quality detection control device, which includes: an industrial computer, a soundproof room, a brake pad conveying component, a defective screening component, an audio generating component, an audio collector, and a programmable logic controller (PLC). The programmable logic controller (PLC) controls the operation of the audio generating component, the brake pad conveying component, and the defective screening component through input and output signals, so as to realize the quality detection of the brake pad to be detected at the correct detection position in the soundproof room and generate an audio signal; the programmable logic controller (PLC) is also used to receive the quality feedback signal generated by the industrial computer. Compared with the manual detection of brake pad quality in the prior art, the invention adopts a fully automatic method to replace manual detection, so that the brake pad detection device can efficiently and accurately detect defective brake pads, which greatly improves the work efficiency and ensures the detection quality.

[0008] In the process of inspecting the produced brake pads by the above-mentioned and similar quality inspection and control devices, since the brake sensor line will produce different types of brake pads in the actual production process, it can obtain a lot of data in the production process. However, when the data is processed, analyzed and recorded in a single way, the data processing flow will be increased, resulting in a decrease in the real-time performance of the data. At the same time, due to the different types of brake pads, in the process of adjusting the production parameters of the sensor line, if only relying on data preprocessing, it will result in the same type of brake pads being produced in the process of producing different types of brake pads, which will lead to production failure. Summary of the invention

[0009] The purpose of the present invention is to provide an intelligent online quality detection and tracing method and system for brake sensor lines to solve the problems raised in the above background technology.

[0010] To achieve the above object, the present invention provides the following technical solution: an online quality detection and tracing method for an intelligent brake sensor line, the online quality detection and tracing method comprising:

[0011] S1: Setting up a sensor network, acquiring and processing sensor data in the production process, and testing the sensor data at the same time. If the test passes, the next step S2 is executed. Otherwise, the early warning mechanism is triggered.

[0012] S2: Establishing a data traceability model, and recording and storing information data during the production process of the brake sensor line, wherein the information data includes production parameters, sensor data, quality inspection data, operation logs, alarm information and production batch information;

[0013] S3: obtaining the real-time analysis result of the brake sensor line, and comparing the real-time analysis result with the preset quality standard. When the real-time analysis result does not pass the quality standard, obtaining the information data of the brake sensor line production process through the data traceability model, and adjusting the production parameters or improving the production process according to the information data. Otherwise, the production line of the brake sensor line continues to operate.

[0014] Furthermore, the triggering process of the early warning mechanism includes:

[0015] S1.1: fusing different sensor data on the brake sensor line production line to obtain material performance information of the brake sensor line, and preprocessing the material performance information;

[0016] S1.2: According to the material performance information, a warning dynamic threshold is set, and the data analysis result of the material performance information is compared with the warning dynamic threshold. When the data analysis result exceeds the warning dynamic threshold range, the warning mechanism is triggered, otherwise, the warning mechanism is not triggered.

[0017] Furthermore, a process of obtaining and preprocessing the material performance information includes:

[0018] SA1.1.1: The different sensor data on the brake sensor line are fused through the Kalman filter to obtain the state estimation value of the sensor, specifically:

[0019]

[0020] in: is the estimated value of the state at time k, is the one-step prediction state estimate at time k, is the Kalman gain, is the measured value at time k, is the observation matrix at time k;

[0021] SA1.1.2: Fuse the state estimation value of the sensor with the multimodal data to obtain fused data, specifically:

[0022]

[0023] in: is the fused data at time k, is the multimodal data acquired at time k, is the estimated value of the state at time k;

[0024] SA1.1.3: Obtain material property information of the brake sensor line from the fused data, and obtain first feature data from the material property information of the brake sensor line, specifically:

[0025]

[0026] in: is the first feature data, is the left singular vector matrix, is a diagonal matrix, is the right singular vector matrix, is the fused data at time k, is the transposed matrix of the right singular vector matrix;

[0027] SA1.1.4: Perform data cleaning, data normalization and data smoothing on the first feature data in sequence.

[0028] Furthermore, another process of obtaining and preprocessing the fused data further includes:

[0029] SB1.1.1: Preprocess the sensor data and multimodal data on the brake sensor line production line, and construct the MM-CNN model. The preprocessed sensor data and multimodal data on the brake sensor line production line are used as the input of the MM-CNN model, and the fusion feature vector is obtained as output, specifically:

[0030]

[0031] in: is the fusion feature vector, is the preprocessed sensor data, is the preprocessed multimodal data;

[0032] SB1.1.2: Obtain material performance information of the brake sensor line production line from the fused features through the attention mechanism, specifically:

[0033]

[0034] in: is the fusion feature vector, is the weight matrix, is the feature weight matrix, is the bias vector, is the activation function;

[0035] SB1.1.3: Filter outliers in the data enhancement module of the MM-CNN model, perform regularization processing in the hidden layer of the MM-CNN model, and perform data normalization processing in the input layer of the MM-CNN model.

[0036] Furthermore, the data analysis result of the material performance information is compared with the early warning dynamic threshold, and the triggering state of the early warning mechanism is determined according to the comparison result, including:

[0037] S1.2.1: Set a warning dynamic threshold according to the change status of the material performance information;

[0038] S1.2.2: Compare the data analysis result of the material performance information with the early warning dynamic threshold. When the data analysis result exceeds the early warning range of the early warning dynamic threshold, execute the next step S1.2.3. Otherwise, execute step S2.

[0039] S1.2.3: Determine the alarm level according to the extent to which the data analysis result exceeds the warning range of the warning dynamic threshold, and the alarm level includes slight abnormality, moderate abnormality and severe abnormality. The slight abnormality is 5% exceeding the warning range of the warning dynamic threshold, the moderate abnormality is 5%-10% exceeding the warning range of the warning dynamic threshold, and the severe abnormality is more than 10% exceeding the warning range of the warning dynamic threshold;

[0040] S1.2.4: According to the alarm level, adjust the working mode of the brake sensor line production system and send an alarm signal, specifically:

[0041] When the alarm level is a slight abnormality, the brake sensor line production system automatically corrects and does not send an alarm signal;

[0042] When the alarm level is a medium abnormality or a serious abnormality, the brake sensor line production system enters a monitoring mode and sends an alarm signal at the same time.

[0043] Furthermore, the formula for obtaining the early warning dynamic threshold is specifically:

[0044]

[0045] in: is the mean of the material data at time t, is the size of the time window, is the material data value at the nth time point, is the standard deviation of the material data at time t, , is the index number of the time, is the dynamic threshold for early warning, is the safety factor.

[0046] An intelligent online quality detection and tracing system for brake sensor lines, the online quality detection and tracing system comprising:

[0047] An acquisition module is used to acquire sensor data on the production line of the brake sensor line;

[0048] A judgment module sets a warning dynamic threshold and judges the relationship between the warning dynamic threshold and the sensor data;

[0049] An early warning module determines the operating state of the brake sensor line production system according to the judgment relationship between the early warning dynamic threshold and the sensor data;

[0050] A storage module, which records and stores information data during the production process of the brake sensor line through a data tracing model;

[0051] The traceability module adjusts the production parameters or improves the production process through the storage module according to the operating status of the brake sensor line production system.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] First, the present invention monitors and analyzes sensor data, production parameters, quality inspection data, etc. in the production process of brake sensor lines in real time, so as to timely discover abnormal conditions in the production process and take corresponding measures to make adjustments, thereby improving product quality;

[0054] Second, the present invention can quickly locate the problem by setting up an early warning mechanism and a data tracing model, thereby reducing the troubleshooting time, avoiding unnecessary downtime and rework, and improving the overall efficiency of the production line. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a flow chart of the online quality detection and tracing method of the present invention;

[0056] Figure 2 A schematic diagram of a process for obtaining and processing material property information of a brake sensor line according to the present invention;

[0057] Figure 3 Another schematic diagram of a process for obtaining and processing material property information of a brake sensor line according to the present invention;

[0058] Figure 4 It is a system block diagram of the online quality detection and tracing system of the present invention. DETAILED DESCRIPTION

[0059] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] Example 1

[0061] refer to Figure 1 This embodiment provides an online quality detection and tracing method for an intelligent brake sensor line, and the online quality detection and tracing method includes the following steps:

[0062] Step S1: According to the specific needs of the brake sensor line production process, a sensor network is set up on the brake sensor line production line. For example, strain sensors for measuring stress distribution, thermocouples for measuring temperature distribution, accelerometers for detecting vibration and noise, and optical sensors for detecting surface defects of materials can be selected. At the same time, in the process of laying out the sensors, it is necessary to ensure that all key links can be covered, such as raw material storage, processing, finished product inspection, etc. It is worth noting that the sensors should be installed in a position that can accurately reflect changes in material properties.

[0063] In this embodiment, the setting of the sensor network is specifically described from three aspects: raw material storage, processing process and finished product inspection, specifically:

[0064] In the process of raw materials entering the warehouse, the goal is to ensure that the quality of raw materials meets production requirements through the sensor network. The types of sensors involved in this process include: optical sensors (used to check the appearance of raw materials, such as whether there are scratches, cracks and other defects), mechanical sensors (used to measure the thickness, hardness and other mechanical properties of raw materials), electrical sensors (used to detect the conductivity or other electrical properties of raw materials). The installation location can be selected according to the handling of raw materials. For example: before the raw materials enter the warehouse, install optical sensors for surface quality inspection. Install mechanical sensors near the raw material unloading area to detect thickness and hardness. Install electrical sensors in the raw material storage area for electrical performance testing.

[0065] During the processing, the goal is to monitor the key parameters of the processing process to ensure stable product quality. The types of sensors involved in this process include: temperature sensors (used to monitor temperature changes during heating), pressure sensors (used to monitor pressure changes during calendering or pressing), vibration sensors (used to monitor whether the equipment is running smoothly and detect potential faults in time), image sensors (used to monitor the surface quality of the product online), and acoustic sensors (used to monitor the sound of equipment operation and detect abnormal noises). The installation location can be selected according to the processing device, for example: install a temperature sensor near the heating furnace to monitor the heating temperature. Install a pressure sensor on the calender or pressing equipment to monitor the applied pressure. Install vibration sensors at key parts of the production equipment to monitor vibration conditions. Install image sensors at the visual inspection station of the production line to monitor the surface quality of the product. Install acoustic sensors around the production line to monitor the sound of equipment operation.

[0066] During the finished product inspection process, the goal is to ensure that the finished product meets the predetermined quality standards. The types of sensors involved in this process include: optical sensors (used for finished product appearance inspection to ensure that there are no obvious defects), mechanical sensors (used for finished product tensile strength, bending strength and other mechanical performance tests), electrical sensors (used to detect finished product electrical properties such as resistance), thermal sensors (used to detect finished product performance in high or low temperature environments). The installation location can be selected according to the performance of the inspection, for example: install optical sensors at the finished product inspection station to conduct finished product appearance inspection. Install mechanical sensors at the mechanical performance test station to test the mechanical properties of the finished product. Install electrical sensors at the electrical performance test station to test the electrical properties of the finished product. Install thermal sensors at the environmental simulation test station to test the performance of the finished product in extreme environments.

[0067] It is worth noting that after the sensor network is installed according to the above settings, various physical properties and material analysis data can be collected in real time through the sensor network, including but not limited to optical, mechanical, thermal, electrical and other parameters. At the same time, the data from different sensors are fused to ensure the consistency and integrity of multi-modal data, and the fused data are pre-processed accordingly.

[0068] Furthermore, after obtaining the pre-processed data, the pre-processed data is tested, that is, when the test passes, the next step S2 is executed, otherwise, the early warning mechanism is triggered. The details are as follows:

[0069] Step S1.1: Fuse the different sensor data on the brake sensor line production line accordingly, and obtain the material performance information of the brake sensor line from the fused sensor data, and pre-process the obtained material performance information. Figure 2, as follows:

[0070] Step SA1.1.1: In this embodiment, the Kalman filter is used to fuse the different sensor data on the brake sensor line production line to improve the consistency and reliability of the data, and then obtain the state estimation value of the sensor, specifically:

[0071]

[0072] in: is the estimated value of the state at time k, is the one-step prediction state estimate at time k, is the Kalman gain, is the measured value at time k, is the observation matrix at time k.

[0073] In the specific implementation process, the optical sensor is used to check whether there are defects on the surface of the raw material. The data obtained during this working process is [0.98, 0.99, 0.98, 0.97], where 0.98 means no obvious defects and 0.97 means slight scratches. The mechanical sensor is used to detect the thickness and hardness of the raw material. The data obtained during this working process are [0.5mm, 0.49mm, 0.51mm, 0.5mm] (thickness) and [70 HRC, 69 HRC, 71 HRC, 70 HRC] (hardness). The thermal sensor is used to monitor the temperature during the heating process. The data obtained during this working process is [120°C, 122°C, 121°C, 120°C]. The electrical sensor is used to detect the conductivity of the raw material. The data obtained during this working process is [0.01Ω·m, 0.011Ω·m, 0.01Ω·m, 0.01Ω·m].

[0074] Furthermore, in a production line of a brake sensor line in this embodiment, the surface finish, thickness, hardness and temperature of the raw materials are monitored. The following data are obtained from the sensors on the production line:

[0075] Optical sensor: detects several tiny scratches on the surface of the raw material.

[0076] Mechanical sensor: The measured thickness of the raw material is 0.5 mm and the hardness is 70 HRC.

[0077] Thermal sensor: registers a temperature of 120°C in the heating furnace.

[0078] Electrical sensor: The resistivity of the raw material detected is 0.01 Ω·m.

[0079] At the same time, image data obtained from cameras on the production line showed that there were some tiny scratches on the surface of the raw materials, but the overall appearance was good.

[0080] Specifically, through the formula for obtaining the state estimation value of the sensor in this embodiment, the fused mechanical and thermal data can be obtained as: [0.5, 70, 120].

[0081] Step SA1.1.2: Transform the sensor state estimate Fuse with multimodal data to obtain fused data, specifically:

[0082]

[0083] in: is the fused data at time k, is the multimodal data acquired at time k, is the estimated value of the state at time k.

[0084] It is worth noting that the multimodal data in this embodiment is specifically described by taking image data as an example.

[0085] Step SA1.1.3: Extract the material performance information of the brake sensor line, such as surface finish, strength, hardness, temperature, etc., from the fused data obtained in step SA1.1.2. That is, extract the first feature data from the fused data by principal component analysis. In other words, obtain the material performance information of the brake sensor line, and obtain the first feature data from the material performance information of the brake sensor line, specifically:

[0086]

[0087] in: is the first feature data, is the left singular vector matrix, is a diagonal matrix, is the right singular vector matrix, is the fused data at time k, is the transpose of the right singular vector matrix.

[0088] Furthermore, in this embodiment, the first characteristic data obtained are surface finish, hardness and temperature, where the left singular vector matrix is [u1,u2,u3], a diagonal matrix is [σ1,σ2,σ3]. Specifically, the data u1, u2 and u3 are the main eigenvectors, and the data σ1, σ2 and σ3 are the corresponding singular values.

[0089] Step SA1.1.4: The first feature data obtained in step SA1.1.3 is sequentially cleaned, normalized and smoothed to facilitate data analysis in subsequent steps. Specifically, data cleaning in this embodiment is to remove invalid data, outliers and noise, data normalization is to convert data of different dimensions into the same dimension, and data smoothing is to eliminate high-frequency noise in the data through a filter.

[0090] Furthermore, when performing data cleaning, this embodiment takes the removal of abnormal values ​​in mechanical sensor data as an example for specific description. The specific calculation formula for judging the abnormal value is:

[0091]

[0092] in: is the value of the i-th data point in the data set, is the mean of the data set, is the standard deviation of the data set.

[0093] That is to say, after the above data is judged as outliers, the cleaned data is [0.5, 0.5, 0.5, 0.5].

[0094] Furthermore, after data cleaning, the mechanical sensor data is further normalized, specifically:

[0095]

[0096] in: is the normalized value of the ith data point in the data set, is the value of the i-th data point in the data set, is the minimum value in the data set, is the maximum value in the data set.

[0097] That is to say, after the above data is normalized, the normalized data is [0, 0.5, 1, 0.5].

[0098] Furthermore, after the data is normalized, the mechanical sensor data is smoothed by a moving average filter, specifically:

[0099]

[0100] in: is the value of the jth data point in the data set, is the smoothing window width, is the index of the data point in the dataset, The index of the data point whose smoothing value is currently being calculated.

[0101] That is to say, after the above data is smoothed, the smoothed data is [0.5, 0.5, 0.5, 0.5].

[0102] Step S1.2: According to the material performance information processed in step SA1.1.4, set the early warning dynamic threshold, and compare the data analysis results of the material performance information with the early warning dynamic threshold. Specifically, when the data analysis results exceed the early warning dynamic threshold range, the early warning mechanism is triggered, otherwise, the early warning mechanism is not triggered. The details are as follows:

[0103] Step S1.2.1: According to the change state of the material performance information processed in step SA1.1.4, set the early warning dynamic threshold. Specifically, in this embodiment, the early warning dynamic threshold is regularly updated through the rolling window method, and the mean and standard deviation of the data are re-obtained. The size of the window and the formula for obtaining the early warning dynamic threshold are specifically:

[0104]

[0105] in: is the mean of the material data at time t, is the size of the time window, is the material data value at the nth time point, is the standard deviation of the material data at time t, , is the index number of the time, is the dynamic threshold for early warning, is the safety factor.

[0106] Furthermore, the safety factor in this embodiment is The size of is set to 2 or 3. In the specific implementation process, the mean and standard deviation are calculated based on the data of 30 time points:

[0107] The time window size is: .

[0108] The hardness data in the material data is: .

[0109] That is, its mean is: .

[0110] The standard deviation is: .

[0111] At the same time, the safety factor in this embodiment The size is set to 3, then the warning dynamic threshold The size is: .

[0112] Step S1.2.2: Compare the data analysis results of the material performance information with the early warning dynamic threshold set in step S1.2.1. When the data analysis results exceed the early warning range of the early warning dynamic threshold, execute the next step S1.2.3; otherwise, execute step S2.

[0113] That is to say, when the data analysis result obtained in real time shows that the hardness is 76HRC, it means that the hardness data 76HRC obtained at this time is not within the range of 67HRC to 73HRC mentioned above, and the next step S1.2.3 is executed at this time.

[0114] Step S1.2.3: Determine the alarm level based on the extent to which the data analysis results exceed the warning range of the warning dynamic threshold. The alarm levels include slight abnormalities, moderate abnormalities, and severe abnormalities. Slight abnormalities are those that exceed the warning range of the warning dynamic threshold by 5%, moderate abnormalities are those that exceed the warning range of the warning dynamic threshold by 5%-10%, and severe abnormalities exceed the warning range of the warning dynamic threshold by more than 10%.

[0115] That is to say, when the data analysis result does not exceed the warning range of the warning dynamic threshold or exceeds the warning range of the warning dynamic threshold by less than 5%, the alarm will not be triggered at this time.

[0116] When the data analysis results exceed the warning range of the warning dynamic threshold by 5%, the alarm level triggered at this time is a minor abnormality.

[0117] When the data analysis results exceed the warning range of the warning dynamic threshold by 5%-10%, the alarm level triggered at this time is medium abnormality.

[0118] When the data analysis results exceed the warning range of the warning dynamic threshold by more than 10%, the alarm level triggered at this time is a serious abnormality.

[0119] Step S1.2.4: According to the alarm level, adjust the working mode of the brake sensor line production system and send an alarm signal, specifically:

[0120] When the alarm level is slightly abnormal, the brake sensor line production system automatically corrects and does not send an alarm signal.

[0121] When the alarm level is medium abnormal or severe abnormal, the brake sensor line production system enters the monitoring mode and sends an alarm signal at the same time.

[0122] Step S2: According to the database management system, such as a relational database (such as MySQL, PostgreSQL) or a NoSQL database (such as MongoDB, Cassandra), and according to the business needs, the table structure is designed to establish a data traceability model. That is to say, through the established data traceability model, the information data in the production process of the brake sensor line is recorded and stored, and the information data includes production parameters (such as temperature, pressure, speed and other process parameters), sensor data (data obtained from various sensors, such as optical sensors, mechanical sensors, thermal sensors, electrical sensors, etc.), quality inspection data (product inspection data obtained through quality control links, such as dimensional measurement, hardness testing, surface finish, etc.), operation log (recording the operator's operation behavior and timestamp), alarm information (recording the alarm information issued by the system and its cause) and production batch information (including production date, batch number, production line number and other identification information).

[0123] Step S3: Obtain the real-time analysis results of the brake sensor line, that is, collect various indicator data of the brake sensor line in real time through sensors and other monitoring equipment, and analyze and process the various indicator data through the operation process of step S1, and then compare them with the preset quality standards. When the real-time analysis results do not pass the quality standards, the information data of the brake sensor line production process is obtained through the data traceability model, and the production parameters are adjusted or the production process is improved according to the information data. Otherwise, the production line of the brake sensor line continues to operate.

[0124] It is worth noting that the preset quality standards here are to set a series of quality standards or thresholds based on industry standards, customer needs and production process requirements. In other words, there is no clear value and it can be set according to actual needs.

[0125] During the implementation process, the real-time data obtained from the sensors include:

[0126] Surface finish: 0.98.

[0127] Thickness: 0.5mm.

[0128] Hardness: 70HRC.

[0129] Temperature: 120℃.

[0130] Electrical conductivity: 0.01Ω·m.

[0131] After the above data is analyzed and processed through the operation flow of step S1, the following data is obtained:

[0132] Surface finish: 0.97.

[0133] Thickness:0.5.

[0134] Hardness:70.

[0135] Temperature:120.

[0136] Electrical conductivity: 0.01.

[0137] The preset quality standards set according to actual needs are:

[0138] Surface finish: ≥0.97.

[0139] Thickness: 0.49mm≤Thickness≤0.5mm.

[0140] Hardness: 69HRC≤Hardness≤71HRC.

[0141] Temperature: 120℃≤Temperature≤122℃.

[0142] Electrical conductivity: ≤0.011Ω·m.

[0143] After comparing the above analyzed data with the preset quality standards, the results are as follows:

[0144] Surface finish: 0.97<0.98 (failed).

[0145] Thickness: 0.55∈[0.49,0.51] (through).

[0146] Hardness: 70∈[69,71] (pass).

[0147] temperature: 120∈[120,122] (passed).

[0148] Electrical conductivity: 0.01≤0.011 (pass).

[0149] That is, the surface finish feature does not pass quality standards and appropriate action needs to be taken.

[0150] Example 2

[0151] refer to Figure 3 This embodiment provides an intelligent online quality detection and tracing method for brake sensor lines. The basic process is the same as that of Embodiment 1. The difference is that in step S1.1, the data of different sensors on the brake sensor line production line are fused to obtain the material performance information of the brake sensor line, and the material performance information is preprocessed. The preprocessing can also be performed through the following steps, specifically including:

[0152] Step SB1.1.1: Preprocess the sensor data and multimodal data on the brake sensor line production line, build a deep learning model for multimodal data fusion, and use the preprocessed sensor data and multimodal data on the brake sensor line production line as input to the deep learning model, and automatically extract and fuse features through the deep learning model.

[0153] Specifically, in this embodiment, the sensor data and multimodal data collected on the brake sensor line production line are preliminarily cleaned and normalized to remove invalid data, outliers and noise, and convert data of different dimensions into the same dimension.

[0154] In the specific implementation process, the data of the optical sensor is [0.98, 0.99, 0.98, 0.97], where 0.98 means no obvious defects and 0.97 means slight scratches. The data of the mechanical sensor are [0.5mm, 0.49mm, 0.51mm, 0.5mm] (thickness) and [70 HRC, 69 HRC, 71 HRC, 70 HRC] (hardness). The data of the thermal sensor is [120°C, 122°C, 121°C, 120°C]. The data of the electrical sensor is [0.01Ω·m, 0.011Ω·m, 0.01Ω·m, 0.01Ω·m].

[0155] That is to say, according to the above sensor data, it can be known that the thickness in the mechanical sensor data is:

[0156]

[0157] The hardness in the force sensor data is:

[0158]

[0159] The temperature in the force sensor data is:

[0160]

[0161] The resistivity in the force sensor data is:

[0162]

[0163] Furthermore, by constructing a multimodal convolutional neural network MM-CNN model with an attention mechanism, the sensor data and multimodal data on the brake sensor line production line after the above preprocessing are used as the input of the MM-CNN model, and the fusion feature vector is obtained as the output, specifically:

[0164]

[0165] in: is the fusion feature vector, is the preprocessed sensor data, The preprocessed multimodal data.

[0166] That is to say, after the above data are fused, the output fused feature vector is: [0.3, 0.4, 0.6, 0.5, 0.2, 0.8, 0.7, 0.3, 0.1, 0.9].

[0167] Step SB1.1.2: Obtain material performance information of the brake sensor line production line from the fused features through the attention mechanism, specifically:

[0168]

[0169] in: is the fusion feature vector, is the weight matrix, is the feature weight matrix, is the bias vector, is the activation function.

[0170] That is to say, the material property information is [0.4, 0.6, 0.8, 0.5, 0.3, 0.9, 0.7, 0.4, 0.2, 0.1].

[0171] Step SB1.1.3: Filter outliers in the data enhancement module of the MM-CNN model, perform regularization in the hidden layer of the MM-CNN model, and perform data normalization in the input layer of the MM-CNN model.

[0172] That is to say, the data output by the data enhancement module is: [0.5, 0.1, 1, 0.5], [0, 1.1, 0.5, 0]. The data after regularization processing is: [0.4, 0.6, 0.8, 0.5, 0.3, 0.9, 0.7, 0.4, 0.2, 0.1]. The data after data normalization processing is: [0.5, 0, 1, 0.5], [0, 1, 0.5, 0].

[0173] Example 3

[0174] refer to Figure 4, this embodiment provides an intelligent online quality detection and tracing system for brake sensor lines, which includes an acquisition module, a judgment module, an early warning module, a storage module and a tracing module. In this embodiment, the acquisition module acquires sensor data on the brake sensor line production line and sends the acquired sensor data to the judgment module. The judgment module compares the sensor data sent by the acquisition module with the early warning dynamic threshold through the set early warning dynamic threshold, and sends the comparison result to the early warning module. The early warning module determines the operating status of the brake sensor line production system by receiving the judgment result sent by the judgment module, and sends the operating status result of the brake sensor line production system to the tracing module. The storage module records and stores the information data in the production process of the brake sensor line through the established data tracing model. The tracing module retrieves the information data in the production process of the brake sensor line stored in the storage module by receiving the operating status result of the brake sensor line production system sent by the early warning module, and adjusts the production parameters or improves the production process according to the stored information data in the production process of the brake sensor line.

[0175] Specifically, the specific detection and tracing process of the online quality detection and tracing system in this embodiment is as follows:

[0176] Step M1: The acquisition module acquires sensor data on the production line of the brake sensor line through the sensor network set on the production line of the brake sensor line, and sends the acquired sensor data to the judgment module so that the judgment module makes corresponding data judgments based on the sensor data.

[0177] Step M2: The judgment module compares the sensor data sent by the acquisition module with the warning dynamic threshold through the set warning dynamic threshold. When the sensor data is within the range of the warning dynamic threshold, the warning state is not triggered. When the sensor data is not within the range of the warning dynamic threshold, the warning state is triggered and the next step M3 is executed at the same time.

[0178] Step M3: When the warning state is triggered, the operating state of the brake sensor line production system is adjusted according to the level of the warning state. That is, when the alarm level is a slight abnormality, the brake sensor line production system automatically corrects and does not send an alarm signal. When the alarm level is a medium abnormality or a serious abnormality, the brake sensor line production system enters the monitoring mode and sends an alarm signal at the same time. It is worth noting that in this embodiment, the operating state of the brake sensor line production system will be sent to the traceability module to facilitate the traceability of its overall operation process.

[0179] Step M4: The established data traceability model records and stores the information data in the production process of the brake sensor line, and sends the recorded and stored information data to the traceability module so that the traceability module can perform a reliable traceability.

[0180] Step M5: When the operating state of the brake sensor line production system changes, it indicates that a corresponding fault has occurred in the brake sensor line production system. At this time, the traceability module retrieves the information data of the brake sensor line production process stored in the storage module, and can adjust the production parameters or improve the production process according to the stored information data of the brake sensor line production process.

[0181] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is limited by the attached embodiments and their equivalents.

Claims

1. An intelligent online quality detection and tracing method for brake sensor lines, characterized in that: The online quality detection and tracing method includes: S1: Setting up a sensor network, acquiring and processing sensor data in the production process, and testing the sensor data at the same time. If the test passes, the next step S2 is executed. Otherwise, the early warning mechanism is triggered. S2: Establishing a data traceability model, and recording and storing information data during the production process of the brake sensor line, wherein the information data includes production parameters, sensor data, quality inspection data, operation logs, alarm information and production batch information; S3: obtaining the real-time analysis result of the brake sensor line, and comparing the real-time analysis result with the preset quality standard. When the real-time analysis result does not pass the quality standard, obtaining the information data of the brake sensor line production process through the data traceability model, and adjusting the production parameters or improving the production process according to the information data. Otherwise, the production line of the brake sensor line continues to operate.

2. The method for online quality detection and tracing of an intelligent brake sensor line according to claim 1 is characterized in that: The triggering process of the early warning mechanism includes: S1.1: fusing different sensor data on the brake sensor line production line to obtain material performance information of the brake sensor line, and preprocessing the material performance information; S1.2: According to the material performance information, a warning dynamic threshold is set, and the data analysis result of the material performance information is compared with the warning dynamic threshold. When the data analysis result exceeds the warning dynamic threshold range, the warning mechanism is triggered, otherwise, the warning mechanism is not triggered.

3. The method for online quality detection and tracing of an intelligent brake sensor line according to claim 2 is characterized in that: A process of obtaining and preprocessing the material performance information includes: SA1.1.1: The different sensor data on the brake sensor line are fused through the Kalman filter to obtain the state estimation value of the sensor, specifically: ,in: is the estimated value of the state at time k, is the one-step prediction state estimate at time k, is the Kalman gain, is the measured value at time k, is the observation matrix at time k; SA1.1.2: Fuse the state estimation value of the sensor with the multimodal data to obtain fused data, specifically: ,in: is the fused data at time k, is the multimodal data acquired at time k, is the estimated value of the state at time k; SA1.1.3: Obtain material property information of the brake sensor line from the fused data, and obtain first feature data from the material property information of the brake sensor line, specifically: ,in: is the first feature data, is the left singular vector matrix, is a diagonal matrix, is the right singular vector matrix, is the fused data at time k, is the transposed matrix of the right singular vector matrix; SA1.1.4: Perform data cleaning, data normalization and data smoothing on the first feature data in sequence.

4. The method for online quality detection and tracing of an intelligent brake sensor line according to claim 2 is characterized in that: Another process of obtaining and preprocessing the fused data also includes: SB1.1.1: Preprocess the sensor data and multimodal data on the brake sensor line production line, and construct the MM-CNN model. The preprocessed sensor data and multimodal data on the brake sensor line production line are used as the input of the MM-CNN model, and the fusion feature vector is obtained as output, specifically: ,in: is the fusion feature vector, is the preprocessed sensor data, is the preprocessed multimodal data; SB1.1.2: Obtain material performance information of the brake sensor line production line from the fused features through the attention mechanism, specifically: ,in: is the fusion feature vector, is the weight matrix, is the feature weight matrix, is the bias vector, is the activation function; SB1.1.3: Filter outliers in the data enhancement module of the MM-CNN model, perform regularization processing in the hidden layer of the MM-CNN model, and perform data normalization processing in the input layer of the MM-CNN model.

5. The method for online quality detection and tracing of an intelligent brake sensor line according to claim 2, 3 or 4, characterized in that: Comparing the data analysis results of the material performance information with the early warning dynamic threshold, and determining the triggering state of the early warning mechanism according to the comparison results, including: S1.2.1: Set a warning dynamic threshold according to the change status of the material performance information; S1.2.2: Compare the data analysis result of the material performance information with the early warning dynamic threshold. When the data analysis result exceeds the early warning range of the early warning dynamic threshold, execute the next step S1.2.

3. Otherwise, execute step S2. S1.2.3: Determine the alarm level according to the extent to which the data analysis result exceeds the warning range of the warning dynamic threshold, and the alarm level includes slight abnormality, moderate abnormality and severe abnormality. The slight abnormality is 5% exceeding the warning range of the warning dynamic threshold, the moderate abnormality is 5%-10% exceeding the warning range of the warning dynamic threshold, and the severe abnormality is more than 10% exceeding the warning range of the warning dynamic threshold; S1.2.4: According to the alarm level, adjust the working mode of the brake sensor line production system and send an alarm signal, specifically: When the alarm level is a slight abnormality, the brake sensor line production system automatically corrects and does not send an alarm signal; When the alarm level is a medium abnormality or a serious abnormality, the brake sensor line production system enters a monitoring mode and sends an alarm signal at the same time.

6. The method for online quality detection and tracing of an intelligent brake sensor line according to claim 5 is characterized in that: The formula for obtaining the early warning dynamic threshold is specifically: ,in: is the mean of the material data at time t, is the size of the time window, is the material data value at the nth time point, is the standard deviation of the material data at time t, , is the index number of the time, is the dynamic threshold for early warning, is the safety factor.

7. An intelligent online quality detection and tracing system for brake sensor lines, characterized in that: The online quality inspection and traceability system includes: An acquisition module is used to acquire sensor data on the production line of the brake sensor line; A judgment module sets a warning dynamic threshold and judges the relationship between the warning dynamic threshold and the sensor data; An early warning module determines the operating state of the brake sensor line production system according to the judgment relationship between the early warning dynamic threshold and the sensor data; A storage module, which records and stores information data during the production process of the brake sensor line through a data tracing model; The traceability module adjusts the production parameters or improves the production process through the storage module according to the operating status of the brake sensor line production system.

Images