Automobile part quality early warning method and device, computer equipment and readable storage medium

Abstract

The invention relates to an automobile part quality early warning method and device, computer equipment and a computer readable storage medium. The method comprises the following steps: according to historical quality defect data of an automobile part, carrying out early warning detection on the automobile part under a preset early warning dimension to obtain an early warning detection result; the preset early warning dimension comprises at least one of the number of defective products, the growth rate of the defective products and the predicted change rate of the defective products; the historical quality defect data comprises the number of defective products of the automobile parts in a plurality of periods; carrying out fusion processing on the early warning detection results under each early warning dimension to obtain a comprehensive early warning detection result of the automobile part; and performing graded quality early warning on the automobile parts according to the comprehensive early warning detection result. By adopting the method, the early warning accuracy can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of vehicle quality management, and in particular to a quality early warning method, device, computer equipment and computer-readable storage medium for automobile parts. Background Art

[0002] With the rapid development of the automotive industry, defective products in the automotive market, such as the outbreak of quality problems, will consume a lot of labor costs in isolation and analysis; if defective parts flow into the market, it may cause economic losses to automobile manufacturers, dealers and related industrial chains. Therefore, it is necessary to provide timely and effective early warning of defective product conditions so that quality management personnel can take corresponding measures in advance to reduce potential economic losses.

[0003] Traditional market quality analysis methods for automotive parts often rely on manual experience and simple statistical data. Warning results are obtained by comparing preset warning thresholds with statistical data, resulting in relatively simple warning results and warning benchmarks, resulting in low warning accuracy. Summary of the invention

[0004] Based on this, it is necessary to provide a quality warning method, device, computer equipment, computer-readable storage medium and computer program product for automobile parts that can improve the accuracy of warning in response to the above technical problems.

[0005] In a first aspect, the present application provides a quality early warning method for automobile parts, comprising:

[0006] According to the historical quality defect data of the automobile parts, early warning detection is performed on the automobile parts under the preset early warning dimension to obtain an early warning detection result; the preset early warning dimension includes at least one of the number of defective products, the growth rate of defective products and the predicted change rate of defective products; the historical quality defect data includes the number of defective products of the automobile parts in multiple cycles;

[0007] Performing fusion processing on the early warning detection results under each early warning dimension to obtain a comprehensive early warning detection result of the automobile part;

[0008] According to the comprehensive early warning detection results, graded quality early warning is carried out for the automobile parts.

[0009] In one embodiment, the step of performing early warning detection on the automobile parts under a preset early warning dimension based on the historical quality defect data of the automobile parts to obtain the early warning detection result includes:

[0010] Determining the cumulative number of defective products of the automotive parts based on the historical quality defect data;

[0011] According to the cumulative number of defective products and the number of defective products in each cycle, the early warning detection result of the automobile parts in the dimension of the number of defective products is obtained.

[0012] In one embodiment, obtaining the early warning detection result of the automobile parts in the dimension of the number of defective products according to the cumulative number of defective products and the number of defective products in each cycle includes:

[0013] According to the number of defective products in each of the cycles, determine the growth amount of defective products in the cycle closest to the current moment relative to the next closest cycle, as well as the changing trend of the number of defective products in multiple cycles closest to the current moment; when the cumulative number of defective products is greater than the quantity threshold, the growth amount of defective products is greater than the growth threshold, or the changing trend is a continuous growth, determine that the early warning detection result of the automotive parts in the defective product quantity dimension is a triggering early warning; the growth threshold is determined based on the average number of defective products in each cycle.

[0014] In one embodiment, the step of performing early warning detection on the automobile parts under a preset early warning dimension based on the historical quality defect data of the automobile parts to obtain the early warning detection result includes:

[0015] According to the historical quality defect data, the defective product growth rate of each cycle and the average defective product growth rate of each cycle are determined; from each cycle, a target cycle in which the defective product growth rate is greater than the average defective product growth rate is determined; when the number of the target cycles is greater than a preset target cycle number threshold, the early warning detection result of the automotive part under the defective product growth rate dimension is determined to trigger an early warning.

[0016] In one embodiment, the step of performing early warning detection on the automobile parts under a preset early warning dimension based on the historical quality defect data of the automobile parts to obtain the early warning detection result includes:

[0017] The quality defect data of the period closest to the current moment is input into a pre-trained defective product quantity prediction model to obtain a defective product quantity prediction result for the next period of the period closest to the current moment; the pre-trained defective product prediction model is obtained by training an initial long short-term memory neural network through historical quality defect data of a variety of automobile parts; a secondary difference processing is performed on the defective product quantity prediction result to obtain a defective product change rate for the next period; when the defective product change rate for the next period exceeds a preset warning value, it is determined that the warning detection result of the automobile part under the dimension of the predicted defective product change rate is a triggering warning.

[0018] In one embodiment, the quality early warning method for automobile parts described in the above embodiment further includes:

[0019] Based on the historical quality defect data, the mean of the defective product growth rate and the variance of the defective product growth rate are determined; based on the time period length of a single cycle, the mean and variance of the defective product growth rate, and the risk value algorithm, the preset warning value is determined.

[0020] In one embodiment, the step of providing a graded quality warning for each of the automobile parts according to the comprehensive warning detection result includes:

[0021] Obtain a graded quality warning result according to the comprehensive warning detection result and the preset graded warning rule; and generate a heat map corresponding to the graded quality warning result according to the graded quality warning result;

[0022] The heat map is displayed.

[0023] In a second aspect, the present application also provides a quality early warning device for automobile parts, comprising:

[0024] an early warning determination module, configured to perform early warning detection on the automobile parts under preset early warning dimensions according to historical quality defect data of the automobile parts, and obtain early warning detection results; the preset early warning dimensions include at least one of the number of defective products, the growth rate of defective products, and the predicted change rate of defective products; the historical quality defect data include the number of defective products of the automobile parts in multiple cycles;

[0025] A result fusion module is used to fuse the warning detection results under various warning dimensions to obtain a comprehensive warning detection result of the automobile parts;

[0026] The graded warning module is used to provide graded quality warning for the automobile parts according to the comprehensive warning detection results.

[0027] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0028] According to the historical quality defect data of the automobile parts, early warning detection is performed on the automobile parts under the preset early warning dimension to obtain an early warning detection result; the preset early warning dimension includes at least one of the number of defective products, the growth rate of defective products and the predicted change rate of defective products; the historical quality defect data includes the number of defective products of the automobile parts in multiple cycles;

[0029] Performing fusion processing on the early warning detection results under each early warning dimension to obtain a comprehensive early warning detection result of the automobile part;

[0030] According to the comprehensive early warning detection results, graded quality early warning is carried out for the automobile parts.

[0031] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0032] According to the historical quality defect data of the automobile parts, early warning detection is performed on the automobile parts under the preset early warning dimension to obtain an early warning detection result; the preset early warning dimension includes at least one of the number of defective products, the growth rate of defective products and the predicted change rate of defective products; the historical quality defect data includes the number of defective products of the automobile parts in multiple cycles;

[0033] Performing fusion processing on the early warning detection results under each early warning dimension to obtain a comprehensive early warning detection result of the automobile part;

[0034] According to the comprehensive early warning detection results, graded quality early warning is carried out for the automobile parts.

[0035] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0036] According to the historical quality defect data of the automobile parts, early warning detection is performed on the automobile parts under the preset early warning dimension to obtain an early warning detection result; the preset early warning dimension includes at least one of the number of defective products, the growth rate of defective products and the predicted change rate of defective products; the historical quality defect data includes the number of defective products of the automobile parts in multiple cycles;

[0037] Performing fusion processing on the early warning detection results under each early warning dimension to obtain a comprehensive early warning detection result of the automobile part;

[0038] According to the comprehensive early warning detection results, graded quality early warning is carried out for the automobile parts.

[0039] The above-mentioned quality warning method, device, computer equipment, computer-readable storage medium and computer program product for automobile parts, by performing warning detection on automobile parts under preset warning dimensions according to historical quality defect data of automobile parts, obtains warning detection results, and realizes multi-faceted warning detection of historical quality defect data of automobile parts, wherein the preset warning dimension includes at least one of the number of defective products, the growth rate of defective products and the predicted change rate of defective products, and the historical quality defect data includes the number of defective products of automobile parts in multiple cycles, which not only considers the quality of automobile parts in the historical time and the change of defective products, but also considers the change of predicted defective products, and can take corresponding measures according to the prediction results in time. Then, the warning detection results under each warning dimension are fused to obtain the comprehensive warning detection results of automobile parts, and the final comprehensive warning detection results are obtained after the warning detection results from multiple aspects are fused, avoiding the one-sidedness of the warning, and further according to the comprehensive warning detection results, the automobile parts are graded for quality warning, and the graded warning can improve the effect of the warning, and facilitate targeted response according to the graded quality warning and take corresponding measures. In summary, the method realizes multi-dimensional warning detection for automobile parts, which can improve the accuracy of the warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 A schematic diagram of a process of a quality early warning method for automobile parts in one embodiment;

[0042] Figure 2 A schematic diagram of a quantity threshold calculation process in one embodiment;

[0043] Figure 3 A schematic diagram of a defective product quantity variation curve showing an increase in volatility in one embodiment;

[0044] Figure 4 It is a schematic diagram of a relationship curve between the cumulative defective rate (cumulative bad rate) and time in one embodiment;

[0045] Figure 5 A schematic diagram showing test results obtained through training in an embodiment;

[0046] Figure 6 A schematic diagram of a preprocessing process of historical quality defect data in one embodiment;

[0047] Figure 7 A schematic diagram of a process for determining a defective product change rate in the next cycle in one embodiment;

[0048] Figure 8 A schematic diagram of a result of triggering an early warning due to a defective product change rate in the next cycle in an embodiment;

[0049] Fig. 9 A schematic diagram of a heat map of early warning scores in an embodiment;

[0050] Fig.10 A schematic flow chart of a method for early warning of a defect of an automobile part in another embodiment;

[0051] Fig.11 A schematic diagram of the structure of a bad classification warning model for the automobile parts market in one embodiment;

[0052] Fig.12 A schematic diagram of a calculation process of the cumulative bad growth rate and the mean value in one embodiment;

[0053] Fig.13 A schematic diagram of a verification warning result of a market bad classification warning model in an embodiment;

[0054] Fig.14 It is a schematic diagram of the bad warning result of the 2020 EVCT gearbox in one embodiment;

[0055] Fig.15 is a structural block diagram of a quality early warning device for automobile parts in one embodiment;

[0056] Fig.16 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0058] In one embodiment, Figure 1 As shown, a quality early warning method for automobile parts is provided. This embodiment takes the method applied to a terminal as an example for illustration. It can be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0059] Step S102, based on the historical quality defect data of the automobile parts, early warning detection is performed on the automobile parts under the preset early warning dimension to obtain early warning detection results.

[0060] Among them, the preset warning dimensions include at least one of the number of defective products, the defective product growth rate and the predicted defective product change rate.

[0061] Among them, historical quality defect data includes the number of defective auto parts in multiple cycles. The cycle can be set according to actual statistical needs, such as setting a cycle of one month, one week or one day. Historical quality defect data can be related data on defective auto parts in the market data from any historical moment to the current moment. The data features are production and purchase date, failure date, mileage, model year, frame number, engine, transmission number, diagnosis content, authorized store information, vehicle type, model, part number, sales volume, etc.

[0062] There may be many types of automobile parts, and each type of automobile part is applicable to the quality early warning method for automobile parts proposed in this embodiment.

[0063] Optionally, the system obtains historical quality defect data of automobile parts, performs preprocessing such as data cleaning on the acquired historical quality defect data, and performs early warning detection on automobile parts under preset dimensions based on the preprocessed historical quality defect data to obtain corresponding early warning detection results.

[0064] Step S104, fusing the warning detection results under each warning dimension to obtain a comprehensive warning detection result of the automobile parts.

[0065] Among them, the comprehensive early warning detection result can be a final early warning result for the quality of automobile parts that integrates the early warning detection results under various early warning dimensions, which is more comprehensive.

[0066] Optionally, the system integrates the warning detection results under each warning dimension. For example, the warning detection results under each dimension are determined by a scoring system, and each warning detection result corresponds to its own score. The scores of all warning detection results are added together to obtain a comprehensive warning detection result for automobile parts.

[0067] Step S106, performing graded quality warning on automobile parts according to the comprehensive warning detection results.

[0068] Among them, graded quality warnings can be warnings of different levels for different quality levels.

[0069] Optionally, the system determines the quality grade of the detected auto parts according to the comprehensive warning test results, determines the corresponding warning method according to the quality grade, and performs graded quality warning according to the corresponding warning method. For example, if the comprehensive quality test result of the auto parts is severe, a red warning is issued, and if the comprehensive quality test result is mild, a yellow prompt is issued, etc.

[0070] In the above-mentioned quality early warning method for automobile parts, early warning detection is performed on automobile parts under preset early warning dimensions according to the historical quality defect data of automobile parts to obtain early warning detection results, thereby realizing multi-faceted early warning detection of historical quality defect data of automobile parts, wherein the preset early warning dimensions include at least one of the number of defective products, the growth rate of defective products and the predicted change rate of defective products, and the historical quality defect data include the number of defective products of automobile parts in multiple cycles, which not only considers the quality of automobile parts in the historical time and the change of defective products, but also considers the change of predicted defective products, and can take corresponding measures according to the prediction results in time. Then, the early warning detection results under each early warning dimension are fused to obtain the comprehensive early warning detection results of automobile parts, and the final comprehensive early warning detection results are obtained after the early warning detection results from multiple aspects are fused, thereby avoiding the one-sidedness of the early warning, and further according to the comprehensive early warning detection results, the automobile parts are graded for quality early warning, and the graded early warning can improve the effect of the early warning, and facilitate targeted response according to the graded quality early warning and take corresponding measures. In summary, the method realizes early warning detection under multiple dimensions for automobile parts, which can improve the accuracy of the early warning.

[0071] In an exemplary embodiment, step S102 performs early warning detection on automobile parts under preset early warning dimensions based on historical quality defect data of automobile parts to obtain early warning detection results, including:

[0072] Based on historical quality defect data, the cumulative number of defective auto parts is determined; based on the cumulative number of defective parts and the number of defective parts in each cycle, the early warning detection results of auto parts in the dimension of defective number are obtained.

[0073] The cumulative number of defective products may be the sum of the number of defective products counted in the historical time period corresponding to the historical quality defect data.

[0074] Optionally, the system calculates the cumulative number of defective products of automobile parts based on the total number of defective products in each cycle in the historical quality defective product data, and issues an early warning for the cumulative number of defective products and the number of defective products in each cycle according to the preset early warning rules, thereby obtaining an early warning detection result of automobile parts in the dimension of the number of defective products. For example, the early warning rule may be to compare the cumulative number of defective products and the number of defective products in each cycle with a preset defective product threshold, and determine the early warning detection result based on the comparison result.

[0075] In this embodiment, when early warning detection is performed on automobiles in the dimension of defective product quantity, early warning is performed by considering both the cumulative number of defective products and the number of defective products in each cycle, thereby further improving the accuracy of early warning.

[0076] In an exemplary embodiment, the steps of the above embodiment obtain early warning detection results of automobile parts in the dimension of defective product quantity according to the cumulative defective product quantity and the defective product quantity in each cycle, including:

[0077] According to the number of defective products in each cycle, determine the growth of defective products in the cycle closest to the current moment relative to the next closest cycle, as well as the changing trend of the number of defective products in multiple cycles closest to the current moment; when the cumulative number of defective products is greater than the quantity threshold, the growth of defective products is greater than the growth threshold, or the changing trend is continuous growth, determine that the early warning detection result of automotive parts in the dimension of defective product quantity is a triggering warning.

[0078] The defective product growth amount may be the growth amount of the defective products in any period relative to the defective products in the period before the period.

[0079] The changing trend of the number of defective products may be a change in the number of defective products as each cycle changes, such as an increasing trend over time, a fluctuating trend over time, etc.

[0080] The quantity threshold and the growth threshold can be set and flexibly adjusted according to the actual warning needs. The quantity threshold can be set as follows: Figure 2 As shown, a schematic diagram of the quantity threshold calculation process is provided. The cumulative number of defective pieces in the figure is equivalent to the cumulative number of defective products in this embodiment, and the cumulative number of defective pieces warning value is equivalent to the quantity threshold of this embodiment. Cumulative defective quantity threshold = 0.3% × (year model launch time ÷ warranty period) × cumulative sales (refer to the market measures of peer companies to determine the limit. If the incidence rate is greater than 0.3%, active market measures should be considered to respond). The date (year-month) is recorded as ,in is the start date of the statistics, is the deadline for the statistics, is the number of defective parts of a component in the corresponding month (cycle), The corresponding launch time and sales volume are and , the warranty period is months, the cumulative defective quantity threshold can be written as:

[0081]

[0082] in, It is an adjustable parameter. Here we use the industry experience parameter When the cumulative number of defective products is greater than the quantity threshold, that is When , the auto part is warned. The growth threshold is determined based on the average number of defective products in each cycle. When the growth of defective products in the cycle closest to the current moment relative to the next closest cycle is greater than the growth threshold, a warning is triggered. The corresponding judgment formula is:

[0083]

[0084] in, is an adjustable parameter, is the cumulative number of defectives, that is, when the number of defectives in a single cycle exceeds the average number of defectives in each cycle times, triggering an early warning.

[0085] Optionally, the system determines the difference between the number of defective products in the cycle closest to the current moment and the number of defective products in the cycle second closest to the current moment based on the number of defective products in each cycle, determines the increase in the number of defective products in the cycle closest to the current moment relative to the second closest cycle, and further determines the changing trend of the number of defective products in multiple cycles closest to the current moment. When the cumulative number of defective products is greater than the quantity threshold, the increase in the defective products is greater than the increase threshold, or the changing trend is continuously increasing, the warning detection result of the automotive parts in the defective product quantity dimension is determined to trigger a warning. For example, when the changing trend of the number of defective products in three or more consecutive cycles shows an increasing trend, a warning is triggered.

[0086] It should be noted that in this embodiment, the comparison of the cumulative number of defective products with the quantity threshold to trigger an early warning is aimed at the overall quality level of each automotive part. If an early warning is triggered, it may be due to defects in the part design itself, insufficient overall quality assurance capabilities of the parts supplier, etc. The early warning is based on the growth amount and growth threshold of defective products in each cycle, as well as the change trend of the number of defective products in multiple cycles closest to the current moment, in order to target defective products in recent production batches. If an early warning is triggered, it may be due to changes in the parts supplier, etc.

[0087] In this embodiment, the cumulative number of defective products and the quantity threshold are compared to trigger an early warning, taking into account the overall quality level of automobile parts since leaving the factory. The growth amount of defective products in each cycle and the growth threshold, as well as the changing trend of the number of defective products in multiple cycles closest to the current moment, are used to issue an early warning. This is to check the recent quality level of automobile parts, thereby achieving more accurate early warnings and quickly determining the causes of the early warnings based on the early warning results so that corresponding measures can be taken.

[0088] In an exemplary embodiment, step S104 performs early warning detection on automobile parts under preset early warning dimensions based on historical quality defect data of automobile parts to obtain early warning detection results, including:

[0089] Based on historical quality defect data, determine the defective product growth rate in each cycle and the average defective product growth rate in each cycle; determine the target cycle in which the defective product growth rate is greater than the average defective product growth rate from each cycle; when the number of target cycles is greater than the preset target cycle number threshold, determine the early warning detection result of the automotive parts in the defective product growth rate dimension as triggering an early warning.

[0090] The defective product growth rate may refer to the growth ratio of the number of defective products in any period relative to the number of defective products in the previous period. The calculation method may be the difference between the number of defective products in the period and the number of defective products in the previous period, and the ratio of the difference to the number of defective products in the previous period is used as the defective product growth rate of the period.

[0091] Among them, the target cycle quantity threshold can be set and adjusted according to actual warning needs.

[0092] Optionally, the system calculates the defective product growth rate of each cycle and the average defective product growth rate of each cycle based on the historical quality defect data, compares the defective product growth rate of each cycle with the average defective product growth rate, obtains a comparison result, and determines the target cycle in which the defective product growth rate is greater than the average defective product growth rate from each cycle based on the comparison result. When the number of target cycles is greater than the preset target cycle number threshold, the early warning detection result of the automotive parts under the defective product growth rate dimension is determined to trigger an early warning. For example, historical quality defect data within half a year is selected, and the defective product growth rate and the average defective product growth rate in the past five months are calculated. If the defective product growth rate is greater than the average defective product growth rate for three or more months out of five months, an early warning is triggered.

[0093] It should be noted that this embodiment determines whether to trigger an early warning by determining the target number of cycles in which the defective product growth rate is greater than the average defective product growth rate. The derivation basis of the early warning rule is that the above-mentioned embodiment provides early warning for the cumulative number of defective products, which can effectively provide early warning for parts with excessively high cumulative number of defective products, frequent growth, and sharp growth in recent times, but lacks early warning capabilities for parts with fluctuating numbers of defective products, such as Figure 3 As shown in the figure, a schematic diagram of the defective product quantity (number of defective pieces) variation curve showing an increase in volatility is provided. In order to deal with this situation, it is necessary to use the defective product growth rate for judgment. Since the defective rate of automotive parts generally follows the Weibull distribution, its cumulative defective rate in the first 12 months can be approximated as a linear function, as shown in Figure 4 As shown in the figure, the relationship curve between the cumulative defective rate (cumulative defective rate) and time is provided. By taking the derivative of the cumulative defective rate in the first 12 months, a constant approximate to the slope can be obtained. ; Take the cumulative defective rate in the past six months. Due to the interval of the horizontal axis, directly differentiate the data to get the defective growth rate For parts with fluctuating defective quantity growth, when the defective quantity increases, there is at least one corresponding growth rate. Greater than , the argument is as follows: Assume is the cumulative defective rate for the corresponding month (period), is the mean growth rate of defective products, is the growth rate of defective products in the corresponding month, then:

[0094]

[0095]

[0096]

[0097]

[0098] For parts with fluctuating defective counts, when the number of defectives decreases, the growth rate is a negative number. According to the drawer principle, when the number of defective products increases, there must be at least one growth rate. ,otherwise:

[0099] With the formula Contradiction, proof complete.

[0100] In this embodiment, the defective product growth rate is used to implement early warning for changes in the number of defective products of automobile parts that show increasing volatility. This makes up for the fact that the early warning rules based on the defective product quantity dimension are unable to provide early warning detection for automobile parts with increasing volatility, and further improves the accuracy of quality early warning for automobile parts.

[0101] In an exemplary embodiment, step S106 performs early warning detection on automobile parts under preset early warning dimensions based on historical quality defect data of automobile parts to obtain early warning detection results, including:

[0102] The quality defect data of the cycle closest to the current moment is input into the pre-trained defective product quantity prediction model to obtain the defective product quantity prediction result of the next cycle closest to the current moment; the defective product quantity prediction result is subjected to secondary difference processing to obtain the defective product change rate of the next cycle; when the defective product change rate of the next cycle exceeds the preset warning value, the warning detection result of the automobile parts in the dimension of the predicted defective product change rate is determined to trigger the warning.

[0103] Among them, the pre-trained defective product prediction model is obtained by training the initial long short-term memory neural network through the historical quality defect data of various automobile parts, and the data set obtained from the historical quality defect data is divided into 80% as a training set and 20% as a test set for training. Among them, the long short-term memory neural network is a special recurrent neural network (Recurrent Neural Network, RNN). RNN is a neural network used to process sequence data. Compared with general neural networks, RNN can process data with sequence changes. The horizontal axis of the bad data in the automobile market is time, which belongs to time series data, so RNN-type neural networks can be used for analysis and prediction. Compared with RNN, the advantage of LSTM is that it can remember long-term information and is suitable for processing and predicting important events with long intervals and delays in time series. Therefore, LSTM is selected to predict the number of defective parts of automobile parts. The hyperparameter settings of the initial long short-term memory neural network in this embodiment are shown in Table 1:

[0104] Table 1

[0105]

[0106] The defective quantity prediction model for the 2020 model year 12513 automobile part is trained, such as Figure 5 The following is a schematic diagram showing the test results obtained through training, where the Actual curve is the actual defect quantity data for each month, and the Predicted curve is the predicted defect quantity for each month output by the defect quantity prediction model.

[0107] Among them, difference is to calculate the difference between adjacent elements in the sequence, which can help eliminate trends and seasonality in the data, making the data more suitable for linear modeling. The first difference is to calculate the difference between two adjacent time points; and the second difference is to perform a second difference based on the first difference.

[0108] Optionally, after the system obtains the historical quality defect data, it needs to preprocess the historical quality defect data, such as Figure 6As shown in the figure, the preprocessing process of historical quality defect data is provided. The past defective product data in the market comes from the company's internal database. The data features are production and purchase date, failure date, mileage, model year, frame number, engine, transmission number, diagnosis content, authorized store information, vehicle type, model, part number, sales volume, etc. First, data cleaning is performed. Since the model is based on the failure date for early warning, the output is the early warning value of a certain component of the corresponding model year, so only the failure date, model year and main part number are retained after processing. First, the data containing blank values ​​is removed, and then the abnormal data such as the date filled in incorrectly (mileage, part number) are removed. Then, the cleaned data is preprocessed. After extracting and integrating information such as model year, part number, and failure date, the defective product data table is finally obtained, as shown in Table 2 (taking part 16450 as an example). This table needs to be converted into a four-layer data structure of "part number-model year-date-number of defective products" in python (a computer language), which reduces the computer memory usage and facilitates indexing in subsequent processing.

[0109] Table 2

[0110]

[0111] In addition, if Figure 7 As shown, a flow chart of obtaining the defective product change rate (predicted defective change rate) of the next cycle is provided. The quality defect data of the cycle closest to the current moment after preprocessing is input into the pre-trained defective product quantity prediction model to obtain the defective product quantity prediction result of the next cycle of the cycle closest to the current moment, and the defective product quantity prediction result is subjected to secondary difference processing to obtain the defective product change rate of the next cycle, as shown in FIG. Figure 8 As shown, a schematic diagram of the result of triggering an early warning based on the defective product change rate (predicted defective change rate) in the next cycle is provided. When the defective product change rate in the next cycle exceeds the preset early warning value (VaR warning line), the early warning detection result of the automotive parts under the dimension of predicted defective product change rate is determined to trigger an early warning.

[0112] In this embodiment, a long short-term memory neural network is introduced to predict the number of defective products in the next cycle. By analyzing the prediction results, an early warning can be issued in advance, thereby improving the timeliness of the early warning and taking corresponding measures in a timely manner.

[0113] In an exemplary embodiment, the quality early warning method for automobile parts described in the above embodiment further includes:

[0114] Based on historical quality defect data, determine the mean of the defective product growth rate and the variance of the defective product growth rate; based on the time period length of a single cycle, the mean and variance of the defective product growth rate, and the risk value algorithm, determine the preset warning value.

[0115] Among them, the preset warning value is the VaR (Value at Risk, risk value model) value, and the confidence interval is divided for the prediction result according to the confidence level. When the predicted defective product change rate exceeds the confidence interval, the warning is triggered. Under a certain probability level (confidence level), the maximum possible loss of the value of a financial asset or securities portfolio in a specific period in the future. is the confidence level, is the loss within the time period, and is the maximum possible loss. The VaR model can be expressed as:

[0116]

[0117] Among them, in the defective field of the automobile market, VaR means the maximum defective growth rate of the automobile parts in a specific period in the future under a given confidence level (note that the defective growth rate here is different from the cumulative defective growth rate in the above embodiment. The latter is obtained by the difference of the cumulative defective rate and is a positive number; the former is obtained by the difference of the defective rate, which can be positive or negative, and reflects the degree of change of the defective rate. For the convenience of distinction, it is referred to as the defective change rate below). For example, take , For 3 months, we can calculate , which indicates that in the next 3 months, The probability of defective product change rate is guaranteed Less than .

[0118] Optionally, the mean of the defective product growth rate and the variance of the defective product growth rate are determined based on historical quality defect data; the preset warning value VaR is determined based on the time period length of a single cycle, the mean and variance of the defective product growth rate, and the risk value algorithm. The corresponding formulas include:

[0119]

[0120]

[0121]

[0122] in, It is a standard normal distribution Quantile, are the mean and variance of the defective growth rate, It is the length of a single cycle. For the historical simulation method, the bad growth rates are sorted from large to small, and the 5% quantile value is selected as the VaR value. The VaR values ​​of the parameter method and the historical simulation method are weighted and averaged.

[0123] In this embodiment, the value of the VaR model is calculated by the risk value algorithm as the preset warning value, which can reduce the error of the warning for the defective product change rate and further improve the accuracy of the quality warning.

[0124] In an exemplary embodiment, step S106 performs graded quality warning for each automobile part according to the comprehensive warning detection result, including:

[0125] According to the comprehensive warning detection results and the preset graded warning rules, the graded quality warning results are obtained; according to the graded quality warning results, a heat map corresponding to the graded quality warning results is generated; and the heat map is displayed.

[0126] Among them, the hierarchical warning rule can be the correspondence between the comprehensive warning detection results and the warning categories, as shown in Table 3, which provides the specific content of the preset hierarchical rules:

[0127] Table 3

[0128]

[0129] Among them, the score P represents the score corresponding to the comprehensive warning detection result, and the category represents the corresponding warning category.

[0130] Among them, heat map can be Heatmap is a data visualization technology that is used to represent the numerical density or intensity of data through color changes. It is usually used to analyze and display the distribution of large amounts of data to help people quickly identify patterns, trends or outliers in the data.

[0131] Optionally, the system obtains a graded quality warning result based on the comprehensive warning test result and preset graded warning rules. For example, five warning triggering conditions are described in the above multiple embodiments, and a scoring system is used to assign points to each warning trigger. The warning test result is scored 2 points for each warning trigger. The comprehensive warning test result is the sum of multiple warning test results, with an upper limit of 10 points. Corresponding to Table 3, it can be seen that if the comprehensive warning test result is 0 to 2 points, the graded quality warning result is a reminder, if the comprehensive warning test result is 2 to 6 points, the graded quality warning result is a warning, and if the comprehensive warning test result is 6 to 10 points, the graded quality warning result is a warning. Furthermore, the system generates a heat map corresponding to the graded quality warning results based on the graded quality warning results of automobile parts, and displays the heat map through the system's display module, such as Fig. 9 As shown, a schematic diagram of the early warning score heat map is provided.

[0132] In this embodiment, by introducing a heat map to display the graded quality warning results of the comprehensive warning detection results, technical personnel can intuitively determine the different levels of warning situations from the heat map, and take corresponding measures, thereby realizing the visualization of the comprehensive warning detection results.

[0133] In an exemplary embodiment, Fig.10 As shown, another method for early warning of bad automobile parts is provided, including:

[0134] Step 1, data preprocessing: Use Python's numpy and pandas libraries to clean and preprocess the market bad data, store the data in the data structure of "part number-year-month-date-number of faults", and build a three-layer structure of the bad warning model based on the deep learning algorithm, risk value model, and derivative difference method in the field of artificial intelligence: quantity layer, bad rate layer, and LSTM-VaR layer to achieve graded warning of bad auto parts. Among them, the quantity layer performs three independent logical judgments: whether the cumulative number of bad parts (number of defective products) exceeds the threshold, whether the number of bad parts has increased continuously, and whether it has increased sharply; the growth rate layer mainly supplements the shortcomings of the quantity layer, and differentiates the bad cumulative rate to determine whether there is a fluctuating upward trend; the LSTM-VaR layer sets a warning line according to the confidence level, and determines whether there will be a greater bad risk in the future according to the prediction results. The above three layers all adopt a scoring system, and the graded warning effect is displayed in the form of a heat map according to different scores. In general, the implementation of the model is divided into four sub-parts: data preprocessing, bad analysis, graded warning, and warning visualization. In subsequent calculations, the launch date and sales volume of different model years are needed. In the database, the model year, launch date and sales volume data are directly obtained and integrated into a subtable.

[0135] Among them, the market's past bad data (historical quality defect data) comes from the company's internal database. The data features are production and purchase date, failure date, mileage, model year, frame number, engine, transmission number, diagnostic content, authorized store information, vehicle type, model, part number, sales volume, etc.

[0136] Optionally, data cleaning is performed first. Since the model is based on the fault date for early warning, the output is the early warning value of a certain component of the corresponding model year, so only the fault date, model year and main part number are retained after processing. First, the data containing blank values ​​are removed, and then the abnormal data such as incorrect date filling (wrongly filled in as mileage, part number) are removed. Considering that some parts are defective in a small number of sporadic types (occasionally one or two defective parts appear), in order not to affect the calculation of the subsequent growth rate layer and LSTM-VaR layer, the data with an average monthly defective number of less than (adjustable parameter, 0.5 is selected here) and no continuous defective parts in the recent period are removed to complete the data screening and cleaning. Then, the cleaned data is preprocessed. After extracting and integrating information such as model year, part number, and fault date, the defective data table is finally obtained.

[0137] Step 2: Establishment of the bad warning model: The bad classification warning model architecture of the automotive parts market is as follows: Fig.11 As shown. The preprocessed data (the monthly defective number of a certain component of a certain model year) is input into the model, and will enter the quantity layer, defective rate layer, and LSTM-VaR layer in sequence. Each layer calculates the data through different algorithms, judges and scores the calculation results according to the established rules, adds up the scores of each layer (the early warning detection results under each dimension), and finally outputs the early warning score of the data (comprehensive early warning detection results).

[0138] Among them, the early warning functions of each layer of the model complement each other: the quantity layer is mainly responsible for judging the growth of the number of defects in the near future and issuing early warnings for the recent outbreak of defects; the defect rate layer can determine whether the component has fluctuating growth. If it happens to be in the decline period after the peak, the quantity layer cannot issue an early warning, and the defect rate layer makes up for this; the LSTM-VaR layer uses the variance and quantile of past data to give a warning line based on the confidence level, and issues an early warning based on whether the prediction result touches the line, which supplements the singleness of the first two layers in the time dimension, making the early warning effect more timely. Finally, in the calculation formula of each layer, some adjustable parameters are involved, so that the early warning benchmark can be adjusted according to the actual situation of the defective data and the quality target.

[0139] Optionally, regarding the functional principle of the quantity layer: Among the early warning methods, judging by quantity is the most basic and intuitive method. For defective auto parts, it is to judge whether the number of defective parts in the recent period is on an upward trend and whether the cumulative number of defective parts exceeds the threshold. Based on the above judgment, this layer establishes the following three early warning rules.

[0140] The first judgment is based on the cumulative number of defective parts. The cumulative number of defective parts warning value = 0.3% × (year of model launch time ÷ warranty period) × cumulative sales (refer to the market measures of peer companies to judge the limit. If the incidence rate is greater than 0.3%, active market measures should be considered). When the cumulative number of defective parts (the cumulative number of defective products) is greater than the warning value (quantity threshold), the component is immediately warned. The second judgment is based on the growth of the number of defective parts in multiple periods and considers the number of defective parts (changing trend) in the last three months. If the number of defective parts continues to grow, the component is warned.

[0141] The third case is based on the growth of the number of defective parts in a single period. If the growth of the number of defective parts in a single period is greater than the monthly average number of defective parts, the component will be warned. Among them, is an adjustable parameter.

[0142] In addition, for the above three judgment situations, the reasons for each situation being judged as a warning are as follows: For the first situation, it is mainly for the overall quality level of parts and components. If a warning is triggered, it may be due to defects in the part design itself, insufficient overall quality assurance capabilities of the parts supplier, etc. For the second and third situations, it is mainly for defects in recent production batches. If a warning is triggered, it may be due to changes in the parts supplier, etc.

[0143] Optionally, regarding the function and principle of the defective rate layer, the early warning items of the quantity layer can effectively warn of parts with excessively high cumulative defective numbers, frequent growth, and sharp growth in the recent period, but lack the ability to warn of parts with fluctuating defective numbers. In order to deal with this situation, it is necessary to use the cumulative defective growth rate for judgment, that is, the second layer - the defective rate layer. Since the defective rate of automotive parts generally obeys the Weibull distribution, its cumulative defective rate in the first 12 months can be approximated as a linear function. By taking the derivative of the cumulative defective rate in the first 12 months, a constant approximate to the slope can be obtained; taking the cumulative defective rate in the past six months, due to the interval of the horizontal axis, the data can be directly differentiated to obtain the defective growth rate.

[0144] This layer selects the defective data of the past six months and calculates the cumulative defective growth rate of the past five periods. If the growth rate of three or more periods is greater than the average, the component is warned. Fig.12 As shown in Figure 1, a schematic diagram of the calculation process of the cumulative defective growth rate and mean is provided. In summary, the defective rate layer has a good early warning effect on parts with fluctuating defective parts numbers, and can make up for the shortcomings of the quantity layer.

[0145] Optionally, the function and principle of the LSTM-VaR layer. The quantity layer and the defect rate layer are already quite comprehensive in identifying various types of defective situations. However, the data based on these two layers are only updated to the statistical deadline, resulting in them mostly issuing warnings when the defective problems have already erupted on a large scale. In this case, quality management personnel can often only take measures to stop bleeding, which are essentially remedial measures. From a cost perspective, the cost of remedial measures is much higher than that of preventive measures. In order to improve this situation, this layer introduces a long short-term memory neural network (LSTM) to predict the number of defective pieces. By analyzing the prediction results, early warnings can be issued in advance, thereby overcoming the shortcomings of the first two layers in terms of the timeliness of early warnings. The prediction results are often uncertain, and directly issuing early warnings on the prediction results will bring large errors. Therefore, this layer additionally introduces the Value at Risk (VaR) model to divide the confidence interval for the prediction results according to the confidence level. Only when the predicted defective change rate exceeds the confidence interval, the layer will issue an early warning. Through the introduction of the VaR model, the error of the early warning result can be greatly reduced. First, use the past defective rate to calculate the variance and mean, and get the VaR value according to the formula. Then use LSTM to predict the number of defective items for 3 months, and then perform secondary difference on the predicted data to get the predicted defective change rate. Finally, compare the predicted defective change rate and VaR value to determine whether to issue an early warning.

[0146] Step 3, defect warning and visualization. Parts of each model year go through 5 warning items in total in the quantity layer, growth rate layer and LSTM-VaR layer: Cumulative number of defective parts > warning value? Continuous growth in the number of defective parts? Significant growth in the number of defective parts? Cumulative growth rate of defective parts > average growth rate? Defective change rate > VaR value? The score of each item is 2, and the upper limit of the warning score is 10. Every time a part triggers an item, the warning score is increased. Obviously, the higher the warning score of a part, the higher the severity of the recent defect, the more attention the quality management personnel need, and even the necessary countermeasures need to be taken. Therefore, this model divides the categories of parts according to different scores and performs graded warnings on the parts. According to the scores of the parts, the seaborn library of python is used to draw the warning heat map.

[0147] Step 4, model verification: Take the currently used early warning model as the benchmark for comparative verification. The data set uses the engine defect warranty data of ACC10 models 19, 20, 21 and 22 from March 2019 to August 2022. The early warning time node is August 2022. Fig.13 As shown, a schematic diagram of the verification and early warning results of this model is provided.

[0148] Among the total 472 parts, 66 parts triggered the warning project. Among them, the parts with a score of 2 accounted for 77.55%, belonging to the "reminder" category; the parts with scores of 4 and 6 accounted for 22.45%, belonging to the "warning" category. In the verification process, the warning data calculated by this model and the benchmark are shown in Table 4, where ○ means that the part is only warned by this model, and the benchmark model fails to warn; ● means that the part is only warned by the benchmark model, and this model fails to warn; ▲ means that the part is warned by both the benchmark and this model.

[0149] Table 4

[0150]

[0151] As can be seen from Table 4, this model has achieved full coverage of the warning parts for the baseline model, and the number of warning parts has increased from 33 to 66 compared with the baseline, and the warning sensitivity has been greatly improved. For parts like 12513-2020 (2020 cylinder head blind hole plug), the baseline model cannot warn of such fluctuating rising defects, while this model can accurately identify and warn through the defective rate layer. Similarly, for parts whose defective number has an increasing trend in the future, the baseline model cannot warn them. This model predicts the number of defective parts in the next three months through the LSTM-VaR layer, and finds that the number of defects in this part may increase significantly in the future, thereby issuing a warning and intervening in advance. In the verification results, this model identified a total of 10 parts with a growth trend in the future through the LSTM-VaR layer, of which 2 were not identifiable by the baseline model and the quantity layer and defective rate layer of this model. Taking the 2020 EVCT gearbox as an example, if Fig.14 As shown, a schematic diagram of the adverse warning results of the 2020 EVCT (Electric Variable Continuously Variable Transmission) gearbox is provided.

[0152] In this embodiment, the warning sensitivity and timeliness of the model of this embodiment are greatly improved compared with the baseline model. It can make effective and timely warnings for parts with fluctuating defective numbers and a future growth trend, making up for the shortcomings of the baseline model; the warning results are visualized in the form of heat maps, and graded warnings are used to facilitate quality management personnel to analyze and judge; for the situation of partial over-warning in this model, the warning threshold can be adjusted by adjusting the model parameters, so that the model performance reaches the optimal level under the expectation of quality goals.

[0153] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0154] Based on the same inventive concept, the embodiment of the present application also provides a quality warning device for automobile parts for implementing the quality warning method for automobile parts involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the embodiments of one or more quality warning devices for automobile parts provided below can refer to the limitations of the quality warning method for automobile parts above, and will not be repeated here.

[0155] In an exemplary embodiment, Fig.15 As shown, a quality warning device 1500 for automobile parts is provided, comprising: a warning determination module 1501, a result fusion module 1502 and a graded warning module 1503, wherein:

[0156] The warning determination module 1501 is used to perform warning detection on automobile parts under preset warning dimensions based on historical quality defect data of automobile parts to obtain warning detection results; the preset warning dimensions include at least one of the number of defective products, the growth rate of defective products and the predicted rate of change of defective products; the historical quality defect data include the number of defective products of automobile parts in multiple cycles.

[0157] The result fusion module 1502 is used to fuse the warning detection results under various warning dimensions to obtain comprehensive warning detection results of automobile parts.

[0158] The graded warning module 1503 is used to provide graded quality warnings for automobile parts based on the comprehensive warning detection results.

[0159] Furthermore, in one embodiment, the early warning determination module 1501 is also used to determine the cumulative number of defective products of automobile parts based on historical quality defect data; based on the cumulative number of defective products and the number of defective products in each cycle, obtain the early warning detection results of automobile parts in the dimension of defective product quantity.

[0160] Furthermore, in one embodiment, the warning determination module 1501 is also used to determine the growth amount of defective products in the cycle closest to the current moment relative to the next closest cycle, and the changing trend of the number of defective products in multiple cycles closest to the current moment based on the number of defective products in each cycle; when the cumulative number of defective products is greater than the quantity threshold, the growth amount of defective products is greater than the growth threshold, or the changing trend is continuous growth, determine that the warning detection result of the automobile parts in the defective product quantity dimension is a triggering warning; the growth threshold is determined based on the average number of defective products in each cycle.

[0161] Furthermore, in one embodiment, the warning determination module 1501 is also used to determine the defective product growth rate of each cycle and the average defective product growth rate of each cycle based on historical quality defect data; determine the target cycle in which the defective product growth rate is greater than the average defective product growth rate from each cycle; and when the number of target cycles is greater than a preset target cycle number threshold, determine that the warning detection result of the automotive parts in the defective product growth rate dimension is a triggering warning.

[0162] Furthermore, in one embodiment, the warning determination module 1501 is also used to input the quality defect data of the cycle closest to the current moment into a pre-trained defective product quantity prediction model to obtain a defective product quantity prediction result for the next cycle closest to the current moment; the pre-trained defective product prediction model is obtained by training an initial long short-term memory neural network with historical quality defect data of a variety of automobile parts; a secondary difference processing is performed on the defective product quantity prediction result to obtain a defective product change rate for the next cycle; when the defective product change rate for the next cycle exceeds a preset warning value, the warning detection result of the automobile part in the dimension of the predicted defective product change rate is determined to trigger a warning.

[0163] Furthermore, in one embodiment, the warning determination module 1501 is also used to determine the mean of the defective product growth rate and the variance of the defective product growth rate based on historical quality defect data; and to determine a preset warning value based on the time period length of a single cycle, the mean and variance of the defective product growth rate, and the risk value algorithm.

[0164] Furthermore, in one embodiment, the graded warning module 1503 is also used to obtain a graded quality warning result based on the comprehensive warning detection result and preset graded warning rules; generate a heat map corresponding to the graded quality warning result based on the graded quality warning result; and display the heat map.

[0165] Each module in the above-mentioned quality warning device for automobile parts can be implemented in whole or in part by software, hardware or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each module.

[0166] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Fig.16 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store historical defect quality data, early warning detection results under various early warning dimensions and other data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a quality early warning method for automotive parts is implemented.

[0167] Those skilled in the art will understand that Fig.16 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0168] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0169] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0170] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0171] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0172] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0173] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A quality early warning method for automobile parts, characterized in that: The method comprises: According to the historical quality defect data of the automobile parts, early warning detection is performed on the automobile parts under the preset early warning dimension to obtain an early warning detection result; the preset early warning dimension includes at least one of the number of defective products, the growth rate of defective products and the predicted change rate of defective products; the historical quality defect data includes the number of defective products of the automobile parts in multiple cycles; Performing fusion processing on the early warning detection results under each early warning dimension to obtain a comprehensive early warning detection result of the automobile part; According to the comprehensive early warning detection results, graded quality early warning is carried out for the automobile parts.

2. The method according to claim 1, characterized in that According to the historical quality defect data of the automobile parts, the automobile parts are subjected to early warning detection under the preset early warning dimension to obtain early warning detection results, including: Determining the cumulative number of defective products of the automotive parts based on the historical quality defect data; According to the cumulative number of defective products and the number of defective products in each cycle, the early warning detection result of the automobile parts in the dimension of the number of defective products is obtained.

3. The method according to claim 2, characterized in that The step of obtaining the early warning detection result of the automobile parts in the dimension of the number of defective products according to the cumulative number of defective products and the number of defective products in each cycle includes: According to the number of defective products in each of the cycles, determine the growth of defective products in the cycle closest to the current moment relative to the next closest cycle, and the change trend of the number of defective products in multiple cycles closest to the current moment; When the cumulative number of defective products is greater than the quantity threshold, the growth amount of defective products is greater than the growth amount threshold, or the change trend is continuous growth, it is determined that the early warning detection result of the automotive parts under the defective product quantity dimension is a triggering early warning; the growth amount threshold is determined based on the average defective product quantity in each cycle.

4. The method according to claim 1, characterized in that: According to the historical quality defect data of the automobile parts, the automobile parts are subjected to early warning detection under the preset early warning dimension to obtain early warning detection results, including: Determine the defective product growth rate for each of the cycles and the average defective product growth rate for each cycle based on the historical quality defect data; Determine, from each of the cycles, a target cycle in which the defective product growth rate is greater than the average defective product growth rate; When the number of the target cycles is greater than a preset target cycle number threshold, it is determined that the warning detection result of the automobile part in the defective product growth rate dimension is a triggering warning.

5. The method according to claim 1, characterized in that According to the historical quality defect data of the automobile parts, the automobile parts are subjected to early warning detection under the preset early warning dimension to obtain early warning detection results, including: Inputting the quality defect data of the cycle closest to the current moment into a pre-trained defective product quantity prediction model to obtain a defective product quantity prediction result for the next cycle of the cycle closest to the current moment; the pre-trained defective product prediction model is obtained by training an initial long short-term memory neural network with historical quality defect data of a plurality of the automobile parts; Performing secondary difference processing on the defective product quantity prediction result to obtain the defective product change rate of the next cycle; In the case that the defective product change rate in the next cycle exceeds a preset warning value, it is determined that the warning detection result of the automobile part under the dimension of the predicted defective product change rate is a triggering warning.

6. The method according to claim 5, characterized in that The method further comprises: Determine the mean of the defective product growth rate and the variance of the defective product growth rate based on the historical quality defect data; The preset warning value is determined according to the time period length of a single cycle, the mean and variance of the defective product growth rate, and the risk value algorithm.

7. The method according to claim 1, characterized in that According to the comprehensive early warning test results, a graded quality early warning is performed on each of the automobile parts, including: Obtaining a graded quality warning result according to the comprehensive warning detection result and the preset graded warning rules; According to the graded quality warning result, generating a heat map corresponding to the graded quality warning result; The heat map is displayed.

8. A quality warning device for automobile parts, characterized in that: The device comprises: an early warning determination module, configured to perform early warning detection on the automobile parts under preset early warning dimensions according to historical quality defect data of the automobile parts, and obtain early warning detection results; the preset early warning dimensions include at least one of the number of defective products, the growth rate of defective products, and the predicted change rate of defective products; the historical quality defect data include the number of defective products of the automobile parts in multiple cycles; A result fusion module is used to fuse the warning detection results under various warning dimensions to obtain a comprehensive warning detection result of the automobile parts; The graded warning module is used to provide graded quality warning for the automobile parts according to the comprehensive warning detection results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images