Air conditioner commodity inspection data analysis method and device

Abstract

The invention provides an air conditioner commodity inspection data analysis method and device. The method comprises the steps that commodity inspection data sets of an air conditioner under different working conditions in the commodity inspection process are acquired; cleaning the acquired commodity inspection data set, and removing missing values and abnormal values exceeding a preset threshold value; discretization processing is carried out on the cleaned data, format conversion is carried out on each item of data according to a discretization result, and a transaction item set is generated; mining a frequent item set in the transaction item set by adopting an association rule mining algorithm, performing defect association rule extraction on the frequent item set, and generating a high-order frequent item set association rule according to order progressive increase combination; and matching a production line abnormal combination through a high-order frequent item set association rule, and outputting an air conditioner commodity inspection quality analysis result. The problem that air conditioner product commodity inspection data defects caused by complex factor interleaving are difficult to effectively recognize due to the fact that the prior art depends on a single performance parameter threshold comparison mechanism is solved, and the relevance defect source of the production line technology, materials or assembly links can be located more accurately.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data analysis, in particular to an air conditioner commercial inspection data analysis method and device. BACKGROUND

[0002] In the air conditioner manufacturing industry, commercial inspection is the last comprehensive quality detection link before product delivery, and is also one of the most important links in the entire quality inspection process. Through running the air conditioner prototype under simulated actual operating conditions such as refrigeration and heating, the commercial inspection link collects multiple key performance parameters of the product under specific conditions, and compares them with the preset standards or specifications, which constitutes an important basis for determining whether the final performance quality of the product is qualified. The commercial inspection production line that runs for a long time continuously records production data and accumulates massive commercial inspection process data sets related to various models of products.

[0003] At present, the defect judgment and analysis of air conditioner product commercial inspection data mainly relies on the threshold comparison mechanism of a single performance parameter. However, practice shows that most defects of air conditioner products are usually associated with specific process flow, material source and other multiple factors, and there is strong coupling between air conditioner product operating parameters. Abnormal phenomena often show coordinated changes in multiple parameters. This judgment method that only focuses on a single parameter not only cannot effectively identify systematic problems caused by complex factors, but also cannot quickly trace the root cause of the problem.

[0004] Therefore, there is an urgent need for an air conditioner commercial inspection data analysis method that can handle massive commercial inspection data to achieve systematic quality positioning and provide quantifiable, data-driven optimization directions for quality control and efficiency improvement of air conditioner commercial inspection. SUMMARY

[0005] In view of the deficiencies in the prior art, the present application provides an air conditioner commercial inspection data analysis method and device, which aims to solve the problem that the prior art relies on a single performance parameter threshold comparison mechanism and cannot effectively identify air conditioner product commercial inspection data defect problems caused by complex factors. At the same time, for massive commercial inspection data, an optimization analysis scheme for systematic quality positioning is provided to improve the efficiency and accuracy of defect judgment.

[0006] According to an embodiment of the present application, an air conditioner commercial inspection data analysis method comprises the following steps:

[0007] Obtain commercial inspection data sets of the air conditioner under different working conditions in the commercial inspection process;

[0008] Clean the obtained commercial inspection data sets, remove missing values and abnormal values exceeding the preset threshold;

[0009] Discretize the cleaned data, and convert the format of each data according to the discretization result to generate a transaction item set;

[0010] screening out a target transaction item set based on a preset item set filtering condition, and iteratively pruning and screening the target transaction item set through an association rule mining algorithm, to generate high-order frequent item sets and association rules layer by layer, until a higher-order frequent item set cannot be generated, and the iteration is terminated and the final frequent item set association rule is obtained;

[0011] The frequent item set association rule is matched with an abnormal combination of a production line, and an air conditioner inspection quality analysis result is output.

[0012] Further, the inspection data set of the air conditioner under different working conditions in the inspection process includes: under the refrigeration working condition or the heating working condition, the operating parameters of the air conditioner system and the operating characteristic parameters corresponding to the core components of the air conditioner; wherein the operating parameters of the air conditioner system include high-pressure side pressure Phigh, low-pressure side pressure Plow, discharge temperature Tdis, phase voltage V, and phase current I.

[0013] Further, the cleaned data is discretized: the cleaned data is discretized according to the obtained data type, and at least one of the following discretization methods is selected:

[0014] The discrete type data is encoded through a Boolean variable mapping;

[0015] and / or the continuous type data is discretized according to a preset working condition threshold value;

[0016] and / or the time series type data is discretized according to a set time period.

[0017] Further, the transaction item set includes a data item combination composed of air conditioner operating parameters, core components and associated defect types.

[0018] Further, the target transaction item set is screened out based on the preset item set filtering condition, including:

[0019] According to the batch product experience value and the current inspection environment temperature and humidity, the minimum support threshold and the minimum confidence threshold are set;

[0020] The support degree of each data in the transaction item set in the transaction item set is calculated;

[0021] Based on the support degree, the confidence degree of each data in the transaction item set in the transaction item set is calculated;

[0022] Based on the calculated support degree and confidence degree, the lift degree of each data in the transaction item set in the transaction item set is calculated;

[0023] The data in the transaction item set with a lift degree greater than 1 and a support degree greater than the minimum support threshold is screened out, and a frequent item set is obtained.

[0024] Further, the target transaction item set is iteratively pruned and screened layer by layer by the association rule mining algorithm to generate high-order frequent item set association rules, comprising the following steps:

[0025] Step S1, mining frequent item sets in the target transaction item set by the association rule mining algorithm;

[0026] Step S2, generating high-order frequent item sets by incrementally combining the frequent item sets layer by layer through iterative pruning and screening;

[0027] Step S3, generating all possible association rules for all frequent item sets and calculating the support of each data;

[0028] Step S4, when no new high-order frequent item set meeting the preset threshold condition can be generated, terminating the iteration to obtain the final frequent item set association rule.

[0029] Further, the step S2 generates high-order frequent item sets by incrementally combining the frequent item sets layer by layer through iterative pruning and screening as follows: taking the frequent first item set as the starting point, generating the second item set, the third item set, and so on until the Kth item set, and when a new order frequent item set is generated, screening out data with support less than the minimum support threshold.

[0030] Further, in step S4, the preset condition is that the support of each data is greater than the minimum support threshold, or the support of each data is greater than the minimum support threshold and the confidence is greater than the minimum confidence threshold.

[0031] Further, the core components of the air conditioner include a compressor, an evaporator, a condenser, a four-way valve, an expansion valve, a fan, a power module, and a driving module.

[0032] On the other hand, according to an embodiment of the present application, an air conditioner commercial inspection data analysis device is also provided, comprising:

[0033] An acquisition module is configured to acquire a commercial inspection data set of an air conditioner under different working conditions in a commercial inspection process;

[0034] A cleaning module is configured to clean the acquired commercial inspection data set to remove missing values and abnormal values exceeding a preset threshold;

[0035] A discretization module is configured to perform discretization processing on the cleaned data, and perform format conversion on each item of data according to the discretization result to generate a transaction item set;

[0036] The mining module is used for screening a target transaction item set based on a preset item set filtering condition, and iteratively pruning and screening the target transaction item set through an association rule mining algorithm, and generating high-order frequent item sets and association rules thereof layer by layer until a higher-order frequent item set cannot be generated, and then terminating iteration and obtaining final frequent item set association rules.

[0037] The output module is used for matching a production line abnormal combination through the frequent item set association rules, and outputting an air conditioner commercial inspection quality analysis result.

[0038] Compared with the prior art, the present application has the following beneficial effects:

[0039] The present application unifies multi-source heterogeneous data (such as pressure, temperature, voltage and current and core component parameters) into transaction item sets through discretization processing, combines with an association rule algorithm to mine the implicit combination relationship between parameters, and generates high-order rules according to the order increasing combination, effectively identifies multi-component coupling faults and systematic defects, greatly reduces the complex abnormal omission rate, and improves the accuracy of the commercial inspection result; through the frequent item set association rule matching production line abnormal combination, the commercial inspection data is converted into a process improvement basis that can be landed, such as directly positioning to the assembly link or component supplier problem, not only significantly shortens the quality analysis cycle, avoids the blindness and inefficiency of manual investigation, but also can accurately locate the systematic and related defects of the production line process, materials or assembly link.

[0040] The present application also adopts a dynamic threshold setting mechanism, and based on batch experience values and environmental temperature and humidity, the support degree and confidence degree can be dynamically adjusted, combined with an iterative pruning strategy, the calculation complexity is effectively reduced, and the analysis efficiency is improved. Finally, the originally isolated and complex commercial inspection operation data is converted into high-order association rules with a clear quality orientation, greatly mines and improves the potential value of the commercial inspection data, and provides strong data-driven support for the quality control and process optimization of the air conditioner product. BRIEF DESCRIPTION OF DRAWINGS

[0041] The drawings incorporated into the specification and forming a part thereof, illustrate embodiments consistent with the present application and, together with the specification, serve to explain the principles of the application.

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the embodiments or related art description will be briefly introduced. Obviously, for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0043] Figure 1 A flowchart of an air conditioner commercial inspection data analysis method according to an embodiment of the present application;

[0044] Figure 2 A schematic diagram of the air conditioner commercial inspection data association rule mining process provided by the embodiment of the present application is shown. DETAILED DESCRIPTION

[0045] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described below in connection with the drawings of the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application.

[0046] In the analysis of commercial inspection data of air conditioner products, current defect determination generally relies on the threshold comparison of a single parameter, and less consideration is given to the combined association of multiple parameters. However, parameter abnormalities often occur concurrently, and there is an inherent correlation between parameters, which is usually caused by systematic defects triggered by specific process flows or material sources. Therefore, based on the above background, by deeply mining the implicit association between commercial inspection test data and comprehensively utilizing multi-source data, the rapid positioning of product systematic defects is realized, thereby providing data-driven decision support for quality control and efficiency improvement. For this purpose, the present application focuses on the causal relationship between high-frequency parameter combinations and defects, and proposes an air conditioner commercial inspection data analysis method suitable for various types of air conditioners.

[0047] Please refer to Figure 1 On the basis of the above technical background, based on the problems mentioned in the background art, the embodiment provides an air conditioner commercial inspection data analysis method, comprising the following steps:

[0048] Obtaining commercial inspection data sets of air conditioners under different working conditions in the commercial inspection process;

[0049] Cleaning the obtained commercial inspection data sets, removing missing values and abnormal values exceeding the preset threshold;

[0050] Discretizing the cleaned data, and converting the format of each data according to the discretization result to generate a transaction item set;

[0051] Filtering the frequent item set in the transaction item set based on the preset filtering condition, extracting defect association rules from the frequent item set by using an association rule mining algorithm, and generating high-order frequent item set association rules in ascending order;

[0052] Matching the high-order frequent item set association rules with abnormal combinations on the production line, and outputting the air conditioner commercial inspection quality analysis result.

[0053] The embodiment obtains a raw inspection data set generated after a batch of air conditioners in a preset different test condition completes inspection in the inspection process, the data set contains the detection results of each air conditioner on each performance index, then the obtained raw inspection data set is preprocessed, and data records containing missing values of key performance indexes (i.e., invalid inspection records) are identified and removed. Meanwhile, for each performance index, according to a preset threshold range, abnormal value data points significantly exceeding the range are detected and removed, wherein the specific threshold range is determined based on product standards and historical experience, so as to remove missing items and dirty data items that do not obviously conform to common sense or physical rules. This process can effectively eliminate the interference of noise and invalid information.

[0054] The index data after cleaning is discretized, and the discretization interval or category division standard is predefined based on engineering experience and quality control targets; all performance index states of each air conditioner (corresponding to a record) after discretization are combined and converted according to the "attribute state" format (for example, "low refrigerating capacity" and "high exhaust temperature Tdis"). The inspection result of each air conditioner is converted into a "transaction" composed of multiple "items". The transaction set of all air conditioners constitutes a "transaction item set" for association rule mining. Discretization converts index data into a symbolic form suitable for association rule mining, and builds a standard analysis input structure through format conversion.

[0055] The association rule mining algorithm (such as the Apriori algorithm) is applied to calculate the support of all possible item sets in the transaction item set, and the frequent item sets meeting the preset filtering conditions are screened out. These frequent item sets represent performance index state combinations that often appear simultaneously in the inspection data, which are potential quality problem patterns, then the association rules of the frequent item sets are extracted, and the association rules represent the association between performance defects and other index states, i.e., "defect association rules". The association rule mining algorithm generates high-order frequent item sets by incremental combination, and also extracts association rules, and the mining process can discover complex defect combination patterns involving multiple performance index states.

[0056] For the generated high-order frequent item set association rules, specific subdivision analysis is performed on the abnormal combination item sets pointed by the rules according to production batches and supplier batches. The analysis result can go beyond simple pass / fail judgment, reveal deep and multi-factor related causes of quality problems, and provide direct data-driven support for accurately positioning production bottlenecks, optimizing process parameters, improving component quality, or adjusting inspection strategies. We will describe the method of the embodiment in detail as follows.

[0057] In the specific implementation process, taking the inspection data set analysis of a certain time period of a certain type of household R32 refrigerant variable frequency 2P hanging machine product as an example, the method is described in detail:

[0058] A plurality of different types of sensors are deployed inside the inspection room to collect sensor data of the air conditioner product under refrigeration or heating conditions during the inspection process. The real-time data is transmitted to the host computer system through the adapter board at a transmission frequency of 1 time per second to obtain an inspection data set D within a certain period of time. The inspection data set D includes characteristic parameters such as high-pressure side pressure Phigh, low-pressure side pressure Plow, discharge temperature Tdis, phase voltage V, phase current I, and the like of the air conditioning system under different working conditions, and also includes operating characteristic parameters of core components such as the air conditioner compressor, evaporator, condenser, four-way valve, expansion valve, fan, power module and drive module. The obtained inspection data set D is preprocessed to eliminate missing items and invalid data items that do not conform to common sense or physical rules in the data set.

[0059] The cleaned data is processed according to the data type using at least one of the following discrete methods: performing Boolean variable mapping on categorical data, binning and discretizing continuous data according to a pre-set working condition threshold, and segmenting and discretizing time series data according to a set time period.

[0060] Example 1: For categorical data, it can be directly mapped to a Boolean type

[0061] Fault: {Normal, Defect = Expansion valve blockage};

[0062] Barcode writing: {True, Force};

[0063] Inspection result: {Pass, NG}; ……

[0064] Example 2: For continuous data, binning value setting is performed according to the specific performance process parameters of the product, and the high-pressure side pressure Phigh, low-pressure side pressure Plow, and discharge temperature Tdis are classified and processed according to the following threshold standards:

[0065] Table 1: Binning and discretization based on product performance process

[0066]

[0067] Please refer to Figure 2 , Figure 2 The air conditioner inspection data association rule mining process diagram provided for this embodiment.

[0068] According to the pre-set discretization rule, format conversion is performed on all data items in the inspection data set D to generate a standardized discrete item set. Conversion examples are as follows:

[0069] Trans_Dx = {‘Phigh normal’, ‘Plow normal’, ‘Tdis normal’, …, ‘Pass’};

[0070] Trans_Dy = { 'Phigh high', 'Plow low', 'Tdis high', 'Defect = poor condenser heat dissipation', 'NG'} ;

[0071] Trans_Dz = { 'Phigh high', 'Plow low', 'Tdis high', 'Defect = expansion valve blockage', 'NG'} ;...

[0072] After the format conversion, the association rule mining algorithm (the Apriori algorithm is preferred in this embodiment) is used to mine and extract defect association rules, and the purpose is to find the potential and frequent association patterns between different parameter abnormal combinations and specific defects (such as expansion valve blockage). The specific steps are as follows:

[0073] In combination with the specific model characteristics of the target product, the process experience value (for example, the commercial inspection defect rate of this model product is usually lower than 5%), and the influence of typical environmental temperature and humidity parameters, the minimum support threshold Sup_min = 1% and the minimum confidence threshold Con_min = 70% of the algorithm are set in advance. These thresholds are the key standards for screening association rules that have statistical significance and are potentially useful.

[0074] Taking the matter A→B as an example, where A is an NG item set containing 'Phigh high', denoted as A = {Phigh high}; B is an NG item set containing 'expansion valve blockage', denoted as B = {expansion valve blockage};

[0075] Calculate the support: Where Sup(A→B) represents the frequency of the occurrence of matter A and matter B at the same time in the entire data set.

[0076] Calculate the confidence: Where Con(A→B) represents the probability of the occurrence of matter B when matter A occurs; Sup(A) represents the support of matter A in the entire data set.

[0077] Based on the support and the confidence, the lift is calculated:

[0078] If Lif(A→B) > 1, it indicates that matter A and matter B are positively correlated;

[0079] If Lif(A→B) = 1, it indicates that A and B are not associated;

[0080] If Lif(A→B) < 1, it indicates that A and B are negatively correlated;

[0081] In this embodiment, Lif > 1, which proves that Phigh high and expansion valve blockage failure have significant positive correlation.

[0082] The NG item set is scanned for the first time by using the set Sup_min and Con_min thresholds, and single defect or low correlation item sets with support less than the minimum threshold of 1% and confidence less than the minimum threshold of 70% are screened out, to obtain a candidate item set C1.

[0083] For the candidate item set C1, the first-order item set L1 satisfying the threshold is obtained through the connection and pruning steps of the Apriori algorithm, and the candidate item set C2 is generated through the connection operation starting from the first-order item set L1, the second-order item set L2 satisfying the threshold is obtained by scanning the candidate item set C2 twice and performing pruning screening, and higher-order item sets are generated through layer-by-layer combination according to the increasing order:

[0084] L1: {Phigh high bias → expansion valve blockage}

[0085] L2: {Phigh high bias, Plow low bias → expansion valve blockage}

[0086] L3: {Phigh high bias, Plow low bias, Tdis high bias → expansion valve blockage}

[0087] When a high-order rule is identified, for example, the parameter combination “Phigh too high” and “Plow too low” accompanied by “Tdis high bias” appear at the same time, it may indicate that there is a “expansion valve blockage” fault, the support of the rule reaches 12%, and the confidence is as high as 85%. According to the above Apriori algorithm process, the mining steps are repeatedly performed for all main defect types (such as “compressor seizure”, “refrigerant leakage” and the like) in the data set, and when no new high-order frequent item set satisfying the preset threshold condition can be generated, the iteration is terminated and the complete air conditioner quality defect association rule set is finally constructed. The constructed association rule set is integrated into the manufacturing execution system (MES), and the system can monitor the parameter combination of each detection station on the production line in real time. Once a defect item set (such as the “expansion valve blockage” combination pointed to by the L3 rule described above) defined in the rule library with high support (> Sup_min), high confidence (> Con_min) and significant lift greater than 1 (indicating positive correlation) is triggered, the MES system immediately issues an alarm.

[0088] Through the air conditioner commercial inspection data analysis method proposed in the present application, especially the cooperative application based on the discretization data conversion and the Apriori association rule mining, the inherent association rule of complex multi-dimensional detection data is deeply mined. The method not only can construct a high-precision defect prediction rule library for MES real-time monitoring and early warning, significantly improve the defect detection rate and abnormal response speed, but also can support in-depth tracing of high-frequency faults (by batch / supplier), and provide strong data support for continuous improvement of process and supply chain quality management.

[0089] More importantly, the scheme has the dynamic characteristics of intelligent adaptation to environmental changes. For example, during the operation in the seasonal hot and humid environment (ambient temperature > 35℃, humidity > 80%), the system monitors that the support of the original "insulation failure" defect rule has significantly climbed from the benchmark of 0.7% to 2.1%. At this time, the system can dynamically adjust the minimum support threshold Sup_min of this specific rule according to the preset strategy (such as from 1% to 1.5%), and re-verify and screen the rules based on the new threshold. This dynamic threshold updating mechanism effectively overcomes the lack of adaptability of the static threshold in extreme environments, ensures that the fault detection rate of the system can still be stably maintained above 90% even in harsh conditions, and guarantees the reliability and robustness of the commercial inspection analysis.

[0090] In summary, the present application realizes the precise diagnosis and prediction of air conditioner defects through discretization conversion and association rule mining. The constructed frequent item set association rules can identify multi-parameter combination fault modes in real time, significantly improving the quality inspection efficiency; combined with the dynamic threshold adjustment mechanism, it can adapt to seasonal and environmental changes, guarantee the stability of the fault detection rate, and provide data support for process optimization on the production line.

[0091] Finally, it should be pointed out that the above examples are only used to illustrate the technical solutions of the present application and not to limit it. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the purpose and scope of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. An air conditioner inspection data analysis method, characterized by, The method comprises the following steps: obtaining a commercial inspection data set of the air conditioner under different working conditions in a commercial inspection process; cleaning the obtained commercial inspection data set, removing missing values and abnormal values exceeding a preset threshold; discretizing the cleaned data, and converting the format of each data according to the discretization result to generate a transaction item set; mining a frequent item set in the transaction item set by using an association rule mining algorithm, extracting defect association rules from the frequent item set, and generating high-order frequent item set association rules in an order-increasing combination; matching the high-order frequent item set association rules to an abnormal combination of a production line, and outputting an air conditioner commercial inspection quality analysis result.

2. The air conditioner inspection data analysis method of claim 1, wherein, The commercial inspection data set of the air conditioner under different working conditions in the commercial inspection process comprises: under a refrigeration working condition or a heating working condition, operating parameters of an air conditioner system and corresponding operating characteristic parameters of core components of the air conditioner; wherein the operating parameters of the air conditioner system comprise a high-pressure side pressure Phigh, a low-pressure side pressure Plow, a discharge temperature Tdis, a phase voltage V, and a phase current I.

3. The air conditioner inspection data analysis method of claim 1, wherein, The discretization processing of the cleaned data comprises: selecting at least one of the following discretization methods according to the type of the obtained data for processing; performing Boolean variable mapping on discrete type data; performing binning discretization on continuous type data according to a preset working condition threshold; segmenting and discretizing time series type data according to a set time period.

4. The air conditioner inspection data analysis method of claim 1, wherein, The transaction item set comprises a data item combination of the operating parameters of the air conditioner system, the corresponding operating characteristic parameters of the core components, and a state type.

5. The air conditioner inspection data analysis method of claim 4, wherein, The mining of the frequent item set in the transaction item set by using the association rule mining algorithm comprises: setting a minimum support threshold and a minimum confidence threshold according to batch product experience values and current commercial inspection environment temperature and humidity; calculating the support of each data in the transaction item set in the transaction item set; based on the support, calculating the confidence of each data in the transaction item set in the transaction item set; based on the calculated support and confidence, calculating the lift of each data in the transaction item set in the transaction item set; screening data with a lift greater than 1 and a support greater than the minimum support threshold in the transaction item set to obtain a frequent item set.

6. The air conditioner inspection data analysis method of claim 5, wherein, The defect association rule extraction from the frequent item set and the generation of high-order frequent item set association rules in an order-increasing combination comprise the following steps: Step S1: iteratively pruning and screening the frequent item set, and generating a higher-order frequent item set in an order-increasing combination layer by layer; Step S2: generating possible defect association rules for all frequent item sets, and calculating the support of each data; Step S3: when a new high-order frequent item set that meets a preset threshold condition cannot be generated, terminating the iteration to obtain a final high-order frequent item set association rule.

7. The air conditioner inspection data analysis method of claim 6, wherein, In step S1, the iteratively pruning and screening of the frequent item set and the generation of a higher-order frequent item set in an order-increasing combination layer by layer comprise: starting from a frequent first item set, generating a frequent second item set, a frequent third item set, and so on until a frequent Kth item set, and removing data with a support less than the minimum support threshold when a new-order frequent item set is generated.

8. The air conditioner inspection data analysis method of claim 6, wherein, In step S3, the preset threshold condition is that the support degree of each data is greater than a minimum support threshold.

9. The air conditioner inspection data analysis method of claim 2, wherein, The air conditioner core components include a compressor, an evaporator, a condenser, a four-way valve, an expansion valve, a fan, a power module, and a driving module.

10. An air conditioner inspection data analysis device characterized by comprising: Comprise: An acquisition module is configured to acquire inspection data sets of an air conditioner under different working conditions in a commercial inspection process; A cleaning module is configured to clean the acquired inspection data sets, remove missing values, and remove abnormal values exceeding a preset threshold; A discrete module is configured to discretize the cleaned data, convert the format of each data item according to the discretization result, and generate a transaction item set; A mining module is configured to filter a frequent item set in the transaction item set based on a preset filtering condition, extract a defect association rule from the frequent item set using an association rule mining algorithm, and generate a high-order frequent item set association rule by increasing the order combination; An output module is configured to match a production line abnormal combination through the high-order frequent item set association rule and output an air conditioner inspection quality analysis result.

Images