CN102176239A - Kitchenware fault diagnosing method based on case-based reasoning - Google Patents

Abstract

The invention discloses a kitchenware fault diagnosing method based on case-based reasoning, which comprises the following steps: a kitchenware fault problem is submitted to obtain fault symptom information, the fault symptom information is transmitted to a fault diagnosing server of an enterprise for retrieving and screening of kitchenware fault cases so as to obtain a target fault case, and the target fault case is transmitted back to a client, and a maintenance man or a user maintains the kitchenware according to a maintenance scheme in the target fault cases; if the kitchenware fault can be solved, the fault case is studied; otherwise, the step (5) is carried out to revise the fault case to obtain a new kitchenware fault case for maintenance till the kitchenware fault is solved. The kitchenware fault diagnosing method disclosed by the invention utilizes CBR (Case-Base Reasoning) to diagnose the kitchenware faults, and adopts the strategic concept of classifying the fault cases for retrieval, therefore, the method is beneficial to not only reduction of the difficulty in representing and organizing cases and creating a case base, but also great improvement of the case retrieving efficiency, and has higher adaptability compared with the traditional RBR (Rule-Based Reasoning) or MBR (Model-Based Reasoning) expert systems.

Description

A Case-Based Reasoning Method for Fault Diagnosis of Kitchenware

technical field

The invention belongs to the technical field of artificial intelligence fault diagnosis, and more specifically, relates to a method for fault diagnosis of kitchen utensils based on case reasoning.

Background technique

Kitchen utensils, as a necessary daily necessities in the family, with the improvement of people's living standards, the requirements for kitchen utensils are also increasing. However, in today's highly homogenized products, many products are placed together in the same way, and even high-quality products cannot form an advantage. With the continuous improvement of production quality and technical service requirements, people's selection requirements for household kitchen utensils are also rising. The homogenization of kitchenware product technology and product image makes it impossible for enterprises to obtain differentiated competitive advantages in products. If kitchenware companies want to gain survival advantages in the fierce market competition, they must establish their own unique core competitiveness.

To establish one's own unique core competitiveness, one has to constantly look for new places to make one's own products. The innovation of fault diagnosis technology is one of the key points of new products. First of all, timely fault diagnosis can increase the service life of kitchen utensils; secondly, kitchen utensil accidents may cause unnecessary losses, and in serious cases endanger the lives and properties of users, so fault diagnosis is extremely important.

The fault diagnosis system using artificial intelligence technology has become the development trend of kitchenware fault diagnosis. At present, in the field of kitchenware maintenance, most of the traditional fault diagnosis adopts RBR (rule based reasoning, rule-based reasoning), MBR (model based reasoning, model reasoning) ) expert system technology. Since these traditional expert systems are model-driven (model-based diagnostic methods use deep knowledge such as the structure, behavior, and function models of the diagnostic object for diagnostic reasoning), there are problems in model construction, information acquisition, and information processing. Serious deficiencies, there are some insurmountable shortcomings, such as the difficulty of extracting the rules of system domain knowledge; the creation and management of rule bases and pattern bases is complex and arduous; the rules and patterns in the reasoning process are difficult to accurately select, etc.

Contents of the invention

The purpose of the present invention is to overcome the deficiencies of the prior art and provide a case-based reasoning-based fault diagnosis method for kitchen utensils with strong adaptability and high efficiency.

In order to achieve the above object, the present invention is based on case reasoning kitchen utensils fault diagnosis method, it is characterized in that comprising the following steps:

(1) Submit the problem of kitchen utensil failure

The maintenance personnel or users input the fault symptom information on the client computer according to the symptoms displayed by the current faulty kitchenware, as the input of the faulty problem of the kitchenware; the fault symptom information is transmitted to the fault diagnosis server of the enterprise through the Internet;

(2) Retrieval of failure cases of kitchen utensils

The fault diagnosis server retrieves a set of candidate cases similar to the fault symptom information by using the nearest neighbor method of the gray relational degree to calculate the similarity from its kitchen fault case database:

(3) Screening of kitchen utensil failure cases

If the similarity of a certain case in the candidate case set obtained in step (2) is high enough, and the similarity of other cases is low, it is not necessary to perform case screening, and it is directly used as the target failure case, and enters step (4) ;Otherwise call the case screening module, carry out case screening through the case screening module, obtain the target failure case, and then enter step (4);

(4), maintenance plan reuse

The fault diagnosis server sends the target fault case back to the client computer through the Internet, and the maintenance personnel or users repair the kitchen utensils according to the maintenance plan in the target fault case. If the fault can be solved, go to step (6) for fault case study; Go to step (5) to correct the failure case;

(5), fault case correction

The fault diagnosis server invokes the case correction auxiliary system. Through human-computer interaction, the technicians of the enterprise correct the fault cases to obtain new kitchen utensil fault cases, and then send them back to the client through the Internet. Repair the kitchen utensils in the maintenance plan until the kitchen utensils failure is solved, then transfer to step (6) for failure case study;

(6) Fault case study

The fault diagnosis server adds new kitchen utensil failure cases to the kitchen utensil failure case database; in the case study auxiliary system, the ratio of the kitchen utensil failure cases to solve the actual failure is judged, and if it is less than a certain value, it is deleted.

The purpose of the present invention is achieved like this.

Due to the large number of kitchen utensils, the complex relationship between the fault phenomenon and the fault cause, and the many types of faults, the case-based method is more convenient than the traditional The expert systems of RBR and MBR have stronger adaptability. Using CBR (Case-Base Reasoning, case-based reasoning) for kitchenware fault diagnosis, through the semi-structured description of kitchenware faults, the incomplete and inconsistent structural description of kitchenware faults is avoided. On the basis of analyzing and researching the characteristics of knowledge in the field of kitchenware fault diagnosis and a large number of fault maintenance logs, the key technology of the case-based reasoning method is discussed. It not only reduces the difficulty of case representation, organization and construction of case database, but also greatly improves the efficiency of case retrieval. Simultaneously, the present invention also has the following characteristics based on the case-based reasoning kitchen utensil diagnosis method:

(1) The diagnosis of kitchen utensils becomes easier, and fault diagnosis can still be carried out when there is no kitchen utensil model; because the kitchen utensil failure case library is constantly growing, even a case reasoning system with only a small number of cases at the beginning can also run; based on The case-based reasoning kitchenware diagnosis method can quickly provide kitchenware failure solutions without having to reason from scratch every time; case-based reasoning provides users with specific kitchenware failure cases, which are easy to understand; by acquiring new cases, the failure case library can be obtained from different Learn new knowledge in the field, and easy to maintain.

(2), the present invention proposes that in the case-based reasoning fault diagnosis method, the case secondary retrieval adopts the gray relational degree to measure the similarity of the cases, applies the gray dominance analysis to obtain the candidate case set, and the obtained candidate cases are more accurate.

Description of drawings

Fig. 1 is the flow chart of the kitchen utensils fault diagnosis method based on case reasoning of the present invention;

Fig. 2 is a classification structure diagram of a gas stove failure case library;

Fig. 3 is a secondary index structure diagram of the gas stove failure case library.

Detailed ways

Specific embodiments of the present invention will be described below in conjunction with the accompanying drawings, so that those skilled in the art can better understand the present invention. It should be noted that in the following description, when detailed descriptions of known functions and designs may dilute the main content of the present invention, these descriptions will be omitted here.

Example

Fig. 1 is a flow chart of the kitchenware fault diagnosis method based on case reasoning in the present invention.

1. Submit the problem of kitchen utensil failure

As shown in Fig. 1, in the present embodiment, in the kitchenware fault diagnosis method based on case reasoning, firstly, the kitchenware fault problem is submitted, that is, step (1): maintenance personnel or users show symptoms according to the current faulty kitchenware, in Input the failure symptom information on the client computer as the input of the kitchen utensil failure problem. The fault symptom information is selected and obtained according to the fault symptom options provided by the fault diagnosis server, so that a standard input fault symptom information can be obtained, which is convenient for retrieval of kitchen utensil fault cases. Of course, it can also be generated by inputting fault phenomena by maintenance personnel or users, and automatically identifying them by the fault diagnosis server.

The fault symptom information input by maintenance personnel or users is transmitted to the fault diagnosis server of the enterprise through the Internet.

2. Retrieval of failure cases of kitchen utensils

Then, search for kitchenware failure cases, that is, step (2): the fault diagnosis server uses the nearest neighbor method of gray correlation degree to calculate the similarity level from its kitchenware failure case database, and retrieves information related to failure symptoms. Similar set of candidate cases.

In this embodiment, the kitchen utensil failure case is represented by a quaternion, that is, the kitchen utensil failure case C=(E, S, A, P).

Among them: the tuple E={e ₁ , e ₂ ,..., e _r } of kitchenware failure description is a finite non-empty set, and e _j (j=1, 2,..., r) represents a piece of description information, for example: case number , Kitchenware model, fault location, maintenance personnel, maintenance date and effect evaluation.

The symptom tuple S={s ₁ , s ₂ ,...,s _m } of the kitchenware failure case is a finite non-empty set, and the failure symptoms are divided into qualitative failure symptoms and quantitative failure symptoms, failure symptoms s _j = {f _j , d _j }(j=1, 2,..., m), f _j is a qualitative fault symptom, such as: when the gas stove is difficult to ignite, gas leaks or has a bad smell, d _j represents the fault symptom The confidence degree of the symptom is used to illustrate the severity of the fact of the fault symptom; d _j is the quantitative fault symptom, such as gas, temperature, etc., d _j represents the converted value of the actual measured value of these parameters, and d _j is in the interval [ 0, 1]; in the field of kitchenware fault diagnosis, some qualitative fault symptoms are definite quantities, that is, only manifested as "yes" or "no", such as circuit faults, which can be handled by binary logic, ""Yes" means "1", and no means "0". However, more qualitative failure symptoms are often some vague quantities, such as "serious" and "very serious" ignition difficulties, "a little" and "serious" instability of air supply pipeline leakage, etc. There are certain deficiencies in using the binary logic method to deal with it. For this reason, the method of fault severity can be used to describe it, and a value can be assigned to indicate its severity, as shown in Table 1:

Fault severity Malfunction symptom value Very serious 0.9 serious 0.7

Medium 0.5 generally 0.3 slight 0.1 normal 0

Table 1

The cause tuple A={a ₁ , a ₂ ..., a _l } of kitchenware failure is a finite non-empty set, a ₁ , a ₂ , ..., a _r is one of the causes, and a certain failure can be caused by l failure causes;

The maintenance plan tuple P={p ₁ , p ₂ ,...,p _o } of the kitchenware failure case is a finite non-empty set; p ₁ , p ₂ ,..., p _r is one of the solutions.

In this embodiment, the retrieval of kitchen utensil failure cases includes two sub-processes:

a. First-level retrieval: determine similar case sets. Firstly, according to the fault symptom information of the kitchen utensil fault target case, the abstract case to which it belongs is determined. By comparing the fault symptom information of the kitchen utensil fault target case with the fault symptom information in each abstract case, according to the size of the similarity, we get The specific case represented by the abstract case with the largest similarity with the fault symptom information of the target case of kitchen utensils, completes the first-level index, and if there is no similar abstract case, then directly retrieves the 0th class abstract case (the 0th class case is Refers to the outlier cases and other cases that have no similarities, and become a case independently; secondly, through the key words such as kitchenware failure system and key failure characteristics input by the user, from the results of the first-level index, the key words matching the key are retrieved. Cases of word conditions, get similar case sets, and complete the second-level index.

b. Second level retrieval: determine candidate case sets. In the similar case set, the nearest neighbor method based on the gray relational degree is used to find the candidate case most similar to the target case of kitchen utensil failure by calculating the similarity between the fault symptom information of the target case and the fault symptom information of the cases in the similar case set set.

In the case-based reasoning kitchenware fault diagnosis method proposed by the present invention, the case secondary retrieval is to calculate the similarity of kitchenware fault cases by adopting the nearest neighbor method based on the gray relational degree. The similarity calculation method of the gray correlation degree calculates the gray correlation coefficient of the failure characteristic information of the target case of the kitchen utensils failure and the corresponding symptom index of the candidate case set, that is, the local similarity, and then calculates the overall similarity by taking the average value of the local similarity.

Figure 2 is a classification structure diagram of the gas stove fault case library.

In this embodiment, when organizing the gas stove fault case library, the present invention adopts a hierarchical structure, such as dividing the gas stove fault case library into an ignition system fault case library, an air leakage fault case library, and an abnormal flame case library, each part It can be further divided into several sub-parts. For example, the ignition system failure case library can also be divided into gas supply pipeline case library, igniter case library, and flame hole failure case library, as shown in Figure 4. Among them, the lowest case library (such as the gas supply pipeline case library) stores specific cases.

Fig. 3 is a secondary index structure diagram of the gas stove failure case library.

In this embodiment, the cases in each case base are grouped into categories according to the degree of similarity, and the cases in each category have a high degree of similarity, while the similarities between cases in different categories are relatively small , this partitioning method keeps the number of cases in each category relatively small. Construct an abstract case for each category that can represent all the cases of this category, and build an index for the cases in each category to identify which abstract case these specific cases belong to.

Assuming that there are M cases in the gas stove failure case library, through cluster analysis, they can be divided into M clusters, that is, M abstract cases. Each cluster contains several specific gas stove cases, and abstract cases are used to represent the distribution characteristics of case features in each cluster, and these M abstract cases are used as the first-level index. For each specific case, an index is established according to one or more characteristic attributes of the gas stove failure case (gas stove failure location, key failure characteristics), thereby forming a second-level index, as shown in Figure 5.

The abstract case is just a case that represents the characteristic situation of the cases in this category. It does not exist in the gas stove failure case library, nor is it a complete case. There is only a gas stove failure symptom set, but no solution set.

In this embodiment, the hierarchical retrieval is

(1) Determine the abstract case set that is similar to the fault symptom information of the gas stove fault target case: according to the user's input of the fault symptom information of the target case, determine the abstract case that is most similar to the gas stove fault target case fault symptom information set.

(2) Determine the specific case base represented by the best abstract case: perform similarity matching calculation on the fault symptom information of the gas stove fault target case and the obtained symptom information in the abstract case set, according to a certain minimum similarity threshold , find an abstract case that is most similar to the gas stove failure target case, and retrieve the concrete case library represented by this abstract case; if the similarity of all abstract cases is lower than this threshold, then retrieve the 0th class abstraction The specific case that the case represents.

(3) Determine the similar case set: according to the key feature attribute information input by the user, retrieve the similar case set that meets the requirements from the specific case database obtained above.

(4) Determine the candidate case set, calculate the similarity between the fault symptom information of the gas stove fault target case and all the case symptom information in the similar case set by the nearest neighbor method based on the gray correlation degree, and determine the most similar to the gas stove fault target case set of candidate cases.

Explanation: Type 0 cases here refer to outlier cases, and outlier cases are often core cases. They cannot be discarded and should be regarded as a separate category (category 0). Outlier cases are not similar to other cases.

3. Screening of kitchen failure cases

Then carry out the case screening of kitchen utensil failure, that is, step (3): if the similarity of a certain case in the candidate case set obtained in step (2) is high enough, and the similarity of other cases is low, then no case screening is required, and the Take it as the target fault case, and go to step (4); otherwise, call the case screening module, and use the case screening module to screen cases to get the target fault case, and then go to step (4).

Case screening is to retrieve a group of gas stove candidate cases, through the case screening auxiliary system, with the participation of users, the user further checks and confirms the gas stove failure symptoms, and eliminates the candidate cases that do not meet the requirements. Case screening strategy: If the similarity of the candidate case set is high enough, no case screening will be performed; if the similarity between the candidate case set and the target case is generally low, the candidate case set will be screened out from the candidate case set by means of man-machine combination. Example of user selection criteria. Just as there are many cases of ignition failure caused by gas stoves, all cases will be displayed in the candidate case set, such as battery case, high-voltage output wire case, electrode needle case, electrode needle-to-ignition bracket distance case, igniter circuit board case, The user can exclude according to the characteristics of the gas stove failure target cases. For example, if the battery is confirmed to be charged, the battery case that the gas stove fails to ignite due to insufficient battery voltage can be screened out.

4. Reuse and revision of maintenance plan

Maintenance plan reuse: the fault diagnosis server sends the target fault case back to the client through the Internet, and the maintenance personnel or users repair the kitchen utensils according to the maintenance plan in the target fault case. Learning; otherwise, go to step (5) for fault case correction.

In this embodiment, the reuse of the maintenance plan is to use the experience of solving the source case to solve the gas stove failure target case. The present invention adopts the method of troubleshooting the gas stove for reuse. Reapply the problem-solving methods in the gas stove failure case library to the gas stove failure target case. For example, in the case of gas stove fault ignition failure, the cause of battery failure is ruled out by case screening, and then other fault solutions saved in the gas stove fault case library are used, such as replacing new wires to eliminate faults, cleaning electrode needles, and replacing electrode needles. Adjust the distance from the ignition bracket to 3-4mm, and replace the igniter circuit board in time.

5. Fault case correction

Fault case correction: the fault diagnosis server calls the case correction auxiliary system. Through human-computer interaction, the technical personnel of the enterprise correct the fault case and obtain a new kitchenware fault case, and then send it back to the client computer through the Internet, maintenance personnel or users Repair the kitchen utensils according to the maintenance plan in the target failure case until the failure of the kitchen utensils is solved, then turn to step (6) for failure case study.

Sometimes the ignition failure is not caused by one fault, so one troubleshooting method cannot solve the fault, so the case modification module is used to modify the troubleshooting plan, combining two or more troubleshooting solutions until the fault is solved.

6. Fault case correction

Fault case learning: the fault diagnosis server adds new kitchen fault cases to the kitchen fault case library; in the case learning auxiliary system, the proportion of kitchen fault cases that solve actual faults is judged, and if it is less than a certain value, it is deleted.

After the case retrieval process, the CBR system selects a case that most closely matches as the solution recommended by the system. Under normal circumstances, this recommended solution is suitable. However, if the retrieved cases are not close enough to the gas stove fault target case and cannot meet the solution requirements of the gas stove fault target case, it is necessary to correct the case, that is, to use the retrieved gas stove fault candidate case as a template to modify to meet the The case of the gas stove failure target case. The problem characteristics described in the target case of gas stove failure are adjusted and modified to make it more accurately reflect the essential characteristics of the problem and approach the actual situation. In essence, it is a process of re-recognizing the case and solving the problem. The adjustment of the case is relatively complicated, and it should be completed under the guidance of experts in combination with domain knowledge and expert experience, and achieve the expected correction effect.

Case study is to evaluate and screen new cases after they are generated. Otherwise, the quality of cases in the case base will decrease, and the scale of the case base will expand rapidly, thereby reducing the reasoning efficiency of the system. The following strategy can be used to control its learning behavior: conduct value analysis on the new gas stove case formed in the diagnosis process, that is, calculate its similarity with all the old cases corresponding to the case base. Only when all the similarities are less than a given threshold, the new case is considered to have high value and is allowed to be added to the gas stove failure case database: otherwise, it will not be added. For the cases that already exist in the gas stove failure case library, the use value of each case can be calculated, that is, the number of times each case is cited, and combined with the time of the case, the reference frequency of each case can be obtained. When this value is less than When the value exceeds a certain threshold, it means that the case is not very meaningful, and the case can be deleted, so that the case base can be streamlined. In this way, each case in the case library can be guaranteed to meet the needs of specific users and maintain a certain usage rate.

Although the illustrative specific embodiments of the present invention have been described above, so that those skilled in the art can understand the present invention, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the art, As long as various changes are within the spirit and scope of the present invention defined and determined by the appended claims, these changes are obvious, and all inventions and creations using the concept of the present invention are included in the protection list.

Claims (2)

1. kitchen tools method for diagnosing faults based on reasoning by cases is characterized in that may further comprise the steps:

(1), the kitchen tools failure problems is submitted to

Maintenance personal or user are according to the disease million that shows when the prior fault kitchen tools, and input fault sign information on client computer is as the input of kitchen tools failure problems; Failure symptom information by internet transmission to the fault diagnosis server of enterprise;

(2), kitchen tools fault case retrieval

Fault diagnosis server calculates the secondary grading search method of similarity by the nearest neighbor method that adopts grey relational grade from its kitchen tools fault case storehouse, retrieve the candidate casebook similar to failure symptom information:

(3), kitchen tools fault case screening

If the similarity through a certain case in (2) candidate's casebook of obtaining of step is enough high, and other case similarities are all lower, then need not carry out the case screening, directly with it as the target faults case, entered for (4) step; Otherwise call case screening module, carry out the case screening, obtain the target faults case, entered for (4) step again by case screening module;

(4), maintenance program is reused

Fault diagnosis server is passed the target faults case on the client computer back by the internet, and maintenance personal or user keep in repair kitchen tools according to the maintenance program in the target faults case, if can solve the kitchen tools fault, entering for (6) step carries out fault case study; Otherwise changing for (5) step over to carries out the fault case correction;

(5), fault case correction

Fault diagnosis server calls case correction backup system, pass through interactive means, the technician of enterprise carries out the fault case correction, obtain new kitchen tools fault case, and then by passing back on the client computer by the internet, maintenance personal or user keep in repair kitchen tools according to the maintenance program in the target faults case, and till solving the kitchen tools fault, changing for (6) step over to carries out fault case study;

(6), fault case study

Fault diagnosis server joins new kitchen tools fault case in the kitchen tools fault case storehouse; At the case assisted learning system ratio that the kitchen tools fault case solves physical fault is judged, if less than certain value, then with its deletion.

2. the kitchen tools method for diagnosing faults based on reasoning by cases according to claim 1 is characterized in that, described kitchen tools fault case is:

Kitchen tools fault case C=(E, S, A, P), wherein: kitchen tools fault explanation tuple E={e ₁, e ₂..., e _rBe that a finite nonempty set is closed e _j(j=1,2 ..., r) descriptive information of expression;

The disease million tuple S={s of kitchen tools fault case ₁, s ₂, s _mBe that a finite nonempty set is closed, fault disease million is divided into qualitative fault disease million and quantitative fault disease million, fault disease million s _j={ f _j, d _j(j=1,2 ..., m), f _jBe qualitative fault disease million, be used for illustrating the order of severity of fault disease million facts, d _jIt is quantitative fault disease million;

The origin cause of formation tuple A={a of kitchen tools fault ₁, a ₂..., a _lBe that a finite nonempty set is closed a ₁, a ₂..., a _rBe a certain origin cause of formation wherein, a certain fault can be by l fault cause;

The maintenance program tuple P={p of kitchen tools fault case ₁, p ₂..., p _oBe that a finite nonempty set is closed; p ₁, p ₂..., p _rBe a certain solution wherein.

Images