JPH04274455A - Image forming device provided with self-diagnostic restoration system - Google Patents

Abstract

PURPOSE:To more accurately discriminate the presence or absence of a fault because it is discriminated while dealing with a fuzzy qualitative value, to diagnose the fault based on the fuzzy qualitative value of a parameter obtained by considering a deterioration amount previously stored and to accomplish more accurate diagnosis. CONSTITUTION:Since the fuzzy qualitative value of the parameter changed according to the deterioration amount of every previously set timing of diagnosing the fault is stored, the qualitative value is read out in each diagnosing timing. Then, condition data detected by sensors 1a, 1b and 1c is made to be the fuzzy qualitative value so that the presence or absence of the fault is decided according to the value. As a result, when the fault is found, the fault is diagnosed by setting the parameter value changed according to the deterioration amount as an initial value. The diagnosed result is compared with the condition data which is made to be the fuzzy qualitative value so as to identify the cause of the fault. Corresponding actuators 6a, 6b and 6c are actuated according to the identified cause of the fault so that the fault is restored.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a self-diagnosis and/or self-repair system. More specifically, in addition to utilizing artificial intelligence and knowledge engineering, which have been actively researched in recent years, we also employ fuzzy reasoning to self-diagnose the deterioration status and operating status of equipment, and self-repair as necessary. It is about a system that can

0002

[Prior Art] In the field of development of precision machinery and industrial machinery, artificial intelligence (AI) has recently been introduced in order to save labor in maintenance work and extend automatic operation.
Research into expert systems using so-called AI technology is actively being conducted. Some expert systems can self-diagnose whether a failure has occurred in a device and self-repair the failure that has occurred.

[0003] However, in conventional fault diagnosis systems using expert systems, (A) the knowledge is not versatile and fault diagnosis cannot be performed for various targets; (B)
) The inability to diagnose unknown faults; (C) As the target becomes more complex, the amount of knowledge required for fault diagnosis increases explosively, making it difficult to implement; (D) Difficult to acquire knowledge. , etc. limitations were pointed out.

More specifically, conventional automatic adjustment systems and fault diagnosis systems basically operate corresponding actuators based on the output of a certain sensor. In other words, a type of automatic adjustment and failure diagnosis was performed using a predetermined combination of sensors and actuators. Therefore, basically, a certain sensor corresponds to a specific actuator, and the relationship between the two is fixed. Therefore, (a) the relationship between sensor parameters and actuator parameters must be numerically specified. (b) For the reason (a) above, the relationship between the sensor parameters and the actuator parameters is strongly dependent on the object, and therefore lacks versatility and cannot be used for various objects. (c) The relationship between the parameters of each sensor or the parameters of each actuator is unrelated to control, and only simple control based on the relationship between the corresponding sensor parameters and actuator parameters can be performed, and failures that can be dealt with is limited in advance, and unknown failures cannot be handled. (d) Due to the reason in (c) above, there are problems such as the inability to predict the side effects on the parameters of other actuators that may occur due to the manipulation of the parameters of any actuator.

As described above, in conventional automatic adjustment systems and fault diagnosis systems, predictive failure A is performed based on set A of sensor A and actuator A, and predictive failure B is performed based on set B of sensor B and actuator B. The predicted failure C is a set of sensor C and actuator C.
Fault diagnosis was performed based on each independent sensor and actuator set, and fault repair was performed based on the diagnosis.

[0006]

[Problems to be Solved by the Invention] Therefore, as a technology related to the present invention, the applicants of the present application first took up an image forming apparatus as a target machine, and developed a new self-diagnosis and /or proposed a system that performs self-repair (Japanese Patent Application No. 2-252111 or 2003)
-252216).

The qualitative reasoning used in the self-diagnosis and/or self-repair systems proposed above is complete as a method for determining qualitative transitions from a group of equations and an initial state. On the other hand, since reasoning is done qualitatively, in other words, in a symbolic form of expression, there is an unavoidable fate that does not allow ambiguous expressions as a state expression of the target system (machine). There is. This is insufficient as a method for diagnosing and repairing failures by handling "ambiguous information" that is often encountered in maintenance activities, such as information such as "this may be normal or may be abnormal" regarding the machine's status. It is.

[0008] Furthermore, when considering integrating failure diagnosis using deterioration and failure history information of individual parts constituting a machine, it is necessary to use some other expression method for the qualitative inference proposed previously. Unless we consider an inference method that adds a method to handle ambiguous information by adding together the logic that has been used in the past, it is not possible to construct a more complete fault diagnosis and/or fault repair system.

[0009] Therefore, the inventor of the present application has combined fuzzy theory, which is a theory that deals with ambiguity mathematically, with the qualitative reasoning used in the above-mentioned self-diagnosis and/or self-repair systems. Invented advanced self-diagnosis and/or self-healing systems. The specific purpose of this invention is to develop reasoning that allows for ambiguity from a maintenance standpoint, and to be able to self-diagnose the state of the device using this reasoning, and to An object of the present invention is to provide an image forming apparatus having a repairable system.

0010

[Means for Solving the Problems] An image forming apparatus equipped with a self-diagnosis repair system, in which the image forming apparatus is expressed as a combination of a plurality of elements, and the behavior or attribute of each element and the connection relationship between each element. qualitative data qualitatively expressed using parameters, a target model storage means in which membership functions of each parameter and fault diagnosis knowledge are stored, and predetermined fault diagnosis timings for predetermined parts constituting the image forming apparatus a deterioration data storage means for predicting the amount of deterioration of the image forming apparatus and storing fuzzy qualitative values of parameters changed according to the predicted deterioration amount; a plurality of sensor means, a plurality of actuator means capable of changing the functional state of the image forming apparatus, and state data detected by the sensor means at a predetermined failure diagnosis timing is stored in a target model storage means. data conversion means for converting into fuzzy qualitative values using membership functions of parameters; failure determination means for determining the presence or absence of a failure by comparing the conversion output of the data conversion means with qualitative data stored in the target model storage means; In response to the failure determining means determining that there is a failure, the fuzzy qualitative value of the parameter at the fault diagnosis timing is read from the target model storage means, and the qualitative data and A fault diagnosis means that diagnoses the state of the image forming apparatus based on fault diagnosis knowledge and outputs the diagnosis result in an expression that includes ambiguity; and an image forming means that outputs fuzzy qualitative values converted by the data conversion means from the fault diagnosis means. cause identification means for identifying the cause of failure by comparing it with the state of the device; and selectively actuating the plurality of actuator means to eliminate the cause of failure based on the cause of failure being identified by the cause identification means; The invention is characterized in that it includes a repair means for causing the damage to occur.

[0011]

[Operation] The self-diagnosis repair system according to the present invention predicts the amount of deterioration of parts at each predetermined failure diagnosis timing, for example, every month, and changes in parameter values based on the predicted amount of deterioration are expressed as fuzzy qualitative values. is remembered in Then, at a predetermined fault diagnosis timing, the data detected by the plurality of sensor means is converted into fuzzy qualitative values using a parameter membership function,
The presence or absence of a failure is determined based on the fuzzy qualitative value. Furthermore, if it is determined that there is a failure, the fuzzy qualitative value of the parameter at that failure diagnosis timing is read out, and the failure diagnosis is performed using that value as an initial value. Diagnosis results are output with ambiguous expressions.

Furthermore, the fuzzy qualitative value converted from the detection output of the sensor means is compared with the results obtained in the fault diagnosis to identify the cause of the fault. That is, when a plurality of failure cause candidates are selected as a result of failure diagnosis, which failure cause candidate is more consistent with the actual state of the device is examined. Therefore, the actual fault condition is output more accurately. Then, the failure is repaired by operating a predetermined one of the plurality of actuators by the repair means.

[0013]

[Example] Fuzzy Qualitative Reasoning First, fuzzy qualitative reasoning, which is newly developed reasoning with ambiguity necessary for self-diagnosis, will be explained. (1) Fuzzy qualitative value In the qualitative reasoning used in the self-diagnosis and/or self-repair system related to the applicant's earlier application, the concept of quantitative space and qualitative value is used as a method for symbolically expressing the value of a variable. is used. A quantity space is a finite set in which a set of real numbers is symbolically expressed by landmarks, which are physically meaningful characteristic values, and intervals surrounded by these landmarks. Therefore, a qualitative value can only take either a boundary value or an interval value.

When considering a self-diagnosis and/or self-repair system using an ideal target machine as an example, it would be meaningful to determine whether a certain value is a boundary value or an interval value. However, in the real world, when we consider converting quantitative values obtained through measurement into qualitative expressions and making inferences based on them, we simply choose between boundary values and interval values. There is a risk that it may lack validity. This is because, as mentioned above, "ambiguous information" can be found in conservation activities, and situations may actually occur where the information may be a boundary value or an interval value.

[0015] Therefore, in this invention, we decided to consider the application of fuzzy theory. Fuzzy theory is a theory that deals with ambiguity mathematically, and a feature of the representation of sets in fuzzy theory is that the intermediate state of whether or not an element belongs to a set is
0.0 to 1.0 as the degree to which the element belongs to the set
It is expressed as a decimal value up to . By using this representation format, it is possible to represent intermediate states that cannot be represented using conventional sets. In fuzzy theory, a function that defines the degree (grade) that a certain element belongs to a certain set is called a "membership function."

By introducing the concept of a fuzzy set expressed using this membership function, it becomes possible to express qualitatively with added ambiguity. That is, in this invention, the value of a variable is expressed as a set of a conventional qualitative value and a degree (grade) belonging to that qualitative value. This type of expression will be referred to as a "fuzzy qualitative value." For example, if the value of a certain variable in the quantity space shown in Fig. 1 is expressed as a fuzzy qualitative value, for example, (Normal: 0.4, (Normal, nil)):
0.6) etc.

By using this method of expressing fuzzy qualitative values, it becomes possible to tolerate ambiguity when using sensor information for inference. In other words, in the previously proposed fault diagnosis system using qualitative inference, a certain range is determined as the normal value range, and if the quantitative value obtained from the sensor is within that range, that value is called the "normal value." Qualitative inference was made based on the fact that it was on the landmark of ``.

In contrast, in the present invention, a membership function is used for converting quantitative values obtained from sensors into qualitative expressions. When using membership functions, as shown in FIG. 2, "Normal values (hereinafter abbreviated as N)",
Larger than normal value (Normal, nil; hereinafter, N,
(abbreviated as nil)", "Less than normal value (0, Norm
A membership function such as "al; hereinafter abbreviated as 0, N" is determined in advance. Then, by mapping the measured values obtained from the sensor onto this space shown in FIG. 2, the measured quantitative value of the sensor is converted into a qualitative expression that allows ambiguity. (2) Operation rules for fuzzy qualitative values The algebraic operations for fuzzy qualitative values consist of algebraic operation rules for qualitative reasoning and grade calculations. This will be explained using a specific example. For example, Zf = Xf × Yf Xf = (N: 0.8, (N, nil): 0.2
) Yf = ((0, N) 0.7, N: 0.3), and the relationship between landmarks is explained using (Xf, Yf, Zf) = (N, N, N) as a specific example. do. First, all possible qualitative values of Zf are listed based on the relationship between the respective qualitative values of Xf and Yf (in this example, N and (N, nil) for Xf, etc.) and landmarks. At that time, the grade for the qualitative value of Zf is the smaller of the grades for the qualitative values of Xf and Yf. Specifically, Xf
Yf Zf N
×(0,N)=(0,N)
: 0.8, 0.7=0.7
N × N = N
: 0.8, 0
．． 3=0.3 (N, nil) × (0, N)=
(0,N)N(N,nil): 0.2, 0.7=
0.2 (N, nil) × N
=(N, nil): 0.
2, 0.3=0.2 Furthermore, when two or more grades for the qualitative value of Zf are determined, the maximum value among them is adopted. In the above example, in the first and third equations, Z
For the qualitative value of f (0, N), 0.7 and 0, respectively.
．． 2 can be found, the maximum value of which is 0.7 as Z
The grade is the qualitative value of f (0, n). Similarly, Zf
Also for the qualitative value N, in the second and third equations, 0.
There are two options, 3 and 0.2, so choose the maximum value of 0.3.

From the above calculation, the fuzzy qualitative value of Zf is Zf=((0,N):0.7,N:0.3,(
N, nil): 0.2). Furthermore, standardization is performed so that when the grades are added together, the sum becomes 1. For standardization, the grade of each qualitative value is set to 1.2 (however,
This is done by dividing by 1.2=0.7+0.3+0.2). As a result of normalization, Zf = ((0, N): 0.58, N: 0.25, (
N, nil): 0.17). (3) Inferential reasoning basically uses the propagation method. This is an algorithm that starts with a parameter whose value has already been determined and propagates the value of that parameter sequentially to other parameters using the relationships between the parameters to determine the parameters of the entire system.

[0020] The propagation method uses a method in which an undetermined parameter in a ternary or binary relationship is determined by the above-mentioned fuzzy operation rule using already determined parameters and the relationship therebetween. The specific method of inference will become clearer in the specific examples described below. System Configuration FIG. 3 is a block diagram showing the configuration of an embodiment of the present invention. This system includes a plurality of sensors 1a installed on the target machine (specifically, a small electrophotographic copying machine, etc.).
, 1b, 1c and a plurality of actuators 6a, 6b, 6c for changing the operating state of the target machine.

The plurality of sensors 1a, 1b, and 1c are each for detecting changes in the elements of the target machine or the related states between the machine elements caused by the operation of the target machine. The information taken in from the plurality of sensors 1a, 1b, 1c is amplified by the amplifier circuit 2, and then sent to the A/D
The analog signal is converted into a digital signal by the conversion circuit 3, and is provided to the digital signal/FQ value (fuzzy qualitative value) conversion section 11. Digital signal/FQ value converter 11
is a part that converts the digital signal given from the A/D conversion circuit 3 into a fuzzy qualitative value. That is, the digital signal is divided into qualitative values (for example, none (0), small (0, N), normal (N), large (N, nil),
any of the four symbols) and grade (0
．． It has a conversion function that expresses it by a numerical value from 0 to 1.0). By converting the signals given from the sensors 1a, 1b, and 1c into qualitative information represented by fuzzy qualitative values, functional evaluation can be performed more accurately, as will be described later.

The deterioration data calculation section 12 also includes a timer 19 for providing usage time data to the deterioration data calculation section 12.
, a short-term simulation section 13, a fault diagnosis section 14, a target model storage section 15, and a fault simulation section 16. The deterioration data calculation unit 12
This is the part for calculating changes in the target machine over time. The details of the calculation method will be described later. The short-term simulation section 13 is a section for simulating the current state of the target machine. The fault diagnosis section 14 is a section for performing functional evaluation using the fuzzy qualitative values provided from the digital signal/FQ value conversion section 11, identifying fault symptoms, and deriving the cause of the fault from the fault symptoms. The processing step (failure diagnosis step) for deriving the cause of the failure from the failure symptoms identified by the failure diagnosis unit 14 includes the self-diagnosis and/or self-repair disclosed in the above-mentioned specification of the applicant's earlier application. This is done based on the inference used in the system (non-fuzzy inference that does not use fuzzy inference).

[0023] The target model storage unit 15 stores a "substantive model" (see Table 1 to be explained later) in which the target machine is viewed from a physical perspective and is represented by parameters as a combination of a plurality of elements at the substantive level, and each parameter. "Parameter model" expressed as a join tree (see Figure 8, explained later),
The initial value of each parameter, the membership function of the functional parameter (as shown in Figure 2), the fuzzy qualitative value of the parameter of interest that changes depending on the amount of deterioration at each predetermined failure diagnosis timing, and the degree to which failure phenomena occur ( (see Table 1 to be described later), failure determination reference values, failure candidate knowledge, etc. are stored in advance. Target model storage unit 15
This knowledge stored in the digital signal/FQ
It is utilized when the value conversion unit 11, deterioration data calculation unit 12, short-term simulation unit 13, or failure diagnosis unit 14 performs processing. In addition, failure simulation section 1
6 is a part that performs failure simulation in collaboration with the short-term simulation unit 13 and the failure diagnosis unit 14.

One of the features of this embodiment and the present invention is that the system is equipped with two components: a deterioration data calculation section 12 and a short-term simulation section 13. Next, these two components will be explained in more detail. (1) Deterioration Data Calculation Unit The deterioration data calculation unit 12 picks up parts related to failure diagnosis among the parts constituting the target machine, and calculates the parameter values that change as a result of deterioration progressing over time for each part. Calculate the degree to which failure phenomena occur.

For parts related to failure diagnosis, fuzzy qualitative values of parameters and the degree of occurrence of failure phenomena are predicted and set in advance for each part or group of related parts. For example, Table 1
The deterioration data shown in is set. This deterioration data is stored in the target model storage section 15. The deterioration data calculation unit 12 obtains the elapsed period of the machine up to the present time by integrating the time data given from the timer 19, and by applying that period to Table 1, the deterioration data (fuzzy qualitative parameter of the parameter) of the component of interest is calculated. value and degree of occurrence of failure phenomenon).

[0026]

[Table 1]

(2) Short-term simulation section Short-term simulation (hereinafter referred to as SSIM) is a simulation for determining the current state of the target machine. The state of a machine is represented by a set of physical quantities that represent the attributes of the individual parts that make up the target machine. SSIM is performed on a parametric model (see FIG. 8) in which these physical quantities are related by qualitative equations. The inference method of SSIM uses the above-mentioned fuzzy qualitative inference. In addition, a propagation method is used as an algorithm for fuzzy qualitative inference. The method of propagation using this propagation method will be explained below.

The inference is based on constant parameters, parameters determined by values obtained from the sensor, and deterioration data calculation unit 1.
The process begins with the values of the parameters determined by 2 being determined. When propagating, first, (1)
Among the parameters in a ternary relationship (+, -, ×, etc.),
If two terms have already been determined, decide the remaining term.

(2) If one of the parameters in the binary relationship (=) is determined, the other is determined. The above propagation method is repeated until the values of all parameters are determined. As a result, the SSIM determines the state of the entire target machine, in other words, all parameter values.

Next, a procedure for inferring diagnosis results performed in the deterioration data calculation section 12, short-term simulation section 13, fault diagnosis section 14, and fault simulation section 16 will be explained. Fault diagnosis by fuzzy qualitative inference using deterioration information Referring to FIGS. 4, 5, and 6, inference for fault diagnosis is performed in the following steps.

[0031] In advance, the current state of the target machine is sensed by a sensor installed on the target machine (step S1), and the values of each parameter obtained by the sensing are converted into fuzzy qualitative values (step S2). Sensing of the parameter value in step S1 is of course performed by a sensor, if a sensor is provided (for example, as described later, the light amount Hl of a halogen lamp is measured by an AE sensor, etc.). )but,
If a sensor is not provided, a method may be adopted in which a serviceman or the like manually measures the condition of the target machine and inputs the measured value into the system. Further, the fuzzy qualitative value conversion in step S2 is performed by mapping the measured quantitative value onto the membership function (such as the one shown in FIG. 2) of each parameter stored in advance in the target model storage unit 15. .

(1) Calculation of deterioration state Next, the deterioration state up to the present time is calculated (step S3). In calculating this deterioration state, as described above, the time data given from the timer 19 is
2, and the integrated value is accumulated over a predetermined elapsed period T1, T2.
,..., apply to Table 1 and calculate the period T1
, T2, . . . and the degree of occurrence of the phenomenon are read out. In addition,
The timer 19 itself may be of a time accumulation type and its output value may be used as is, or the timer 19 may not be provided and the usage time may be manually input by a service person or the like.

Instead of Table 1 above, Table 1 may have a simple configuration in which only the fuzzy qualitative values of the parameters are recorded, and only the fuzzy qualitative values of the parameters may be calculated as the deterioration state. . (2) SSIM A simulation is performed using the parameter values obtained in (1) above as initial conditions, and the parameter values of the entire target machine after a predetermined period of time are determined.

Specifically, the values of the predetermined parameters obtained in step S3 are placed on the parameter model (
In step S4), the values of all parameters of the target machine are determined by propagation on the parameter model using the propagation method, and a current state model of the target machine is generated (step S5). (3) Failure determination Next, among the sensor values that have been converted into fuzzy qualitative values in advance in step S2, the sensor values of the functional parameters are checked to determine whether or not there is a failure (step S6).

Evaluation of the sensor values of the functional parameters is performed by comparing them with failure determination reference values stored in the target model storage section 15 in advance. If the sensor value of the functional parameter is determined to be normal, the process proceeds to step S7, and the parameter value on the parameter model obtained in step S5 and the actual parameter value sensed or inputted in step S1 are subjected to fuzzy qualitative analysis in step S2. The converted values are compared to determine the degree of agreement between the parameter model and the actual state of the target machine.

As a result of the matching degree evaluation in step S7, if they match, the process ends. Furthermore, even if they do not match, the machine continues to operate, giving priority to the function parameter determined to be normal in step S6. However, since the model and sensor values do not match and there is still a possibility of failure, a message is displayed on a display device, etc. (step S8).
). There are various possible message display formats, but for example, as mentioned above, if the functional parameters are normal but the models do not match, it is possible that the sensor that measures the functional parameters is malfunctioning. ``This may be the case.''

(4) Failure Diagnosis If a failure is determined in the above (3), failure candidates are derived from the failure symptoms (step S9). A plurality of failure candidates are stored in advance in the target model storage unit 15 (FIG. 3). Tracing is performed on the parameter model from the parameter values of the functional parameters, and a corresponding fault candidate is selected and derived from a plurality of fault candidates stored in advance. Alternatively, failure candidates may be determined by the reasoning explained in the applicant's earlier application (non-fuzzy reasoning that does not use fuzzy theory).

(5) SSIM For each fault candidate derived in step S9, the SSI
M is performed to create a failure model. More specifically, for each fault candidate, the fault conditions and the parameter values obtained in step S3 are placed on the parameter model as initial conditions (step S10), the parameter model is traced by the propagation method, and the target machine is A state model is generated (step S11).

[0039] In this way, a failure model is created. (6) Identification of failure causes Ranking and narrowing down of failure causes is performed based on the sensor information and the degree of occurrence of the phenomenon determined in step S3 described in (1). More specifically, the degree of matching of the model is evaluated based on the parameter values of the state model and the value obtained by converting the sensing value obtained in step S2 into a fuzzy qualitative value (step S12), and the evaluation result of the model is The degree of occurrence of the phenomenon is re-evaluated and prioritized (step S13).

[0040] Note that, omitting the process of step S13,
A simple method may be used in which the cause of the failure is identified only based on the degree of coincidence of the parameter values performed in step S12. Fault diagnosis is completed by the above reasoning procedure. Furthermore, upon completion of the failure diagnosis, work may be performed to add and/or modify the failure history information. and,
Thereafter, it is determined whether the failure can be repaired by parameter manipulation (step S14), and if it is repairable by parameter manipulation, the repair is performed (step S15).
If it cannot be repaired by parameter manipulation, for example, if a halogen lamp burns out in an electrophotographic copying machine, the process ends as it is because it cannot be repaired. The repair operation in step S15 is performed in the repair planning section 17, which will be described below.

Returning to FIG. 3, the remaining structural blocks will be explained. When there is a failure, the repair planning unit 17
deducing a repair plan for repairing the failure;
This is a component for deriving a repair plan. The inference of the repair plan and the derivation of the repair work are based on the previously proposed self-diagnosis and/or
Alternatively, similar to the reasoning in self-healing systems, non-fuzzy qualitative reasoning that does not use fuzzy theory is utilized.

The repair work output from the repair planning section 17 is converted into a digital signal by the symbol/digital signal conversion section 18. The converted digital signal is converted from a digital signal to an analog signal by a D/A conversion circuit 4, amplified by an amplifier circuit 5, and given to a plurality of actuators 6a, 6b, 6c. 6c is selectively activated to perform repair work. Explanation using a specific target machine as an example Configuration and status of a specific target machine Next, this system will be applied to an image forming apparatus, more specifically a small electrophotographic copying machine, as a specific target machine. This will be explained using an example.

FIG. 7 is an illustrative diagram of a small electrophotographic copying machine as a specific target machine. In FIG. 7, 21 is a photosensitive drum, 22 is a main charger, 23 is a halogen lamp for illuminating the original, 24 is a developing device, and 25 is a transfer charger. This specific target machine is provided with, for example, three sensors 1a, 1b, and 1c. That is, the sensor 1a is an AE sensor for measuring the amount of light incident on the photoreceptor drum 21, the sensor 1b is a surface potential sensor for measuring the surface potential of the photoreceptor drum 21, and the sensor 1c
is a densitometer for measuring the density of an image copied onto paper.

Although not shown in FIG. 7, three types of actuators are provided. That is, a main charging volume VR1 for changing the main charging voltage of the photoreceptor drum 21, a lamp volume AVR for controlling the light amount of the halogen lamp 23, and a transfer voltage between the photoreceptor drum 21 and copy paper. Transfer volume V
Three volumes called R2 are provided as actuators.

By the way, considering the electrophotographic copying machine shown in FIG. 7 from a physical perspective, the electrophotographic copying machine is expressed at a physical level as a combination of a plurality of elements, and the behavior and attributes of each element and the relationships between each element are explained. If the coupling relationship is qualitatively expressed using parameters, it will be as shown in Table 2. The expression format shown in Table 2 will be referred to as a "substantive model." Furthermore, the representation shown in FIG. 8 in which the entity model is abstracted and expressed as a connection tree of each parameter will be referred to as a "parameter model."

[0046] The "substantive model" and the "parameter model" will be collectively referred to as the "object model." The "target model" is qualitative data that is generally common to image forming apparatuses and is also utilized for failure repair, which will be described later. The contents of this entity model and parameter model are as follows:
It is stored in the target model storage unit 15 (see FIG. 3).

[0047]

[Table 2]

In the physical model shown in Table 2 or the parametric model shown in FIG. 8, the parameters Hl, D, Vn, β, Vb, γ0, ζ, As
p, sensor parameters obtained by the sensor X, Vs,
Os and parameters that may deteriorate are each set by predicting the relationship between the elapsed period and the deterioration state, and are stored in the target model storage unit 15. For example, in the case of the parameter Hl, this relationship is as shown in Table 3.

[0049]

[Table 3]

According to the relationship shown in Table 3, the fuzzy qualitative value of Hl and the degree of occurrence of the failure phenomenon (more specifically, in this example, The degree is the degree to which the phenomenon of halogen lamp disconnection occurs.) can be immediately determined. Next, several examples will be explained based on the above explanation. Example 1: Example of normal condition after 1 month (T1) since new product The values obtained by the sensors 1a, 1b, and 1c are converted into fuzzy qualitative values in advance.

Specifically, the amount of light measured using the sensor 1a is used as the value of the parameter X. Further, the sensor 1b measures the surface potential Vs after exposure, and the sensor 1c measures the toner density Os on the output paper. Then, when each measured parameter value is mapped onto a fuzzy qualitative quantity space set for each parameter as shown in FIG. 2, the following sensor parameter values are obtained, for example.

[0052]X=((0,N):0.1,N:0.9)
Vs = (N: 1.0) Os = ((0, N): 0.1, N: 0.9) (1) Examine the current state of deterioration and its influence Integrate the time information given from the timer 19 Therefore, the age of the machine T
=T1 is found. By applying the elapsed period T1 to Table 3, Hl=(N:1.0) p(HlCut)=0.9 is obtained. However, P(HlCut)=0.9 means that the degree of occurrence of the phenomenon of halogen lamp disconnection is 0.9 times.

(2) Using SSIM, infer the current state of the entire machine using the results of (1).
l = (N: 1.0)) and the initial condition (in this case, the deterioration state is not calculated for parameters other than Hl, so all remaining parameters are set to (N: 1.0)). Perform SSIM based on SSIM results,
The parameter values are as follows: Total parameter value = (N
:1.0).

H1=(N:1.0) γ0=(N:1
．． 0)D=(N:1.0)Vt=
(N: 1.0) β = (N: 1.0) ζ
=(N:1.0)Vn=(N:1.0)
Asp=(N:1.0)Vb=(N:1.0) (3) Failure determination The following knowledge is used as failure determination criteria. This knowledge is stored in advance in the target model storage unit 15 (see FIG. 3).

(a) Machine function evaluation N value ≧0.5 → Normal → Compare model agreement > 0.5 →
Abnormality→Failure diagnosis (b) Comparison of model and actual measured value Matching degree of machine≧0.5→No failure>0.5→Display “Possibility of sensor abnormality” etc. In the case of this example, (a) Regarding the functional evaluation of the machine, it is sufficient to compare the Os value, so the measured value of Os is converted into a fuzzy qualitative value Os = ((0, N): 0.1, N: 0.9
) to compare. In this case, since the N value is 0.9, N≧0.5, and (a) the functional evaluation is normal.

Next, (b) a comparison is made between the model and actual measured values. Each sensor value is compared with the model resulting from the above (2), and the degree of agreement between the two is determined based on the following conditions. Matching degree = max (min (grade of each item)) As a result, the following result is obtained.

Sensor value
Model value Min value X: (0, N
) 0.1
0 N 0.9
1.0 0.9

max0.9 Vs:
N 1.0 1.0
1.0
ma
x1.0 Os: (0, N) 0.1
0 0
N 0.9 1.0
0.9
ma
x0.9 Also, the degree of coincidence of the entire machine is as follows.

[0058] Matching degree of the entire machine = average matching degree of each sensor = (0.9
+1.0+0.9)/3=0.93 When determining the degree of agreement for the entire machine, instead of taking the average of the degrees of agreement of each sensor, the minimum value is taken, and the degree of agreement is calculated under more severe conditions. You can ask for it. If the above failure determination criterion (b) is applied here, the average degree of coincidence exceeds 0.5, so it is determined to be normal. That is, both (a) and (b) are normal and can continue to be used. In the above example, when (a) the functional evaluation of the machine is normal, and (b) the comparison result between the model and the actual measurement value is abnormal (in FIG. 5, when proceeding from step S7 to step S8), considers the possibility that there is an abnormality in the sensor that measures the parameter related to functional evaluation, that is, Os, and displays this fact. Example 2: Same as example 1, in the first month (T1) after the new product,
Example of failure: Assume that the values of sensor parameters X, Vs, and Os are as follows.

X = (0:0,9, (0,N):0.1)Vs=(
N: 0.1, (N, nil): 0.9) Os = (N: 0
．． 1, (N, nil): 0.9) Also, the parameter Hl
(1) As a result of examining the current state of deterioration and its effects,
From Table 3, Hl=(N:1.0) P(HlCut)=0.9 can be found.

Under such conditions, (2) SSIM
All parameter values of the model obtained = (N: 1
．． 0). (3) Failure determination (a) Functional evaluation of the machine From the fuzzy qualitative value of Os, N=0.1 and N<0.
5, indicating a malfunction. Therefore, it is diagnosed as a failure.

(4) Derivation of Fault Candidates The following information is stored in advance in the target model storage unit 15 as fault candidates. HlCut: Hl = 0 HlOut: Hl = (0, N) VtOut: Vt = (0, N) PaperOut: ζ = (0, N) VbOut: Vb = (N, nil) TonerOut: γ0 = (0, N) MCOut:V
n=(0,N) Note that Out means defective.

[0062] Therefore, the functional abnormality in which Os obtained in (3) above is large, that is, Os = (N, nil), is shown in Fig. 8.
Figure 9 is obtained by tracing on the parameter model shown in Figure 9. In Figure 9, parameters with upward arrows have values that may have changed to (N, nil), and parameters with downward arrows have values that may have changed to (0, N) or (0). Parameters that may have changed or are not marked with an arrow have their values unchanged. As a result, HlCut and HlOut are selected as failure candidates from among the failure candidates.

(5) Execute SSIM on the two fault candidates obtained in (4) above, and infer the fault state at that time. That is, SSIM is performed on HlOut to obtain the following two models. Hl = ((0, N): 1.0) D = (N: 1.0) X = (0: 1.0)
or ((0,N):1.0)β =
(N: 1.0) Vs = ((N, nil): 1.0) Yn = (N
:1.0) Vb = (N:1.0) γ0 = (N:1.0) Ds = ((N, nil):1.0) Vt = (N
: 1.0) ζ = (N: 1.0) Os = ((N, nil): 1.0) Asp = (N:
1.0) Sp = ((N, nil): 1.0) There are two types because two values of X occur. This means that the value of X is Hl minus D, but H
Since l = (0, N) and D = (N), if we subtract the normal from those smaller than the normal, the result of the fuzzy operation is either the one smaller than the normal remains, or 0
This is because there are two possible cases:

Furthermore, when SSIM is applied to HlCut, the following two models are obtained. Hl = (0:1.0) D = (N:1.0) X = (0:1.0)
or ((0,N):1.0)β =
(N: 1.0) Vs = ((N, nil): 1.0) Yn = (N
:1.0) Vb = (N:1.0) γ0 = (N:1.0) Ds = ((N, nil):1.0) Vt = (N
: 1.0) ζ = (N: 1.0) Os = ((N, nil): 1.0) Asp = (N:
1.0) Sp = ((N, nil): 1.0) (6) Identification of failure cause Degree of agreement between the model and sensor and degree of occurrence of the phenomenon 2
From this point on, rank and narrow down the causes of failure.

(i) Degree of agreement between model and sensor A.
HlOut X=0 B.
HlOut X=(0,N)
: 0.9
X: 0.1 Vs
: 0.9
Vs: 0.9 Os:
0.9 Os
: 0.9 Overall 0.9
whole
0.63C. HlCut X=0
D. HlCut X=(0,N
) X: 0.9
X: 0.1
Vs: 0.9
Vs: 0.9
Os: 0.9
Os: 0.9 Overall 0.9
The degree of agreement between the sensor and the model for each fault candidate obtained as a result of performing SSIM with a total of 0.63(5) is determined as described above.

The model obtained in (5) has a failure cause (for example, HlOut as HlOut, Hl=((
0, N): 1.0)) is set, and the influence of the cause of the failure on other parameters is traced. Therefore, the higher the degree of agreement between the value indicated by the sensor of interest and the corresponding parameter value on the model, the closer the current state of the device is to that model.
In other words, it means that the cause of failure assumed to derive the model is likely to be the current cause of failure.

In this example, HlOut, HlCu
The degree of agreement between the t model and the sensor is 0.9, and both are considered to be the cause. For the above four models,
The ranking is as follows. Ranking 1. A, C: 0.9 2. B, D: 0.63 In response to the case where multiple causes are derived in this way, the priority order of the failure causes is stored in the target model storage unit 15 in FIG. You may also do this. In this embodiment, the following operations are performed to further narrow down the search results.

(ii) Degree of occurrence of phenomenon Taking into account the degree of occurrence of the phenomenon derived at the stage of examining the influence of deterioration in (1), multiply the value indicating the degree of coincidence by the value indicating the degree, as shown below. Narrow down your search accordingly. p(HlOut)=1 p(HlCut)=0.9 Rank 1. A: 0.9 x 1.0 = 0.9
Normalize to 1.0 2.
C: 0.9 x 0.9 = 0.81 normalized
0.9 3. B: 0.63×1.
0=0.63 normalized to 0.7
4. D: 0.63 x 0.9 = 0.57 The higher the calculated value of 0.63 (degree of matching) x (degree) after normalization, the higher the priority, so it can be seen that HlOut is the most suspicious. Therefore, the repair will be performed by changing the light intensity of the halogen lamp.

Example 3: Continued use, nth month (nT1)
Fuzzy qualitative values for each parameter are determined from the actual measured values obtained from the sensor. X=(N:0.8,(0,N):0.2)Vs=(N:
0.9, (N, nil): 0.1) Os = (N: 0.9
, (N, nil): 0.1) (1) Examine the current state of deterioration and its effects By applying nT1 to Table 3, H
l=(N:0.915, (0,N):0.085)p(
HlCut)=1.24 is obtained. (2) Using SSIM, infer the current state of the entire machine using the results of (1). From the results of (2) above and the initial conditions, Hl = (N: 0.915, (0, N): 0.085 )D
= (N: 1.0) β = (N: 1.0) Vn = (N: 1.0) Vb = (N: 1.0) γ0 = (N: 1.0) Vt = (N: 1 .0) ζ = (N: 1.0) Asp = (N: 1.0) When SSIM is performed under these conditions, the following results are obtained.

Hl=(N:0.915, (0,N):0.085)D
= (N: 1.0) X = (N: 0.915, (0, N): 0.085)
or (N: 0.915, 0: 0.085) β = (N: 1.0) Vs = (N: 0.915, (N, nil): 0.085
) Vn=(N:1.0) Vb=(N:1.0) γ0=(N:1.0) Ds=(N:0.915, (N, nil):0.085
) Vt = (N: 1.0) ζ = (N: 1.0) Os = (N: 0.915, (N, nil): 0.085
) Asp=(N:1.0) Sp=(N:0.915, (N, nil):0.085
) From this result, it can be seen that the image may become darker (Os increases) as an effect of Hl deterioration. Also,
From the changes in each parameter value, it is generally possible to infer the combined effects of deterioration of multiple parts and, as a result, failures that occur in a cascading manner. (3) Failure determination (a) From the fuzzy qualitative value of machine functional evaluation Os, N=0.9, N≧0
．． Since it is 5, it is normal.

(b) Comparison of model and actual measured values Each sensor value is compared with the model resulting from the above (2) to determine the degree of agreement between the two. A. X=(N:0.915, (0,N):0.
085) Model Matching degree of each sensor X: 0.8 Vs: 0.8 Os: 0.8 Overall 0.87 B. X=(N: 0.915, 0:0.085)
The degree of agreement of each sensor in the model X: 0.8 Vs: 0.9 Os: 0.9 Overall 0.87 Therefore, it can be seen that it seems to match model A or B. In conclusion, the function is normal, and although the machine has deteriorated to the model A or B condition, it is determined that it can be used normally.

Example 4: When a failure occurs under the same conditions as Example 3, the following data is obtained as a sensor value. X = (N: 0.1, (0, N): 0.9) Vs = (
(N: nil): 1.0) Os = ((N: nil): 1.0) Also, from Table 3, the same results as (1) and (2) in Example 3 above were obtained. do. (3) Failure determination (a) Functional evaluation of the machine From the fuzzy qualitative value of Os, N=0 and N<0.5, which is abnormal. Therefore, it is judged to be a failure. (4) Derivation of fault candidates The fault candidates stored in advance are the same as those described in (4) of Example 2. Therefore, by tracing Os=(N, nil) on the parameter model shown in Figure 8, HlCu
t HlOut is obtained as a failure candidate.

(5) Fault simulation HlCut using SSIM
, HlOut, we obtain the same result as in Example 2 (5). In addition, in this example, for the sake of simplicity, the explanation is made assuming that only the exposed portion deteriorates, so the failure simulation result of this example 4 is the same as that of example 2.
Generally, the results will not be the same because other parameter values change due to the effects of deterioration. For example, if ζ=(N: 0.8, (0, N): 0.2) as an effect of deterioration of the output section, the following result is obtained for HlOut.

Hl=((0,N):1.0)D=(
N: 1.0) X = ((0, N): 1.0) or (0: 1.0) β = (N: 1.0) Vs = ((N, nil): 1.0) Vn = (N: 1.0) Vb = (N: 1.0) γ0 = (N: 1.0) Ds = ((N, nil): 1.0) Vt = (N: 1.0) ζ = (N: 0.8, (0, N): 0.2) Os=(
(0, N): 0.2, N: 0.2, (N, nil): 0
．． 8) Normalize to ((0, N): 0.18, N: 0.18, (
(N, nil): 0.8) Asp = (N: 1.0) Sp = ((N, nil): 1.0) (6) Identification of failure cause The degree of agreement between the model and the sensor and the degree of occurrence of the phenomenon 2
From this point on, rank and narrow down the causes of failure.

(i) Degree of agreement between model and sensor A.
HlOut X=0 B.
HlOut X=(0,N)
: 0.9
X: 0.1 Vs
: 1.0
Vs: 1.0 Os:
1.0Os
: 1.0 Overall 0.9
7 whole
0.7 C. HlCut X=0
D. HlCut X=(0,N)
X: 0.9
X: 0.1
Vs: 1.0
Vs: 1.0Os
: 1.0
Os: 1.0 overall
0.97 overall
0.7 rank 1. A, C: 0.97 2. B, D: 0.7 (i) Degree of occurrence of phenomenon p(HlOut) = 1 p(HlCut) = 1.24 Ranking 1. C: 0.97×1.24=1.2
0 Normalized 1.0 2.
A: 0.97 x 1.0 = 0.97 normalized
0.81 3. D: 0.7 ×
1.24=0.87 normalized to 0.73
4. B: 0.7 x 1.0 = 0.
7 Normalized to 0.58 Therefore, C is the most suspicious. Also, at least HlCut is questionable. However, since HlCut means that the halogen lamp is burnt out and repair is impossible, no repair is performed.

Note that whether or not repair is possible is determined by dividing the cause of the failure into repairable causes and the non-recoverable causes of failure and storing them in the target model storage unit 15 in FIG. can do. Furthermore, if the cause of the failure that cannot be repaired is determined with the highest priority, this may be displayed on the display unit.

[0077]

Effects of the Invention As described above, the present invention includes a target model storage means, a deterioration data storage means, a plurality of sensor means,
Since the configuration includes a plurality of actuator means, data conversion means, failure determination means, failure diagnosis means, cause identification means, and repair means, failures can be detected taking into account deterioration data of the components constituting the image forming apparatus. It is possible to provide an image forming apparatus that can automatically perform diagnosis, save labor in repair work, and realize long-term automatic operation.

Furthermore, since the deterioration data is predicted with parameter values that change according to the amount of deterioration and is stored in advance as fuzzy qualitative values, failure diagnosis that takes deterioration data into consideration can be performed extremely quickly with a relatively simple program. . Furthermore, when diagnosing a fault, the cause of the fault is identified while dealing with fuzzy values that have ambiguity, rather than making an either-or decision, making it possible to perform a more accurate fault diagnosis than in the past. Furthermore, it is possible to provide a highly autonomous image forming apparatus that automatically performs appropriate repair according to the diagnosis results.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing an example of a qualitative quantity space.

FIG. 2 is a diagram illustrating an example of a membership function that converts sensor information into fuzzy qualitative values.

FIG. 3 is a block diagram showing the configuration of an embodiment of the present invention.

FIG. 4 is a part of a flowchart showing an inference procedure for fault diagnosis in this embodiment.

FIG. 5 is a part of a flowchart showing an inference procedure for fault diagnosis in this embodiment.

FIG. 6 is a part of a flowchart showing an inference procedure for fault diagnosis in this embodiment.

FIG. 7 is an illustrative diagram of a small electrophotographic copying machine as a specific target machine.

FIG. 8 is a diagram showing a parameter model of the electrophotographic copying machine according to this embodiment.

FIG. 9 is a diagram illustrating a state in which the cause of a functional abnormality in which Os is large is traced on a parameter model in order to derive a failure candidate.

[Explanation of symbols]

1a, 1b, 1c Sensor 6a, 6b, 6c Actuator 11
Digital signal/FQ value conversion section 12 Deterioration data calculation section 13 Short-term simulation section

Claims (1)

[Claims] 1. An image forming apparatus equipped with a self-diagnosis repair system, wherein the image forming apparatus is expressed as a combination of a plurality of elements, and the behavior or attribute of each element and the connection relationship between each element are expressed using parameters. a target model storage means in which qualitative data qualitatively expressed, membership functions of each parameter, and failure diagnosis knowledge are stored; a deterioration data storage means storing a fuzzy qualitative value of a parameter predicted and changed according to the predicted deterioration amount; A sensor means, a plurality of actuator means capable of changing the functional state of the image forming apparatus, and membership of parameters stored in a target model storage means based on state data detected by the sensor means at a predetermined failure diagnosis timing. A data conversion means for converting into a fuzzy qualitative value using a function, a failure determination means for determining the presence or absence of a failure by comparing the conversion output of the data conversion means with qualitative data stored in a target model storage means, In response to determining that there is a fault, reading out fuzzy qualitative values of the parameters at the fault diagnosis timing from the target model storage means;
A failure diagnosis means and a data conversion means that diagnose the state of the image forming apparatus based on the qualitative data and failure diagnosis knowledge stored in the target model storage means using the value as an initial condition, and output the diagnosis result in an expression containing ambiguity. cause identification means for identifying the cause of the failure by comparing the fuzzy qualitative value converted by the image forming apparatus with the status of the image forming apparatus output from the failure diagnosis means; An image forming apparatus comprising: repair means for selectively operating the plurality of actuator means to eliminate the cause.