#Java 8, 312 308 304 301 294 290 bytes

Java 8, 312 308 304 301 294 290 bytes

#Java 8, 312 308 304 301 294 bytes

s->{String r="",x="([AEIOU])",y="([BGNPRST])",z="0O1I2R3E4A5S6G7T8B9P";for(int c:s.getBytes())r+=c!=78&c!=85?z.charAt((c=z.indexOf(c)+(c<58?1:-1))<0?0:c):"";return s.matches("(("+x+y+")*"+x+"?)|(("+y+x+")*"+y+"?)|([0-9]*)")?r.replaceAll(x+"(?="+x+")","$1N").replaceAll(y+"(?="+y+")","$1U"):"";}

Try it here.

s->{                                   // Method with String parameter and String return-type
  String r="",                         //  Result-String
    x="([AEIOU])",y="([BGNPRST])",     //  Two temp Strings for the validation-regex
    z="0O1I2R3E4A5S6G7T8B9P";          //  And a temp-String for the mapping
  for(int c:s.getBytes())              //  Loop over the characters of the input-String
    r+=                                //   Append to the result-String:
       c!=78&c!=85?                    //    If the character is not 'N' nor 'U':
        z.charAt(                      //     Get the character from the temp-String `z`
         (c=z.indexOf(c)+              //      by getting the character before or after the current character
            +(c<58?1:-1))              //      based on whether it's a digit or not
             <0?0:c)                   //      and a 0-check to prevent errors on incorrect input like '!@#'
       :                               //    Else:
        "";                            //     Append nothing
                                       //  End of loop (implicit / single-line body)
  return s.matches("(("+x+y+")*"+x+"?)|(("+y+x+")*"+y+"?)|([0-9]*)")?
                                       //  If the input is valid
                                       //  (Only containing the vowels and consonants of `x` and `y`, without any adjacent ones. Or only containing digits)
    r                                  //   Return the result
     .replaceAll(x+"(?="+x+")","$1N")  //    after we've added 'N's if necessary
     .replaceAll(y+"(?="+y+")","$1U")  //    and 'U's if necessary
   :"";                                //  Or return an Empty String if invalid
}                                      // End of method

(([AEIOU][BGNPRST])*[AEIOU]?)|(([BGNPRST][AEIOU])*[BGNPRST]?)|([0-9]*)

#Java 8, 312 308 304 301 294 290 bytes

s->{String r="",x="([AEIOU])",y="([BGNPRST])",z="0O1I2R3E4A5S6G7T8B9P";for(int c:s.getBytes())r+=c!=78&c!=85?z.charAt((c=z.indexOf(c)+(c<58?1:-1))<0?0:c):"";return s.matches("(("+x+y+")*"+x+"?)|(("+y+x+")*"+y+"?)|\\d*")?r.replaceAll(x+"(?="+x+")","$1N").replaceAll(y+"(?="+y+")","$1U"):"";}

Try it here.

s->{                                   // Method with String parameter and String return-type
  String r="",                         //  Result-String
    x="([AEIOU])",y="([BGNPRST])",     //  Two temp Strings for the validation-regex
    z="0O1I2R3E4A5S6G7T8B9P";          //  And a temp-String for the mapping
  for(int c:s.getBytes())              //  Loop over the characters of the input-String
    r+=                                //   Append to the result-String:
       c!=78&c!=85?                    //    If the character is not 'N' nor 'U':
        z.charAt(                      //     Get the character from the temp-String `z`
         (c=z.indexOf(c)+              //      by getting the character before or after the current character
            +(c<58?1:-1))              //      based on whether it's a digit or not
             <0?0:c)                   //      and a 0-check to prevent errors on incorrect input like '!@#'
       :                               //    Else:
        "";                            //     Append nothing
                                       //  End of loop (implicit / single-line body)
  return s.matches("(("+x+y+")*"+x+"?)|(("+y+x+")*"+y+"?)|\\d*")?
                                       //  If the input is valid
                                       //  (Only containing the vowels and consonants of `x` and `y`, without any adjacent ones. Or only containing digits)
    r                                  //   Return the result
     .replaceAll(x+"(?="+x+")","$1N")  //    after we've added 'N's if necessary
     .replaceAll(y+"(?="+y+")","$1U")  //    and 'U's if necessary
   :"";                                //  Or return an Empty String if invalid
}                                      // End of method

(([AEIOU][BGNPRST])*[AEIOU]?)|(([BGNPRST][AEIOU])*[BGNPRST]?)|\\d*

#Java 8, 312 308 304 301 bytes

s->{String r="",x="([AEIOU])",y="([BGNPRST])",z="0O1I2R3E4A5S6G7T8B9P";for(int c:s.replaceAll("[NU]","").getBytes())r+=z.charAt((c=z.indexOf(c)+(c<58?1:-1))<0?0:c);return s.matches("(("+x+y+")*"+x+"?)|(("+y+x+")*"+y+"?)|([0-9]*)")?r.replaceAll(x+"(?="+x+")","$1N").replaceAll(y+"(?="+y+")","$1U"):"";}

-4 bytes by removing ^ and $ from the regex, because Java's String#matches already validates the entire String.
-4 bytes for a bug-fix (doesn't happen often that the byte-count decreases when I fix a bug in my code.. :D)

Try it here.

s->{                                   // Method with String parameter and String return-type
  String r="",                         //  Result-String
    x="([AEIOU])",y="([BGNPRST])",     //  Two temp Strings for the validation-regex
    z="0O1I2R3E4A5S6G7T8B9P";          //  And a temp-String for the mapping
  for(int c:s.replaceAll("[NU]","").getBytes())
                                        //  Loop over the characters of the input-String excluding all 'N' and 'U'
    r+=z.charAt(                       //   Get the character from the temp-String `z`
        (c=z.indexOf(c)+               //    by getting the character before or after the current character
           +(c<58?1:-1))               //    based on whether it's a digit or not
            <0?0:c);                   //    and a 0-check to prevent errors on incorrect input like '!@#'
                                       //  End of loop (implicit / single-line body)
  return s.matches("(("+x+y+")*"+x+"?)|(("+y+x+")*"+y+"?)|([0-9]*)")?
                                       //  If the input is valid
                                       //  (Only containing the vowels and consonants of `x` and `y`, without any adjacent ones. Or only containing digits)
    r                                  //   Return the result
     .replaceAll(x+"(?="+x+")","$1N")  //    after we've added 'N's if necessary
     .replaceAll(y+"(?="+y+")","$1U")  //    and 'U's if necessary
   :"";                                //  Or return an Empty String if invalid
}                                      // End of method

#Java 8, 312 308 304 301 294 bytes

s->{String r="",x="([AEIOU])",y="([BGNPRST])",z="0O1I2R3E4A5S6G7T8B9P";for(int c:s.getBytes())r+=c!=78&c!=85?z.charAt((c=z.indexOf(c)+(c<58?1:-1))<0?0:c):"";return s.matches("(("+x+y+")*"+x+"?)|(("+y+x+")*"+y+"?)|([0-9]*)")?r.replaceAll(x+"(?="+x+")","$1N").replaceAll(y+"(?="+y+")","$1U"):"";}

-4 bytes (308 → 304) for a bug-fix (doesn't happen often that the byte-count decreases when I fix a bug in my code.. :D)

Try it here.

s->{                                   // Method with String parameter and String return-type
  String r="",                         //  Result-String
    x="([AEIOU])",y="([BGNPRST])",     //  Two temp Strings for the validation-regex
    z="0O1I2R3E4A5S6G7T8B9P";          //  And a temp-String for the mapping
  for(int c:s.getBytes())              //  Loop over the characters of the input-String
    r+=                                //   Append to the result-String:
       c!=78&c!=85?                    //    If the character is not 'N' nor 'U':
        z.charAt(                      //     Get the character from the temp-String `z`
         (c=z.indexOf(c)+              //      by getting the character before or after the current character
            +(c<58?1:-1))              //      based on whether it's a digit or not
             <0?0:c)                   //      and a 0-check to prevent errors on incorrect input like '!@#'
       :                               //    Else:
        "";                            //     Append nothing
                                       //  End of loop (implicit / single-line body)
  return s.matches("(("+x+y+")*"+x+"?)|(("+y+x+")*"+y+"?)|([0-9]*)")?
                                       //  If the input is valid
                                       //  (Only containing the vowels and consonants of `x` and `y`, without any adjacent ones. Or only containing digits)
    r                                  //   Return the result
     .replaceAll(x+"(?="+x+")","$1N")  //    after we've added 'N's if necessary
     .replaceAll(y+"(?="+y+")","$1U")  //    and 'U's if necessary
   :"";                                //  Or return an Empty String if invalid
}                                      // End of method